United States	Policy , Planning,	EPA 230-R-95-007
Environmental Protection	And Evaluation	September 1995
Agency	(2127)
l>EPA Integrating The Environment
And The Economy:
Proceedings Of June 1994
Association Of Environmental
And Resource Economists

Integrating the Environment and the Economy:
Sustainable Development and
Economic/Ecological Modelling
1994 Association of Environmental and Resource
Economists Workshop
This report has been approved for publication as an EPA document. Mention of products or
services should not be construed as approval or endorsement. The views expressed are those of
the authors and do not necessarily reflect those of EPA.
Boulder, Colorado
June 5-6,1994

The 1994 AERE Workshop was held June 5th and 6th in Boulder, Colorado. The topic was
Integrating the Environment and the Economy: Sustainable Development and
Economic/Ecological Modelling. Keynote addresses were given by John Hartwick of
Queens University and Michael Toman from Resources for the Future. Michael's talk was
entitled "Neoclassical Economics and Sustainability," and was based on papers by Michael;
and Michael, John Pezzy and Jeffrey Krautkraemer. John spoke on "Sustainability and
Constant Consumption Paths in Open Economies with Exhaustible Resources."
Session topics included Sustainability : Some Basics; Sustainability: Extensions and Issues;
Issues in Environmental Accounting; and Economic/Ecological Modelling and Ecosystem
There were almost ninety participants, and my perception is that most found the workshop
either productive, enjoyable, or both. I both enjoyed it and learned a lot. The weather was
great, the hotel nice, and the food good. The presentations were great. Those of you who
were not there missed all of the site-specific amenities, but can still enjoy the papers. I
recommend them.
The papers by Bishop and Woodward; and Hrubovack, LeBlanc, and Eakin are revisions of
the manuscripts that were presented at the AERE Workshop. Due to copyright considerations,
only abstracts are included for the following papers: Pezzey; Toman, Pezzey, and
Krautkraemer; Gottfried, Wear, and Lee; Silvestre; and Albers.
Neither the conference nor this EPA volume would have been possible without generous
sponsors. These include the Environmental Protection Agency, the National Oceanic and
Atmospheric Administration, the U.S. Department of Agriculture, and the University of
Colorado. Thanks also goes to the AERE Workshop committee members, Betsy David, Anne
Grambsch, Mary Jo Kealy, Bob Leeworthy, Michael LeBlanc, and Kathy Segerson. Great on-
site help was provided by four Ph.D. students in the Economics Department at the University
of Colorado. Kate Carson, Kathleen Greer, Amanda Lee, and Charles Rossmann; each is
specializing in environmental economics.
Edward Morey

Papers presented at the 1994 Workshop Conference
Michael Toman
John Hartwick
John Pezzey
Peter Kennedy
Joaquim Silvestre
Richard Bishop
Richard Woodward
Anne Grambsch
Gregory Michaels
Kelly Eakin
Daniel Hellerstein
James Hrubovcak
Michael LeBlanc
Nancy Bockstael
Jackie Geoghegan
Robert Gottfried
Robert Lee
David Wear
Heidi Albers
David Simpson
Roger Sedjo
John Reid
Neoclassical Economics arid Sustainability
Sustainability and Constant Consumption Paths
in Open, Economies with Exhaustible Resources
The Optimal Sustainable Depletion of Non-
Renewable Resources
Rethinking Sustainability
An Efficiency Argument for Sustainable Use
Intergenerational Welfare Economics and
Environmental Policy
The United Nations Integrated Environmental
and Economic Accounting system: An
Environmental Economics Perspective
Environmental Accounting: The Impact of
Some Issues Related to Ecological and
Economic Modeling of Ecosystem "Landscapes"
Landscapes, Ecosystem Value, and
People and Parks: Economic Managment of
Khao Yai National Park Thailand
Valuing Biodiversity for Use in Pharmaceutical

ENR 93-14 REV
Michael A Toman, Resources for the Future
John Pezzey, Department of Economics, University College London
Jeffrey Krautkraemer, Department of Economics Washington State University
Resources for the Future
1616 P Street, NW
Washington, DC 20036
Tel: 202-328-5091
Fax: 202-939-3460
October 1993, revised April 1994
©1993 Resources for die Future
All rights reserved. No portion of this paper may be reproduced
without permission of the author.
Discussion papers are materials circulated for information and discussion.
They have not undergone formal peer review as have RFF books and studies.
This paper is forthcoming in the Handbook of Enviromental Economics, edited by Daniel Bromley and
published by Blackwell. The authors are grateful to Geir Asheim, Edward Barbier, Richard Howarth,
David Pearce and Tom Tietenberg for helpful advice during the preparation of this paper. The paper also
benefited from the assistance of Mary Elizabeth Calhoon and Kay Murphy. Pezzey's research was
supported by the UK Centre for Economics and Environmental Development and the Economic and Social
Research Council.

The issue of "sustainability" figures prominently in contemporary discussions of natural
resource and environmental management and economic development. However, the concept is
not easily defined and is interpreted differently by economists, ecologists, philosophers, and
others. Even among economists there are significant differences of interpretation. Some treat
sustainability as not much more than another way of espousing economic efficiency in the
management of services derived from the natural endowment. Others claim that conventional
economic efficiency criteria are inadequate for addressing sustainability concerns.
Our aims in this paper are to identify the issues that seem to be most salient in formal
economic analysis of sustainability, and to review economic growth theory that bears on these
issues. In the latter effort we focus mostly on literature within the methodological mainstream of
neoclassical economics, though the studies do not always maintain all the common assumptions of
neoclassical theory. We first draw together arguments from economics, ecology and philosophy
to briefly describe what seem to be the most important issues m addressing sustainability. Armed
with this characterization, we then review several categories of studies related to economic
advance, natural resource use, and environmental preservation over time. We include both
representative-agent models and overlapping-generations models in the review. The concluding
section of the paper summarizes our discussion and offers an overall assessment of the literature.

Economics and "Sustainability": Balancing Trade-offs
and Imperatives
Michael A. Toman
ABSTRACT. The concept of '"sustainability"
has been increasingly invoked in scholarly and
public policy debates. Discussion has been
hampered, however, by uncertainty and lack of
uniformity in the meaning of sustainability. This
paper seeks to identify some common ground
among economists, ecologists, and environmen-
tal ethicists. Two issues seem salient: require-
ments for intergenerational equity and the defi-
nition of "social capital" to be provided to
future generations. A concept of "safe minimum
standard, " which has received at least some
recognition in the ecology, philosophy, and eco-
nomics literatures, may provide the beginnings
of a common ground for debate about sus-
tainability. (JEL Q2)
The concept that use of natural re-
sources, environmental services, and eco-
logical systems somehow should be "sus-
tainable" has become one of the most
widely invoked and debated ideas in the
area of resource and environmental man-
agement. It was a basic theme in the 1992
"Earth Summit," the United Nations Con-
ference on Environment and Development
(UNCED), and in the World Bank's 1992
World Development Report on environment
and development. It is an issue discussed
not just in professional journals but also in
newspaper articles and in basic textbooks
(see, e.g., Pearce and Turner 1990 and Tie-
tenberg 1992). It is a principle behind the
founding of a professional organization, the
International Society for Ecological Eco-
nomics, many of whose members question
the sufficiency or even the validity of con-
ventional economic approaches to resource
and environmental management problems.
Despite the frequency with which the
term is invoked, the concept of sustainabil-
ity y remains surprisingly ambiguous. It is
clear from examining various usages of the
term that writers have very different mean-
ings in mind.1 For example, the use of the
term in the 1992 World Development Re-
port seems to refer primarily to the applica-
tion of existing neoclassical principles of
efficient resource and environmental man-
agement in developing countries. This is
very different than the ideas expressed by
Herman Daly (see, e.g., Daly 1990, 1991),
who argues that use ("throughput") of
energy and materials must be sharply cur-
tailed to avoid ecological catastrophe. Sus-
tainability also is interpreted very differ-
ently by many economists, who see the
natural environment as one of many fungi-
ble assets that can be deployed in satisfying
human demands, and by many ecologists
and ethicists, who express 'greater concern
for both ecological integrity and the inter-
ests of future generations (compare Ehrlich
1989 and Solow 1993a, 1993b, for example).
The goal of this paper is to provide some
vocabulary and grammar that may be useful
for this ongoing debate among economists,
ecologists, and ethicists. We begin, as do
many others, with the statement about sus-
tainability from the report of the "Brundt-
Senior Fellow, Resources for the Future.
Earlier versions of this paper were presented at,
meetings of the International Society for Ecological
Economics and the American Economic Association,
and at seminars at the World Bank, the Agency for
International Development, and the University of
Maryland. I owe a large debt to Pierre Crosson, Bryan
Norton, and John Pezzey, whose insights played a
substantial role in clarifying my understanding of the
issues raised in the paper. I also appreciate helpful
conversations with Geir Asheim, Doug Bohi, Allen
Kneese, and Jeff Krautkraemer, and perceptive com-
ments by Tom Tietenberg, Scott Gordon, Tim Bren-
nan, arid an anonymous referee on earlier drafts.
* See also Pezzey (1989) and Pearce, Markandya,
and Barbier (1989), who catalogue scores of some-
times vague and conflicting sustainability definitions.
Dixon and Fallen (1989) discuss how sustainability has
been transformed from a condition on steady-state
management of specific resources to an expression of
broad ecological concerns.
Lend Economics • November 1994 1 70(4): 399-413
Toman, Michael, "Economics and 'Sustainability': Balancing Trade-Offs and
Imperatives." LAND ECONOMICS, Volume 70, Number 4 (November, 1994). Reprinted by
permission of The University of Wisconsin Press.

Land Economics
November 1994
land Commission, " the World Commis-
sion on Environment and Development
(WCED). That report described sustainable
development as "development that meets
the needs of the present without compro-
mising the ability of future generations to
meet their own needs" (WCED 1987, 43).
The threat to future generations perceived
in the report arise from potentially large-
scale and irreversible degradation of natu-
ral systems in the course of global eco-
nomic development, particularly in poorer
The Brundtland statement thus focuses
attention on two issues that seem to be
central themes in any conception of sus-
tainability: the nature of the current genera-
tion's responsibility to future generations,
and the degree of substitutability between
"natural capital" and other forms of social
capital-physical investment and invest-
ment in knowledge and institutions as em-
bodied in human capital.2 The next two
sections of the paper examine alternative
views on these two issues to show how they
lead to different conceptions of sustain-
ability. In the fourth section of the paper
these alternative conceptions are related to
each other through a "two-tier" model of
resource management based on the idea of
"safe minimum standard." The fifth and
last section of the paper contains conclud-
ing remarks.
There is an enormous literature, span-
ning over two millennia, on concepts of dis-
tributive justice including fairness across
generations. Unfortunately, there is not yet
a conception of distributive justice that
commands wide intellectual support. Nev-
ertheless, there are several points of view
that have attracted considerable attention
in discussions of sustainability. 3 The dis-
cussion that follows emphasizes issues of
intergenerational fairness even though
these issues cannot be entirely divorced
from the subject of the next section, substi-
tution possibilities among components of
society's wealth endowment.
One fundamental partitioning of justice
concepts separates theories based on max-
imization of an independently defined good
(teleological theories) from theories based
more on innate rights and obligations (de-
ontological theories). A further categoriza-
tion can be made based on theories that em-
phasize the current generation and its
immediate descendants-"presentist" the-
ories-and theories that put greater empha-
sis on the "further future." Yet another
distinction, particularly in nonpresentist
theories of justice, concerns justice con-
cepts that emphasize individuals and more
"organicist" conceptions that put greater
weight on community interests.
The typical criterion of discounted inter-
temporal welfare maximization in applied
welfare economics occupies one point in
the continuum of alternative justice con-
ceptions. This criterion not only empha-
sizes preference satisfaction over rights; it
also is highly presentist, since with any pos-
itive intergenerational discount rate the
welfare of individuals living one generation
in the future is scarcely relevant to current
decision making. Many writers have sug-
gested that the presentist focus of the
present-value (PV) criterion implies an
influence of the current generation over
the circumstances of its more distant de-
scendants that seems, at least intuitively,
to be ethically questionable (Kneese and
JIn emphasizing these themes we are placing our-
selves within the anthropocentric stream of debate
about sustainability, in which the needs and wants of
people are central, as opposed to an "eccentric" per-
spective that asserts the intrinsic worth of the natural
environment. We also are sidestepping, without in any
way minimizing, the issue of how the state of the envi-
ronment may be connected to income distribution
within generations-in particular, connections be-
tween poverty and environmental degradation. See
Pearce, Barbier, and Markandya (1990) and World
Bank (1992) for discussion of these issues. Finally,
we consider sustainability primarily in the context of
resource management to meet identified human needs;
as opposed to the broader "co-evolutionary" perspec-
tive discussed in Norgaard (1988), which emphasizes
the mutual interactions between social actions and
* See Pearce and Turner (1990, chap. 15) for a com-
pact summary; Pezzey (1992) provides a wide-ranging
survey of motivations for considering sustainability.

Toman: Economics and "Sustainability"
Schulze 1985; Norton 1982, 1984, 1989;
Parfit 1983b; Page 1977, 1983, 1988).
The debate over the ethical implications
of the PV criterion is long-standing and in-
volves a number of considerations that of-
ten seem to be misunderstood. One basic
issue in this debate is the relationship be-
tween the PV criterion and the broader
concept of intergenerational economic ef-
ficiency as defined by the Pareto criterion,
which requires only that it be impossible
to improve the welfare of members of one
generation without reducing the welfare of
members of some other generation. This
notion of "no waste" seems desirable in
any intergenerational welfare criterion, at
least to those who give some weight to the
importance of individual preference satis-
faction. The difficulty with the PV criterion
thus is not that it requires Pareto efficiency,
but rather that it puts weight on the welfare
of the current generation in the social wel-
fare function that some regard as excessive.
As Page (1977, 1988) points out, there
are infinitely many intergenerational social
orderings consistent with the Pareto prin-
ciple that allow for different sets of inter-
generational welfare weights without the
"dictatorship" of the current generation
embodied in the present value criterion. A
number of analysts have explored other so-
cial welfare criteria that preserve the Pareto
principle without imposing the preferences
of the current generation on future genera-
This issue has been carefully considered
in a series of papers by Howarth and Nor-
gaard (see Howarth and Norgaard 1990,
1992, 1993 and Howarth 1991a, 1991b). Us-
ing an overlapping generations framework,
they argue that the problem of intergenera-
tional equity must be viewed as a problem
of ethics that is distinct from economic
efficiency in the Pareto sense. They fur-
ther argue that the intergenerational equity
problem should be approached as one that
involves a fair distribution of property
rights between current and future genera-
tions. This argument is a simple but power-
ful intergenerational extension of a stan-
dard result in welfare economics: "The
choice of distribution of income is the same
as the choice of an allocation of endow-
ments, and this in turn is equivalent to
choosing a particular welfare function"
(Varian 1984, 209; see also Bromley 1989).
In particular, Howarth and Norgaard show
that while purely "egoistic" utility con-
cerns will motivate some savings to benefit
the (short-term) future (since people live
more than one period and may also have
concerns for their own immediate descen-
dants), purely egoistic savings will not in
general be adequate to optimize a social
welfare function that includes more altruis-
tic concerns (e.g., the well-being of the en-
tire next generation or individuals further
into the future), Howarth's and Norgaard's
arguments also have important implications
for analyses of environmental valuation,
discount rates, and policy design (e.g., pol-
lution taxation), since all of these ate af-
fected by the income distribution.
Howarth and Norgaard do not investi-
gate the range of intergenerational social
welfare functions that might plausibly be in-
voked in connection with intergenerational
equity. In their analysis they are concerned
primarily with the egalitarian "maximin"
criterion discussed below as an alternative
to maximizing the present value of utility
Streams.5 In addition, trying to achieve
intergenerational equity solely through
savings that transfer endowments across
4 See in particular Page (1977), Pearce (1983), and
Burton (1993) for discussions of intergenerational dis-
counting. These analyses suggest that a positive dis-
count rate to reflect the growth of the economy is com-
patible with a zero rate of pure time preference in the
social welfare function on ethical grounds. The argu-
ments in Sandler and Smith (1976, 1977, 1982), Bishop
(1977), and Cabe (1982) indicate "that the assumption
of a uniform discount rate may not be consistent with
intertemporal Pareto efficiency, particularly with in-
tertemporal public goods.
5 Howarth (1992) derives this social welfare crite-
rion from a more restricted maximin ethic between just
parents and their children. He shows that if parental
altruism extends only to the direct consumption of the
next generation, there is no assurance that utility lev-
els will be maintained or increase over time; but if the
current generation is concerned about the capacity of
its descendants to exercise their bequest motive as
well, the result is concern about the equity of welfare
across all generations.

Land Economics
November 1994
generations may not always be effective.
Randall and Farmer (1993) argue that when
the two-generation analyses of Howarth
and Norgaard are extended to a setting with
three or more generations, a kind of Coa-
sian result obtains: the ultimate equilibrium
allocation is not that sensitive to the initial
distribution of property rights. Randall and
Farmer argue for an approach to sustain -
ability based on preservation rules like the
safe minimum standard discussed subse-
quently in this paper.
The problem of intergenerational equity
has received considerable attention in the
economics literature through the applica-
tion of a Rawlsian (1971) "maximin" con-
cept of intergenerational rights (see, e.g.,
Solow 1974, 1986 and Norton 1989, as well
as the work by Howarth and Norgaard
cited above). The Rawlsian approach has
been criticized as posing too harsh a trade-
off between equity and welfare maximiza-
tion, since a strict application of the Rawl-
sian criterion leads to the outcome that all
generations must be equally well (or badly)
off-that is, there is no scope for the cur-
rent generation to pursue improvements in
future conditions. However, more recent
analyses of the Rawlsian social welfare
problem suggest that this trade-off need not
be so harshly drawn. In particular, Asheim
(1988, 1991) shows that when individual
preferences include some altruistic concern
for immediate descendants, but there is
also a social agreement to follow a Rawl-
sian ethic involving concern for the indefi-
nite future, it is possible within the context
of social welfare maximization to have eco-
nomic growth coupled with a requirement
that future generations be no worse off than
the present.
As Pezzey (1989, 1994a) points out,
there are a number of alternatives to the
maximin criterion for social welfare order -
ings that could be used to reflect intergener-
ational equity concerns. Pezzey (1994b)
analyzes in some detail the implications of
a criterion based on the maximization of the
present value of per-capita utility subject to
an ethical constraint that per-capita utility
not decline over time. Like Asheim, Pezzey
finds that this criterion allows for concern
for future welfare without necessarily sacri-
ficing all growth possibilities. A weaker
version of this criterion would accord inter-
generational equity (as indicated by nonde-
clining utility over time) some finite weight
in the social welfare function, allowing for
well-defined trade-offs between maximum
present value and fairness (see, e.g.,
Broome 1992).
The discussion thus far has concerned
mainly individualistic conceptions of what
is good or right. Even the individualistic
point of view gives rise to deep contro-
versy. On the one hand, critics raise objec-
tions to the capacity of utilitarianism, or
even the concept of human preferences, to
adequately describe human interests (see,
e.g.; Sen 1982; Parfit 1983b; Sagoff 1988;
and Norton 1992).6 Defenders of deontolog-
ical theory, on the other hand, point out
the difficulties in assigning rights to future
generations (e.g., Broome 1991). Even
those who do not necessarily espouse utili-
tarianism agree that there are some deep
logical difficulties in assigning standing to
"potential" future persons whose circum-
stances not only are largely unknown to the
present generation but also are endogenous
to the set of choices made by the current
generation (see, e.g., Baier 1984; Barry
1977; Gelding 1972; Passmore 1974; and
Parfit 1983a).
One approach to this problem has been
the development of organicist arguments
that invoke an obligation to the entire con-
text of future human life-the species as
a whole, and the ecological systems that
surround it—rather than just to potential
future individuals (see, e.g., Leopold 1949;
Lovelock 1988; Callicott 1989; Norton
6Some critics argue that the conventional approach
to specifying preference orderings in economics is de-
ficient on both empirical and moral grounds, since it
does not distinguish "lower" or "higher" impulses,
or "self-interest" and "community-motivatd" inter-
ests. The solution, it is argued, is some hierarchical
representation of preferences. However, Brennan
(1989) argues that this approach does not really solve
any problems associated with conventional preference
reasoning in economics; and in particular, that moral
deficiencies associated with the outcomes of economic
logic should be directly confronted as such, rather
than attempting to reframe that logic.

Toman: Economics and "Sustainability"
1982, 1986, 1989; Page 1983, 1991; Nash
1989; Weiss 1989). This "stewardship" per-
spective emphasizes the safeguarding of the
large-scale ecological processes that sup-
port all facets of human life, from biological
survival to cultural existence. The steward-
ship perspective does not deny the rele-
vance of human preferences, but it asserts
the existence of larger societal concerns
that members of society will feel (in vary-
ing degrees) beyond individualistic prefer-
The organicist position raises the inter-
esting and as-yet unanswered question of
whether there are important social values
that simply cannot be captured in an indi-
vidualistic resource valuation, no matter
how broad and sophisticated the valuation
methods are. The difficulty in addressing
this issue is that the two perspectives are
based on different fundamental axioms.
The organicist position seems to avoid
some of the difficulties in extending indi-
vidualistic fairness concepts to intergenera-
tional circumstances. On the other hand, a
nonindividualistic perspective is a two-
edged sword in that many of humankind's
most cherished economic, political, and
other social institutions derive fundamen-
tally from giving high respect to individual
rights. Organicism without constraints
leads to supremacy of the group over the
individual, a form of social order that his-
tory shows to be very dangerous and de-
structive. The two-tier system described
subsequently in the paper seeks to provide
a venue for considering the balance be-
tween individual trade-offs and social im-
Assuming one accepts some obligation
to consider the well-being of future genera-
tions, what bundles of social capital should
succeeding generations make available to
their descendants? The answer to this ques-
tion depends critically on one's assump-
tions regarding the degree of substitu-
tability between the services provided by
natural capital (material resources, waste
absorption, other ecological functions, aes-
thetic and cultural values) and other forms
of capital (plant, equipment; knowledge,
skills, social institutions).
One view, to which many economists
would be inclined, is that all resources are
relatively fungible sources of well-being.
This view appears to be influenced heavily
by a number of classic and more recent ap-
plications of aggregate growth models with
natural resources. A number of familiar
theorems come out of this literature. In the
standard growth model without natural re-
source constraints, the modified Golden
Rule indicates that per-capita consumption
and utility will grow over time provided the
economy is not already saturated with ca-
pital. Clearly, sustainability presents no
challenge in this world, even with positive
discounting of future utilities. The same
outcome obtains with natural resources
provided these resources are in some sense
"augmentable" —capable of being renewed
or of having damages offset by compensa-
tory investments (for a recent exposition of
this see van Geldrop and Withagen 1993).
Even with exhaustible resources or some
other irreversible degradation of the ser-
vices provided by the natural environment
(such as accumulative pollution), it is possi-
ble for consumption and welfare to grow if
there is sufficient substitutability between
natural resources and capital accumulation,
or technical progress sufficient to offset the
depletion/degradation of natural resource
services (Dasgupta and Heal 1974; Solow
1974, 1986; Stiglitz 1974; Baumol 1986;
Dasgupta and Maler 1991; see also the sur-
veys in Asheim 1989, Pezzey 1992, and To-
man, Pezzey, and Krautkraemer forth-
From this point of view, then, large-
scale damages to ecosystems such as degra-
dation of environmental quality, loss of
species diversity, or destabilization from
global warming are not intrinsically unac-
ceptable. The question is whether compen-
satory investments for future generations in
other forms of capital are feasible and are
undertaken. This is the essence of the argu-
ment advanced by Solow (1986) and Maler
(1991), based on previous work by Hart-
wick (1977), that investments of resource

Land Economics
November 1994
rents in other forms of capital provide the
means to sustain consumption possibili-
ties over time. Investments in human
knowledge, techniques of production and
social organization are especially pertinent
in humankind's efforts to outrace any in-
creases in the scarcity of services provided
by the natural environment.7
An alternative view, embraced by many
ecologists and some economists, is that
such compensatory investments often are
infeasible as well as ethically indefensible.
Physical laws are seen as limiting the extent
to which other resources can be substituted
for scarce natural resources or ecological
degradation. In particular, physical capital
cannot be substituted for scarce energy
without limit because there are minimum
energy requirements for accomplishing any
transformation of matter. In addition, be-
cause matter is conserved, waste is an in-
herent part of any economic activity; and
natural limits may constrain the capacity of
the environment to process these wastes.8
Healthy ecosystems, including those that
provide genetic diversity in relatively un-
managed environments, offer resilience
against unexpected changes that preserve
options for future generations.9 For natural
life-support systems no practical substi-
tutes are possible, and degradation may be
irreversible. In such cases (and perhaps in
others as well), compensation cannot be
meaningfully specified.10
The question of physical scale is central
to this debate. If substitutability is rela-
tively easy, then the total scale of human
activity relative to the natural environment
is of limited significance relative to efficient
use of resources and, depending on one's
ethical perspective, the adequacy of soci-
ety's total savings for the future. The no-
tion of "carrying capacity," so often in-
voked in sustainability debates, then would
be at most ephemeral and at worst mean-
ingless outside its traditional ecological us-
age. Critics of this view turn the entire ar-
gument around by claiming that physical
limits cannot be ignored and then putting
much more emphasis on scale issues (see,
e.g., Goodland, Daly, and El Serafy 1991
and Costanza 1991).
A related issue that sometimes is over-
looked is the distinction between local and
global impacts when considering substitu-
tion possibilities. Local resource depletion
and ecological degradation, while often
having serious consequences, may be more
easily compensated for by trade, economic
diversification, and migration than regional
7As pointed out recently by Asheim (1994) and
Pezzey (1994b), Hartwick's reinvestment rule has
been widely misinterpreted as an instant test of the
future sustainability of an arbitrary economy. Al-
though an economy with constant utility over time
must satisfy the HartWick Rule (as Hartwick proved),
observing that investment currently happens to be
greater than or equal to the resource rent measured at
market prices does not imply that at least the current
level of utility can be maintained by imposing Hart-
wick's Rule from now onwards. The intuition behind
this result is that an economy which is depleting its
natural resources too fast for sustainability will drive
resource prices and hence resource rents too low, and
investment at such a level does not ensure sustainabil-
ity. The correct indicator of permanent sustainability
would be resource rents as measured by shadow prices
which reflect the sustainability constraint (which in-
cludes the constraint of the current resource stock).
This poses a challenge for those interested in devel-
oping empirical indicators of sustainable development.
8Concern over these issues its the economics litera-
ture has been expressed by Ayres and Kneese (1969),
Kneese, Ayres, and d'Arge (1971), Ayres and Miller
(1980), Perrings (1986), Anderson (1987), Barbier and
Markandya (1990), Gross and Veendorp (1990), Victor
(1991), Daly (1992), Townsend (1992), and Common
and Perrings (1992); see also the survey in Toman,
Pezzey and Krautkraemer (forthcoming).
9A related argument at the macro level is that envi-
ronmental quality may complement capital growth as
a source of economic progress, particularly for poorer
countries (Pearce, Barbier, and Markandya 1990).
• 10The importance of the substitutability issue can
be illustrated in connection with the debate over allo-
cating responsibility for greenhouse gas control. If one
accepts the view that investments in adaptation to cli-
mate change have limited scope for effectiveness, then
the atmosphere's capacity to absorb greenhouse gases
also is a depletable resource with limited substitution
potential. In this case cumulative past greenhouse gas
emissions can be a simple metric for assessing a fair
distribution of control obligation: greater cumulative
emissions by industrialized countries imply greater re-
sponsibility. However, if one sees the investment in
economic productive capacity and thus in global adap
tive capacity by industrial nations as having provided
significant benefits that do compensate for depletion
of the atmosphere's capacity for greenhouse gas
absorption, then the responsibility of industrialized
countries is less clear-cut.

Toman: Economics and "Sustainability"
or global adversities. On the other hand,
trade distortions (e.g., discrimination
against manufactured exports by devel-
oping countries) may limit national capaci-
ties to develop sustainably, and individual
countries may appear to develop sus-
tainably by "exporting" unsustainable re-
source use to other nations that supply ma-
The discussion in this section and the
previous one suggests that, at the risk of
some caricature, three alternative polar
conceptions of sustainability can be iden-
1.	Neoclassical presentism. This posi-
tion does not place much emphasis on
sustainability as an issue distinct from
efficient resource use. The standard
present value criterion is adopted for
intergenerational welfare compari-
sons, and natural capital scarcity is
assumed to be remediable (given ap-
propriate price signals and incentives)
through substitution and technical ad-
2.	Neoclassical egalitarianism. This
view is the same as ( 1) with respect to
assumptions about managing natural
capital scarcity, but it also maintains
a concern about a potential shortfall
in total savings for the future that is
not encompassed in the present value
3.	Ecological organicism. In contrast to
(1) and (2), this view emphasizes lim-
its on substitution between natural
capital and other assets. Like (2), this
view includes a concern for intergen-
erational fairness, but that concern is
not entirely individualistic; it also en-
compasses concerns for ecological
systems and the human species as a
whole. "
To be sure, views on sustainability that
are composites of these positions also can
be defined. The model discussed in the next
section allows for a continuum of views
about intergenerational fairness and re-
source substitutability.
In this section a simple conceptual
framework is outlined that can be used in
considering how individualistic resource
trade-offs might be balanced against social
imperatives for safeguarding against large-
scale, irreversible degradation of natural
capital. The framework is not intended to
imply a specific decision rule. Instead, its
purpose is to indicate the implications of
different sustainability conceptions and to
provide some common ground for consider-
ation of differences in conceptions among
economists, ecologists, and ethicists. In
broad outline, the framework is a two-tier
system in which standard economic trade-
offs (market and nonmarket) guide resource
assessment and management when the po-
tential consequences are small and revers-
ible, but these trade-offs increasingly are
complemented or even superseded by so-
cially determined limits for ecological pres-
ervation as the potential consequences
become larger and more irreversible. The
framework is an extension of the logic of
safe minimum standard promulgated by Ci-
riacy-Wantrup (1952) and Bishop (1978).
Variants of this two-tier approach have
been suggested by a number of writers from
different disciplines (see, e.g., Norton
1982, 1992; Page 1983, 1991; and Randall
To begin the discussion, suppose for
simplicity that all potential human impacts
on the natural environment can be charac-
terized by their prospective "cost" and "ir-
reversibility." Prospective cost can be in-
terpreted in several ways. It can be thought
of as an (individualistic) economic measure
of expected opportunity cost, as an ecologi-
cal measure of predicted physical impact,
or as some hybrid of individualistic or or-
ganicist concerns including social values
like political freedom and justice. The
11 It would be possible to identify a fourth position,
ecological presentism, but this view could be inter-
nally contradictory and in any event it seems to hold
little interest.

Land Economics
November 1994
framework does not require a particular
definition of cost, though some precision on
what is counted as a cost is needed in prac-
tice when interpreting alternative concep-
tions of the safe minimum standard.
Similarly, irreversibility can be seen in
terms of an ecological assessment of sys-
tem function or as an economic construct
involving the feasibility of restorative or
compensating investment. Economic irre-
versibility here is taken to be the same as
nonsubstitutability. Of course, consider-
able uncertainty exists regarding both the
cost and irreversibility of particular human
impacts. This uncertainty is in fact central
to the concept of safe minimum standard.
One question that needs to be addressed
is why two metrics are needed for gauging
impacts and determining social responses.
Economists are accustomed to valuing con-
sequences of irreversibility in an uncertain
setting (see, e.g., Krutilla 1967; Krutilla
and Fisher 1985; and Fisher and Hanemann
1987), so this dimension to some extent is
redundant. Indeed, the prospective cost
measure could be thought of as including
premiums reflecting risks that can be mone-
tized. The concept of systemic scale in eco-
logical research also may forge links be-
tween the severity and irreversibility of
impacts (Norton and Ulanowicz 1992). This
research suggests that damages to ecologi-
cal systems that are larger in spatial scale
or higher up in the hierarchy of natural pro-
cesses—more complex, consisting of more
component subsystems—is both more
harmful and harder to reverse because of
the complexity and slower time of adapta-
tion in these systems.
Nevertheless, there are reasons for dis-
tinguishing the metrics. Monetizing all ir-
reversibility suggests that compensatory
investment for any environmental degra-
dation is feasible and ethical.12 This seems
debatable, as already noted. Analytically,
it rules out by assumption the ecological
organicist position on sustainability defined
above. To avoid this, we must retain both
the cost and irreversibility dimensions.
The cost and irreversibility dimensions
can be brought together in a single "sample
universe" as shown in Figure l.13 Individu-
als can, in this theory, locate different im-
pacts on the natural environment (e.g., a
5-degree global mean temperature rise or
a 50 percent loss of tropical forest) in the
square, depending on their own assess-
ments of cost and irreversibility. Because
of uncertainties, these assessments will re-
flect subjective judgments including atti-
tudes toward known or potential risks (in
other words, the cost and irreversibility as-
sessments generally will not reflect just
subjective mean or median values). Individ-
ual judgments inherently will reflect not just
factual information but also personal values
about the nature of the obligation to future
generations. A variety of social institutions,
notably the political process, education,
and mass communication, presumably gen-
erate some synthesis of individual impact
assessments at the societal level. The syn-
thesis is dynamic in that it reflects a variety
of forms of social learning (e.g., improve-
ments in production technique and social
We can now combine this construct with
an extension of the safe minimum standard
logic to indicate how individualistic trade-
offs and social imperatives regarding the
natural environment might be balanced.
The safe minimum standard originally was
developed in the context of individual spe-
cies preservation (see Bishop 1978 and Ciri-
acy-Wantrup 1952). The logic in this setting
is that standard benefit-cost comparisons
may be inadequate if the long-term cost of
species loss is highly uncertain (in the
Knightian sense of having probabilities that
are difficult to gauge) but possibly quite
substantial. Proponents of a safe minimum
standard argue that with low information
but high potential asymmetry in the loss
function, the evenhanded assessment of
benefit-cost analysis should give way to a
greater presumption in favor of species
12 This discussion leaves aside important practical
problems of measurement that arise in any approach
to irreversibility.
11 This diagrammatic approach was originally devel-
oped by Bryan Norton (see Norton 1992). The figure
shown here is an adaptation of Norton's schema.

Toman: Economics and "Sustainability"
Ecological and
human catastrophe
of ecological
Moral imperatives for
resource and
ecosystem protection
Free play of
individual incentives
and resource tradeoffs
^	Increasing nonsubstitutability/	Low-cost, easily
irreversibility of ecological	reversed effects
preservation unless society judges that the
cost of preservation is "intolerable. 1,4
In Figure 1 we extend this logic to a con-
tinuum of potential impacts on the natural
environment in the following way. First,
impacts in the lower-right portion of the
box involve both modest cost and a high
degree of reversibility. In this area there is
little threat of substantial lasting damage to
the interests of future generations, and it is
reasonable to rely upon individualistic valu-
ations and trade-offs as reflected in ben-
efit-cost analysis. Individual incentives
for efficient resource use can be achieved
through markets and incentive-based poli-
cies to correct "conventional" external-
Toward the upper-right comer of the box
the costs become higher but still are rela-
tively reversible. Here the primary concern
in addition to efficient resource use might
be to ensure that the current generation
meets obligations to the future through gen-
eral compensation for environmental degra-
dation. On the other hand, impacts located
toward the lower-left corner of the box are
relatively irreversible but low in cost, so
they presumably can be absorbed without
too much detrimental effect on the future.
It is in considering impacts toward the
upper-left comer of Figure 1 that the safe
minimum standard assumes prominence.
Here the long-term costs are likely to be
high and substitution options likely to
be low, making the impacts irreversible.
Moreover, uncertainty is likely to be sub-
stantial since the impacts in question in-
volve large-scale ecological systems and
functions that remain poorly understood.
Under these conditions even individual-
istic, presentist valuations can provide a
considerable impetus toward resource
preservation. However, the logic of the
safe minimum standard suggests that this
impetus alone may not fully satisfy reason-
able obligations to future generations, par-
ticularly when the negative effects involve
14 See Bishop (1979) and Smith and Krutilla (1979)
as well as Castle and Berrens (1993) for further discus-
sion of the distinction between the safe minimum stan-
dard and benefit-cost analysis. This reasoning is an-
other way of highlighting the need for considering cost
and irreversibility as distinct metrics of impact.

Land Economics
November 1994
large-scale ecological systems and long
gestation periods. One can imagine that the
closer one moves to the northwest comer
of the box, the more entirely individualis-
tic valuation ciriteria are supplemented by
other expressions of community interest in
the form of a priori social rules of a "consti-
tutional" nature for preserving natural
capital. This is illustrated by the fuzzy
demarcation line in Figure 1. Such socially
determined criteria could be changed if the
members of society deem the cost of pre-
serving natural capital to be excessive, but
a higher burden of proof would be placed
on arguments favoring acceptance of high-
cost, irreversible impacts than on accep-
tance of smaller impacts.
As already noted, individual perceptions
of natural impacts and thus individual as-
sessments of where the fuzzy line should
be located depend strongly on individual
values and knowledge. Figure 1 can be used
to illustrate the different positions on sus-
tainability summarized in the previous sec-
tion of the paper. Generally speaking, ecol-
ogists with a primary concern for natural
function and resilience might be more in-
clined than economists to emphasize the
irreversibility dimension and to draw a
more vertical fuzzy line, limiting even
lower-cost irreversible effects; economists
with greater concern for cost and more con-
fidence in substitutability might be more in-
clined toward a horizontal line. Neoclassi-
cal presentists might put little or no area to
the northwest of the dividing line (or even
dismiss the whole construct), while ecologi-
cal organicists would take a contrary view.
Neoclassical egalitarians might take a mid-
dle ground, drawing a close to horizontal
line but placing more area above it to limit
high-cost burdens on future generations.
It should be emphasized again that there
is a distinct difference between the safe
minimum standard approach and the stan-
dard prescriptions of resource and environ-
mental economics, which involve getting
accurate valuations of resources in bene-
fit-cost assessments and using economic in-
centives to achieve efficient allocations of
resources given these valuations. Whether
a resource-protection criterion is estab-
lished through application of the safe mini-
mum standard concept or entirely by trade-
offs through cost-benefit analyses, that
criterion can be achieved cost-effectively
by using economic incentives. However,
for impacts on the natural environment that
are uncertain but may be large and irrevers-
ible, the safe minimum standard posits an
alternative to relying just on comparisons
of expected economic benefits and costs for
developing resource-protection criteria. 15 It
places greater emphasis on scale issues in-
volving potential damages to the natural
system than on the sacrifices experienced
from curbing ecological impacts, which are
seen as likely to be smaller and more
readily reversible. On the other hand, the
arguments in this section do not require
that either the safe minimum standard as a
social decision rule, or individual prefer-
ences for environmental preservation, be
rigidly hierarchical. The safe minimum
standard can be seen as a social compact
for expressing agreed-upon moral senti-
ments in the face of high ecological uncer-
tainty and potential loss asymmetry, even
with egoistic consumption, bequest, and
time Deferences that are entirely neoclas-
The arguments in this section are some-,
what similar to those developed by Vatn
and Bromley (1994) regarding environmen-
tal decision making and economic valua-
tion. Briefly, these authors argue that large-
scale environmental assets or risks are
inherently difficult to value meaningfully in
a conventional economic sense. This is not
just because of limited information about
these assets and risks, which causes indi-
vidual preferences to be poorly defined, but
also because large-scale environmental con-
15 See also Pezzey (1989, 1994a), who shows with
a simple example that efficient management of exter-
nalities over time may not generate sustainable welfare
"Tim Brennan suggests (in private communica-
tion) that the safe minimum standard also can be seen
as a social decision strategy that economizes on costly
information-gathering and enforcement activities rela-
tive to theoretically preferred marginal evaluations
and policies.

Toman: Economics and "Sustainability"
siderations are bound up in social mores
that condition individual preferences. Vatn
and Bromley argue that people must be
seen as dualistic, behaving as citizens as
well as consumers, and that many social
institutions for environmental manage-
ment—including the norms surrounding
government of the environment—must be
seen as ways that societies have attempted
to circumvent the informational and "con-
textual" problems surrounding individual-
istic valuation. This point of view justifies
in particular the imposition of safe mini-
mum standards determined through politi-
cal discourse and other complex social pro-
Sustainability ultimately is intimately
wrapped up with human values and institu-
tions, not just ecological functions. An en-
tirely ecological definition of sustainability
is inadequate; guidance for social decision
making also is required. It must be recog-
nized that human behavior and social de-
cision processes are complex, just as eco-
logical processes are. At the same time,
economic analysis without adequate eco-
logical underpinnings also can be mis-
leading. The sustainability debate also
should remind economists to carefully dis-
tinguish between efficient allocations of re-
sources—the standard focus of economic
theory-and socially optimal allocations
that may reflect other intergenerational (as
well as intragenerational) equity concerns:
The tension between ecological and eco-
nomic perspectives on sustainability sug-
gests several ways in which both econo-
mists and ecologists could adapt their
research emphases and methodologies to
make the best use of interdisciplinary con-
tributions. For ecologists, the challenges
include providing information on ecological
conditions in a form that could be used in
economic assessment.17 Ecologists also
must recognize the importance of human
behavior, particularly behavior in response
to economic incentives-a factor often
given short shrift in ecological impact anal-
yses. Economists for their part could ex-
pand analyses of resource values to con-
sider the function and value of ecological
systems as a whole, making greater use of
ecological information in the process. Both
methodological research and case studies
are needed to synthesize ecological and
economic perspectives. Research by econ-
omists and other social scientists (psychol-
ogists and anthropologists) also could help
to improve understanding of how future
generations might value different attributes
of natural environments.
From the standpoint of economic the-
ory, an important direction for further re-
search is the consideration of how both
physical limits and ethical' constraints on
resource use may affect the time paths and
shadow values of natural capital stocks, rel-
ative to the results found in standard the-
ory. The literature on economic growth
with natural resources is beginning to ad-
dress these issues, and there is a lot of basic
methodology that can be exploited for this
One example is the work by Asheim
(1988, 1991) and Pezzey (1989, 1994a,
1994b) alluded to earlier. Asheim shows
that if we accept the idea of two-tiered so-
cial preferences, in which individuals have
limited altruism for the next generation
but also subscribe to a broader conception
of intergenerational social justice, socially
preferred outcomes can promote justice
without sacrificing growth. In particular,
this argument provides a more basic justi-
fication for the criterion of nondecreasing
utility assumed in Pezzey's sustainability
Another set of examples concerns the is-
sue of resource substitution. A number of
17	Carpenter (1992) argues that the current state of
biophysical measurement for assessing the sustainabil-
ity of human impacts on ecological systems is too
weak to effectively operationalize the concept of natu-
ral capital; only gross unsustainability can be detected.
18	For furtlier discussion see Toman, Pezzey, and
Krautkraemer (forthcoming).
19	Because of the obvious importance of uncertainty
in dealing with long-term environmental change, for a
complete analysis it is necessary to explicitly reflect
this uncertainty in social welfare orderings. This issue
is tackled in Asheim and Brekke (1993).

Land Economics
November 1994
papers have explored the consequences for
present-vahie-maximizing paths of includ-
ing stocks in utility functions as a reflection
of some sort of "amenity" value (see, e.g.,
Krautkraemer 1985, 1988 and Tahvonen
and Kuuluvainen 1993). In these analyses,
preservation of some positive level of envi-
ronmental attribute is not assured; achiev-
ing preservation in the steady state requires
some combination of large initial capital
accumulation and unbounded disutility
from environmental degradation. Barbier
and Markandya (1990), in particular, con-
sider the consequences of requiring a
threshold level of environmental preserva-
tion to stave off irreversible environmental
disaster. Common and Perrings (1992) go
further in discussing the basic differences
between economic and ecological sus-
tainabihty, and the difficulties in bringing
these ideas together in a single model.
Despite its continued abuse as a buzz-
word in policy debates, the concept of sus-
tainability is becoming better established as
a consequence of studies in economics,
ecology, philosophy, and other discipfines.
With a better understanding of the interdis-
ciplinary theoretical issues, and a better
empirical understanding of both ecological
conditions and social values, sustainability
also can evolve to the point of offering
more concrete guidance for social policy.
Anderson, C. L. 1987. "The Production Pro-
cess: Inputs and Wastes. Journal of Envi-
ronmental Economics and Management 14
(Mar.): 1-12.
Asheim, G. B. 1988. "Rawlsian Intergenera-
tional Justice as a Markov-Perfect Equilib-
rium in a Resource Technology." Review of
Economic Studies 55 (July):469-84.
	. 1989. "Intergenerational Conflicts in the
Management of Natural and Environmental
Resources: Sustainability in a Growth-
Theoretic Perspective." Center for Applied
Research, Draft Manuscript, Norwegian
School of Economics and Business Adminis-
tration (Oct.).
	. 1991. "Unjust Intergenerational Allo-
cations. " Journal of Economic Theory 54
	. 1994. "Net National Product as an Indi-
cator of Sustainability." Scandinavian Jour-
nal of Economics 96 (June):257-65.
Asheim, G. B., and K. A. Brekke. 1993. "Sus-
tainability When Resource Management Has
Stochastic Consequences." Central Bureau
of Statistics, Discussion Paper No. 86, Oslo
Ayres, R. U., and A. V. Kneese. 1969. "Produc-
tion, Consumption, and Externalities."
American Economic Review 69 (June):
Ayres, R. U., and S. Miller. 1980. "The Role of
Technological Change." Journal of Envi-
ronmental Economics and Management 7
Baier, A. 1984. "For the Sake of Future Genera-
tions." In Earthbound: New Introductory Es-
says in Environmental Ethics, ed. T. Regan.
New York: Random House.
Barbier, E. B., and A. Markandya. 1991. "The
Conditions for Achieving Environmentally
Sustainable Development." European Eco-
nomic Review 34 (May):659-69.
Barry, B. 1977. "Justice Between Genera-
tions." In Law, Morality, and Society: Es-
says in Honour of H. L. A. Hart, eds. P. M. S.
Hacker and S. J. Rax. Oxford: Clarendon
Press. "
Baumol, W. J. 1986, "On the Possibility of Con-
tinuing Expansion of Finite Resources."
Kyklos 39167-79.
Bishop, R. C. 1977. "Intertemporal and Inter-
generational Pareto Efficiency: A Com-
ment." Journal of Environmental Economics
and Management 4 (Sept.) :247-57.
—. 1978. "Endangered Species and Uncer-
tainty: The Economics of the Safe Minimum
Standard." American Journal of Agricultural
Economics 60 (Feb.): 10-18.
—. 1979. "Endangered Species, Irrevers-
ibilty and Uncertainty: A Reply." American
Journal of Agricultural Economics 61 (May):
Brennan, T. J. 1989. "A Methodological Assess-
ment of Multiple Utility Frameworks." Eco-
nomics and Philosophy 5:189-208.
Bromley, D. W. 1989. "Entitlements, Missing
Markets and Environmental Uncertainty."
Journal of Environmental Economics and
Management 17 (Sept.): 181-94.
Broome, J. 1991. The Intergenerational Aspects
of Climate Change. Bristol, UK: University
of Bristol.
	. 1992. Counting the Cost of Global
Warming. Cambridge: White Horse Press.
Burton, P. S. 1993. "Intertemporal Preferences

Toman: Economics and "Sustainability"
and Intergenerational Equity Considerations
in Optimal Resource Harvesting." Journal of
Environmental Economics and Management
24 (Mar.): 119-32.
Cabe, R. A. 1982. "Intertemporal and Intergen-
erational Pareto Efficiency: An Extended
Theorem." Journal of Environmental Eco-
nomics and Management 9 (Dec.):355-60.
Callicott, J.B, 1989. In Defense of the Land
Ethic. Albany: State University of New York
Carpenter, R.A. 1992. "Can Sustainability Be
Measured." Paper presented at the 2nd
Meeting of the International Society for Eco-
logical Economics, Stockholm, Sweden,
Stockholm University, August 3-6.
Castle, E. N., and R. P. Berrens. 1993. "Endan-
gered Species, Economic Analysis, and the
Safe Minimum Standard." Northwest Envi-
ronmental Journal 9:108-30.
Ciriacy-Wantrup, S. V. 1952. Resource Conser-
vation. Berkeley: University of California
Common, M., and C. Perrings. 1992. "Towards
an Ecological Economics of Sustainability."
Ecological Economics 6 (July):7-34.
Costanza, R., ed. 1991. Ecological Economics:
The Science and Management of Sustainabil-
ity. New York Columbia University Press.
Daly, H. 1990. "Toward Some Operational Prin-
ciples of Sustainable Development." Ecolog-
ical Economics 2 (Apr.): 1-6.
	. 1991. "Ecological Economics and Sus-
tainable Development From Concept to Pol-
icy." Environment Department Working Pa-
per No. 1991-24, The World Bank.
	. 1992. "Is the Entropy Law Relevant to
the Economics of Natural Resource Scar-
city?—Yes, of Course It Is!" Journal of En-
vironmental Economics and Management 23
(July) :91-95.
Dasgupta, P. S., and G. M. Heal. 1974. "The
Optimal Depletion of Exhaustible Re-
sources." In Review of Economic Studies,
Symposium on the Economics of Exhaustible
Resources. Edinburgh, Scotland Longman
Group Limited.
Dasgupta, P. S., and K.-G. Maler. 1991. "The
Environment and Emerging Development Is-
sues." In Proceedings of the World Bank An-
nual Conference on Development Economics
1990. Washington, DC: The World Bank.
Dixon, J. A., and L. A. Fallon. 1989. "The Con-
cept of Sustainability: Origins, Extensions,
and Usefulness for Policy." Society and Nat-
ural Resources 2:73-84.
Ehrlich, P. R. 1989. "The Limits to Substitu-
tion: Meta-Resource Depletion and a New
Economic-Ecological Paradigm." Ecological
Economics 1 (Feb.):9-16.
Fisher, A. C., and W. M. Hanemann. 1987.
"Quasi-Option Value: Some Misconceptions
Dispelled.'" Journal of Environmental Eco-
nomics and Management 14 (June): 183-90.
Gelding, Martin P. 1972. "Obligations to Future
Generations." The Monist 56:85-99.
Goodland, R., H. Daly, and S. El Serafy, eds.
1991. "Environmental Sustainable Economic
Development Building on Brundtland." En-
vironment Department, Environment Work-
ing Paper No 46, The World Bank (July).
Gross, L. S., and E. C. H. Veendorp. 1990.
"Growth With Exhaustible Resources and
a Materials-Balance Production Function."
Natural Resource Modeling 4 (Winter):
Hartwick, J. M. 1977. "Intergenerational Equity
and the Investing of Rents from Exhaustible
Resources." American Economic Review 67
Howarth, R.B. 1991a. 'Intergenerational Com-
petitive Equilibria under Technological
Uncertainty and an Exhaustible Resource
Constraint." Journal of Environmental Eco-
nomics and Management 21 (Nov.):225-
	. 1991b. "Intertemporal Equilibria and
Exhaustible Resources: An Overlapping
Generations Approach." Ecological Eco-
nomics 4 (Dec.):237-52.
	. 1992. "Intergenerational Justice and the
Chain of Obligation." Environmental Values
1 (Summer): 133-40.
Howarth, R. B., and R. B. Norgaard. 1990. "In-
tergenerational Resource Rights, Efficiency
and Social Optimality." Land Economics 66
(Feb.): 1-11.
	. 1992. "Environmental Valuation under
Sustainability." American Economic Review
82 (May):473-77.
	. 1993. "Intergenerational Transfers and
the Social Discount Rate'." Environmental
and Resource Economics 3 (Aug.):337-58.
Kneese, A. V., and W.D. Schulze. 1985.
"Ethics and Environmental Economics." In
Handbook of Natural Resource and Energy
Economics, Vol. 1, eds. A. V. Kneese and
J.L. Sweeney. Amsterdam: North-Holland.
Kneese, A., R. Ayres, and R. d'Arge. 1971.
Economics and the Environment. Baltimore,
MD: Johns Hopkins University Ress for Re-
sources for the Future.
Krautkraemer, J. A. 1985. "Optimal Growth,
Resource Amenities and the Preservation

Land Economics
November 1994
of Natural Environments. " Review of Eco-
nomic Studies 52 (Jan.): 153-70.
	. 1988. "The Rate of Discount and the
Preservation of Natural Environments."
Natural Resource Modeling 2 (Winter):
Krutilla, J. V. 1967. "Conservation Recon-
sidered." American Economic Review 54
(Sept. ):777-86.
Krutilla, J. V., and A. C. Fisher. 1985. The Eco-
nomics of Natural Environments: Studies in
the Valuation of Commodity and Amenity
Resources, 2d ed. Washington, DC: Re-
sources for the Future.
Leopold, A. 1949. A Sand County Almanac.
New York: Oxford University Press.
Lovelock, J. 1988. The Ages of Gaia. New
York: Norton.
Maler, K.-G. 1991. "National Accounts and En-
vironmental Resources." Environmental and
Resource Economics 1:1-16.
Nash, J. A. 1989. "Ecological Integrity and
Christian Political Responsibility." Theology
and Public Policy 1 (Fall):32-48.
Norgaard, R. B. 1988. "Sustainable Develop-
ment: A Co-Evolutionary View. " Futures 20
Norton, B. G. 1982. "Environmental Ethics and
the Rights of Future Generations. " Environ-
mental Ethics 4 (Winter):319-30.
	. 1984. "Environmental Ethics and Weak
Anthropocentrism." Environmental Ethics 6
(Summer): 131-48.
	. 1986. "On the Inherent Danger of Un-
devaluing Species." In The Preservation of
Species, ed. B. G. Norton. Princeton, NJ:
Princeton University Press.
	. 1989. "Intergenerational Equity and
Environmental Decisions: A Model Using
Rawls' Veil of Ignorance." Ecological Eco-
nomics 1 (May): 137-59.
	. 1992. "Sustainability, Human Welfare,
and Ecosystem Health." Environmental Val
ues 1 (Summer):97-lll.
Norton, B. G., and R. E. Ulanowicz. 1992.
"Scale and Biodiversity Policy: A Hierarchi-
cal Approach." Ambio 21 (May):244-49.
Page, T. 1977. Conservation and Economic
Efficiency. Baltimore, MD: Johns Hopkins
University Press for Resources for the Fu-
	. 1983. "Intergenerational Justice as
Opportunity." Energy and the Future, eds.
D. MacLean and P. G. Brown. Totowa, NJ:
Rowman and Littlefield.
	. 1988. "Intergenerational Equity and the
Social Rate of Discount." In Enviromental
Resource and Applied Welfare Economics,
ed. V. K. Smith. Washington, DC: Resources
for the Future.
	. 1991. "Sustainability and the Problem
of Valuation." In Ecological Economics: The
Science and Management of Sustainability,
ed. R. Constanza. New York: Columbia Uni-
versity Press.
Parfit, D. 1983a. "Energy Policy and the Further
Future: The Identity Problem." In Energy
and the Future, eds. D. MacLean and P. B.
Brown. Totowa, NJ: Rowman and Littlefield.
	. 1983b. "Energy Policy and the Further
Future: The Social Discount Rate." In En-
ergy and the Future, eds. D. MacLean and
P.B. Brown. Totowa, NJ: Rowman and Lit-
Passmore, J. 1974. Man's Responsibility for Na-
ture. New York: Scribner's.
Pearce, D.W. 1983. "Ethics, Irreversibility, Fu-
ture Generations and the Social Rate of DiS'
count." International Journal of Environ-
mental Studies 21:67-86.
Pearce, D., and R. Turner. 1990. Economics
of Natural Resources and the Environment.
Baltimore, MD: Johns Hopkins University
Pearce, D., A. Markandya, and E. Barbier.
1989.	Blueprint for a Green Economy. Lon-
don: Earthscan.
Pearce, D. W., E. Barbier, and A. Markandya.
1990.	Sustainable Development: Economics
and Environment in the Third World. Lon-
don: Earthscan.
Perrings, C. 1986. "Conservation of Mass and
Instability in a Dynamic Economy-Environ-
ment System.!' Journal of Environmental
Economics and Management 13 (Sept.): 199-
Pezzey, J. 1989. Economic Analysis of Sustain-
able Growth and Sustainable Development.
World Bank, Environment Department
Working Paper No. 15. Now reprinted as
Pezzey, J. 1992. Sustainable Development
Concepts: An Economic Analysis. World
Bank Environment paper No. 2. Washington,
DC: World Bank.
	. 1992. "Sustainability: An Interdisci-
plinary Guide," Environmental Values 1
(Mar): 321-62.
	. 1994a. "Sustainability, Intergenera-
tional Equity, and Environmental Policy."
Draft manuscript, University College Lon-
don (April).
	. 1994b. "The Optimal Sustainable De-
pletion of Nonrenewable Resources."
Draft manuscript, University College Lon-
don (May).
Randall. A. 1986. "Human Preferences, Eco-

Toman: Economics and "Sustainability"
nomics, and the Preservation of Species. "
In The Preservation of Species, ed. B. G.
Norton. Princeton, NJ: Princeton University
Randall, A., and M. C. Farmer. 1993. "Policies
for Sustainability: Lessons from an Overlap-
ping Generations Model." Draft manuscript,
Columbus: The Ohio State University (Jan.).
Rawls, J. 1971. A Theory of Justice. Cambridge:
Harvard University Press.
Sagoff, M. 1988. The Economy of the Earth.
New York: Cambridge University Press.
Sandier, T., and V. K. Smith. 1976. Intertem-
poral and Intergenerational Pareto Effi-
ciency." Journal of Environmental Econom-
ics and Management 2 (Feb.): 151-59.
	. 1977. "Intertemporal and Intergenera-
tional Pareto Efficiency Revisited." Journal
of Environmental Economics and Manage-
ment 4 (Sept.):252-57.
	. 1982. "Intertemporal and Intergenera-
tional Pareto Efficiency: A Reconsideration
of Recent Extensions." Journal of Environ-
mental Economics and Management 9
Sen, A. K. 1982. "Approaches to the Choice
of Discount Rates for Social Benefit-Cost
Analysis." In Discounting for Time and Risk
in Energy Policy, eds. R. C. Lind et al. Wash-
ington, DC: Resources for the Future.
Smith, V. K, and J. V. Krutilla. 1979. "Endan-
gered Species, Irreversibilities, and Uncer-
tainty: A Comment." American Journal of
Agricultural Economics 61 (May):371-75.
Solow, R.M. 1974. "Intergenerational Equity
and Exhaustible Resources." In Review of
Economic Studies, Symposium on the Eco-
nomics of Exhaustible Resources. Edin-
burgh, Scotland: Longman Group Limited.
	. 1986. "On the Intergenerational Alloca-
tion of Natural Resources." Scandinavian
Journal of Economics 88 (June): 141-49.
	. 1993a. "An Almost Practical Step To-
ward Sustainability." Resources Policy 19
(Sept.): 162-72.
	. 1993b. "Sustainability: An Economist's
Perspective." In Selected Readings in Envi-
ronmental Economics, ed. R. Dorfman and
N. Dorfman, 3rd ed. New York: Norton.
Stiglitz, J. 1974. "Growth with Exhaustible
Natural Resources: Efficient and Optimal
Growth Paths." In Review of Economic
Studies, Symposium on the Economics of Ex-
haustible Resources. Edinburgh, Scotland:
Longman Group Limited.
Tahvonen, 0., and J. Kuuluvainen. 1993. "Eco-
nomic Growth, Pollution, and Renewable
Resources." Journal of Environmental Eco-
nomics and Management 24 (Mar.): 101-18.
Tietenberg, T. 1992. Environmental and Natural
Resource Economics, 3rd ed. New York:
Harper Collins.
Toman, M. A., J. Pezzey, and J. Krautkraemer.
Forthcoming, "Neoclassical Economic
Growth Theory and 'Sustainability." In
Handbook of Environmental Economics, ed.
D. W. Bromley. Oxford: Basil Blackwell.
Townsend, K. N. 1992. "Is the Entropy Law
Relevant to the Economics of Natural Re-
source Scarcity? Comment." Journal of En-
vironmental Economics and Management 23
van Geldrop, J., and C. Withagen. 1993, ' 'Natu-
ral Capital and Sustainability." Draft Manu-
script, Department of Mathematics and
Computing, University of Technology, Eind-
hoven, Netherlands.
Varian, H. 1984. Microeconomic Analysis, 2d
ed. New York: Norton.
Vatn, A., and D. W. Bromley, 1994. "Choices
without Prices without Apologies." Journal
of Environmental Economics and Manage-
ment 26 (Mar.): 129-48.
Victor, P. A. 1991. "Indicators of Sustainable
Development: Some Lessons from Capital
Theory." Ecological Economics 4 (Dec.):
Weiss, E. B. 1989. In Fairness to Future Gen-
erations. Dobbs Ferry, NY: Transnational
World Bank. 1992. World Development Report
1992: Development and the. Environment.
New York: Oxford University Press.
World Commission on Environment and Devel-
opment (WCED). 1987. Our Common Fu-
ture. New York: Oxford University Press.

John M. Hartwick*
Queen's University
Sustainability and Constant Consumption Paths in Open Economies
with Exhaustible Resources
We review some of the historical background to the capital theory approach to
sustainability. We then turn to sustainability in a group of countries trading flows from an
exhaustible resource. We derive an adjusted invest-resource-rents rule which leaves each
country, in a group of trading countries, on a constant consumption path. Oil importers invest
a fraction (greater than unity) of the rents ascribable to the current use of their own oil stocks
and oil exporters invest a fraction (less than unity) of the rents ascribable to their current use of
their own oil stocks. Each country's value of imports equals its value of exports. In a partial
equilibrium model of a small open oil exporting country, we observe that the exact invest-
resource-rents does leave the country's consumption constant over time.
(for AERE Conference, Boulder, Colorado, June 5, 1994)
* This paper was written while I was a visitor at the Center for Economic Studies, University
of Munich. Thanks to them for hospitality and support and to Ray Rees for helpful discussions.
The SSHRCC, Government of Canada, also provided support. Geir Asheim and Cees Withagen
provided invaluable comments on an earlier draft. An earlier version was presented in seminars
at the Universities of Munich and Tilburg.

Sustainability and Constant Consumption Paths in Open Economies
with Exhaustible Resources
I nlroducl ion
Since there are at least three good surveys of theoretical aspects of sustainability (namely,
Solow [1991], Hammond [1994], Pezzey and Toman [1994]) available, I will not attempt a
cannibalized fourth. Instead I will make some brief general remarks about the background of
theory of sustainability and then turn to an area of current research, namely open economy
aspects of sustainability. With this approach I can still present the references to the literature
which I know about, an invaluable part of a good survey, and also introduce a reader to the core
of the theory because I need this material as the stalk to graft on my open economy analysis.
There are at least three distinct ideas tied up in the economic theory of sustainability
which I am dealing with today. There is first the idea that if exhaustible resource stocks are
depleted today in the course of producing final goods, one need not immediately contemplate a
permanent shrinkage in future production possibilities because producible or machine capital can
be "over-accumulated" in order to "compensate" for the current reduction in the stock of natural
capital. This idea is mentioned in Pigou [1935] and in Hayek [1941; p.88]. An important variant
of this idea is of "over-accumulating" knowledge capital in order to "balance-off' the current
diminution of the stock of natural capital (Robson [1980]). More generally, technical progress
may allow smaller and smaller flows from exhaustible resources to maintain say a non-shrinking
set of production possibilities (as in for example Stiglitz [1974]). The second idea that comes
to mind is that sustainability suggests non-shrinking production possibilities as time passes. A
simple indicator of non-shrinking production possibilities is of course the observed aggregate

consumption level not declining over time.1 For the case of multiple consumption goods one
turns to THE utility of the current consumption vector not declining over time. Rawls maximin
criterion, a moral injunction, is a polar case in this line of thought, and in part inspired the
classic Solow [1974] paper on sustainability. The injunction is of course: do for others who will .
occupy period t+1 what we would have preferred back in t-1, what others who occupied period
t-1 would do for us, the occupants of period t. The third idea involves linking the first two ideas
together. The simplest variant is of course: consume at a level which results in no shrinking of
one's "capital". For an individual, this is not too difficult to contemplate since everything can
be measured in dollars but at the level of the nation satisfactory measures of what "capital" is
being maintained intact are generally elusive. Hicks [1942] and Pigou [1941] debated aspects of
the meaning of "maintaining capital intact". This was a final exchange in the long-running debate
on the links between capital and national accounting, a debate in which Pigou and Hayek sparred
"and Hicks assisted in clarifying matters. A primary legacy was Hicks' [1939; Chapt. 13] notion
that INCOME be defined as POTENTIAL CONSUMPTION which if "withdrawn" from current
* tji 2
production leaves capital ffltECt. In Solow [1974], the problem of measuring "capital intact" was
reduced to, given oil stocks being run down in accord with the Hotelling efficiency condition,
and given the level of consumption unchanging, how much K is currently needed to "support"
this program, at least for another period. This is in fact one kind of investment balancing off
1 Asheim [1988] [1991] has axiomitized the concept of non-declining U(C) in an economy
with exhaustible resources. There is no simple way to rank two distinct efficient candidate paths.
See also Pezzey [1993].
This leads to the idea that Net National Product be defined as some sort of "interest" on
"national wealth" (Samuelson [1961], Weitzman [1976], Kemp and Long [1982], Lozada [1992],
Asheim [1994] and Hartwick [1994]).

disinvestment in another stock. Dixit, Hammond, Hoel [1980] have labelled such paths as those
with "zero net investment". Such paths are not in general those in which aggregate capital value
(national wealth) are remaining constant because zero net investment is essentially changes in
quantities of stocks at prevailing prices and changes in national wealth comprise both quantity
changes and price changes - a chain rule calculation. The constant consumption model of Solow
[1974] is of course a zero net investment model but it is not a constant wealth model. It is an
increasing national wealth model. More on this bclOW,3
When the stock of natural capital is regenerating itself as with say fish stocks, forest
stocks, and environmental capital stocks, the notion of preserving capital intact is straight-
forward in the steady state. In fact, the term sustainable yield has been around in the economics
of the fishery and forestry much longer than it has been in the discussion of how any economy
is performing (as in, for example, the Brundtland Report). There remains however the question
of what course of action to take along the approach to the steady state (the transient trajectory)
with renewable resources in the economy. If one is wedded to a constant consumption path over
ail time, then the investing of resource rents is the appropriate strategy off the steady state
trajectory (Hartwick [1978], Becker [1982] Hamilton [1994]). This result contains the not-new
suggestion that the exhaustible resource use problem is a special case of the renewable resource
use problem in the sense that in the former, the economy has only a transient path to occupy.
We now turn to some detailed analysis on constant consumption paths in open economies.
' I am indebted to Geir Asheim for clarifying this in conversation.

Onen Economy Considerations
Consider splitting a closed economy with exhaustible resources, enjoying constant
consumption over time, into two countries, one importing some oil (the exhaustible resource)
from the other. We observe below that if each country saves exactly the resource rents
ascribable to load resource stock flows, the importer's consumption level will be declining and
the exporter's will be increasing (Asheim [1986]). We can describe this as the importer under-
saving and the exporter over-saving relative to levels for constant consumption paths. Below,
we characterize adjustment weights on each country's own resource rents which "neutralizes"
the importer's under-saving and the exporter's over-saving. With "corrected" local savings
levels, each country "ends up" on a constant consumption path, an intergenerational equity path
(Solow [1974]).
The under and over saving takes the form of price changes on oil trade flows - opposite
in sign but equal for each country. The adjustment weights on local own resource rents appear
in offsets to these "capital gains" terms and one characterization is as an r percent rule on certain
oil flows, not values, r is the rate of return, equal to the marginal product of capital in our
model. We will work with an almost symmetric split of the one world into two countries. This
makes the exposition straightforward and allows us to detour around special cases with corner
solutions. The reader can easily develop the analysis for not nearly symmetric splits of the one
world and for more than two countries. We comment on this in detail.
Under exact investment of resource rents, each country's change in consumption turns
out to equal the exhaustible resource flow traded multiplied by its current price change. Thus
the consumption shifts in each country can be interpreted as an adverse terms of trade shift for

oil importers and a favorable terms of trade shift for oil exporters. This becomes clear when we
set out a model of an oil exporting nation facing constant world prices and interest rates, at the
end. No terms of trade effects or consumption "wedges" are observed under the exact invest
resource rents strategy. Thus "over-saving" and "under-saving" under exact savings of own oil
use rents in the two country model are a consequence of endogenous terms of trade shifts,
induced, of course, via oil price changes. The oil price changes are a consequence of asset
equilibrium in the market for oil stocks (Hotelling's Rule). Our partial equilibrium model at the
end has constant world oil prices; the r percent changes in resource rents operate via endogenous
extraction cost shifts.
The Model
We look first at the structure of a closed one world economy. It has S(t) tons of say oil
left at date t. - S(t) = R(t) will be used in production of Q(t) equal to F(K(t), R(t)) at date t.
K(t) is non-depreciating machine capital. F(*) is homogeneous of degree 1 in inputs and K(t) and
R(t) are smoothly substitutable. F(*) is concave in its arguments. (Existence of constant
consumption paths over infinite time requires F(*) to be Cobb-Douglas (see Solow [1974],
Dasgupta and Mitra [1983] and Hamilton [1993]).) Population N, constant, only Consumes.4 We
postulate the savings-investment rule (invest resource rents):
K(t) « X(t)R(t)FR(t)	(i)
where X(t) moves exogenously through time, say near unity. F*(t) is the derivative 3F( )/3R. We
* This is not an issue with a Cobb-Douglas production function but otherwise, putting N, a
constant in the production function can introduce complicated scale effects as the economy's
level of aggregate output, Q(t), changes over time.

also take dynamic efficiency in exhaustible resource use as given, that is (Hotelling r% Rule):
= FK(t).	( 2 )
Ji(t) 1KV
current consumption C(t) is given by C(t) = F(*) - K. If one differentiates this expression with
respect to time, and does the same for (1), and one uses (1) and (2), one obtains (see the
C(t) " (i-\(t))R(t) F,(t)'	<3)
The central case of investing exactly exhaustible resource rents (namely X = 1) yields C = 0
Hartwick [1977]. See also GXtCQSionS* in Dixit, Hammond, Hoel [1980] and cairns [1986].).
Consider the value of aggregate capital or national wealth W(t) in this economy at date
t. We define W(t) = K(t) + S(t)FR(t). Observe that W(t) = K(t) + S(t)FR(t) + FR(t)S(t) and
W(t) is the change in wealth (aggregate capital value) in the economy at date t. The following
result can be derived. If the economy is efficient, has net investment zero, and has constant
returns to scale in F(K,R) then
C + W(t) = W(t)FK(t)
or C + W(t) is the interest flow from current wealth W(t). The demonstration requires simple
substitution, i.e. C = F(*)-K, F(") = KFg + RFr, K = RFr, etc. This result is quite Hicksian
since the income flow on the left is interest on capital on the right. The "logic" of Hicks'
position suggests that the left hand side is net national product in this economy. Asheim [1994]
seems to espouse this view. W(t) includes capital gains S(t)FR(t) on oil stocks and these terms
'These include extending consumption C to a vector in U(C). Then U(C) remains constant
and extending our two capital goods K and S to many capital goods. The C=0 result was proved
as an if and only if theorem. Our investing resource rents can be interpreted as aggregate or
combined investment being zero. Another extension was to treat this combined investment as
positive and constant.

have not been included in NNP in the modem stream of thought in national accounting, although
some observers recommend land revaluations be placed in NNP (see Hartwick [1992] and
references there). It turns out that these identical capital gains are in the WFK term on the right.
This suggests taht there is a more basic relation lying within ours above (it is C - K Kk) and that
the claims for C + W(t), with its capital gains on current oil stocks, as the 'formula" for NNP
suspect. We end this discussion with the observation that W(t) above is not constant for the
Solow [1974] constant consumption, zero net investment model. Thus maintaining capital value
constant (capital "intact"?) is a separate matter from maintaining consumption constant over time
or maintaining aggregate investment zero over time.
We now split the one world economy (X = 1) into two price-taking, trading countries
We set K, (t) = K2(t), given K, (t) + K2(t) = K(t) above. We set N, = N2 with Nb + N2 = N,
above. We make country 1 (CI) less endowed with oil stocks, that is Sj(t) < S2(t) with S, +
S2= S(t), above. We assume Sj(t) = S2(t) so that country 1 will import €(t) a small amount
of R(t) at each date. Since Kj = K2, efficiency requires that Ri(t) + €(t) = R2(t) where Ri(t)
is use of exhaustible resource from stock S^t). World prices are given from the one large
country scenario earlier.
(a) The oil importer (CI)
We have the output balance
C,(t) - F(K,(t), R,(t) + €(t» - K,(t) - €F^(t)	(4)
where €E FR(t) is payment for oil imports, G (t), and K, (t) is own investment in K, (t). In keeping
with each country "covering off the economic depreciation of its own oil stock S^t), we have

Ki(t) = X,(t) Rj(t) FR(t)	(5)
where Xt(t) is a fraction, endogenous and presumably near unity for €(t), small. Our task is to
characterize Xj(t) since (5) represents the "adjusted" invest resource rents rule. We also have
These derivatives will be the spree as those in (2). If one differentiates (4) and
(5) with respect to time and combines them, and uses (5) and (2), one obtains (see the procedure
in the Appendix):
C-(,) " (To® F"(t) - €(t*"(t)-
It follows that Ci(t) = 0 if
= FK(0,
_ c	(7)
where A,(R,) - (1 -X,)R,(t) ' (Recall that FR/FR = FK.) This condition for Cj = 0 is an
r percent rule in quantities, since FK(t) is the "rate of interest" here and €(t) and (1-X,(t)) K,0)
are quantities of oil. This r percent rule defines the time path of x,(t) and when combined with
(4) becomes the adjusted invest resource rents rule.6 Observe that if x,(t) = 1, then we would
have the unadjusted invest resource rents rule and (6) would become
c. = -€(t)FK(t).
This is a rendering of the result in Asheim [1986], namely, if country i invests its resource
rents, its C^t) will not be constant. In this case, importers Cl's Cj(t) is declining because it is
"under-saving" in revering its own economic depreciation in its stock S, 0) and in paying for
imports, €(t). Thus X,(0) must be greater than 1 and decrease toward 1 as time passes.
6 Asheim [1986] and Asheim [1994a] contain expressions for country i's savings to cause
Ci to remain constant. Their appearance and derivation are quite different from our adjusted
resource rents expressions yielding Cj = 0.

Observe that € (t)FE(t) is a quantity traded € (t) multiplied by a price change FR(t) and
is thus a terms-of-trade effect. € (t)FR(t) equals € (t)FR(t)FK(t). Hence the current decline of
C|(() from C|(0). given Xj(t) set at 1 is |ol € (t)FR(s)FK(s)ds where C^O) is a constant of
integration. Since e(t) = -se(t) where Sg(t) is the decline in C2's stock resulting from
exporting €(t), we have7
C,(0) - C,(t) - - JJ Se(s)F,(s)F,.(s) constant returns to scale in F(0, and
efficiency, one gets Cj + Wj(t) = W^Fj^t) or Cj + W(t) is interest on own wealth. This
balance relation simplifes to Cj = KjFk +	This contrasts with the closed economy
analogue in which C equalled KFk alone. Thus	- 1)R!Fr is income "withdrawn" from KjFk
to pay for the oil imports in Cj. The constant Cj is less than interest on local K. The capital
gains on oil stocks S^Fj^t) in W, again cancel with such gains in YV, (!) Kk and this suggests that
Cj + Wj(t) is not a satisfactory "formula" for NNP in this economy. More on defining NNP
(b) The oil exporter (C2)
C2'S situation is the mirror image of that of the oil importer. Now C2's savings to
replace her current oil use are	where R^t) is current oil extracted in C2. R2(t) -
7 The term - | J S(s)Fa(s)ds figured prominently in Hartwick [1994]. It was a key
measure of wealth. The analogous expression for machine capital was also prominent. See also
Solow [1986]. Here we are dealing with a gap between two flows, C^O) and Cl(t), not stocks.
Hence the appearance of FK(s) under the integral.

£ (t) is used in production in C2. Hence C2's replacement rule is
K2(t) = X2(t)R2(t)FR(t).	(8)
C2's value balance relation is
C2(t) = F(K2(t),R2(t) - €(t)) - X2(t)R2(t)FR(t) + €(t)FR(t).	(9)
We now differential (8) and (9) with respect to time, combine them, use (8) and (2) and obtain
(see the procedure in the Appendix):
e'(t) = (i-x-,(t))RI(e) F«(t) * €(t)t«(,)-	<10)
This is the same as (6) with a sign change. (10) yields our principal savings rule result, now for
C2, namely C2(t) = 0 if
. FK(t)	(")
€(t) "w
where	(11) characterizes the time path	in the investment
rule in (8). The rule is the same as that for CI in (7) except in our case Mt) will be less than
unity, and will increase toward unity as time passes.	was above unity and declined toward
unity as time passed.)
For X^t) set equal to 1.0, C2(t) > 0 by current capital gains € (t)FR(t). C2 is in fact
over-saving relative to a constant consumption scenario, and for this case
Cj(t) - 0,(0) - 6(s)F„(s)ds
• - jj Se(s)FR(s)FK(s)ds.
Our crucial adjustment terms X,(t) and ^(t) are, in view of (7) and (11), not independent.

(7) and (11) imply
_	- A,(Rt) = 0.	(12)
(12) indicates, roughly speaking, that for the case Xi = Xj = 1, Cl's under-saving matches C2's
over-saving. Xj(t) and X2(t) ( 9^ 1) in (12) reflect this balancedness of the adjustments for over-
and under-saving between our two countries. In fact X,(0 -l - i- Xj(t) because kj(t) + K2(t)
= K(t) where k(t) is investment in the closed economy case and Ri(t) + R2(t) = R(t).
Again for C2's wealth defined in W2(t) = K2(t) + S2(t)FR(t) we can obtain C2 + W2(t)
= W2(t)FR(t), i.e. the left hand side is interest on local wealth. Again capital gains on oil stocks
cancel on both sides to leave C2(t) = K2(t)FK(t) +	The oil exporter enjoys
a constant level of consumption above the income from interest on K2(t) because it receives extra
income from exporting oil. (Note that (l-X^t)) is positive.)
Corner Solutions and More than Two Countries
We have characterized the savings-investment rule which yields constant consumption
paths for our two-country, trading world with an essential exhaustible resource. It is an adjusted
invest-resource-rents rule. Our framework was two almost identical countries. This made trade
flows small so that neither country was specialized and the two country assumption allowed us
to sign the oil flows from exporter to importer. Clearly no part of our calculations depended on
our assumption of K: = K2 and S: SS S2 with S: < S2. Suppose, however, that C2 owned all
the oil. In this case RiFr is zero and weigting this by Xj does not yield more saving. (An
approach for this case is for C2 to have x*(t) = 1 and to transfer € (t)FR(t) to CI in order to
have Cj = C2 = 0. This was proposed by Asheim [1986].) However, as long as own oil use

Ri(t) is infinitesimally positive, XjRiFr (= Kj) can be defied and our two-country results go
through. (We require C, (() and Xj(t) to remain positive.) Thus as long as each country holds'
some positive stock S,(() at t, our adjusted saving-investment rule is relevant. (We require that
each country owns sufficient capital K to have income to pay for imports of oil in order to rule,
out comer solutions.)
With say three countries, the pattern of oil flows in trade becomes more complicated.
Suppose CI is an oil importer and C2 and C3 are potential exporters, being equally 'over"
endowed with oil stocks. Suppose Kj(0) = K2(0) = K3(0). In this case CI should import equal
amounts from both C2 and C3. It is not complicated to use our above reasoning to obtain
appropriate x»(t), x2(t) and x3(t) for this case. Our X(t) adjustment factors "work" for the many-
country case. Note, also, that standard national accounting procedures "work" for each country
in the trading system. In particular the value of exports equals the value of imports for each
country. Also domestic NNP in each nation equals consumption C^t) plus domestically financed
investment. That is, C^t) +	+ Xj(t) - Mi(t) is NNPi(t) for country i, where
is investment in i generated from current domestic production, X,(() is current
exports and M^t) is current imports. All components are denominated in the numeraire
commodity price, namely final goods output X,(() - M(() equals zero in our framework. In the
two country "example", M, (t) were oil imports and F,C) cl(t) Xt(t)R,(t)FR(t) were exports
of the final good. This yields NNPj(t) in value-added in CI as Fi(') - Mj(t). Note that
F1(K1,R1+ €) here is gross of oil import flow €. Hence F,() €FR(t) is Cl's valued-added
derived from domestic factors of production. Hence F,() - Mj(t) is domestic valued-added and
equals Cl's NNP(t).

In C2, NNP2(t) = C2(t) + X2(t)R2(t)FR(t) + X2(t) - M2(t). Given C2(t) in (9), it follows
that NNP2(t) = F2(K2,R2- £ ) + X2-M2 is value-added and X2-M2 = 0. In each country, the value"
of exports equals the value of imports in "free trade". World NNP equals NNPj(t) + NNP2(t)
which in turn equals world value-added F(K,R) * F(K,+K2, Rt+Ra) = F.^R. + G)
+F2(K2,R2 - €).
An Oil Exporter Facing Constant Prices and Interest Rates
Our analysis above involved two country trade with endogenous prices, including the
marginal product of capital, the interest rate. These prices were changing over time. Consider
the case of a price-taking "oil republic" (OR) a country living off exports of oil. This is an
autonomous problem. World oil prices will be constant at p per ton and the OR will have
unchanging extraction costs. e(R) for R tons currently extracted from its stock, S(t). We assume
e(0) = 0 and eR » de/dR > 0 and a d2e/dR2 > 0. There is a constant population (say just
consuming so that e(R) has no labor costs in it) and extraction is pursued to maximize discounted
net profit. Hence
P-CrCR) = f	(13)
is satisfied (the Hotelling r% efficiency rule), r is the constant discount (interest) rate. We
assume that the elders in this OR invest R(0- Ep~®r(0] abroad each period and live off current
interest income rH(t) plus current producer surplus L(t) = pR(t) - e(R(t)) - R(t)-[p-eR(t)]. That
is consumption
C(t) = rH(t) + ut).	(14)

Since interest rH(t) is being drawn off wealth abroad period by period, we have
H(t) = J J [p-eR(s)]R(s)ds + H(0). Thus4 H(t) = |>e,(t)]R(t) * If one differentiates (14) with
respect to time and uses (13) and H = [p-eR]R, one obtains C(t) = 0. Hence investing oil rents
abroad and living off the current interest on such, plus current producer surplus, yields a
constant consumption path.9 When S(t) declines to zero at say T, there will be H(T) dollars
invested abroad and C(T) will equal rH(T) which will be the same value as was being enjoyed
up to T. Clearly this policy of efficiently extracting oil and accumulating rent, net of interest,
abroad is a savings-consumption strategy identical with selling off Sq at market price
V(S°) = j; [pR * (t) - e(R * (t))]e"rtdt at t=0 and setting C(t) = rV(So). (*'s indicate optimal
values.) This is true because there are no market imperfections or uncertainties in our set-up,
and the problem is autonomous.
Our autonomous, constant price and interest rate model for a single oil exporter differs
from that for exporter C2 in our two country model in the sense that oil prices heed by C2
varied over time and generated terms of trade changes in €(t)FR(t). We had to "neutralize" these
capital gains enjoyed by C2 with an adjusted invest resource rents savings rule. The constant oil
price p eliminated capital gains in our autonomous model of the OR. In both models agents were
acting with perfect foresight so that they could anticipate price and interest rate changes and
optimize appropriately.
8	H(t) is another instance of the index number mentioned in footnote 1. Clearly this index
number is cumulative uncompounded or discounted rent. The lack of compounding occurs here
because potential interest accumulation is "neutralized" by the period by period drawing, off of
current interest on the capital value.
9	This argument was set out in detail in Hartwick and Hageman [1993] but no formal
demonstration of C(t) = 0 was given.

Concluding Remarks
There are indeed subtleties in moving from a unitized world system to a system of
countries trading flows from their exhaustible resource stocks and each maintaining consumption
constant over time. We derived the "wedges" that arise when our investment is financed in oil
importing countries by own resource rents and derived adjustment weights for the own savings
(resource rents). Oil importers should save more than resource rents ascribable to their own
exhaustible resource flows and oil exporters should save less than resource rents ascribable to
their own exhaustible resource case. Our subsequent model of a small open oil exporting nation,
a PRICE-TAKER at a constant interest rate and commodity prices, revealed no "wedges" that
were seen in the two country system with endogenous prices. Thus trade introduces subtleties
to the derivation of constant consumption paths because prices are indeed moving over time and
these price change effects show up as endogenous terms of trade effects. Relatively complicated
savings-investment rules are needed in each country to neutralize these terms of trade effects on
the simple invest-resource-rents rule, familiar for closed economies.
With our adjusted savings rule, we have been able to re-construct the closed economy
set-up, given multiple countries in trade. This was our goal. We also noted that no new valuation
issues were met and that traditional NNP measures "go through" in the open economy system.
We were also able to relate constant consumption paths to interest-on-wealth expressions. These
are compelling Hicksian notions of current national "income" being interest on national wealth.
However constant consumption paths are not reflections of constant wealth paths. In no case was
national wealth remaining constant over time.

Appendix: Derivation of Fqnation C.S1
One differentiates C(t) = F(K(t),R(t)) - K(t) to obtain
C= FkK + FrR - K(t)	(Al)
One differentiates equation (1) to obtain
K(t) = \(t)R(t)FR(t) + X(t)FR(t)R(t) + R(t)FR(t)X(t).	(A2)
In Al, for Fk substitute FR(t)/FR from (2) and	for K. Also for K(t) in Al substitute
the expression in A2. Al reduces to C = ________ F-(t), our expression in (3) in the text.
(l-X(t))R(t) R

Asheim, G.B. [1986], "Hartwick's Rule in Open Economies", Canadian Journal of Economics,'
19, pp. 395-402.
Asheim, G.B. [1988], "Rawlsian Intergenerational Justice as a Markov-Perfect Equilibrium in
a Resource Technology" Review of Economic Studies. 55, pp. 469-484.
Asheim, G.B. [1991], "Unjust Intergenerational Allocations", Journal of Economic Theory. 54,
pp. 350-371.
Asheim, G.B. [1994a], "The Concept of 'Net National Product' in an Open Economy",
Asheim, G.B. [1994], 'Net National Product as an Indicator of Sustainability", Scandinavian
Journal of Economics. 96 (2).
Becker, Robert [1982], "Intergenerational Equity: The Capital Environment Trade-Off, Journal
of Environmental Economics and Management. 9, pp. 165-185.
Cairns, Robert [1986], "Intergenerational Equity and Heterogeneous Resources", Scandinavian
Journal of Economics. 88, pp. 401-416.
Dasgupta, S., and T. Mitra [1983], " "Intergenerational Equity and Efficient Allocation of
Exhaustible Resources", International Economic Review 24, 1, pp. 133-153.
Dixit, A., "P. Hammond, M. Heel [1980], 'On Hartwick's Rule for Regular Maximin Paths of
Capital Accumulation and Resource Depletion", Review of Economic Studies, 47, pp.
Hamilton, Kirk [1993], "Sustainability, the Hartwick Rule and Optimal Growth", (mimeo).
Hamilton, Kirk [1993], 'The Greenhouse Effect and Constant Consumption Paths", (mimeo).
Hammond, Peter [1994], "Is There Anything New in the Concept of Sustainable Development?"
in L. Camriglio, L. Pineschi, D. Siniscalo and T. Treves, eds., The Environment After
Rio: International Law and Economics. London: Graham and Trotman.
Hartwick J.M. [1977], "Intergenerational Equity and the Investing of Rents from Exhaustible
Resources", American Economic Review, 66, pp. 972-974.
Hartwick J.M. [1978], "Investing Returns from Depleting Renewable Resource Stocks and
Intergeneration Equity", Economics Letters 1, pp. 85-88.

Hartwick, J.M. [1992], "Deforestation and National Accounting", Environmental and Resource
Economics 2, pp. 513-521.
Hartwick, J.M. and A.P. Hageman [1993], "Economic Depreciation of Mineral Stocks and the
Contribution of El Serafy" in Ernst Lutz, ed. Toward Improved Accounting for the
Environment Washington, D.C., World Bank.
Hartwick, J.M. [1994], "National Wealth and Net National Product", Scandinavian Journal of
Economics 96 (2), pp. 253-256.
Hayek, F.A. [1941], The Pure Theory of Canital. London: Routledge and Kegan Paul.
Hicks, John R. [1939], Value and Capital Oxford: Oxford University Press.
Hicks, J.R. [1942], "Maintaining Capital Intact: A Further Suggestion" Economics 9 , pp. 174-
Kemp, M.C. and N.V. Long [1982], "On the Evaluation of Social Income in a Dynamic
Economy" in G.R. Feiwel, ed. Samnelson and Neoclassical Economics Boston:
Kluwer-Nijhoff, pp. 185-189.
Lozada, Gabriel [1992], "Resource Depletion, National Income Accounting, and the Value of
Optimal Dynamic Programs", Department of Economics, Texas A&M University,
Pezzey, John [1993], "The Optimal Sustainable Depletion of Non-renewable Resources I:
Capital Resource Substitution", Department of Economics, University College,
University of London, (mimeo).
Pezzey, John and M. Toman [1994], "Sustainability and Neoclassical Economic Theory",
Pigou, A.C. [1935], "Net Income and Capital Depletion", Economic Journal 45, pp. 235-241.
Pigou, A.C. [1941], "Maintaining Capital Intact", Economica 8, pp. 271-275.
Robson, Arthur [1980], "Costly Innovation and Natural Resources", International Economic
Review. 21, 1, February, pp. 17-30.
Samuelson Paul A. [1961], "The Evaluation of 'Social Income'" Capital Formation and Wealth"
in F.A. Lutz and D.C. Hague, eds. The Theory of Capital New York: St. Martin's

Solow, R.M. [1974], "Intergenerational Equity and Exhaustible Resources", Review of
Economic Studies, Symposium, pp. 29-45.
Solow, R.M. [1986], "On the Intergenerational Allocation of Resources", Scandinavian Journal
of Economics, 88, pp. 1441-1449.
Solow, R.M. [1991], "Sustainability: An Economist's Perspective", The Eighteenth J. Seward
Johnson Lecture, Marine Policy Center, Woods Hole Oceanographic Institution.
Stiglitz, J.E. [1974], "Growth with Exhaustible Natural Resources: Efficient and Optimal
Growth Paths", Review of Economic Studies. (Symposium on the Economics of
Exhaustible Resources), pp. 123-137.
Weitzman, Martin L. [1976], "On the Welfare Significance of National Product in a Dynamic
Economy", Quarterly Journal of Economics. 90, pp. 156-162.

John Pezzey
Department of Environmental Economics and Environmental
Management, University of York, York YOl 5DD, U.K.
Abstract. In a simple growth model based on capital accumulation, non-
renewable resource depletion and constant technology, the maximum
sustainable utility level is strictly less than net national welfare; and a
growth path may be unsustainable, even though it has rising net wealth.
PV-optimal sustainability ('opsustimality') is defined as maximising the
present value of utility, subject to utility being non-declining forever. The
opsustimal path will either have a finite phase of rising utility, and followed
by a continuous transition to constant utility; or will always have constant
utility. Only on the opsustimal path does non-declining net wealth always
coincide with sustainability. Numerical simulations suggest that a rising
opsustimal path has higher utility than the PV-optimal path at the same time.
A consumption tax can achieve sustainability in a market economy only by
approaching a 100% subsidy, and neither a resource depletion tax nor a
resource stock subsidy can achieve sustainability.
This research was funded by award no. L320-27-3002 of the UK Economic and
Social Research Council's Global Environmental Change programme. I particularly
thank Mike Toman for helpful discussions at all stages of this paper, and I also thank
Geir Asheim, Geoff Lewis, Malcolm Pemberton and David Ulph for many useful
conversations and comments. The usual disclaimer applies.
Copies of the manuscript may be obtained from:
Prof. John Pezzey
University of York
York YOl 5DD

Peter W. Kennedy
Department of Economics
University of Victoria
Victoria, British Columbia
Canada V8W 3P5
First version: September 1993
This version: May 1994
Comments welcome.

This paper argues that sustainability is an inappropriate guiding principle
for the design of policy in a democratic setting. If policy is to be generally
implementable then its design must be consistent with the social setting in
which it is applied. In a democratic setting that permits individuals to act
and vote in accordance with views of distributive Justice that are not
necessarily consistent with sustainability, policy based on a sustainability
criterion will not be generally implementable. The alternative conceptual
framework I propose is based on three key elements: the relaxation of the
distinction between ethics and preferences; the recasting of an
intergenerational equity problem as an intragenerational allocation problem;
and the requirement that Intergenerational resource allocations be
intragenerationally efficient.

"The less you know about it, the better it sounds"
- Robert M. Solow (1991)
This paper argues that su stainability is an inappropriate guiding
principle for the design of policy in a democratic setting. This is not a
fashionable stance to take. The tide of academic literature and political
rhetoric seems to flow overwhelmingly in quite the opposite direction. Indeed,
it is difficult to find a recent policy statement of any kind that does not
make some reference to the term. The continuing absence of a clear definition
of su stainability does not seem to detract from its political appeal. A cynic
might even claim that the ambiguity of the notion is the main source of that
appeal. It is not my intention in this paper to attempt to resolve that
ambiguity by proposing yet another definition of sustainability. My primary
purpose is to argue that an emphasis on sustainability in the design of policy
in a democratic setting is misplaced.
The paper is organized as follows. In the next section I present a case
against the imposition of a sustainability criterion, in the design of policy
in a democratic setting. I then propose in section 3 an alternative approach
to i n t e r g e n e r a t io n a 1 resource allocation. I present a simple illustrative
application of the proposed approach in section 4. Section 5 concludes the
paper with a brief summary and some closing remarks.

Sustainability as a societal goal is based on a particular ethical,
stance. It rests on a view of distributive justice that gives at least some
consideration to the well-being of future humans and/or other elements of the
biosphere. Within that general class of ethical views there are many specific
positions that are consistent with some notion of sustainability. They range
from the purportedly non-anthropocentric view of the deep ecologists [Naess
(1986)] to some form of (anthropocentric) egalitarianism embodied in a
Rawlsian-type intergenerational social welfare function [Solow 1974]. It is
not my intention to present a taxonomy of these ethical positions nor to
debate their relative philosophical merits.* Nor do I intend to examine the
nature of the link between an ethical stance and the definition of
sustainability that it implies. I do not mean to imply that this is a topic
unworthy of examination. Indeed, it is my impression that the clarity of the
debate over competing definitions of sustainability could be enhanced if more
attention was paid to the axiomatic derivation of a proposed definition from.
the particular ethical position underlying it. Consider for example the
disagreement between those who define sustainability to mean the preservation
of the "natural capital stock" [Pearce (1988), Costanza and Daly (1992)] and
those who define it to mean the preservation of the "composite capacity to
produce well-being" [Solow 1992]. This debate has sometimes confused two
distinct Issues. The focus of the debate has been on the degree to which
manufactured and human capital can physically substitute for natural capital
in production. I believe this focus is misplaced. To the extent that there do
exist at least some physical substitution possibilities, there remains the
issue of whether or not it is ethically acceptable for the current generation
to make those substitutions (perhaps irreversibly) without the consent of

future generations. I believe this is the fundamental source of disagreement
in this debate. It is a conflict of ethical positions. If one adopts an
ethical position embodying an obligation to preserve for future generations
the same opportunities to choose that were available to the current generation
then one is obliged to make no irreversible substitutions. Whether or not such
substitutions are physically possible is then irrelevant.3
Making the ethical positions that underlie various definitions of
sustainability more explicit would help to clarify the differences between
them but it would not necessarily lead to a convergence of those definitions.
There is unlikely to arise a consensus about the precise meaning of
sustainability until a consensus is reached about the meaning of distributive
justice. No such consensus seems imminent. The main point I want to make in
this paper is that the absence of a universally accepted notion of
distributive justice has implications more fundamental than ambiguity in the
precise meaning of sustainability. It raises the question of whether
sustainability - no matter how it is defined - is an appropriate guiding
principle for policy at all.
Disagreement over the meaning of distributive Justice can and does extend
beyond the set of ethical positions that are consistent with some notion of
sustainability. Some ethical positions, do not imply sustainability in any
sense. Consider a deliberately extreme example. "It is perfectly just for my
generation to consume the entire resources of the planet at the expense of
future generations because we were here first". I do not think there are many
people who subscribe to this ethic (although I suspect there are some). The
point is that this ethical position and many less extreme ethical positions
are not consistent with sustainability. This means that the adoption of
sustainability as a guiding principle for policy is restrictive. What is the

basis for restricting the set of admissible ethics in this way? I submit that
the restriction is an arbitrary one. This is not objectionable in itself. All
guiding principles are fundamentally arbitrary. My objection to the
restrictiveness of sustainability is based on more pragmatic grounds.
I begin with the assertion that any guiding principle for the design of
policy should be consistent with the social setting in which it is to be
applied. By consistent I mean here that the resource allocation implied by a
policy prescription must be implementable given the social structure, in the
sense that there cannot exist a constitutional mechanism by which the
implementation of the allocation would be blocked. This is admittedly an
arbitrary stance to take. It reflects my own view that policy should be
designed with its eventual implementation squarely in mind. Not everyone may
agree with this position but it seems sensible to me. My purpose is to examine
what this criterion implies for the design of policy in matters of
intergenerational resource allocation.
Whether or not a particular guiding principle is consistent with a given
social structure will of course of depend on the nature of that structure. I
will focus here on democratic structures, and do so at a fairly abstract
level. I will assume a structure in which each agent is free to vote against a
candidate allocation in favor of some alternative if they so wish. This is a
reasonable approximation to a democratic system for the purposes of this
paper. I should stress that it is not the purpose of this paper to advocate
democracy over some other social structure. I focus on democracy only because
it is the system currently in place in many countries.
If a guiding principle for policy is to be consistent with a democratic
structure then it must be respectful of the voting rights of the members of
that democracy. In the democracies with which I am familiar, voting rights are

not restricted to those individuals who subscribe to an ethical position that
is consistent with s u s t a i n a b i 1 i t y . Individuals are permitted to vote
regardless of their ethical position, within certain limits. These limits are
often enshrined in a constitution. For example, in Canada and the United
States, a charter of rights and freedoms restricts the ability of the
collective to violate what are deemed to be the rights of the individual. In
Canada it is illegal to incite hatred of a particular social group. These
restrictions reflect the fact that there are some ethical positions that these
societies have deemed to be unacceptable. Ethical positions that are
inconsistent with some notion of sustainability may some day be included among
them. Currently they are not. To impose sustainability as a guiding principle
for resource allocation is to ignore in principle the voting rights of
individuals who subscribe to those ethical positions. This creates the
potential for the policies formulated under this guiding principle to be
systematically u n i m p 1 e m e n t a b 1 e . This does not mean that democracy is
necessarily inconsistent with sustainability. I have already noted that some
notion of sustainability could possibly be enshrined in a constitution without
necessarily rendering it undemocratic. Even without such a restriction, it is
possible that all Individuals with voting rights might happen to subscribe to
an ethical position that is consistent with sustainability. But to impose a
guiding principle that does not conceptually admit a converse possibility
fails my implementability criterion and is in my view inappropriate.
A conceptual framework for the analysis of intertemporal resource
allocation issues must be able to accommodate conflict among individuals with
voting rights if it is to be generally useful in the guidance of policy
formulation. This rules out the imposition of sustainability as a guiding
principle. More generally, it rules out the imposition of an intergenerational

welfare function on a planning problem designed to guide policy. To do so
implies that all of the agents in the modeled economy subscribe to the ethical
position embodied in the welfare function. This is true regardless of whether
or not the particular welfare function is consistent with sustainability. Such
modeling exercises should be interpreted only as positive analyses of how an
economy of agents with a common ethical position would optimally allocate
resources across generations. They are inadequate for guiding policy in a
realistic democratic setting because by construction they cannot in general
admit differences in ethical positions.
I have argued that it is generally inappropriate to impose sustainability
directly or to assume a particular intergenerational welfare function for the.
purpose of guiding policy in a democracy. So how should one proceed? In the
next section I propose one possible approach. I focus on the question of
intergenerational equity rather than a more general consideration of
distributive justice - that might Include, for example, the perceived rights
of other sentient beings - only because this issue has received the most
attention in the economics literature. The approach I propose could in
principle be extended to encompass broader issues of distributive justice.
There are three key elements of the conceptual framework I advocate for
addressing issues of intergenerational resource allocation in a democratic
setting. The first is the relaxation of the distinction between preferences
and ethics (or "social preferences").S This distinction is sometimes used to
justify the imposition of an intergenerational welfare function that applies

positive weight to future generations in an economy in which agents have
preferences defined only over their private consumption.6 The welfare function
is interpreted as a reflection of an ethical position that is conceptually
distinct from preferences. This distinction may or may not be a
philosophically interesting one; in any case it has little practical relevance
in a democratic setting in which a vote motivated by preferences is treated
equally alongside an observationally equivalent vote motivated by ethics. If a
conceptual framework is to have practical relevance then it cannot rest on a
distinction between ethics and preferences.
The second key element is the recasting of an intergenerational equity
problem as an intragenerational allocation problem. The interests of future
generations can be represented in a democratic setting only to the extent that
current generation agents act as their advocates. This necessitates a focus on
current generation agents. I have already argued that whether this advocacy is
motivated by ethics or preferences is practically irrelevant. The important
point to recognize is that this advocacy reflects some concern for the
well-being of those future generations. The well-being of current generation
agents can depend on the well-being of future generation agents just as surely
as it depends on their own consumption. A transfer of consumption from the
current generation to future generations can potentially make both generations
better off. A conceptual framework that places exclusive focus on private
consumption as the determinant of well-being is inappropriate. An equally
important point to note is that current generation agents can differ in the
degree to which they care about future generations. If these different agents
are entitled to act and vote in a democratic setting then there can arise a
conflict of interests among current generation agents. It is this conflict of
interests that must be accommodated in a conceptual framework for addressing

intergenerational resource allocation issues in a democratic setting. To frame
these issues in terms of a conflict of interest between generations is not
helpful for the purpose of guiding policy.
The third key element of the conceptual framework I advocate is a focus
on intragenerational efficiency in the assessment of an intergenerational
resource allocation. If each agent in the current generation is free to vote
against a candidate intergenerational allocation in favor of some alternative,
then i m p 1 e m e n t a b i 1 i t y of the candidate allocation requires that it be
efficient from the perspective of the current generation. If there exists an
alternative allocation at which all current generation agents are better off
than at the candidate allocation then the candidate allocation would be
unanimously rejected in favor of the alternative. The candidate allocation
cannot be implemented without a suspension of the democratic process.
Intragenerational efficiency is a necessary condition for implementability in
this setting.
It should be noted that intragenerational efficiency does not necessarily
imply intergenerational efficiency. Suppose, for example; that current
generation agents do not care at all about the well-being of future generation
agents, and that there exist two allocations between which current generation
agents are indifferent. If these two allocations have different implications
for the well-being of future generations then imposing intragenerational
efficiency alone will not guarantee intergenerational efficiency: the
allocation In which the future generation is worse off will pass the
intragenerational efficiency screen and could be chosen. But intergenerational
efficiency is only a relevant criterion if the current generations deems it to
be so, and it will be deemed so only if there are current generation agents
who care about future generations. If there is at least one such agent then

i n t e r g e n e r a t io n a 1 efficiency is implied by i n t r a g e n e r a t io n a 1 efficiency.
Therefore, nothing meaningful is lost by focusing exclusively on
intragenerational efficiency.
Intragenerational efficiency will generally not identify a unique social
optimum. There will generally exist a continuum of efficient allocations from
which one must be chosen according to some social choice rule. It should be
stressed that there is nothing internally inconsistent about this. In a
democratic setting the particular voting rule in place will determine which
allocation is chosen and this voting rule is taken as given for the purpose of
guiding implementable policy.
In the section following I present a simple illustrative example of the
approach I have proposed. This example falls far short of a general
formalization of the proposed approach but it does serve to demonstrate that
resource allocation rules consistent with this approach can be very different
from those implied by a sustainability criterion. A secondary purpose of the
example is to highlight a potentially important reason why inefficiency can
arise in intergenerational resource allocation, and that democratically
consistent policy can play in a role in correcting it.
Consider an economy with a sequence of identical generations each
comprising n agents. Each agent in generation t has utility function
u(ct> u ) defined over her own consumption and the utility of her
immediate heir • This representation of Intergenerational altruism has
been used extensively before in various contexts. It reflects

"non-paternalistic altruism" in the sense that utility is derived from the
well-being of another person rather from their consumption. I assume a
specific frictional form that is amenable to closed-form solution:
(1)	ut = log(ct) + 0ut+i
where 0e[O,1) reflects the agent's degree of concern for the well-being of her
heir. I assume initially that |3 is the same for all agents. It should be
stressed that fi does not represent the agent's private rate of time
preference. (In an extended model of multiple-period lived agents a separate
parameter for the rate of time preference would have to be introduced). Each
generation presumes that the preferences of their heirs will be the same as
their OWI1.
Consumption relies on a stream of benefits provided by natural capital,
and this natural capital becomes depleted if over-exploited. The transition
process for natural capital is given by
<2> V, x 
(3)	max log(cQ) + 01^
s. t. u = log(c ) + 0u Vt
_ t	t	t+1
R = (R - nc )(1+3) Vt
t+i	t	t
R given
Recursive substitution for allows this to be reformulated as a standard
infinite horizon dynamic programming problem:
(4)	max 3tlog(c )

s. t. R = (R - nc Hl+6) Vt
t*i	t	t
R_ given
The corresponding Bellman equation is
(5)	W(R ) = max 
economy could be implemented only if democracy is suspended.
Efficiency with heterogeneous agents
I now turn to a case with heterogeneous agents. Suppose there are two
types of agents: strongly altruistic (type 1) agents with an altruism
parameter £ , and weakly altruistic (type 2) agents with an altruism parameter
0 <0 • Let a denote the proportion of strongly altruistic agents. Agents of
type j presume that their heirs will also be of type j. The intragenerational
efficiency frontier for this economy can be derived from a planning problem in
which the utility of a representative type 1 current generation agent is
maximized subject to some lower bound on the utility of a representative type
2 current generation agent:
(8)	max log(c*) + 0xu|
s.t. ¦ log(c^) +	Vt
R = (R - anc1 - (l-a)nc2)(1+5) Vt
t+i	t	t	t
u ¦ u
R„ given
Recursive substitution for U | allows this program to be reformulated as
(9)	max Z" ahog(cj)
{ci> t-ori t
s.t. R » (Rt - anc) - (l-a)ncf)(l+«) Vt
t+i	t	t	t
O2l0g(C!} - "
Rq given
The key to finding a solution to this program is to recognize that at the
optimum the second constraint must be satisfied with minimal use of natural
capital. The linearity of the transition equation in this example makes it
straightforward to exploit this characteristic of the solution. The stock of

natural capital at any point in time can be conceptually split into two
12	12	1
separate stocks and such that R^R^+R^, where R^ provides a consumption
stream for type 1 agents and R^ provides a consumption stream for type 2
agents. It is then possible to find the minimal value of Rq needed to satisfy
the second constraint as the solution to
(10)	= min
0 RJ
s.t. o>g(cj) - a
R*+1 = (R* - n(l-a)c^) (1+5) Vt
This is just the dual of the standard dynamic programming problem described in
(4). It solves for a consumption path given by
(11)	c* - (l-02)RVn(l-a)
Solution of the system then yields the following minimum value for R^:
(12)	Rq = exp^(l-02)(u - 0)/32j
9 = Zr=0^1Og[(1~P2)P2(1+6)t/n(1"a)L The overall planning program can
now be reformulated as
(13)	max l" 0*log(cl)
 tm° 1 1
s.t. R*+i - (rJ - «nc*)(l+a) Vt
K'%- «*p[fi-0a>
Rq given
This is now a standard problem with solution
(14)	cl - (1-fi )Rj/na
»	1 b
Aggregate consumption for the current generation as a whole is given by
(15)	C„ - (l-0t) IR0-R»J ~ (1-eX
Whether or not this level of consumption is sustainable depends on 6, 0^ and
02> and on Rq. That is, the distribution of utility between strongly and
weakly altruistic agents within the current generation, as reflected in RQ»

will generally be important in determining whether or not consumption is
sustainable. In particular, s u s t a i n a b i li t y is less likely if the utility
distribution favors the weakly altruistic over the strongly altruistic. As
noted earlier, the utility distribution that arises in this economy will
depend on the particular voting rule in place. A natural distributional
arrangement to consider is one that provides each group with control over a
share of the natural capital stock proportional to its representation in the
population. This implies current generation aggregate consumption equal to
(16)	Cq = Cl-0i)aR(j + (1-02)(1-oc)Ro
This consumption level is sustainable if and only if + (l-a)|3_] Sl/( 1+6).
That is, if enough agents care enough about their heirs then the efficient
path based on proportional representation is sustainable. Otherwise it is not.
An alternative distributional rule is to vest control of the entire natural
capital stock in the hands of the group that constitutes a majority. The
preferences of this group would then dictate the consumption levels for all
agents in the economy. In this case the efficient path would be sustainable if
(3^1/(1+S) when a>l/2, and if £^1/(1+5) when a
capital stock is characterized by open access. In some cases it is feasible to
assign private property rights over natural capital (such as with some fish
stocks and trees) but this is not always possible. In this section I derive
the Nash equilibrium consumption path when there is open access to natural
capital. I focus on the homogeneous agent case since it illustrates the
consequences of open access most simply.
Each agent k in generation t perceives the following transition
(17)	Rt+1 = (Rt - c* - Cfk)(H-5)
where is consumption by agents other than agent k in period t; it is taken
as given by agent k since there is open access to the natural capital stock. I
confine consideration to rational expectations equilibria. This means that
each agent in the current generation correctly anticipates the equilibrium
implications of her consumption decision for the utility of her heir and
correctly anticipates that all of her descendants will do the same. The choice
problem for agent k in period t can therefore be formulated as
(18)	max log(ct) ~ 0ut^
. s. t R - (R - c* - C"k)(l+«) Vt
t+i	t t t
where	is equilibrium per capita consumption in period t + i. This is not a
standard dynamic programming problem. However, a solution can be found by
positing a time-invariant equilibrium consumption rule of the form C	^•
and verifying that this in fact solves the program.9 Solving the problem in
this way and imposing symmetry yields the following equilibrium aggregate
consumption path:
(19)	Ct - Rtn(l-0)/[0 + (l-0)n]

Comparing this equilibrium path with the efficient path reveals that
equilibrium consumption is too high in early generations. This inefficiency is
due to the —open access to natural capital. Each individual in period t
recognizes that the natural capital she leaves intact for her descendants will
also be available for consumption by the descendants of her fellow citizens.
She cannot protect the legacy she leaves. Recognition of this fact leads her
to leave less than she otherwise would.10 It should be noted that this result
is sensitive to the form of the utility function. The inefficiency could in
principle go the other way: the open access could induce an agent to consume
less than is efficient in an attempt to compensate for the fact that the
legacy she leaves for her heir may be depleted by others. The inefficiency of
the Nash equilibrium implies a role for policy intervention. In this example
policy intervention is needed to reduce the rate of consumption but it is
conceivable that intervention in the opposite direction may be needed. In
either case the appropriate role for policy in a democratic setting is to
ensure efficiency rather than to impose SUStainability. **
In this paper I have argued that su stainability is an inappropriate
guiding principle for the design of policy In a democratic setting. If policy
is to be generally implementable then its design must be consistent with the
social setting in which it is applied. In a democratic setting that permits
individuals to act and vote in accordance with views of distributive justice
that are not necessarily consistent with sustainability, policy based on a
sustainability criterion will not be generally implementable. The alternative
conceptual framework I have proposed is based on three key elements: the

relaxation of the distinction between ethics and preferences; the recasting of
an i n t e r g e n e r a t io n a 1 equity problem as an i n t r a g e n e r a t io n a 1 allocation
problem; and the requirement that intergenerational resource allocations be
intragenerationally efficient.
To recognize that different individuals in a democracy can legitimately
hold different and incompatible views on distributive justice is not to say
that there is no place for continued debate about the meaning of distributive
justice. Such debate is surely valuable. Economists can and should play an
important role in that debate. But is it essential that economists carefully
distinguish between their philosophizing about the meaning of distributive
justice and the more mundane business of guiding implementable policy.

Chichilnisky, Graciela (1993), "What is sustainable development?", mimeo,
Columbia University.
Costanza, Robert and Herman E. Daly (1992), "Natural capital and sustainable
development", Conservation Biology, 6, 37-46.
Hartwick, John (1977), "Intergenerational equity and the investing of rents
from exhaustible resources", American Economic Review, 67, 972-974.
Howarth, Richard B. and Richard B. Norgaard (1992), "Environmental valuation
under sustainable development", American Economic Review, 82, 473-477.
Levhari, David and Leonard J. Mirman (1980), The great fish war: an example
using a dtnamic Cournot-Nash solution, Bell Journal of Economics, 11,
Marglin, Stephen A. (1963), The social rate of discount and the optimal rate
of investment, Quarterly Journal of Economics, 77, 93-111.
Naess, Arne (1986), "The deep ecological movement: some philosophical
aspects", Philosophical Inquiry, 8, 10-31.
Pearce, David (1988), Economics, equity and sustainable development, Futures,
Pearce, David and R. Kerry Turner (1990), Economics of Natural Resources and
the Environment, Johns Hopkins University Press, Baltimore.
Ray, D. (1987), Nonpaternalistic intergenerational altruism, Journal of
Economic Theory, 41, 112-132.
Solow, Robert M. (1974), Intergenerational equity and exhaustible resources,
Review of Economic Studies, Symposium Issue, 29-45.
Solow, Robert M. (1992), "An almost practical step toward sustainability",
Invited Lecture for the Fortieth Anniversary of Resources for the Future,
Resources for the Future: Washington, DC.

I am grateful to Malcolm Rutherford and participants at the 1993 Meeting of the
Canadian Resource and Environmental Economics Study Group for helpful comments
and discussions.

1Pearce and Turner (1990, Chapter 15) provide a concise review of some of the
ethical views that are most commonly cited to motivate sustainability.
See Chichilnisky (1993) for some recent work in this direction.
3As an aside, it seems to me that a requirement to preserve the same
opportunities to choose is impossible to fulfill. The second law of
thermodynamics renders it physically impossible to leave the planet exactly as
we found it over a sufficiently short time interval. To the extent that the
next generation follows the current generation instantaneously (there are a
continuum of generations) then they cannot inherit exactly what the current
generation inherited.
4This of course begs the question of how this unacceptability is decided upon.
Important as this question is, It is not one on which I need comment here. My
scope is more narrow. I am concerned only with the consistency of a guiding
principle with the democratic constitution currently in place. The process by
which that constitution is established or revised is not directly relevant to
that issue.
See Sen (1977) for a discussion of this distinction.
6See Howarth and Norgaard (1992) for an example of such a model.
7See Ray (1987) for a discussion of this representation.
8 While I later allow agents to differ according to the size of their 0, they
could also conceivably differ in their beliefs about what future generation
preferences will look like. It should in principle be possible to extend
consideration to this issue within the same basic framework.

See Levhari and Mirman (1980). It should be noted that this approach does not
guarantee that the posited equilibrium is unique.
10Levhari and Mirman (1980) derive an exactly analogous result in the context
of a fish war between two infinitely-lived national governments.
"Marglin (1963) identifies a different potential source of inefficiency in
intergenerational resource allocation. In a model with paternalistic altruism
Marglin shows that if the consumption level of the next generation as a group
is a public good for current generation agents then there may be too little
saving in the economy due to free-riding. The equilibrium discount rate will
consequently be too high. This public good problem could co-exist with an open
access problem.

An Efficiency Argument for Sustainable Use
Joaquim Silvestre
Working Paper Series No. 94-11
September 1994
Note: The Working Papers of the Department of Economics, University of California, Davis, are preliminary materials
circulated to invite discussion and critical comment. These papers may be freely circulated but to protect their
tentatitve character they are not to be quoted without the permission of the author.
This paper is available in Property Relations, Incentives, and Welfare, edited by John E.
For USA only:
Published by St. Martin's Press, Incorporated
For the world excluding the USA:
Published by Macmillan Press, Ltd. (London)

"An efficiency argument for sustainable use."
Joaquim Silvestre
Deaprtment of Economics
University of California
Davis, CA 95616
Sustainability is often viewed as a moral obligation to future generations.
The paper adds an argument for sustainability that is entirely based on efficiency
and is free from distributional considerations.
Many natural environments admit two uses: (i) a destructive use, where the
environment is converted into a private good, used by (a fraction of) the present
generation: and (ii) a nondestructive use. where the environment is maintained in
its natural form: the environment is thus a public good, useful to both present
and future generations. The nondestructive use can often be defended purely on
efficiency grounds: this is made precise in a quasilinear model of a finite number
of overlapping generations. Efficiency is there equivalent to the maximization of
surplus, i.e., the maximization of the sum of the benefits over generations minus
the sum of costs.
Two qualifications. First, large transfers of wealth from future to
present generations must be physically possible Second, if individual discount the
future, then efficiency requires the maximization, not of the sum of utilities,
but of a discounted sum of utilities. Efficiency can dictate conservation in
Society I and destruction in Society II for two societies that are identical
except that individuals discount the future in Society. This is somewhat
surprising in overlapping generation models.
D61, H41, H43, H82, Q20, Q23, Q26, Q30, Q38.

Anne Grambsch
Economic Analysis and Research Branch
Office of Policy, Planning, and Evaluation
U.S. Environmental Protection Agency
401 M Street, SW
Washington, DC 20460
Phone: 202-260-2782
Fax: 202-260-5732
R Gregory Michaels
Abt Associates Inc.
4800 Montgomery Lane, Suite 600
Bethesda, MD 20814
Phone: 301-913-0537
Fax: 301-652-7530
Paper presented to the 1994 AERE Workshop "Integrating the Environment and the
Economy: Sustainable Development and Economic/Ecological Modeling." June 6.
Boulder, CO.
The views expressed in this paper are the authors' own and do not represent the
official position of the Environmental Protection Agency.

Conventional economic accounting, including accounts for assets, income and product as
well as input-output accounts, is practiced by most nations of the world because it supports
economic policy in several important ways. The national accounts provide measures of a nation's
wealth, summary statistics regarding overall economic performance, and an instantaneous but static
picture of the flows of economic activity. The description provided by the national accounts of the
relationship between outputs of economic processes-the production of goods and services-and
economic inputs supporting these processes is critical for economic analysis and policy. An
understanding of these mechanisms is essential if the government wishes to influence economic
activity predictably. Since their introduction over 50 years ago, the national economic accounts
have evolved to respond to changes in the structure of the economy and the analytic and data needs
of policy makers.
While there is widespread agreement that the standard national economic accounts provide
invaluable information on economic activity, there is also recognition that the standard measures fail
to capture other factors which influence social welfare, such as the quality and quantity of
environmental resources and amenities. Changes in the environment and in natural resources have
not been explicitly included in the economic accounts, principally because ways to measure these
changes monetarily were not apparent and thus integrating them with other entries in the accounts
was impossible. This neglect of environmental and natural resource activity impairs the functions of
the national accounts. First, it fails to include a potentially important category of a nation's wealth
and thus future production and consumption possibilities. Second, it provides an overly optimistic
picture of economic perfomance in that it omits the effect of negative environmental externalities,
such as pollution, on current well-being and the effect of natural resource degradation on future
well-being. Finally, the ability to picture relationships between outputs and inputs is degraded since
the environment and natural resources generate important input and output services that compete
with and substitute for the monetized services that are covered in the conventional accounts.
Integrated environmental and economic accounting systems attempt to address these
shortcomings of the national accounts. Three major objectives of the such integrated systems are to:
1) provide an accounting of the interaction of the economy and the natural environment, 2) address
sustainable development concerns through proper accounting of both man-made and natural assets,
and 3) develop environmentally adjusted measures of product (i.e., a "green" GDP) and income,
which may inform on and serve as guidance toward sustainable development policies. The
particular system proposed by the United Nations (System for Integrated Environmental and
Economic Accounting or SEEA) is designed to expand upon and complement existing economic
accounting systems (Systems of National Accounts or SNA) with regard to costing the use
(depletion) of natural resources in production and to satisfy final demands, and recording net
changes in environmental quality associated with production consumption, and natural events on the
one hand and environmental protection and restoration on the other. Using the SNA as the basic
framework for an integrated environmental and economic system is not meant to lead to an
exclusively economic view of environmental concerns. Rather, it is intended to introduce
environmental issues into mainstream economic analysis and policy making through the use of a
common framework. Ultimately, the integrated system is intended to provide a suitable database for
analyzing sustainable development policies and options.
Revised Draft - Do not cite or quote

It should be noted that extending the framework to incorporate environmental concerns is a
separate issue from the failure of SNA-type aggregates (e.g., GDP) as welfare measures. Gross
Domestic Product is a measure of the market value of economic production; modifyng it to reflect
environmental issues will not make GDP a welfare measure. Further, the SEEA also excludes
phenomena which take place entirely outside the economic system. For example, the generation of
solid waste and gaseous emissions by natural sources and associated assimilation and transformation
by ecosystem processes would not be included in the SEEA. Rather, the architects of the proposed
system believe that such phenomena are better dealt with by complementary biophysical resource
accounts, environmental statistics, and environmental monitoring systems with appropriate linkages
to the SEEA. As a result, the SEEA is primarily concerned with the interactions between the
environment and economic production, value added expenditures, and tangible wealth.
Over the past decade, a series of international workshops and meetings and a growing body
of research and implementation efforts, culminated in the publication of a set of guidelines for
integrated environmental and economic accounting (United Nations, 1993). With a few notable
exceptions, environmental economists have not played a large role in the development of the
accounting framework. Our purpose in this paper is to stimulate a discussion on integrated
environmental and economic accounting within the environmental economics community and to
challenge members to contribute to the implementation effort drawing on the analytical skills and
insights gained from years of studying environmental issues. The first section discusses conceptual
issues associated with implementing an integrated environmental and economic accounting system,
paying particular attention to how standard welfare analysis concepts can be translated into the
measures needed for an accounting effort. The basic structure of the SNA and the proposed
extensions to reflect environmental concerns are summarized in the following sections. Results
from a preliminary implementation of the SEEA for the U.S. are presented in. the fourth section.
Given the major omissions and measurement difficulties, these results shouk^Be taken very
seriously. Rather, they are intended to illustrate the types of protocols that are necessary for
implementing the system, as well as possible adjustments to summary aggregate measures. Finally,
summary conclusions and possible extensions are described in the last section.
Conceptual Issues in Implementing Integrated Environmental and Economic Accounting Systems
Conceptually, the natural environment can be viewed as an asset or reproducible capital
good which provides a flow of goods and services to the economy over time. When economic use
of the "output" of the natural environment results in a permanent or temporary reduction in the
quantity of the asset this quantitative reduction is termed depletion. When use results in a reduction
in the quality of the natural asset this use is termed degradation. Further, economic use of natural
assets also results in feedback effects: depletion of natural resource stocks reduces future flows of
goods and services from the environment, degradation due to the disposal of residuals results in
costs imposed on third parties. In addition, firms and households may be required to make
expenditures for pollution abatement and control. Obtaining a comprehensive picture of advantages
and disadvantages of the economic use of the environment in production activities requires estimates
of all of these items.
Constructing accounting entries which maintain comparability with the SNA requires the
market prices or proxies of market prices, i.e., marginal values exclusive of consumer surplus, and
associated quantities. Market values can be used for those natural assets which are connected with
Revised Draft - Do not cite or quote

actual or potential market transactions, such as subsoil minerals and managed forests. However,
environmental functions of these natural assets (e.g., habitat provision and C02 sequestration) are in
most cases not reflected in the market value of the asset. Directly observable market values for
environmental assets (e.g., air, undisturbed ecosystems) do not exist because there are no market
mechanisms to convert the value of their generated services into observable market prices. One
approach (Peskin, 1989) suggests treating environmental assets as if their services were in fact
marketed. However, since it is necessary to record transactions from both "buyers'" and "sellers'"
points of view, it raises questions regarding the service quantity to be valued (i.e., the current level
of discharges into the environment or the current level of environmental services provided given
existing environmental quality) and the appropriate valuation concept to be applied to this service
quantity. The second issue arises, of course, because there are no market forces driving buyers' and
sellers' marginal valuation to an equilibrium.
The standard macroeconomic analysis of externalities can be used to illustrate these issues.
To simplify the discussion, we assume: 1) there are only two users of an environmental asset, 2)
their uses of the asset are mutually conflicting and 3) there is an insufficient quantity of the asset to
satisify both user's demands. For example, industry may seek to use the air or water to dispose of
wastes, and households may seek to use the air or water to support certain levels of health or
recreation. The more air or water is used to dispose of wastes the less it is able to support
specified levels of health or recreation. Scarcity of the resource ensures nonzero marginal values.
The traditional focus of welfare analysis on maximizing total net social benefits leads to a
determination of optimal environmental asset use where marginal benefit equals marginal cost. The
point we wish to make here is that, at any particular level of air or water quality, there are
reciprocal benefits and costs for each user.
From industry's viewpoint there are costs which have already been incurred due to
government regulations which restrict their access to the asset (i.e., environmental protection costs)
and benefits associated with using the air or water to the current allowable level. This benefit could
be viewed as a potential cost to industry, i.e., the potential costs of future air or water regulations if
they further restrict industry's access beyond current levels. The first type of cost is recorded in the
economic accounts, although their separate identification and reporting as distinct accounting entries
is relatively recent, 1 while the second type of prospective cost is not included in the accounts. As
noted in Peskin (1989), a complete accounting of all sources of income would include such costs
since they measure the value of a nonmarketed factor input. In essence, industry is receiving a
"subsidy" from the environment in the form of unpaid environmental waste disposal services.
For households, there are health and welfare "costs" (damages) or environmental
repercussion coats associated with the current level of air and water quality, i.e., from being denied
access to clean air and water. However, to the extent there has been some improvement in
environmental quality due to abatement efforts, some of these damaging effects have already been
avoided and benefits realized. Conventional accounts implicitly include damages which manifest
themselves in markets (e.g., medical care expenditures) although they are not separately identified.
1 Data on environmental protection expenditures has been collected since 1973. The
Bureau of Economic Analysis has published a series of articles entitled "Pollution
Abatement and Control Expenditures" in the Survey of Current Business, various issues.
Revised Draft - Do not cite or quote

Nonmarket service values of air and water, both potential and realized benefits (damages), are not
accounted for in conventional accounts. Graphically, we can depict the situation as in Figure 1.
Figure 1.





"Increasing Waste Disposal Services
Increasing health/recreational services
Them is a total of OD of the environmental asset. From the industry viewpoint curve aC
represents the marginal benefits of being allowed access to the asset or the potential marginal costs
if they are not allowed access to the asset. As drawn, industry would not use the entire asset (OD).
From the household point of view, Ae represents the marginal benefits of being allowed access to
the asset or the potential marginal "costs" (i.e., damages) if they are not allowed access to the asset
As drawn industry could use OA of the asset without causing any damage to households (i.e., them
is a non-damaging threshold). We assume that government regulations allow OB of the asset to be
used by industry.
Associated with the current levels of asset use are two "prices": the marginal
benefit/cost (b) for households and the marginal benefit/cost (c) for industry. Absent an
equilibrating force such as a market, we would not expect in general for b = c. Using marginal
values, we can define the following three areas:
¦	BDfi: The market valued benefits received by households from current policies (i.e.,
from denying access to industry)
¦	OBhc: The market valued benefits received by industry (or prospective future costs)
from being allowed access to the asset.
¦	OBib: The market valued costs (damages) imposed on households.
Revised Draft - Do not cite or quote

The area BCh represents the actual pollution control costs incurred by industry, which are
already recorded in GDP. These costs are to be separately elaborated in the SEEA, in aggregate and
at the sector level. In extended versions of the SEEA, these costs are separated into external and
internal environmental protection activities and a symmetric input-output table developed.
The SEEA captures the notion of competing uses of environmental assets by distinguishing
the concepts of "costs caused", i.e., costs associated with economic units actually causing or
potentially causing environmental deterioration by their own activities, and "costs borne", i.e., costs
which are borne by economic units independent of whether they have actually caused or potentially
caused environmental deterioration. In benefit-cost terms, costs caused would correspond to costs,
costs borne would correspond to benefits. For example, consider a benefit-cost analysis of a
proposed policy to reduce lead emissions to a specified non-damaging level. The analyst might
estimate pollution abatement control costs which would fall on the industries emitting the lead and
the benefits of expected improvements in human health and welfare which households would enjoy.
For environmental accounting purposes, industry is causing environmental deterioration or a
reduction in the service flow from the natural asset air. Households are bearing the repercussions
associated with the degradation and presumably would be willing to pay (in terms of reduced
consumption) to avoid this burden.
These two valuation concepts correspond to two possible approaches to environmental
accounting: 1) accounts which describe the environmental impacts of economic activities, and 2)
accounts which describe the condition of the natural environment and its effect on human health and
welfare. The latter is a much more complex undertaking since it requires substantially more
information (ecological processes, impacts on health, behavioral responses, etc.) which must take
into account time and space dimensions. An additional complication is that costs borne will
normally require some type of contingent valuation (CV) to estimate the value of adverse health and
welfare effects associated with environmental degradation.
The preferred concepts in SEEA are oriented towards "costs caused", i.e., SEEA focuses on
which economic agents/activities are responsible (accountable) for deteriorating the natural
environment. This focus is driven by both data availability and the relevance of accountability in
integrated policies and stainable development management Recent attempts to implement the
SEEA have focused on quantitative and qualitative asset use associated with economic production
(OB), which is valued at cost (c). Thus, in the SEEA hypothetical (imputed) costs (OBhc) play a
prominent role. The cost of using the natural environment is extended to include costs that would
have been incurred had the environment been used in such a way that its future use was unimpaired
(i.e., the costs that would have been required to maintain natural capital intact). This approach
parallels the treatment of man-made capital in the conventional accounts: consumption of fixed
capital represents the monetary amount necessary to maintain the current level of man-made assets
intact and thus allowing for sustainable fixture income flows. In addition capital consumption
estimates the current costs of using fixed assets in production and could be interpreted as
constituting a payment to the services of man-made capital.
Relatively little emphasis has been placed on the costs borne concept in the SEEA (area
OBib in the diagram). As noted above, determining the impacts of a depleted and degraded natural
environment will be a difficult undertaking. However, the SEEA recognizes that it is important to
take into account the values accorded to the environment by those (industry and households) who
bear the consequences of environmental degradation and depletion. For production activities, costs
Revised Draft - Do not cite or quote

borne are to be estimated using actual or imputed market values. That is, the reduction in the
market value of a natural asset due to quantitative depletion or qualitative degradation (which may
be partly counterbalanced by restoration activities of the government) would be treated as a cost and
integrated into the production accounts of the SEEA. For households, the SEEA acknowledges that
a significant part of the valuation of repercussions associated with a deteriorated natural environment
will require willingness to pay or CV methods.
Many accountants doubt whether it is possible to determine monetary values for preferences
in the absence of markets (see Hueting, 1980, chap. 4.5), and remain skeptical about the feasibihty
of applying CV methods in a national accounting framework. The SEEA also expresses
reservations about CV, stating "Use of the contingent valuation approach in environmental
accounting is still in an exploratory stage. Further research and discussion are needed. The
following proposals therefore provide only a generic framework for further experimentation with
this valuation method and related accounting procedures" (UNSO, 1993). This reluctance on the
part of accountants to use CV is not surprising given the focus of conventional economic accounting
on production, transactions, and costs. Accountants would be the first to acknowledge that the
accounts record market transactions, not values, and hence accounting aggregates are not measures
of welfare.
In summary, both the costs caused and costs borne concepts involve actual costs, which are
recorded in the SNA although not separately identified and imputed environmental costs, which are
recorded as additional cost items in the SEEA. Nonmarket valuation approaches will be required to
estimate imputed environmental costs although the costs borne concept would require additional,
relatively more controversial, alternative valuation concepts such as CV. The SEEA recommends
using maintenance costs to estimate costs caused since the data are more reliable and available and
responsible economic agents are identified and held accountable. It is also similar to approaches
followed for other non-marketed goods and services. For certain non-marketed goods and services
(e.g., subsistence farming agricultural products, own-account production of housing services)
conventional accounts base valuation on prices of similar products which are marketed (e.g., market
prices of agricultural products, market housing rentals). However, where such market information is
lacking, non-market goods and services are valued at cost (e.g., government services).
United Nations System of National Accounts (SNA)
The revised SNA constitutes Version I of the SEEA. The parts of the SNA that form the
conceptual basis for the development of the SEEA are the supply and use table of produced goods
and services and the non-financial asset accounts, which includes both produced (man-made and
natural) and non-produced natural assets. These two segments of the SNA are described below.
Supply and Use accounts
The SNA supply and use accounts record production activities which took place during the
accounting period. The total production by the economy, augmented by production from the rest of
the world (i.e., imports) is then available to be used to satisfy intermediate and final demands. The
supply and use accounts attempt to measure these transactions, recording them from both
transactors' points of view. The supply-use accounting identity is:
Revised Draft - Do not cite or quote

(1) P-M = C/^C + l + Ex
where P = production, M = imports, Ci = intermediate consumption, C = final consumption, I =
gross capital formation (or Investment) and Ex = exports. A second identity defines gross product
or value added (Y) as the difference between total production (P) and intermediate consumption
(2)	Y = P-Ci
When this income identity is substituted into (1), the familiar domestic-product identity emerges:
(3)	Y = C + I +(EX - M)
Asset Accounts
The SNA asset accounts record all stocks and flows associated with changes in those stocks
which are defined as part of the economy. Valuation is normally restricted to market values,
although certain nonmarketed goods and services are included which are valued either on the basis
of prices of similar products and services that are marketed (e.g., owner occupied housing) or at cost
(e.g., government services). Relationships between the environment and the economy are viewed
from an economic perspective only, i.e., the environment is viewed in terms of its use in economic
production. However, a key criticism of conventional accounting is that the use of environment is
not treated as a cost and so is not reflected in summary measures such as Net Domestic Product
The SNA asset accounts categories include opening and closing stocks, capital formation,
other changes in assets, and revaluation (holding gains/losses). These accounts explain changes
between opening and closing stocks associated with flows during the accounting period. For both
produced and nonproduced economic assets, the balances are defined as follows:
(4)	K, = Ko + I- Depr. + OC + Rev
where K, = closing stocks, K<, = opening stocks, I = gross capital formation Depr. = consumption of
fixed capital (or depreciation), OC = other changes in assets, and Rev = revaluation (i.e., holding
gains or losses).
Certain elements of the capital formation account (i.e., gross fixed capital formation and
consumption of fixed capital) intersect with the supply and use accounts described above. Gross
capital formation refers generally to produced assets, although it also includes some additions to
non-produced assets (e.g, reforestation). Gross capital formation is included in calculations of GDP
as shown in equation (3). Subtracting consumption of fixed capital from both sides of equation (3)
yields the net domestic product identity:
(5)	Y« = C + In + (Ex - M)
where Yn = net product and In = I - depr. or net capital formation. Net domestic product may be
considered a measure of Hicksian income (i.e., the maximum amount of income a nation can
consume which will leave the nation as well off at the end of the period as it was at the beginning
Revised Draft - Do not cite or quote

of the period). Hicksian income is thus "sustainable" and many argue that measures such as NDP
represent a first step in developing sustainable development indicators. Consequently, many in the
environmental policy community have focused on "greening" the NDP.
During the recent revisions to the SNA (United Nations, 1992) it was recognized that a more
detailed description of assets was required. This was accomplished in part through the expansion of
the asset boundary. In the revised SNA, the definition of assets was expanded to include all assets
over which ownership rights can be enforced and which provide economic benefits to their owners.
Conceptually, the asset boundary includes natural assets, both those which are owned and managed
or cultivated directly by humans and those which are owned but not managed or cultivated. Within
the revised SNA asset boundary, two types of nonfinancial assets can be distinguished: 1) produced
assets and 2) non-produced assets. Produced assets may be man-made assets (e.g., buildings,
equipment, inventories of harvested crops) or developed natural assets (e.g., cultivated biological
assets such as livestock for breeding, fish stocks, orchards, and timber tracts). Non-produced natural
assets include land, subsoil assets, uncultivated biological assets such as wild fish and forests, and
water resources.
The other changes in assets account is particularly important for environmental analysis
since it contains information on the impact of the environment on natural and other assets. This
account contains economic appearance of non-produced assets (e.g., additions to proven oil reserves,
additions to timber reserves through the logging of virgin forests), natural growth of uncultivated
biological resources, and economic disappearance of non-produced assets (e.g., depletion of subsoil
assets and forests, degradation of non-produced assets). However, these entries are not recorded as
part of the production accounts and therefore do not affect the calculation of GDP or NDP. For
example, if a site is degraded because it is used to dispose of solid waste, the market price of the
natural asset (land) may reflect this degradation which would be recorded as other changes in
assets. Essentially this reduction in the market price of the land is not considered a cost of
The use of the terms "economic appearance" and "economic disappearance", especially with
respect to non-produced assets, reveals one difficulty facing conventional national accountants.
Natural resources "appear", not as a result of economic activity but rather as a result of ecological
processes. Consequently, they are considered "free gifts of nature". Of course since they are "free"
they are presumably available in unlimited quantities. Expenditures to develop these gifts (e.g.,
unproved mineral reserves can be developed into proved mineral reserves) are recorded as gross
capital formation and the natural resource is considered a non-produced economic asset. What is
not clear is how to record the additions to (appearance ol) the natural resource stock itself.
Similarly, these resources "disappear" as they are used up. Unlike appearance, however,
disappearance is clearly tied to economic activity, which suggests that an entry to reflect this
depletion should appear in the production accounts in a way that parallels depreciation of man-made
capital. Initially, the U.S. national accounts did include such entries beginning in 1942.
Dissatisfaction with this asymmetric treatment of natural resources (i.e., entries for depletion but no
entries for additions), led to the removal of depletion from the national accounts in 1947.
A simplified SNA supply and use account with asset balances is depicted below based on
these accounting identities and protocols.
Revised Draft - Do not cite or quote

SNA Supply and Use Accounts with Asset Balances for Economic Assets
(SEEA Version I)

Economic Asset Balances
Rest of
Open Stocks
Economic Supply

Economic uses
Gross Product

Capital consumption


Net Product/ Net Capital


Other Changes
in Assets

Holding gams/losses

Closing Stocks

United Nations System for Integrated Environmental and Economic Accounting (SEEA)
In general, the SEEA advocates following principles and rules established for national
economic accounting systems. For example, the SEEA observes the SNA'S production boundary,
uses SNA methods of analyzing costs and outputs and incorporates the same accounting identities
between supply and use of products and between value added and final demand. This allows the
integration of environmental information into established economic accounting systems. The
possibility of extending the framework to include environmental welfare effects (e.g., damages
associated with the impairment of human, health, recreation, and other aesthetic values) is also
Distinguishing the boundary between economic and ecological systems is difficult and
subject to a substantial amount of controversy. From an ecological point of view, the economy is
part of nature; integrated accounting systems should thus determine ecologically sound balances
between nature and human activities. From an anthropocentric (economic) point of view, the
natural environment is considered only in terms of how it affects human beings, especially in the
context of economic activities; integrated accounting systems should thus retard those natural
functions which are exploited by human beings. The SEEA attempts to reflect a synthesis of the
ecological and economic points of view. That is, the economy is not viewed solely as part of the
environment and the environment is not viewed solely in terms of its economic usefulness. Several,
often complementary, approaches to natural resource and environmental accounting are presented in
Revised Draft - Do not cite or quote

the SEEA with the aim of developing compatible data sets which can be used to analyze
environmental-economic relationship s.
Finally, most accountants believe that it is important to make a distinction between
accounting and analysis. In their view, the accounts should rely to the maximum extent possible on
observed data and not on imputations or modeling. In many cases, modeled output is used to
characterize the environment. This raises the question whether environmental modeling should be
included in the accounts (i.e., considered as a generator of quantity and quality data for the
accounts) or should such modeling be considered analysis which uses the data contained within the
accounts. For example, the U.S. National Income and Product Accounts contain imputations
(modeling) for the value of owner-occupied housing and the national accounts data is used in
macroeconomic models of the U.S. economy (e.g., DRI, Wharton, Jorgenson-Wilcoxen, etc.).
Implementation of the SEEA
The SEEA is designed to be as comprehensive as the data will allow, while maintaining
consistency within the system and close linkage with conventional national economic accounts.
Given the lack of consensus on environmental accounting methods and data constraints,
implementation of the SEEA requires a flexible, "building blocks" approach. Beginning with the
revised SNA (Version I of the SEEA), four stages of implementation are described in the SEEA
Handbook (UNSO, 1993). These are:
1.	Reformatting and disaggregation of the SNA (Version II),
2.	Physical accounting (Version III),
3.	Imputed environmental costs, using alternative valuation methods (Versions
IV. 1-3), and
4.	Possible extensions (Versions V.l-6), including extending the production boundary to
include household activities and environmental services produced by nature, and
input-output analysis.
The fourth stage involves approaches which remain controversial and for which there is no
general consensus on their feasibility and desirability. The SEEA handbook recognizes that they
may become important for particular analyses and briefly covers these possible extensions to the
SEEA. We do not discuss them further in this paper.
In Version II, environment-related monetary flows within the production and asset accounts
are identified and further elaborated. The relevant portions of the supply and use tables of produced
goods and services are disaggregation with respect, to the actual expenditures for: 1) prevention and
restoration of negative environmental impacts associated with economic activities, as defined in the
draft classification of environmental protection activities, and 2) for mitigating the repercussions
associated with a degraded natural environment which encompasses avoidance activities (e.g.,
installation of water purifiers) and treatment of damages caused by environmental deterioration (e.g.,
purchase of additional health and cleaning services). Together the actual expenditures associated
with environment-related activities are called actual environmental costs and comprise environmental
protection costs and repercussion costs. All actual environmental costs are borne by the economic
units financing the expenditures, although they may not have caused the environmental deterioration.
Revised Draft - Do not cite or quote

The SNA classification of non-financial assets is modified to more explicitly reflect natural
assets. In particular, land is broken down to separately identify soil and air is introduced as an
asset, although no monetary value is applied to it (i.e., it is to be used in physical accounting and in
estimating imputed environmental cost. As noted above, within the SNA non-financial asset
accounts, the other changes in assets is particularly important.
The data in other changes in assets are grouped into categories of depletion, degradation (as
reflected in market values), other accumulation (additions to mineral reserves, natural growth of
non-cultivated biota, etc.), and other volume changes (i.e., changes which are due to political,
natural or other non-economic causes which affect the economic system). Thus Version II of the
SEEA (shown below) will look much like the SNA presentation with additional entries detailing the
environment-related information.
SEEA Version II. Supply and Use Accounts with Asset Balances for Economic Assets,
Elaboration of Environmental Protection Costs, Environmental Repercussion Costs,
and Elements of Changes in Other Assets

Economic Asset Balances
Production ROW Consumption
Produced Assets
Open Stocks
Economic Supply
Excluding EP, ER
EP Activities
ER Activities
P.ex.EP M

Economic uses
Excluding EP,ER
EP Expenditures
ER Expenditures
Ci.ex.EP-ER Ex C.ex.EP-ER
Capital consumption
Excluding EP,ER
EP assets
ER assets

Net Product/Net Capital

Other Accumulation

Holding gams/losses

Closing Stocks

Revised Draft - Do not cite or quote

Version III focuses on a physical accounting of the environment. The SEEA physical
accounts are based on the concepts of materials/energy balances and natural resource accounting.
Materials/energy balances show the material input from the natural environment into the economy,
the use and transformation of these inputs in economic activities, and their return to the
environment. Natural resource accounts focus on natural resource stocks, such as biological,
subsoil, and water assets, which are valuable from an economic point of view as well as changes in
the quantitative and qualitative characteristics of those stocks. As noted previously, the SEEA does
not attempt to provide information on the transformation processes which take place entirely within
the natural environment. Nor does the SEEA provide a complete assessment of the transformation
processes within the economy. Rather, the physical information in the SEEA is hmited to recording
flows from natural assets to the economy and residual flows back to the environment at an
aggregate level.
An additional limitation is the lack of spatial detail in the SEEA natural asset accounts (i.e.,
the SEEA is intended to be a national system of accounts). Detailed regional-level accounts, based
on various graphical information systems, are needed to adequately describe the natural
environment. These regional accounts could be hnked to the SEEA to provide a national picture,
although it remains to be seen whether such aggregate accounts yield useful information for
environmental pohcy purposes. Similarly, it would be desirable to describe the flows of natural
resource inputs, products, and residuals in a detailed breakdown by type of input and output.
Unfortunately, existing data on production and consumption activities is usually not sufficiently
detailed to provide this information.
Flows of residuals (pollution) are recorded at the point in time they are generated by a
particular economic activity. Similarly, the impact of these residuals on ambient conditions are
shown only as environmental quality changes over the time period covered by the accounts. The
impacts of many long-term environmental problems such as global climate change, stratospheric
ozone depletion, and accumulation of toxics will thus be recorded when they occur. For example,
the SEEA would show emissions of greenhouse gases which occurred in the last year, the impacts
of climate change would not be recorded until they occurred, which may not happen for many years.
The SEEA is not intended to record or predict future impacts and alone it will not be able to address
many of the concerns surrounding sustainable development. Rather, the SEEA is designed to
provide data to ecological-economic models which would capture the dynamics of environmental
Using materials/energy balances and natural resource accounts for the physical accounts of
the SEEA does not mean that SNA concepts have to be modified. Linkages between the monetary
data in the SNA and the physical data in the SNA can be accomplished by ensuring that
corresponding items in the two systems can follow the same definitions and classifications.
Alternatively, bridging matrices which apphed compatible concepts at the interface between the
SNA and the physical data in the SEEA could be used. This procedure would be necessary when
there is no direct counterpart in the SNA for the physical data in the SEEA.
Presentation of environmental-economic interactions in only physical terms would severely
limit the usefulness of the SEEA. If the SEEA is to truly integrate economic activities and
environment effects, the relative importance of each needs to be determined and results aggregated,
which in turn requires a common metric. Version IV of the SEEA introduces imputed
Revised Draft - Do not cite or quote

environmental costs in order to provide a more comprehensive picture of environmental and
economic interactions. Three different valuation methods are proposed:
1.	Costs borne at market values by industry (Version IV. 1),
2.	Costs caused at maintenance costs (Version IV.2), and
3.	Costs borne at market values by industry and at contingent values by households
(Version IV. 3)
Each approach involves imputing additional costs to economic activities, either through the
rearrangement of existing information in the SNA (Version IV. 1) or by estimating costs using
hypothetical control costs or other non-market (e.g, CV) methods. An additional asset category, non-
produced environmental assets, is also appended to the Version I SEEA table.
Version IV.1. Imputed environmental costs at market values
This version of SEEA involves shifting the depletion and degradation items in the other
changes in asset accounts into the production accounts. That is, the reduction in natural asset
market values associated with depletion and degradation are treated as a cost. Corresponding
positive cost items are imputed to the economic agents which cause the depletion and degradation
and appear in the production column. In general it will be difficult to identify changes in market
values of natural assets due to degradation. The accumulation items are shifted into the capital
formation account and a parallel negative counterpart appears in non-produced environmental assets
(OA.env). This element is intended to reflect the transfer of environmental assets and their services
to economic activities. Two Environmentally adjusted net Domestic Product measures (EDP1) can
be defined as follows:
(6)	EDP1 = C + (I/i + OA.np - OA.env - Depl.np) + (Ex - M)
= C + (In - Depl.np) + (Ex - M)
(7)	EDP2 = C + (Iw + OA.np - OA.env - Depl.np - Depr.np) + (Ex - M)
= C + (In - Depl.np - Degr.np) + (Ex - M)
Version IV.2 Imputed environmental costs at maintenance costs
Maintenance costs have been discussed in the context of costs caused above. The use of
maintenance costs reflects a conservationist view toward the environment. Given the uncertainty
with respect to long-term environmental problems and the potential for irreversible damage a high
degree of risk aversion may be prudent. In this situation many have argued for, at a minimum, the
maintenance of the current level of environmental quality. Maintenance mats are also closely
related to sustainable development concepts, in that they measure the costs that would have been
required to keep the natural environment intact during the accounting period. These costs are
hypothetical since an actual use did take place which affected the environment. Of course,
calculation of depreciation of freed assets is also hypothetical since it is not know whether actual
investments will be made which will maintain the capital stock. Using the maintenance coat
approach in combination with traditional depreciation measures allows for both the maintaining of
income flows and preserving the natural environment intact.
Revised Draft - Do not cite or quote

Ideally determination of maintenance costs should be based on: 1) data which describes
physical changes in the natural environment caused by economic activities, 2) analysis of ambient
conditions to determine whether depletion or degradation is occurring,
3) determination_of non-damaging (sustainable) environmental quality levels (e.g., quantitative
standards), 4) activities (e.g., discharge reductions) needed to meet these standards and 5) an
estimate of the costs associated with these activities. Several types of actions aimed at preventing or
restoring environmental deterioration could be undertaken.
Depletion of natural assets can result in a reduction in economic production. Reducing economic
production or altering consumption patterns can reduce the generation of residuals. Changes in the
composition of output, substitution of inputs, technological change and environmental protection
activities can all prevent deterioration or restore the natural environment. Calculation methods will
depend on the specific activity considered. For example, in the case of pollution, imputed
environmental costs could be based on reductions in net value added or household consumption
expenditures, substitution costs and environmental protection costs. Estimated degradation costs
should be based on the most efficient methods for meeting environmental standards. One alternative
for imputed depletion costs has been proposed by El Seraphy (1989), which allocates part of the
operating surplus for alternative investment.
Imputed environmental costs are associated with the environmental media which directly
receive the residuals generated by economic activities. The ultimate destination of these residuals is
not taken into account. For example, acidic deposition and consequent damages to terrestrial and
aquatic ecosystems due to the emissions of sulfur oxides into the atmosphere by electric utilities are
not recorded. Similarly, unless transported by economic agents outside the territorial boundaries of
the country, the transfer of residuals to another countries is not considered. In Version IV.2 of the
SEEA, there are additional entries, particularly degradation of environmental assets.
Version IV.3 Imputed environmental costs at market and contingent values
The SEEA handbook raises several concerns regarding CV and its use in environmental
accounting. The SEEA provides only a generic framework within which further research,
discussion, and experimentation with CV and related accounting procedures are to be explored.
While the SEEA does not emphasize the use CV, neither does it dismiss the technique outright.
The SEEA suggests that CV questions be posed in terms of specific consumption activities
and expenditures that households would be willing to forego. The SEEA also notes that the number
and order of environmental concerns raised may influence respondents willingness to reduce
consumption. To deal with this problem, the SEEA recommends asking for total willingness to
forego consumption as a first step and then ask for the proportion that respondents would allocate to
alleviating specific environmental-concerns. Finally, households should be willing to reduce their
consumption by at least actual repercussion costs, suggesting that CV studies should focus on
respondents additional willingness to pay beyond the defensive expenditures they currently make.
An alternative approach would be to present households with substitute consumption patterns and
activities which are less environmentally damaging. Differences in expenditures associated with the
offered change in activities could be used to represent the value of lost environmental quality.
Imputed repercussion costs, based on contingent values, are recorded as reduction in
individual consumption and as additional costs of economic activities of households. An extended
concept of household production, as discussed in Version V, would be needed to develop a
Revised Draft - Do not cite or quote

comprehensive picture of the distribution of imputed repercussion costs. To avoid extending the
production boundary of the SNA to include household production, a new row (Shift of
environmental costs) is introduced and imputed repercussion costs are shifted from consumption to
domestic production of industries. This shift allows the SEEA to fully account for the social cost of
environmental degradation.
The table below is based on Version IV.2 and shows the types of changes that could be made to the
basic SEEA. Corresponding entries in the additional column, Non-produced Environmental Assets
(OA.env, Depl.env, Degr.env) and in the production and economic asset accounts can thus explicitly
reflect interactions between environmental assets and economic activities. Corresponding definitions
for EDP1 and EDP2 would be:
(8)	EDP1 = C + (In - Depl.np - Depl.env) + (Ex - M)
(9)	EDP2 " C + (I/l - Depl.np - Depl.env - Depr.np - Degr.env) + (Ex - M)
Pilot Implementation of SEEA for the U.S.
This section outlines the environmental components of the SEEA. These components can
quickly add up to a dizzying array of rows and columns of data to any reader unfamiliar with the
certain conventions of economic accounting in general and the specific organization of the SEEA.
To make it easier to understand the final table that consolidates all of the major SEEA components
achieved in this pilot implementation, the description in this section proceeds component by
component, building up the table until all of the pieces are represented. Keeping this in mind may
help the reader proceed through this demonstration more effectively. Before the final, consolidated
table, four tables are presented. These tables focus on the following in turn: disaggregation of the
accounts to show the role of environmental protection in the economy, adjustments to NDP to
reflect the depletion of natural resources (EDP1), the linkage of EDP1 to asset balances for natural
resources, and adjustments to NDP to integrate environmental degradation into the accounts (EDP2)
and the linkage of EDP2 to balances for environmental assets.
Disaggregation of Economic Accounts
Information that is already in the accounts can provide insights into the role that the
environment and environmental protection play in economic activity. For example, using input-
output analysis and by isolating environmental protection expenditures currently undertaken by
economic agents it is possible to illustrate the contributions of an environmental protection sector to
each of the conventional macroeconomic aggregates.
In the shaded area of Exhibit A the contribution of such a instructed environmental
protection industry to U.S. value-added (GDP) is indicated, from the work of Nestor and Pasurka
(1994). Although the level of environmental protection effort by the U.S. has commonly been
gauged by comparing environmental protection expenditures directly to GDP, such a comparison is
misleading because the two measures are not on equivalent terms. Using the value-added estimate
for the environmental protection is more appropriate. In 1987, the environmental protection sector's
share of value-added was approximately 0.6% ($28 billion).
Revised Draft - Do not cite or quote

SEEA Version IV.2. Supply and Use Accounts with Asset Balances for Economic
Assets, Environmental Protection, Contingent Valuation of the Repercussion Costs of
Households and Capital Accumulation at Maintenance Values

Economic Asset
Open Stocks

Economic Supply

Economic uses


D e p r . p



adjusted net
product: EDP1


adjusted net
product: EDP2

Other Changes
in Volume



closing stocks


Revised Draft - Do not cite or quote

Exhibit A.
Environmental Accounts for the United States, 1987
Environmental Protection Expenditures Separately Identified
($ Millions)

Economic Activities


Rest of

Opening Assets

Fixed Assets







Natural Gas

Economic Supply

Economic Uses

Product GDP
Env. Protection



Net Product: NDP

EDP1: Adjusting NDP to Reflect Natural Resource Depletion
EDP1 is a measure of NDP that has been adjusted for the depletion of marketed natural
resources. In this pilot implementation for the U. S., the focus is on six natural resources. They are
timber, oil, natural gas, coal, selected minerals, and water. These six were judged to be important
because of their value or the sheer volume of their use in economic activities.
The shaded area of Exhibit B highlights the new components added to measure the depletion
of these six natural resources and the resulting estimate of EDP1 in 1987. These figures show how
natural resource adjustments in national economic accounting can present a more pessimistic view
of the economy's performance. For the U.S., the revision is small, only 0.8%. Even though even a
small difference in measures of output can accumulate to a large amount in absolute terms, this
revision still appears to be minor. This finding is not surprising for the U.S. because of the diverse
nature of the economy. Nonetheless, it has been argued that this revision results m a dramatic
downward revision in the rate of return that can be derived from national economic accounts for the
associated industries (Bureau of Economic Analysis 1994). This result may be informative for
national economic policymakers but it probably is not new information for private investors in these
industries who should already be aware that natural resource production or retraction depletes the
assets of the industry. Given the results of this pilot case, it appears that including natural resource
depletion in U.S. national economic accounts matters for keeping them as complete and
comprehensive as possible even if the results do not appear to be significant. On this point, others
Revised Draft - Do not cite or quote

Exhibit B.
Environmental Accounts for the United States, 1987
Resource Depletion
($ Millions)
Economic Activities
Production !
Rest of |
Opening Assets !
Fixed Assets i
Natural Gas

Economic Supply
Economic Uses
Product: GDP
Env. Protection

Net Product: NDP
Environmental Uses
Timber Harvest*
Timber Net GftferiKi''.
Co8lMiirife£ : '
fnrf FHiin ' *•>"
MtBBHl cxawomm::-¦¦¦:.
Miatta! Discoveries
... $175931
> . v . ' ¦'
?/••• ••¦; • " '
¦ mm

t ummi
may disagree. Having the information publicly and widely available, as they would be if the SEEA
were fully implemented, is consistent with an important function of the accounts - to provide access
to a common set of data so that many users can evaluate them and draw their own conclusions.
Revised Draft - Do not cite or quote

Computation of EPD1
The computation of each of the depletion charges presented in Exhibit B is described briefly
below. Greater-details are currently available only in an unpublished document (Abt Associates,
1994).2 In all circumstances, a net price approach was applied to the change in the resource stock
in question. This approach requires information on the opening and closing stocks of the resource,
to infer physical depletion, and an estimate of the net price or its analogue. Each resource is
considered in turn below.
The growing stock of timber for 1987 was interpolated from U.S. Forest Service inventories
conducted in 1986 and 1991. Stocks grew from 756 billion cubic feet to 762 cubic feet. To value
the stocks and harvests, information from competitively bid sales of U.S. Forest Service timber was
used. These values exclude production costs and therefore were taken as estimates of net prices.
The opening and closing stocks of timber were valued at $5,729 billion and $5,758 billion
respective y. The increase in the stock, approximately $29 million, reflects the fact that net natural
growth ($159 million) exceeded removals ($130 million) by this amount. The net differences is
subtracted from NDP to calculate EDP1.
Information on crude oil and natural gas reserves and production were obtained from the
U.S. Department of Energy's Energy Information Administration (EIA, 1988). The net prices of oil
and gas were derived from estimates of "resource values," a net income concept, developed for 1981
by Stauffer and Lennox (1984).3 A central assumption for this derivation was that resource values
as a proportion of revenues were constant between 1981 and 1987. In the calculation of EDP1, only
the depletion of oil and gas stocks is considered.4 Based upon the net price method, oil extinction
was valued at $17.8 billion and natural gas extraction at $11.6 billion. Together they account for
87% of the depletion that constitutes the difference between conventional NDP and EDP1.
Statistics on coal production and reserves were derived from EIA and U.S. Department of
Commerce data (EIA, 1989; U.S. Department of Commerce, 1989). As was the case for oil and
gas, only the extraction of coal figures in the calculation of a depletion charge against conventional
NDP. In the absence of better information on production costs, an estimate of the resource value of
coal in 1987 was calculated from industry accounting data. Net revenue for the coal industry was
estimated by adding operating income and coal production taxes and subtracting income taxes from
their sum. The resulting depletion charge for coal was $532 million.
2	The authors would like to acknowledge the capable assistance of Todd Aagaard in
the compilation of the EDP1 data.
3	Their estimate of resource value is the sum of lease and land acquisition of non-
producing acreage, taxes other than income taxes, royalties, and windfall profits taxes.]
As will be shown below, the SEEA treats discoveries in the asset balances, not in
the measures of flow. The logic is that no production was involved since these are
nonrenewable resources.
Revised Draft - Do not cite or quote

Statistics on production and reserves of more than eighty minerals are routinely collected by
the Bureau of Mines (1988). Some production occurred for the majority of these minerals in 1987
but it was not possible to characterize the depletion of each one because of data constraints. For
example, to preserve confidentiality, the Bureau of Mines did not release data on the domestic
production of fourteen minerals. Furthermore, the Bureau estimated certain essential financial
information only for selected minerals. For these minerals, Bureau of Mines' estimates of taxes and
royalties per unit of minerals, averaged over facility lifetime with a 15% discount rate, were applied
to the 1987 prices of the minerals to calculate their depletion charge. The resulting depletion charge
for twelve minerals was $824 million.5
The U.S. Geological Survey (USGS) publishes estimates of water use every five years. For
eight categories of users (domestic, commercial, irrigation, livestock, industrial, mining, and
thermoelectric power), the USGS estimates total use and consumptive use. Consumptive use means
that the water used dissipated, was incorporated into products or crops, consumed by humans or
livestock, or otherwise removed (USGS, 1988). For this pilot study, the physical depletion of water
was derived from net water use - the difference between water extinction (from surface or ground
sources) and water returned. Data on water prices from the 120 largest metropolitan areas (Arthur
Young & Company, 1988) and on government capital and operating and maintenance expenditures
from the Department of Commerce (reported in EPA, 1990) were used to calculate net water prices.
The average net. price of publicly-supplied water, weighted by categories of use, was $0.09 per 1000
gallons. The estimated depletion of water was $3 billion, reflecting the difference between water
extraction ($10.9 billion) and water returned ($7.9 billion).
Linkage of EDP1 to Asset Balances
An important feature of SEEA is its characterization "of the contribution of the environment
to economic activities. This contribution is depicted as a transfer from the environment to the
economy. Natural resources that have not yet become "economic" (having a net price greater than
zero) are defined as being environmental assets. Only once these natural resources are proven,
which here is equated with being "discovered" do they become economic. At this point, they are
transferred from the environmental asset balance to the economic asset balance. While conventional
accounting would show a gain in wealth with the discovery of a natural resource- like oil, the SEEA
does not. Because the discovery is treated as a transfer, overall wealth stays constant as long as
none of the oil is depleted. For example, in 1987 the discoveries of oil were worth $20 billion. In
Exhibit C, this discovery is shown as a deduction of oil from the environment asset balance and an
increase in non-produced economic assets.
It has already been demonstrated that SEEA only the depletion' of oil and other non-
producer natural resources is considered in adjusting NDP. The balance sheets for non-produced
assets also incorporate this depletion. So, for example, the $17.8 billion depletion charge for oil that
was included in the calculation of EDP 1 is also included in the asset balance for oil. Overall, in
1987, there was a growth in oil as an economic asset of $2.2 billion given discoveries and depletion
of $20 billion and $17.8 billion respectively. Natural gas is shown in analogous terms, with a
transfer from the environment of $8.4 billion. Natural gas, in contrast to oil, declined as an
5 The twelve minerals were aluminum, asbestos, barite, copper, gold, lead,
molybdenum, phosphate rock, potash, silver, sulfur, and zinc.
Revised Draft - Do not cite or quote

Exhibit C.
Environmental Accounts for the United States, 1987
Resource Depletion
($ Millions)

Economic Activities

1 i : Environment 1
Rest of
c Assets
Environment '
Opening Assets
Fixed Assets
Natural Gas

Economic Supply

Economic Uses

Product; GDP
Env. Protection



Net Product: NDP

Enviroomeatal Uses
Timber Harvests
Timber Net Growth
Oii Discoveries
Nat. Gas Extraction
Nat. On Discoveries
Coal Mining
Coal Discoveries
Mineral Extraction
Mineral Discoveries
Water Extraction!
Water Returned

' 1
¦ m
1" »
Net Product EDPI
$3,933,800 J
economic asset, by $3.2 billion. For coal and minerals no discoveries took place to of offset the
depletion charges recorded in the calculation of EDP1.
The changes in water resources are characterized solely within the economic non-produced
asset balances. This classification reflects the explicit judgement that water is a controlled resource
rather than one that exists in the environment. This specification raises an important classification
issue. To date, neither the applications of the SEEA nor the SEEA Handbook provides
unambiguous guidance on how to classify a natural resource like water. The extent to which it
exists in the economic or the environment realm is a question that environmental economics can aid
national accountants in answering.
Revised Draft - Do not cite or quote

In closing this discussion of natural resource commodities, it is useful to emphasize that
focusing on natural resource depletion provides an incomplete picture of the changing status of
natural resource assets. This shortcoming applies as well to several past efforts to adjust GDP for
natural resource_depletion. The calculation of EDP1 in SEE A does not complete the picture, since it
ignores the discovery of non-produced natural assets (but not the production of natural resources like
timber). Only in the SEEA asset balances, both economic and environmental, is there a complete
picture of the overall change in wealth. In this respect, despite the classification issues raised
earlier, the SEEA represents an improvement in the integration of economic and environmental
perspectives for natural resource commodities.
Environmental Degradation, NDP Adjustments, and Asset Balances
In the larger scheme of things, the natural resource adjustments to NDP reflected in EDP1
were not that large. Environmental adjustments are more significant under this particular application
of SEEA. EDP2, shown within the shaded area of Exhibit D, is the result of adjusting conventional
NDP to reflect the costs of controlling residual pollution. The resulting estimate is $3.7 trillion, or
91% of conventional NDP. Since the unabated pollution is characterized as a loss of environmental
assets, overall net capital formation is not only much smaller than under conventional economic
accounting it is negative. This implies a decline in the overall capital (man-made and natural) of the
U.S. There is a growth in produced assets of $247 billion a decline in economic non-produced
assets of $5 billion, and a decline in environment non-produced assets of $328 billion summing to a
decline of $86 billion. The decumulation of capital stock causes tremendous concern when ordinary
capital is involved. If one accepts the definition of environmental assets as part of the capital stock
from which we derive important goods and services, then this decumulation raises the possibility
that something socially undesirable is occurring. Taking these estimates at face value, net manmade
capital accumulation needed to be about 35% higher to avoid a loss in national wealth.
Calculation of EDP2
In this U.S. pilot study of SEEA the degradation of environmental resources is valued using
a maintenance cost approach. For each of three environmental media (land, water, and air), the
costs of controlling the existing level of pollution became the basis for adjusting NDP/EDP1 to
derive EDP2. The level of assumed control is complete meaning that the aggregate maintenance
costs presented here suggest the level of resources necessary to eliminate this pollution entirely.
This assumption may appear to be an extreme one. It does however illustrate the type of
decision that anyone implementing this SEEA version has to make. The SEEA developers provide
little definitive guidance on how to specify this parameter. In effect, it represents one's assumptions
about the level of pollution with which no damages are associated. Environmental policymakers,
much less national income accountants, would be hard pressed to make a clear decision, except
possibly through the use of extensive modelling.
While the zero-pollution assumption may result in an unusually high estimate of the
maintenance costs associated with the level of pollution in 1987, the estimated unit costs themselves
tend to offset this tendency. For land and air pollution, unit costs were estimated using average
costs of control experienced in the past. These costs would probably be lower than the marginal
costs of controlling existing pollution. Water pollution may be the single but large exception. Unit
costs of controlling conventional water pollutants were derived from recent surveys of wastewater
Revised Draft - Do not cite or quote

Exhibit D.
Environmental Accounts for the United States, 1987
Resource Depletion and Environmental Degradation
($ Millions)
Economic Acq viae* I	|	j	;	 Environment

Econonac Assets
Non-Prod, j

Rest of

Opening Assets


Fixed Assets




Timber 1





Natural Gas








Economic Supply

Economic Uses

Produce COP

Env. Protection



Net Product: NOP

Environmental Uses

Timber Harvests

Environmental damages were identified primarily through inventories of polluting residuals.
An important consideration in constructing the SEEA system on a regular basis is the availability
and reliability of these databases. Although most of the data were compiled from EPA data sources,
there is still wide variability in the reliability and regularity in the collection of the source data. If a
more routine implementation of the SEEA is envisioned, these data issues will need to be weighed
in decisions about how the SEEA will be constructed and used.
The calculation of degradation costs for each of the three media are discussed briefly below.
Three categories of degradation of land are considered: soil erosion, hazardous wastes, and
"non-hazardous" wastes.
Estimates of soil erosion were derived from an inventory of rural lands by the Soil
Conservation Service. No estimates of the maintenance costs were identified. For this mason, the
entry is represented by the amount of soil erosion (million tons) times the hypothetical unit cost of
mitigating the erosion.
Although this version of the SEEA does not permit its use, it was possible to derive an
aggregate estimate of the off-site damages associated with soil erosion. This estimate was $14.3
billion, which indicates the amount that society would be willing to give up to control soil erosion.
Under a different version of the SEEA that considers costs borne, this estimate could be used to
adjust EDP2 by adjusting consumption in lieu of or in addition to production.
Hazardous wastes are for the purposes of this pilot study defined by federal regulations.
Any wastes defined as hazardous must be managed in specific ways to protect the environment.
Accordingly, it is assumed that hazardous wastes do not impose damage on the environment given
the additional assumption of full compliance. Consequently, while the volume of hazardous wastes
generated each year is very large (290 million tons in 1987), hazardous wastes do not affect the
calculation of EDP2. The hazardous waste management expenditures are already reflected in the
calculation of conventional GDP and NDP and in the estimate of value-added associated with the
environmental protection industry highlighted above.
Just as hazardous wastes are defined by federal regulation by default so are non-hazardous
wastes. These are wastes that do trot have to be managed according to federal hazardous waste
regulations but which nonetheless may still pose a hazard or some other environmental impact.
More than 8 billion tons of these wastes were generated, mostly from manufacturing. To estimate
maintenance costs, it was assumed that the same level of care was necessary for these wastes as for
hazardous wastes. To avoid any environmental impact, this assumption may not be too
unreasonable since there are indivisibilities in the capital necessary to manage the wastes according
to federal regulations (such as the requirement of liners in landfills). The actual unit costs applied
to non-hazardous wastes reflects the incremental costs of going from the current level of expenditure
on these wastes to what would be comparable for this volume of wastes if federal requirement were
imposed. The resulting aggregate maintenance costs are $33.6 billion, accounting for about 10% of
the adjustment for EDP2.
Revised Draft - Do not cite or quote

Water pollution presented one of the most difficult challenges in calculating degradation.
The difficulty stemmed primarily from the lack of reliable information on emissions of conventional
or toxic pollutants. At the same time, the resulting estimate of aggregate maintenance costs were so
large, accounting for 77% of the adjustment for EDP2. Together, these circumstances are cause for
concern. Consequently this component of EDP2 should be viewed with considerable caution at this
stage of development.
To construct a basic aggregate picture of the amount of conventional and toxic pollutants, it
was necessary to use a variety of often irregular data sources. This approach posed significant
obstacles to an independent verification of the estimated loadings. Ultimately, the physical measures
of the majority of pollutants were not used directly (Biological Oxygen Demand (BOD), Total
Suspended Solids (TSS), nitrogen and toxics) in the estimation of maintenance costs. Only
phosphorus emissions and their estimated control costs were used directly since their volume
determined the number of facilities that needed to be constructed, assuming- that water pollution
appears uniformly in the concentrations for which these facilities are designed.6 Furthermore, given
this same assumption for other pollutants, these same facilities would be sufficient in theory, to
control BOD and 97% of nitrogen. TSS was so much larger in aggregate volume that this approach
seemed inappropriate. As such, the estimated costs of $230 billion reflect the resources necessary to
control phosphorus, BOD, and most of nitrogen under very special circumstances.
This approach to deriving a maintenance cost estimate is unsatisfactory because it applies a
point-source means of control (wastewater treatment facilities) to a problem that sterns largely from
non-point sources, which means that less costly ways to accomplish even a zero-pollution goal are
likely. Ironically, this approach is also unsatisfactory because it may not satisfy the SEEA criterion
of estimating the costs to reduce emissions to. a non-polluting level since TSS would not be
eliminated. In this respect, the maintenance cost estimates may be too low. In practice, then, the
implementation of SEEA with respect to water degradation had severe limitations in this U.S.
Nonetheless there may be some value in the experience. If even just a portion of the $230
billion degradation estimate proves realistic, then at least given a maintenance cost perspective,
water degradation is likely to carry great weight within an SEEA system. For environmental
economists, this may be a frightening prospect because it only reflects the cost side of the issue.
Frightening, that is, unless the as yet unknown estimated damages are least equally large.
Environmental degradation from air pollution may be the most straightforward of all the
ones considered. EPA routinely collects or estimates statistics on the emissions of certain key
pollutants. This SEEA example focuses on TSP, SOx, NOx, VOCs, carbon monoxide (CO), and
lead (Pb). Toxic air pollutants represent a more recently considered phenomenon and are less well-
6 The estimates of control costs per facility were obtained from the National Research
Council (1993).
Revised Draft - Do not cite or quote

Unit costs were estimated using information from published EPA documents, a not-so-small
indication that this information could be compiled on a routine basis to support the implementation
of environmental accounts. The unit costs were derived by dividing the aggregate air pollution
control expenditures in 1987 by the estimated emissions reduction attributable to these expenditures.
For stationary sources, it was possible to use pollutant-specific expenditures in 1987, based on U.S.
EPA (1990), as well as pollutant-specific reductions (U.S. EPA, 1989). Because joint control of
pollutants is more common with mobile sources, pollutant-specific expenditures were not available.
As a result, the total mobile source expenditures for air pollution control were divided by the sum of
all mobile source pollutants (by weight). In Exhibit D, the row labeled TRANSPORT provides the
relevant maintenance costs for mobile sources. Together, the stationary and mobile source
pollutants account for approximately 10% of the EDP2 adjustment.
Environmental Degradation: Conclusion
In sum, the exercise of implementing the SEEA maintenance cost concept of degradation
revealed several types of problems that should be considered more carefully before the results can
be taken seriously - for the light they shed on economic-environmental interactions much less for
any bearing they may have on economic or environmental policy. Nonetheless, these results do not
detract completely from the general impression drawn from this exercise that environmental
degradation poses a serious matter to be addressed by economic accounting. Even discounting the
degradation results by an order of magnitude (to $30 billion), they still appear to be substantial, at
least relative to the depletion estimates ($34 billion) and to current efforts to control pollution
(measured by value-added of $28 billion from the environmental protection industry).
Exhibit E presents the complete, consolidated table of SEEA results from this pilot U.S.
study. Nothing significant has been added. The final components which have been added are the
revaluation rows, for which no data are provided, and the closing asset balances, which summarize
changes presented in earlier discussions of asset changes. Note that no entries are provided in the
closing balances or the opening balances of environment non-produced assets. The SEEA only calls
for measuring changes in these assets not their total values. Omitting total values seems to be more
of a concession to the substantial obstacles posed by estimation than a conclusion about the validity
of the concept. The difficulty may not stem so much from the challenge of making a physical
inventory of the environment. That is indeed possible for certain facets, such as the extent of old
"growth forests. Instead a substantial part of difficulty comes from the challenge of valuation which
as this application has shown, is already very hard when only marginal changes are involved.
Consolidated SEEA Results for the U.S.
Exhibit F presents several key statistics from this pilot SEEA application, some of which
have already been cited above. At a glance, this presentation highlights two environmental
phenomena that are worth further inquiry. One is that environmental depletion adjustments to NDP
stand out far more than natural resource depletion adjustments. The second phenomenon is the
apparent indication that in 1987, the U.S. appeared to be living beyond its means. There is an
apparent decumulation of wealth as indicated by negative net capital formation as well as by the
fact that the final consumption exceeds EDP2. If the current implementation were more than a pilot
effort and if the economic accounts truly encompassed all capital, this particular finding could point
to unsustainable tendencies in U.S. economic activities. As it is, this statistical result merely
suggests that there may be tendencies that are worth worry about and investigating further.
Revised Draft - Do not cite or quote

Exhibit E.
Environmental Accounts for the United States 1987
Resource Depletion and Environmental Degradation
($ Millions)

Economic Acuvitna


le Aaaea

Exhibit F
Comparison of Indicators Based on Conventional and on Environmental Measures

Conventional Accounts
(% of Conventional)
(% of Conventional)
$4,037.8 billion
$4,004.1 billion
$3,704.6 billion
(91 .7%)
Net Capital
$247.1 billion
$213.3 billion
Net Capital
Formation, as
% of NDP or
as % of NDP
or EDP
We return to the three objectives of the SEEA stated at the outset of this paper to determine
whether implementation of the SEEA is appropriate and useful for the U.S. They are to: 1) provide
an accounting of the interaction of the economy and the natural environment, 2) address sustainable
development concerns through proper accounting of both manmade and natural assets, and 3)
develop environmentally adjusted measures of GDP to serve as a guide toward sustainable
development. Since it seems reasonable to assume that these objectives are ones generally shared
by our society, we review each as the means of answering this question.
The SEEA does provide a means for better accounting of the interaction of the economy and
the natural environment, in at least two special ways. First, SEEA is far more complete in showing
ways that the economy can infringe on the environment and how the environment contributes to the
economy than the SNA ever was. When one starts with an accounting system that is so thoroughly
oriented to market transactions and production it is quite an achievement to flesh out a system that
maintains consistency with economic accounting while incorporating the environment. Through the
progression of various versions of SEEA, it is possible to see, as shown in the paper, the concepts
of environmental depletion and degradation transformed from alien concepts that are almost
completely excluded or ignored in conventional accounting to ones that are full-fledged elements of
an accounting framework that actually uses non-market values. Whether national accountants ever
go that far may depend on how much environmental economists get involved in the process, an
issue to which we return below.
The second special way that SEEA provides a better means for depicting environment-
economic interactions is its recognition of natural capital as a legitimate component of a nation's
asset balances. Although this step is a long way from making it possible to track the sustainability
Revised Draft - Do not cite or quote

of a nation's wealth, it is a necessary step. Until natural capital is scrutinized in tandem with
manmade capital, economic accounting will be biased against the preservation of natural capital,
While the pilot study in this paper has demonstrated in a limited way the magnitude of the
difficulties in reliably implementing a system of manmade and natural capital accounting, these
difficulties do not remove the appeal of putting natural and manmade capital on even terms.
This point relates to the second objective of SEEA. Whether better natural capital
accounting can help address sustainable development concerns will depend on how well natural
capital is actually understood. Consequently, whether SEEA is right for the U.S. does not depend
solely on the structure of SEEA itself. SEEA depends critically on the information which it
incorporates. It should be emphasized that SEEA incorporates the environment less than it opens up
the accounting system to better information on linkages between the economy and the environment.
In many ways, SEEA can be seen as a user of environment-economic information rather than as a
generator. For example, the maintenance cost approach demonstrated in this paper depends on
judgments of the non-damaging levels of pollution. National accountants cannot be the arbiters of
such choices. Instead environmental economist, public health specialists, ecologists, and others
could be. This circumstance presents an opportunity for environmental economists. SEEA is like
an empty vessel. It is good enough to use a lot of information that has not yet been fully
developed. Any improvements in understanding the relationships between the environment and the
economy can be incorporated in the SEEA system.
Promising as such developments seem to be, the number of unanswered questions about the
relationship between the environment and the economy is very large. This predicament brings us to
the third objective of SEEA - to develop environmentally adjusted measures of GDP as guidance for
sustainable development. If the characterizations of natural capital and of environmental goods and
services are still so limited how good can any resulting "green GDP" measures be that incorporate
them? We suspect that they may indeed be inadequate but, in the face of GDP measures which turn
an even blinder eye to the environment, we also suspect that improved knowledge lies in the
direction of environmental accounting and not toward past conventions.
Revised Draft - Do not cite or quote

Abt Associates Inc. 1994. Environmental and Natural Resource Accounting: A Compendium of
U.S. Estimates Based on the U.N. SEEA. Draft report for the U.S. Environmental
Protection Agency prepared under contract with IEc Incorporated. March 3.
Arthur Young & Co. 1988. 1988 National water and wastewater rate survey. Arthur Young
National Environmental Consulting Group: New York, New York.
Bureau of Economic Analysis. 1994. Accounting for Mineral Resources: Issues and BEA's Initial
Estimates. Survey of Current Business. April.
Bureau of Mines. 1988. Mineral commodities summaries 1988. U.S. Department of the Interior,
Bureau of Mines.
EIA. 1989. Estimation of U.S. coal reserves by coal type. U.S. Department of Energy, Energy
Information Administration Publication No. EIA-0529. October.
EIA. 1988. U.S. crude oil, natural gas, and natural gas liquids reserves, 1987 annual report. U.S.
Department of Energy, Energy Information Administration.
Nestor, D. V. and C. A. Pasurka. 1994. Environment-Economic Accounting and Indicators of the
importance of Environmental Protection Activities. Revised paper presented to the 1993
Annual Meeting of the Southern Economic Association. March.
NRC. 1993. Managing Wastewater In Coastal Urban Areas. Committee on Wastewater
Management for Coastal Urban Areas, Water Science and Technology Board, Commission
on Engineering and Technical Systems, National Research Council. National Academy
Press: Washington, D.C.
Stauffer, T.R. and F.H. Lennox 1984. Accounting for "Wasting Assets": Income Measurement for
Oil and Mineral-Exporting Rentier States. OPEC Fund for International Development
Vienna, Austria.
U.S. Dept. of Commerce. 1989. National energy accounts. U.S. Department of Commerce, Office
of Business Analysis, Catalog No. OBA-NEA-08. March.
U.S. EPA. 1990. Environmental investments: the cost of a clean environment. U.S. Environmental
Protection Agency, Office of Policy, Planning, and Evaluation, Publication No. EPA-230-11-
90-083 November.
U.S. EPA. 1989. National Air Pollutant Emission Estimates: 1940-1987. U.S. Environmental
Protection Agency, Office of Air Quality Planning and Standards. Publication No. EPA-
450/4-88-022. March.
USGS. 1988. Estimated use of water in the United States in 1985. U.S. Department of the
Interior, Geological Survey, Circular No. 1004.
Revised Draft - Do not cite or quote

Intergenerational Welfare Economics
and Environmental Policy
Richard C. Bishop
Richard T. Woodward
Department of Agricultural Economics
University of Wisconsin-Madison
Paper Presented at the Association of Environmental and
Resource Economists Workshop,
"Integrating the Environment and the Economy: Sustainable
Development and Economic/Ecological Modeling,"
Boulder, Colorado, May 5-6, 1994.

Intergenerational Welfare Economics and Environmental Policy1
Richard C. Bishop and Richard T. Woodward2
Department of Agricultural Economics
University of Wisconsin-Madison
The central "story" of environmental economics is now well established. Market economies do
not, through the unaided guidance of the invisible hand, achieve economic efficiency in the allocation of
many environmental resources. Causes of market failures include externalities, the public good
characteristics of some environmental services, and property rights problems such as open access. Much
has been accomplished by prescribing policies to reduce market failures. Nevertheless, one must ask
whether market-failure based approaches adequately capture the full extent of the environmental issues
facing the world today. Are global warming, worldwide erosion of soils, contamination of groundwater,
losses of biological diversity, destruction of wetlands, overfishing, ozone depletion, rapid exhaustion of
nonrenewable resources, and other such issues only of economic interest when they result from market
failures? In this paper, we argue that defining environmental problems and their solutions within the
market-efficiency framework misses the crux of many of today's environmental problems. A more
complete environmental economics would be based on the dual goals of efficiency and sustainability.
We define an economy as sustainable if each successive generation has per capita economic
opportunities at least as large as those enjoyed by earlier generations. By focusing on "opportunities"
rather than "welfare" or "income," this definition places conditions upon initial endowments that each
1	Paper Presented at the Association of Environmental and Resource Economists Workshop,
"Integrating the Environment and the Economy Sustainable Development and Economic/Ecological
Modeling," Boulder, Colorado, May 5-6, 1994.
2	Please address correspondence to Richard C. Bishop, Department of Agricultural Economics,
University of Wisconsin, Madison, WI 53706; Phone: (608)262-8966; FAX: (608)262-4376; E-Mail:
rcbishop@calshp. cals. wise. edu

generation should receive. Endowments are broadly defined to include not only natural resources but also
the capital, infrastructure, technology, knowledge and institutions that today's generation will pass onto
its children.
In the first major section of this paper, we discuss the implications for welfare economics of
including sustainability as well as efficiency. To accomplish this, three intergenerational theoretical
models are developed. Following Page (1977), the first two are dubbed the "Hardtack World" and the
"Corn World." In the Hardtack World, a finite number of generations divide a single, non-renewable
resource. In the Corn World, an unlimited number of generations exploit a renewable resource. Finally,
we add a world with capital. The problem there is like the hardtack problem but capital formation (and
hence technological progress) are possible. Here, output is produced using a non-renewable resource and
capital. The output in each period can either be consumed or invested. Accumulated capital is productive
in later periods and can be substituted, up to a point at least for diminished stocks of the resource as time
Though these cases are abstract and highly stylized, they serve to illustrate how a basic result of
welfare theory carries over to the intergenerational world. Based on the familiar Edgeworth box
diagrams, any Pareto efficient state of the economy rests on a foundation of initial endowments held by
economic actors. However, as is well known, an infinite number of Pareto efficient states are possible,
each based on a different allocation of initial endowments. While the Edgeworth box itself must be
discarded in favor of a more dynamic representation of the economy, this basic conclusion carries over to
a world with time and more than one generation. The result is an infinite number of possible efficient
time paths for an economy, each depending on a different intergenerational allocation of endowments. In
each of the cases we consider, there are many efficient time paths. Along any of these, it is impossible to
make members of one generation better off without harming members of another. An important
conclusion follows: While there are an infinite number of possible Pareto-efficient time paths, only a
subset of those efficient paths are also sustainable. Achieving efficiency does not guarantee sustainability.
Rather, if society wishes to be both efficient and sustainable, the quest for economic efficiency must be
carried out within what we shall term "sustainability constraints."

We then discuss how the principles of sustainability might be applied in a real world context.
First, we address the basic question of whether sustainability should be a goal of economic analysis or not.
Given the great public interest in sustainability and global environmental issues, it is our conclusion that
economists would be remiss if we left such an important issue aside. A number of important
complications arise in putting sustainability concepts to work. Most of all, the uncertainty associated with
long-term environmental and economic issues makes determining if a particular path is truly sustainable
difficult if not impossible. Uncertainty, even ignorance, of the long-term ramifications of our actions,
make planning for efficiency and sustainability a very inexact task. Faced with this uncertainty, we
discuss two policy options designed to push the economy towards both sustainability and efficiency.
In this section we develop three simple models to discuss the fundamental issues of incorporating
sustainability into the framework of welfare economics. We will demonstrate the importance of
establishing constraints on the economy if society is to ensure that sustainability is achieved. While the
framework presented here is far from general, we believe that extensions of the model can be developed to
form policies for real economies that have a multitude of endowments and outputs. We start, however,
with the most simple case.
The Hardtack World3
What we need to explore the welfare economics of intergenerational resource use is the dynamic
analogue of an Edgeworth box diagram. Our simplest model elaborates a bit on an argument of Norgaard
(1991). Figure 1 illustrates the principles involved. In order to focus on very fundamental issues, this
figure takes the simplest possible intergenerational case: an economic universe consisting of only two
non-overlapping generations with equal populations that exploit a single, non-renewable resource. It is as
if "society" for purposes of welfare analysis consists of two separate groups of people who will be
3 This section draws heavily on Richard C. Bishop and Richard T. Woodward, "Efficiency and
Sustainability in Imperfect Market Systems," Oregon State University, Graduate Faculty in Economics,
Public Lecture Series, Forthcoming.

marooned on a desert island during non-overlapping time periods and only the first generation will have
provisions, composed of a fixed supply of hardtack. The first group must decide how much hardtack to
eat and how much to leave for the second group. We assume that capital per se does not exist and that
there is no technological progress.
Figure 1
The per capita utility of the future generation, Uf, is measured by the vertical axis above the
origin. Likewise the current generation's utility, UCT is measured along on the horizontal axis to the right
of the origin. Positive utility is assumed to be possible only when resources are consumed in positive
quantities. Each point in the graph's northeast quadrant, then, represents a time path of per capita utility
and points along the curve connecting points E and E' thus represents the efficiency frontier for this very
simple world.4
4 We assume that the we assume that levels of utility are directly comparable across generations and
that wealth within each generation is distributed according to that generation's social preferences.
Obviously we are suppressing very important issues here, not the least of which is that Arrow (1963) has
shown that a social ordering which adheres to a few simple rules is impossible.

The other quadrants in the figure help illustrate the derivation of EE'. The allocation of the
resource between the generations is depicted in the southwest quadrant by the constraint with a slope of -1
to reflect its nonrenewable nature. From a slightly different perspective, the constraint pictured in the
southwest quadrant shows the alternative time paths for intergenerational resource endowments. The
curves in the southeast and northwest quadrants show the maximum levels of per capita utility that can be
achieved by the current generation and the future generation, respectively, as a function of resource
consumption. The intragenerational utility functions are assumed to be monotonically increasing in
consumption and concave.
The point in the north-east quadrant that is actually reached depends upon two factors: the
efficiency with which each generation uses the resource, and the distribution of the resource between the
two generations. For example, if the current generation uses resources at point A, where the available
resource is divided equally, and both generations behave efficiently, then per capita utility is F for both
generations. If the current generation uses more than A, there will be so little of the resource left that the
future generation will not be able to achieve a per capita utility level equal to that available to the current
Sustainability can be simply defined in the fully efficient case. It would be achieved if the future
generation achieves a level of per capita utility at least equal to that of the current generation. This
criterion is met here if the current generation uses no more than A of the resource, so that the per capita
resource stock available to the future generation is at least as great as that used by the current generation.
In a fully efficient economy, therefore, sustainability can be defined either in terms of the distribution of
endowments or in terms of outcomes.
The situation becomes slightly more complex if the possibility of intragenerational inefficiency is
admitted. Suppose that the current generation does not achieve efficiency, say because it has a market
economy and market failures are allowed to persist. Then it will enjoy some level of per capita utility
below its utility frontier. Suppose, as a specific case, that it uses D of the resource, but only achieves level
C of per capita utility. This would allow the future generation to achieve only G at a maximum. Since, at
G, the future generation's well-being exceeds that of the present, should we say that the current generation

acted in a manner consistent with sustainability? Surely the answer must be "no." As point G makes
clear, in a world with economic imperfections, it is not satisfactory to define sustainability in terms of
levels of utility. While at G the future generation achieves a higher level of utility than the current
generation, this is true only because of the inefficiencies of the current generation. We see in this simple
example the importance of defining sustainability in terms of endowments, in this case the initial division
of the resource stock. Our simple economy will be sustainable only if resource consumption by the current
generation is less than or equal to A.
Interestingly, our analysis indicates that efficiency and sustainability need not be conflicting
goals. We have demonstrated in the case of two generations and one-dimensional endowments that a
subset of the infinite number of efficient paths is also sustainable. The dual goals of efficiency and
sustainability could be pursued by treating sustainability as a constraint. A society holding both goals
would constrain itself to considering only those efficient paths that are also sustainable. In the simple
world of Figure 1, the sustainability constraint can be simply stated.
Sustainability constraint Rc ^
where S0 is the initial level of the resource. It is straightforward to extend this model to an economy with
n generations. In this case the sustainability constraint for generation g would be given by
Sustainability Constraint: R„ ^	
8 n-g + 1
Even with a large number of generations, it is possible to seek a path that is both efficient and sustainable.
While we must examine this conclusion in more complex models, there is no obvious reason to believe
this basic principle would not apply there as well.5
5 We should candidly admit right here near the beginning that this issue has not been thoroughly
explored in a rigorous fashion. Howarth and Norgaard (1990) and Howarth (1991) have made important
beginnings. Their basic approach, however, is to assume a social welfare function and then investigate its
implications for resource endowments. From our perspective, the dynamic equivalent of an Edgeworth
box would be more useful in defining necessary and sufficient conditions for an efficient and sustainable
equilibrium. Considerable progress has been made on growth models with overlapping generations (see,
for example, Fisher 1992) but to our knowledge such models have yet to included natural resources.

Of course, the Hardtack World has tremendous limitations. It is not fully satisfactory as a
model of sustainability for many reasons, not the least of which is its rather dim view of long run
prospects. So long as the only resource is non-renewable and capital accumulation and technological
progress are ruled out by assumption, sustainability over the indefinite future is infeasible. As the number
of generations increases without limit the sustainability constraint will approach a restriction that none of
the resource be used. Once a renewable resource is introduced, however, this difficulty disappears. Thus,
we move from the Hardtack World to the Corn World.
Efficiency and Sustainability in the Corn World
Let us again consider an economy of non-overlapping generations, Instead of bringing a box of
hardtack like in the preceding model, the resource is a renewable resource, say corn, where the g
generation inherits an initial endowment of seed totaling Sg. The initial endowment of corn can either be
consumed or planted. We presume growth rates are constant for each seed planted so that technology is
constant returns to scale. Each pound of corn will result in a harvest of 1 + r units of corn at the end of a
growing season. We shall assume that each generation lives for one growing season so that the corn
available at the end of the growing season becomes the inheritance of the next generation. To be efficient
all the corn available to each generation g, must either be planted or consumed, none can be lost. To be
sustainable, each generation must plant enough corn, measured as Ig, so as to satisfy
Sustainability constraint I. S 	
8 1+r
If this constraint is just satisfied, each generation will inherit an endowment of at least Sg so that the
opportunities available to the next generation are identical to generation g. If generation g satisfies the
sustainability constraint and is also fully efficient, they will be able to consume
c = ——
8 1+r"
Notice that economic growth is possible in the corn economy. Earlier generations could plant more than
the minimum required by the sustainability constraint and enhance consumption possibilities for later
generations. Later generations could in turn ratchet up the sustainability constraint from the initial level.
This is, of course, a much brighter world than the hardtack world provided that the initial
endowment of corn is adequate. An indefinite number of generations could be supported at the minimal
level set when the first generation arrives or at some higher level if the growth occurs. Furthermore, the

corn world is easily interpreted within a welfare theoretical framework. An infinite number of Pareto
efficient time paths exist. The only requirement is for Pareto efficiency is the one that has already been
stated: each generation must either consume or plant all the corn at its disposal. Then, it would be
impossible to reallocate corn among the generations to make members of one generation better off without
simultaneously making some members of another generation worse off. Some of these time paths would
be heavily skewed in favor of consumption by earlier generations; others would be more egalitarian; and
still others would be skewed in favor of consumption by later generations. Still others might have rising
and falling consumption across the generations. By adopting a sustainability goal, society chooses to limit
itself to the subset of efficient paths that satisfy the sustainability constraint.
Before we begin to try to ferret out conclusions for policy from all this, one more world will be
visited. It is like the hardtack world in that it depends to some extent on a non-renewable resource, but
investment in productive capital will be possible.
Sustainability in an Economy with Resources and Capital
In the corn and hardtack economies discussed above, the endowment of each generation was
limited to a single resource. We now consider the meaning of sustainability in economies with a two-
dimensional endowment, consisting of a resource component, S, and a capital component, K. Some
extensions to higher dimensions will be suggested but not fully developed. The basic idea, however,
remains the same whether we are considering a one-dimensional or an n-dimensional endowment. The
sustainability constraint will restrict economic activities to ensure non-decreasing economic opportunities.
In production each generation g uses up part of its stock of resources, R^, leaving the next
generation with V. = S* " Rs The resources are used as inputs into a general production function f(Kg,
Rg) which is increasing and concave in both terms. The total output, f(Kg, Rg), is either invested in
capital, I, or a consumed, Cg, so that *V. = K + [f(K R) - CJ. The capital stock is presumed to not
®	8	See
depreciate and, once created, cannot be consumed but only used as an input into the production process.
The population is again assumed to be constant, generations do not overlap and the total number of
generations is finite. The implications for an economy with an infinite number of generations will be
discussed below.

Since each generation's welfare is solely a function of consumption, generation g is acting
sustainably if, after producing and consuming, it leaves an endowment of capital and resources sufficient
for all succeeding generations to consume at the level that generation g could have consumed by being
both efficient and sustainable. As an intermediate step, we define the sustainability set, Og(C0), as the set
of all endowment pairs, (Kg,Sg), which are sufficient to allow generations g, g+l,...,T to consume at least
C0. This set can be expressed in symbols as
Og(C„) = {(Kg,Sg): 3 RgX): K + f (K^-C^„ Sg-Rg=Sg+1, (K^,, S^eCV,^)}.
The frontier of this set is the sustainability constraint, Og (Cj). There exists a maximum level of
sustainable consumption C , which is the greatest level of sustainable consumption given the available
resources. The actual endowment of the g^1 generation (Kg,Sg), lies on the sustainability constraint
associated with C*, Og (C*).
Derivation of a two dimensional sustainability constraint
The sustainability constraint in the capital-resource economy is derived using backward
induction. Consider the last generation in a T generation world. The last generation will, presumably,
use up all remaining resources and not invest in capital so that, if they are efficienct, Gj^fCKpS-j.). The last
generation's sustainability constraint associated with a consumption level C0 is the set of all capital-
resource endowments that will allow it to exactly produce C0. This constraint, CC in Figure 2, is simply
an isoquant. If the endowment pair inherited by generation T lies anywhere above CC, then it will have
more than enough total resources to produce C0. If it receives an endowment that falls below the
constraint, then it will not be able to produce C0.

t \
C f(K,S)=Co
-> K
The next step is to derive the sustainability constraint for the second to last generation. Consider
a point on CC, say X in Figure 2. If generation T is going to receive the endowment X, then generation
T-l will have to produce a level of output such that it is able to consume C0 and still leave generation T at
X. Since output, prior to choosing a level of consumption, can be used either for capital formation or
consumption, generation T-l must have an endowment sufficiently large to reach X' in Figure 2.

Figure 3
In Figure 3 three possible endowments A, A' and A" are indicated, all of which would be
sufficient to allow generation T-l to reach X'. Take point A, for example, where generation T-l's
endowment is (KT1, ST_,). By using Rj., of the resource, and taking advantage of its endowment of
capital, Kt_,, generation T-l could produce a total output of IT_,-t-C0. The curve connecting A and X' is a
production possibility frontier indicating the total output that can be produced at different levels of
resource use. The production frontier is concave because as more resource stock is used up (movement
vertically downward), the marginal increase in output declines. To reach X' from A, generation T-l
would have to use R^,. Because additional capital increases the marginal productivity of the resource,
endowments with more capital stock, like A, would require less resource use to reach X'. Endowments
with less capital stock, like A", would require more resource use to reach X'.

Figure 4

\ X a



\X 0, *

—— c

For each point like X, on the T^1 generation's sustainability constraint, therefore, there are a
multitude of possible endowments for the preceding generation that would allow it to consume C0 and still
leave generation T at X. By joining all the feasible endowments associated with X, we obtain a locus of
points labeled with a small x in Figure 4. We then repeat the same operation for another point, Y, and
obtain another locus, y in Figure 4. All the points on each of these two loci indicate endowments that
would allow sustainable consumption of C„ by generation T-l.
Consider two points along the feasible set of points x and y at a given level of capital Ko. Since
either of these two points lead to the same level of consumption in generation T, the upper point, on they
locus, indicated with an a is more than sustainable. That is, if the endowment inherited by generation T-l
were at a, then generation T-l could produce enough to pass on a sustainable endowment to generation T
and consume more than C0. Hence point a lies above 0T.] (C^. If we derived loci of sustainable
endowments similar to x and y for every point on CC, the outer envelope of these curves would be the
sustainability constraint for the T-l^1 generation Of-i (C0).

Figure 5
The resulting sustainability constraint for generation T-l can then be traced out and would take a
form like BB in Figure 5. Following the same procedure, the sustainability constraint for generation T-2
could also be traced out and would look something like AA. Repeating this process over and over again,
the sustainability constraint of the generation,	is found. This locus would be the set of
minimum endowments that generation g would need in order to consume C0 and still leave an endowment
of capital and resources so that generation g+1 and all following generations can also consume C0.
Once Og (C,,) is found, we can compare the actual endowment, (Kg,Sg), with the sustainability
constraint. If we find that the g^1 generation's endowment lies above Og (C0), then a C0 is not optimal, a
higher level of consumption could be sustainably consumed. If we find that the actual endowment lies
below Og (C0), then C0 is not sustainable and a lower level of consumption must be considered. In an
iterative fashion it would be possible to determine the level of consumption C* such that the g^1
generation's endowment lies on Og (C ).
Extensions and generalizations of the multi-dimensional sustainability constraint,
A number of extensions of the above analysis are worth pointing out. First, we can make some
inferences about the economy as we relax the assumptions of finite generations. Much like the hardtack
world above, the world that we have been discussing here may not allow sustainable positive levels of

consumption if the number of generations is infinite. One way to state this is that for any positive level of
consumption C0, any finite endowment (K, S) will become unsustainable (fall below the sustainability
constraint) in some finite number of generations, T . If, for example, the production function f(0 is CES
with an elasticity of substitution less than one6, then the average product of a unit of resource is bounded
from above, making it impossible infinitely sustain a positive level of output (Dasgupta and Heal, 1979).
If the resource is more like that of the corn economy, such that if there is any positive resource
stock S, a level of resource R{ S can be used without diminishing the resource endowment of the
following generation, then sustainability can be achieved even with an infinite number of generations. In
this case the sustainability constraint associated with any finite level of consumption will converge to a
single locus so that Og (Cy = Og+k (C0) = O (C0) for all finite k.7 This capital-corn economy simplifies
the analysis in many ways since if generation g can determine the sustainability constraint on which its
endowment lies, it can determine whether its actions are sustainable by evaluating if the endowment it
passes onto generation g+1 lies on the same constraint. This property is used to analyze indicators of
sustainability in Appendix A.
While we have considered only a two-dimensional endowment, the endowment vector could in
principle be extended to a third or higher order vector. The number of calculations involved in
6	A constant elasticity of substitution (CES) production function in K and R is of the form
f(K,R) =[a,K(<7-,)/0, o*l.
where cr is the elasticity of substitution between K and R If a = 1, and ai+ (*2 = 1, then f(K,R) is Cobb-
7	This can be seen by noting that if the g^1 generation inherits (Sg, K ), then passing on the same
endowment to the next generation would clearly be sustainable, though this might not be optimal.
Nonetheless, for any R>0 and £>0, there is a level of capital, K such that f(K>E)=C- So, for all resource
levels S that yield a positive recharge, R, we know that there is a capital level sufficiently high to support
any consumption level without diminishing the resource stock. Hence, there is an upper bound on the
sustainability constraint, composed of levels of K and S which can produce C0 without diminishing the
resource stock. Because this upper bound to the sustainability constraint exists, it must be the case that a
single constraint exists at or below this upper bound.

calculating such a frontier, however grows exponentially. Hence, the properties of the n-dimensional
sustainability constraint are not explored in this paper.
There are several lessons from the above theoretical models that we will draw on to discuss the
implications of sustainability for economic policy. First and foremost, in each of the models we showed
that to ensure that sustainability is achieved, policy makers must consider the joint objectives of economic
efficiency and sustainability. This may at first appear to be an obvious extension of the Second Welfare
Theorem, and in a sense it is. Yet, when we consider intertemporal problems, as some examples below
will show, very often economists voice only efficiency concerns when sustainability seems to be the central
Secondly, we find that substitution and replenishment are both sources of sustainability in the
long run. In the corn economy, sustainability required that each generation consumed no more than the
recharge to the resource stock. In the capital-resource economy sustainability could be achieved if
attention is given to the degree to which substitution is possible given the economy's productive capacity.
Here too, however, sustainability will not be guaranteed unless the economy operates within the bounds
defined by the sustainability constraint.
Finally, up to a point, sustainability can be achieved through substitution. As we see in the
capital-resource economy in which substitutability is possible, sustainability does not require that the
resource endowment passed from one generation to another be constant. Unless a particular resource is
both essential and non-renewable, a policy that leads to the reduction of that resource is not necessarily
unsustainable. However, unless specific measures are taken to increase other dimensions of the
endowment vector, policies that have the effect of diminishing the resources being passed on to future
generations will threaten sustainability. Here we see that our uncertainty makes defining policies that
pursue both sustainability and efficiency particular troubling. Accurate knowledge of the sustainability
constraint is never available. One policy intended to move the economy towards sustainability despite our
enormous uncertainty is discussed below. We turn now, however, to a more fundamental question.

Should Sustainabilitv Be An Economic Goal?
Advocating sustainability as a policy goal will be viewed with uneasiness by many economists
because of their strong propensity to avoid expressing views on what is fair and what is not. The
widespread interest that sustainability is generating among policy makers, environmental scientists, and
the general public is reason enough to assume, for the remainder of this paper, that making economies
sustainable is a worthy policy goal. We propose to conduct an economic discussion on a "what if basis:
What if sustainability were a goal of economic policy? What would the implications be for environmental
economics? Our case for arguing that this is a meaningful exercise for economists to participate in is
strengthened by the result that efficiency and sustainability need not be conflicting goals. It should be
possible to seek a path that is both efficient and sustainable. Proposed steps that are viewed by their
advocates as promoting sustainability will also have implications for efficiency. As a result, economists
are being drawn into the debate.
As long as we are dealing with potential qualms of our economic colleagues, a second question
also deserves attention. Some economists who will grant that sustainability is the potentially interesting
from a theoretical perspective may still argue that the concept is irrelevant to policy, since economic
growth can be expected to continue into the indefinite future. In the context of sustainability, economic
growth in excess of growth rates in population implies ever expanding economic opportunities for
successive generations. Witness for example, Beckerman's (1992) statement in the context of the debate
over policies to address global warming,
to give priority to highly speculative global environmental issues in general and to global
warming in particular, in the interests of future generations who are likely to be far richer than
we are today, and to take drastic action in pursuit of this goal, however costly it may be in terms
of current living standards, would represent an unjustified sacrifice of the clearly apparent
interests of billions of very poor people today.
In the context of this paper, such statements maybe interpreted as arguing that the sustainability
constraint is not binding. Let us consider this view further using the concept of evolving intergenerational
A look at the relationship between economies and nature makes it hard to escape the feeling
that future generations are in a vulnerable position. Each generation tends to treat as its endowment

virtually all the natural resources that it has the technological and economic means to exploit.
Historically, resource depletion and degradation were limited by technology, labor, and capital constraints.
Exploitation of natural resources on the scale that is feasible today was impossible. The current
generation, in contrast, is using non-renewable resources at an unparalleled rate. Furthermore, renewable
resources are being more and more heavily exploited and degraded on a global scale. There can be little
doubt that future generations will inherit natural resource endowments that are much reduced and much
Societies have historically augmented their natural resource endowments through conquest and
exploration to offset the depletion and degradation of their resources. Certainly, augmentation of resource
endowments will continue to occur, but diminishing returns to efforts in this direction maybe felt. No
more continents filled with nearly virgin resources are available. One has to wonder, for example, how
many more oil producing areas with reserves as large as the Middle-East or even Alaska's North Slope are
available for discovery.
With natural resource and environmental endowments declining, sustainability, if it can be
achieved at all, will depend on increasing non-resource components of the endowment that future
generations will receive. Just as capital can be augmented to makeup for reductions in the resource stock
in the simple economy above, in the real world non-resource components are augmented by processes that
we shall refer to collectively as "social progress". Progress takes many forms: scientific and technological
innovations, improvements in institutions, increases in cultural items (e.g., art and music), and human
and physical capital accumulation. Social progress creates substitution possibilities, reducing or
overcoming the ill-effects of declines in the natural resource endowments. Institutions, such as those
associated with markets, can also play a role, creating incentives for both substitution and social progress.
8 Of course, much can be done to reduce resource depletion and degradation. Still, the point is that
human life as it exists at the current time, appears to be incompatible with increasing or even constant
future resource endowments. Recycling, pollution control, and other approaches are less than perfectly
effective in stemming the tide of depletion and degradation.

In recent decades and centuries, social progress and resource augmentation in many countries
have been more than adequate. The result has been expanding per capita economic opportunities.
Despite reductions in the resource stock, these nations have apparently not violated their sustainability
constraint. Though this is encouraging, sufficient social progress to allow continued growth in per capita
economic opportunities may not be automatic.
Those who followed the "Growth Debate" of the 1970s no doubt find all this familiar. There,
systems scientists and economists debated the prospects for further economic growth.9 One can recast the
conclusions of systems scientists into today's language by saying that they concluded that then-current
economic trends were not sustainable. Economists responded by suggesting that the models developed by
systems scientists were woefully inadequate in portraying the possibilities for social progress and resource
augmentation. The Sustainability Debate of the 1990s has its own nuances, but it is fundamentally a
continuation of the Growth Debate of the 1970s, which in turn can be traced back at least to Malthus.
We do not propose to resolve this debate here. Rather, these are issues about which sensible
people ought to agree to disagree. Those who argue that the current economy is not sustainable ought to
admit that they could be wrong, Perhaps social progress will be adequate to counterbalance depletion and
degradation of natural resource endowments for the foreseeable future. And, those who have more
confidence in social progress should admit that the economy could possibly be on an unsustainable path.
Neither theoretical economic arguments nor empirical evidence are sufficient to justify a definite
conclusion about the sustainability of the time paths on which the earth's economies find themselves.
Accordingly, an investigation of the economic implications of combining efficiency and sustainability
goals could have substantial policy relevance.
An undercurrent in what has just been said about the Sustainability Debate now needs to be
made explicit: Implementation of the concept of sustainability constraints in actual policies would have to
9 Relevant literature is summarized in Hartwick and Olewiler (1986), Chapter 6.

be attempted in a world of extreme uncertainty. Our theoretical efforts here have been conducted under
the assumption of perfect knowledge. In fact, we of the current generation are quite ignorant about how
our use of environmental and other natural resources will affect the economic prospects of future
generations. As has already been emphasized, it is not clear whether the sustainability constraint is even
binding. Earlier generations have a limited basis for judging which resources can be exhausted and
degraded with little or no harm to later generations and which might be extremely valuable.
Furthermore, the nature of the trade-offs between environmental resource components of the endowment
vector and non-resource components are difficult to anticipate. Producing human and physical capital;
science and technology; art, music, and literature; and even social institutions requires that we of the
current generation use natural resources. In any given case, it is difficult to predict whether future
generations will be better off in terms of economic opportunities with more environmental resources or
with more social progress to augment their non-resource endowments. Alternative endowment vectors
(including various levels of natural resource and non-resource components) have highly uncertain
potential economic implications.
The uncertainty associated with these decisions is of an extreme kind, which for convenience,
we might term "ignorance."10 If we think in traditional terms, "risk" is used to characterize situations
where more than one future outcome is possible and where all outcomes are known in terms of their
payoffs and probabilities. "Uncertainty," in the traditional terminology, involves situations with more
than one possible outcome, where payoffs are known, but probabilities are completely unknown. Neither
of these constructs seems quite appropriate here, Under "ignorance," as we shall use that term, not all
possible outcomes are known and payoffs from known outcomes are not always clear. Probabilities are
likely to be inestimable or very tentative at best. 11
10 We believe this term was originally suggested to describe such uncertainty in natural resource
problems in some of the unpublished work of Alan Randall. He used the term in Randall (forthcoming).
Randall and Thomas (1991, p. 15) explicitly suggested that "the problem is one of ignorance rather than
mere risk and uncertainty," thus anticipating the argument made here.
For example, how does one deal with the logical requirement that probabilities summed across all
outcomes must equal unity if some of the possible outcomes are not known?

As is clear from any recent issue of mainstream economic journals, the standard procedure for
dealing with uncertainty involves assuming that outcomes and associated payoffs are known and
probabilities are known at least in subjective terms. It is worth asking whether such approaches are
applicable to ignorance. Perhaps strategies are needed that address ignorance directly, rather than trying
to fit the problem into a risk framework. 12 At any rate, as we move from theory to policy, ignorance must
be explicitly considered.
From Macro-Level Theory to Micro-Level Decision Criteria
Sustainability, as defined in this paper, is a macroeconomic concept. Either an economy, taken
as a whole, is on a sustainable path or it is not. To ask whether a specific macroeconomic alternative is
"sustainable" or not makes sense only in the context of the economy as a whole. A discussion of
macroeconomic issues associated with sustainability and national income accounting are discussed in
Appendix A. In the meantime, some attention needs to be devoted to considering how to go from the
macroeconomic status of the economy vis-^-vis a sustainability constraint to criteria that can be applied to
macroeconomic-level decision making.
In a sense, our goal is to develop microeconomic criteria for specific resource decisions. We
take it for granted that actual decisions relating to sustainability will have to occur in a piecemeal fashion.
In both the public and the private sectors, management of natural resources involves many individual
choices over time. Our task is to explore whether criteria can be developed to judge whether each such
decision is, in some sense, "sustainable."
As we are using the term macroeconomic, Pareto efficiency is also a macroeconomic concept.
An economy as a whole is either on its Pareto frontier or it is not. It is instructive to consider how the
transition from the macroeconomic level to the microeconomic level works for efficiency. That actual
economic decision making about the allocation of specific resources must be piecemeal is taken for
12 perringg (1991) suggests the use of a notion of uncertainty based on Shackle (1952) as an alternative
to standard risk analysis in such situations.

granted. An economist who notes that a Pareto condition is violated in some specific instance, prescribes
policies to make the economy "more efficient" with respect to that specific micro-level problem. Doing so
raises second-best considerations. Given that inefficiencies are present in many sectors of the economy,
applying the Pareto conditions piecemeal is unlikely to be fully optimal and the result of an intervention
intended to improve economic efficiency could actually reduce aggregate social welfare. However,
attempting to fine tune micro-level decision criteria to account for inefficiencies elsewhere is normally not
practical. In practice, the economist hopes that application of simplified efficiency criteria in arriving at
individual public decisions will improve efficiency most of the time.
Similar strategies will be required if sustainability is to be translated into workable criteria at
the micro-level. A decision alternative maybe said to "enhance sustainability" or "make the economy
more sustainable" if it expands the aggregate economic opportunities available to future generations. In a
partial sense, it maybe relatively easy to determine how the policy is affecting a few components of the
endowment. Much more difficult to anticipate, however, are the indirect effects of physical spillovers and
reactions by economic agents to the new policy. As we have pointed out already, economic opportunities
depend on the full endowment vector including non-resource components as well as natural resource
components. If the policy indirectly leads to changes in the economy that affect other components of the
endowment by diminishing their quantity or quality or inhibiting their growth, then, in net, the policy
might have a negative effect on sustainability.
Such complications are analogous to the problem of the second best of efficiency analysis. For
example, consider a policy that would encourage soil conservation and would not, in any identifiable way,
impede social progress. Society might proceed with this intervention in order to enhance sustainability,
only to learn that it led farmers to use more chemicals that contaminated groundwater. Just as sectors of
the economy are interlined by market signals that affect whether a given projector policy is efficient, so
resource and non-resource endowments are linked both in nature and through the economy in complex
ways. Obviously such linkages should be identified and evaluated to the extent possible in considering
whether public or private decisions will enhance sustainability. But, the ability to trace such effects is
likely to be limited in practice. Following the efficiency analogy, the analyst can do little more than hope

that the more obvious effects of choosing alternative courses of action will be sufficient most of the time to
indicate whether those alternatives will enhance or reduce future economic opportunities.
Trade-offs between different components of the vector of endowments must be carefully
considered in judging the sustainability-enhancing potential of a particular choice. Our example of soil
versus groundwater illustrates this well. Suppose that, without a project, future generations living in a
certain region will inherit less soil but purer groundwater. If the soil conservation project is adopted, the
opposite will be true. Which alternative would contribute most to their utility possibilities is not obvious
at first glance. If soil erosion is economically irreversible over relevant time spans, but groundwater could
be purified using known technologies at modest cost, the soil erosion control project might be judged as
contributing positively to sustainability. There are likely to be many judgment calls on such issues.
Confronted with ignorance and the possibility of unexpected consequences of interventions
designed to enhance sustainability some will no doubt decide that the whole problem of sustainability is
intractable and choose to ignore it. The theory of the second best and concerns about economic fairness
have led some to adopt a similar attitude with respect to economic efficiency. Others, whether the issue is
efficiency or sustainability, accept second best problems and ignorance as facts of life, and try to figure out
how humankind might muddle through anyway. As part of the latter group, we will now proceed to
consider policies that might help to achieve sustainability goals.
Two preliminary steps toward practical implementation of the framework developed here will
be discussed. First, we shall consider the Safe Minimum Standard of Conservation for endangered
species, reinterpreting this long discussed concept as a sustainability constraint. Second we turn to global
warming, focusing on how a carbon tax would work in an economy seeking an efficient, sustainable path.
In both of these examples, we emphasize that if sustainability is to be achieved, policies should explicitly
consider this goal. Efficiency based analysis alone will not ensure sustainability.
The Safe Minimum Standard

Extinction of plants and animals is an economic issue because it narrows the biological
diversity upon which current and future generations may depend for the stability and productivity of the
ecosystems within which human activities must be conducted. Furthermore, the earths plants and
animals provide a reservoir of potential new resources to produce food, building materials, aesthetic
enjoyment, energy, paper products, pharmaceuticals, transportation, recreation, and other desired
commodities and services. Maintaining a sufficiently diverse flora and fauna has the potential to
contribute to both economic efficiency and sustainability.
As long as human-caused extinctions were rare, there was little need for concern. Species
diversity was a free gift of nature. At the end of the Twentieth Century, however, species diversity can no
longer be taken for granted. Thousands of species of plants and animals will be lost in the next few
decades unless steps are taken to save them. Such steps, however, would require the commitment of
scarce capital, labor, and natural resources. Thus, on the one hand, massive extinction of living
organisms may limit future economic possibilities. On the other hand, reducing the rate at which
biological diversity is eroding will involve economic costs to the current generation that not only will
harm its members but could conceivably affect the non-environmental endowments of future generations.
In the terms developed here, extinctions threaten efficiency and sustainability, but measures to protect
diversity could also have the potential to threaten both goals. Defining a sustainable, efficient course is
not a simple problem.
The safe minimum standard of conservation (here abbreviated SMS) as originally proposed by
Ciriacy-Wantrup (1952) and further developed by Ciriacy-Wantrup and Phillips (1970), Bishop (1978,
1980) and Randall (1991, 1995). Adopting the SMS strategy as a policy objective would mean avoiding
extinction in day-to-day resource management decisions, Exceptions would occur only where it is
explicitly decided that the costs of avoiding extinction are intolerably large or other social objectives must
take precedence.
Randall (1991, p. 16) has explained the idea this way
The SMS rule places biodiversity beyond the reach of routine trade-offs, whereto give up ninety
cents worth of biodiversity to gain a dollars worth of ground beef is to make a net gain. It also
avoids claiming trump status for biodiversity, permitting some sacrifice of biodiversity in the face

of intolerable costs. But it takes intolerable cost to justify relaxation of the SMS. The idea of
intolerable costs invokes an extraordinary decision process that takes biodiversity seriously by
trying to distinguish costs that are intolerable from those that are merely substantial.
The SMS strategy does not involve a new economic paradigm but is instead a crude step toward
the ideal of a fully efficient, sustainable economy. Because of ignorance about the future and other issues
(Bishop and Woodward 1994), such an ideal is far from attainable. The SMS should be thought of as a
practical strategy to be implemented in lieu of the ideal. The goal of the SMS strategy is to safeguard the
economic opportunities of future generations by preserving some species that will prove useful and
valuable to them and that would otherwise have been lost. The first-best solution to the problem, were it
attainable, would involve an optimal endowment composed of a wide range of species and other resource
and non-resource components. The SMS strategy is intended to push economies in that direction by
augmenting future endowments of species diversity. Under the SMS, we presume that substitution of
other components of the endowment for the species is difficult but not impossible. Costs of protecting a
species become "intolerable" when it is believed that protecting a species might be so restrictive that both
efficiency and sustainability would be inhibited.
The SMS strategy also requires consideration, within the limitations imposed by ignorance, of
the implications of preservation for efficiency and for the non-environmental endowments of future
generations. The social costs of choosing the SMS are important indicators of potential losses in
efficiency and sustainability. Ignorance means that the full benefits of preserving specific species cannot
be known. The higher are costs, however, the more likely they are to exceed benefits, were the latter filly
known. Furthermore, though obviously any generalization would be questionable, one might expect that
the higher are costs, the more disruptive will preservation of species be to social progress and hence to the
non-environmental endowments of future generations. The SMS seeks to increase the future endowments
of biological diversity without large sacrifices in efficiency or social progress.
Social costs here include the out-of-pocket costs for protecting species of plants and animals. For
example, guards may be needed to protect an animal species from poaching. Opportunity costs, reflecting
foregone resource uses, would need to be added in. Such opportunity costs might include, for example,
the timber value of old-growth forests that must remain unharvested to provide habitat. External costs,

such as livestock losses to an endangered predator, may also occur and need to be counted. Against these
costs must be counted any measurable benefits from preservation. Some species, though endangered, may
provide aesthetic enjoyment. Some members of society may hold existence values for preservation of
endangered species of wildlife (Boyle and Bishop 1987; Bowker and Stoll 1988). If so, these should be
counted. Because the long-term benefits that biodiversity may contribute through ecosystem stability and
discovery of new resources are so difficult to anticipate, probably no allowance for them will be possible in
most cases. We stress this problem by defining the net social costs of the SMS as out-of-pocket costs, plus
opportunity costs, plus external costs minus measurable benefits. Measurable benefits are those benefits
that can be expressed in monetary terms with reasonable confidence.
Whether the net social costs of the SMS are within the bounds of acceptability or not is a social
decision that may have to be left to Randall's "extraordinary decision process." What we are asking, in
part, is whether or not it is reasonable for the current generation to be required to make a given level of
sacrifice to enhance the species diversity endowments of future generations. Such decisions involve value
judgments beyond those that most economists are comfortable making. Societies, through the institutions
of government, may have to consider such issues without direct help from economists.
Since the SMS depends upon the current generation's judgment as to what represents
"intolerable" costs, it is nearly inevitable that either too many or too few species will be preserved under
the SMS compared to the ideal. Because of ignorance about which species will ultimately prove valuable
and which will not, to some extent, the wrong species will be saved. Some species that would have turned
out to be of great value to future generations may be lost. Some species that will never be worth anything
either directly or in terms of their contributions as parts of larger ecosystems maybe saved.
Note also that the SMS would only be one of many objectives of policy. As Randall stated in the
quotation presented earlier, the SMS would not have "trump status." Many worthwhile objectives must
vie for economic attention and public resources, and preservation of biodiversity probably would not take
precedence in all cases. Most societies have a policy objective of preventing murder, yet the resources
devoted to this end are not sufficient to prevent all murders. Similarly, if the SMS were an objective of
policy, this would not mean that all extinctions of plants and animals would be prevented. The SMS

policy would help limit extinction of plants and animals to those that can be saved only by bearing
unacceptably high costs or through unacceptable sacrifices in other social objectives.
Randall (1991, 1995) has recently introduced a new and highly original framework for
considering the SMS in the context of public policy formulation. This framework is useful in considering
the relationships between efficiency and what we here term sustainability. Since loss of biodiversity raises
intergenerational ethical questions, Randall reasoned that insights might be gained by considering it in
the context of three major theories of ethics. Randall argued that making social choices based on benefit-
cost analysis can draw some support from all of these schools but none would endorse benefit-cost analysis
in an unqualified way. However, quoting Randall (1995, p.36),". . . it seems that the same general kind
of decision rule - maximize net benefits subject to an SMS constraint -is admissible under the
consequentialist, duty-based, and contractarian reasoning." Since concerns about sustainability are
grounded in ethics, this would appear to be a promising direction for additional work.
Global Warming and the Carbon Tax
The possibility of global warming due to the accumulation of greenhouse gasses in the
atmosphere poses a very real threat to global sustainability. Based on climatic models and some empirical
evidence, scientists believe that emissions of carbon dioxide and other "greenhouse gases" into the
atmosphere is setting in motion a gradual warming of the planet. Though highly speculative, global
climatic models predict that by the end of this decade greenhouse gases that will have accumulated in the
atmosphere will commit future generations to a planet as much as three degrees Celsius warmer than the
climate we enjoy today (Cline, 1992, Table 2. 1). Over the long run, even greater changes in the globe's
climate are predicted. With this increase in temperatures will come a wide range of effects on
humankind. Some effects will be positive, such as regional increases in agricultural production. Other
changes will negatively affect society, such as the destruction of coastal ecosystems and real-estate due to
rising sea levels. In net, it is generally believed that the effects of global warming will impose costs upon
future generations (see, for example, National Academy of Sciences, 1992).
Emissions of greenhouse gases are intimately linked with the economy. Virtually all productive
activities in developed nations use carbon based energy, contributing to the greenhouse warming. To

some extent, our economic activities today are carried out at the expense of the climate of the next
century. Using the language of this paper, our production today diminishes the climatic endowment of
future generations. The greenhouse problem therefore, is fundamentally one of sustainability.
Despite the very long term distributional consequences of global warming, the debate within
economics about how to best address the problem has centered on issues of efficiency. Nordhaus (1993),
for example, refers to the greenhouse effect as "the granddaddy of public goods problems" (p. 18). When
seen in this light, the problem can be reduced to a standard externality problem in which the level of
greenhouse gas production is inefficiently high. Analysis motivated entirely by an efficiency perspective,
however, will fail to address what we see as primarily one of sustainability since, as we have shown above,
pursuit of the efficiency will not necessarily lead to sustainability.
The policy option to address global warming concerns that is most frequently discussed is a tax
on carbon emitted by the burning of fossil fuels. This tax would encourage a reduction in the level of C02
emitted but would impose costs on some sectors of the economy. The costs of a carbon tax policy would
take the form of a reduction in the quantity of goods and services that are produced using carbon based
fuel. These costs would be borne by both current and future generations. The benefits of a carbon tax, on
the other hand, would accrue primarily to the generations of the next century and beyond. As a result of
reductions in greenhouse gas emissions, the planet would warm less than it would have without the policy,
reducing the costs that will be borne by those generations. The efficient level for the tax is where the
marginal benefit of increasing the tax equals its marginal cost. At lower tax rates, the marginal present
value of benefits exceeds costs, at higher rates the marginal cost exceeds the benefits.
Implicitly, such efficiency measures look for the point where the timers from an additional
reduction in gases can no longer payoff the losers. Elsewhere (Woodward and Bishop, forthcoming) we
have argued that standard efficiency analysis of global warming implies a distribution of rights in which
"the current generation has the right to emit endlessly and future generations are obligated to accept the
consequences unless 'they' are capable of compensating 'us'." While efficiency driven policies may
diminish the warming experienced by future generations, under such a policy greenhouse gases would

continue to accumulate and the planet will continue to warm. Efficient policies, therefore, will not
eliminate the threat that the greenhouse effect poses to global sustainability.
As Beckerman argues above, it could be presumed that sustainability is not at risk because other
components of the endowment vector are growing fast enough to more than compensate future generations
for losses in the climatic endowment. In this case only the efficient level of reduction could be justified.
The uncertainty in global warming analysis, however, is extreme. Of a surveyed group of experts, ten
percent estimated that the damages associated with a three degree C warming would be 5.5 percent of
world output or more while another ten percent had a median estimate of zero total loss or less (Nordhaus,
1993, p. 17). With such uncertainty on only one issue, how can we be certain that the multitude of
changes in the endowment vector overtime will in net mean that the sustainability constraint is not
binding? Sustainability may indeed be threatened, and if this is true, steps beyond those that can be
justified on efficiency grounds maybe necessary.
In Woodward and Bishop (forthcoming) we propose that given these uncertainties, a prudent
policy would be to address both efficiency and sustainability. Recognizing that global warming does have
efficiency implications, a carbon tax should be used to reduce emissions at least to the point where the
marginal benefits of a reduction in emissions equals the marginal cost. However, we argue that the
carbon tax offers an opportunity to take "a full step in the direction of sustainability." Since a carbon tax
would generate enormous revenues, we suggest that those revenues should be used to explicitly
compensate future generations by augmenting other components of the endowment vector. This could be
done by improving environmental components of the endowment, stimulating technological progress,
expanding infrastructure, even diminishing the debt burden that we pass on to our children (Bromley,
1989). Moreover, when global warming is seen as threatening sustainability, it might be acceptable to
adopt a policy which reduces emissions beyond the level which follows from efficiency analysis in order to
augment the climatic endowment of future generations. Just as infratemporal distributions cannot be
justified on efficiency grounds, there is no reason to believe that a policy that redistributes
intergenerational endowments would have benefits in excess of the costs.

The issue of global warming demonstrates well the importance of explicitly recognizing
sustainability as a goal within economic analysis. The concern about the greenhouse effect arises not
because we see the problem as reducing our total economic productivity, but because a sense of fairness
and moral responsibility makes the status quo unacceptable to many. As such, while efficiency is
important in discussing any policy alternative, it cannot be the sole criterion on which economists base
their policy recommendations.
We have demonstrated in this paper that, if sustainability is deemed to be a social objective, then
it both can and should be incorporated directly into economic analysis. Economic efficiency does not
necessarily lead to sustainability. To ensure sustainability society must be take care to avoid violating its
sustainability constraints. In theory, sustainability constraints can be determined which establish exactly
the endowments that need to be passed onto the next generation in order to provide them with
opportunities equal to those enjoyed by the present generation. In practice, exact determination of the
sustainability constraint is impossible given the enormous uncertainty that dominates long-term economic
and environmental issues. However, despite our ignorance and the complexity of interactions between the
economy and the environment some guidelines for policy can be established.
The SMS and the carbon tax with associated spending priorities illustrate how sustainability
constraints tight be implemented in practice. Obviously, a fully general constraint would have to cover a
wide range of other resource issues, possibly including contamination of groundwater, soil erosion,
deforestation, ozone depletion, and the like. In each such case, the endowments of future generations
would need to be carefully considered in making resource management policy. Furthermore, a distinct
approach would need to be developed to protect each such resource to give due attention to efficiency as
well as sustainability. Once a more or less general constraint is in place, then it should be possible for
both public and private economic agents to re-optimize to pursue the efficiency goal within the new
regime of intergenerational endowments. In this way, the economy would move toward an efficient and
sustainable path.

Appendix A: Sustainability Constraints and Indicators of Sustainability
While in practice it may be impossible to find the sustainability constraint with precision, the
construct can be used to understand the meaning of sustainability. Consider a capital-resource economy
in which sustainability is possible so that the sustainability constraint converges to a single locus. An
implicit function O (Kg,Sg)=0, can be defined where 0 is a constant and 0* (•) closely approximates
Og(C ). Taking the total differential of O (•), we find that along the sustainability constraint
Hence, an approximate rule for sustainability would be to ensure that
This constraint is similar to many other linear rules for sustainability, such as Hartwick's (1977) rule that
the scarcity rents from a resource should be reinvested to ensure sustainable growth. This relationship is
also similar to the implicit rule that in natural resource accounting studies which estimate the depreciation
of natural resources. What distinguishes the rule derived from the sustainability constraint is that it is
grounded only in the production possibilities of the economy and does not assume that the economy is
operating on the efficiency frontier.
Consider a simple estimate of the net domestic investment (NDI) as might arise from a natural
resource accounting study. Accounting for both the appreciation and depreciation of the capital and
resource sectors, net investment is estimated, NDI=pK-AK+ps-AS. If the estimated value of NDI were
greater than zero, this might be interpreted as indicating that the economy is on a sustainable path since
the value of the total endowment has not diminished. This rule will be consistent with sustainability,
dO* IdO*
however, only if / = Pk/Ps • Since prices in an economy are critically dependent upon the
eK / &S
distribution of endowments across agents in the economy (see Howarth and Norgaard, 1990), it is not
guaranteed that the market prices would be appropriate even in a perfectly functioning market economy.
The problem becomes more severe if the endowment has public good characteristics (e.g., national parks),

markets do not exist for portions of the endowment (e.g., the climate) or other sources of market failure
are present. The sustainability constraint, therefore, provides a useful target for valuation in natural
resource accounting studies.
This framework, therefore, provides a new perspective on economic indicators of sustainability.
By working directly with the sustainability constraint we allow for substitution, but do not presume that
markets provide all necessary information with perfection. Of course the framework is not fully developed
and, would certainly have substantial informational needs, perhaps even more so than standard
environmental accounts (United Nations, 1993). We would argue that the returns might be higher since
such an approach leads directly to the societal sustainability constraint and, therefore, is a better indicator
of the economy's sustainability.

Arrow, Kenneth Joseph. 1963. Social Choice and Individual Values. New Haven, Yale University Press.
Beckerman, W., 1992. "Economic growth and the environment: Whose growth? Whose environment?"
World Development. 20(4):481-96.
Bishop, Richard C. 1978. "Endangered species and uncertainty: the economics of a safe minimum
standard." American Journal of Agricultural Economics 60:10-18.
Bishop, Richard C. 1980. "Endangered species: an economic perspective," Transactions of the 45t,h
North American Wildlife and Natural Resources Conference, pp. 208-218.
Bishop, Richard C. and Richard T. Woodward. 1994. "Efficieny and sustainability in imperfect market
systems." University Graduate Faculty of Economics Lecture Series #6, Oregon State University.
Bowker, J.M. and J. R Stoll 1988. "Use of dichotomous choice nonmarket methods to value the
whooping crane resource." American Journal of Agricultural Economics. 10:372-381.
Bromley, D. W., 1989. Entitlements, missing markets, and environmental uncertainty. Journal of
Environmental Economics and Management, 17(2): 181-94.
Boyle, Kevin J. and Richard C. Bishop, 1987. "Valuing wildlife in benefit cost analysis: a case study
involving endangered species." Water Resource Research. 23:943-50.
Ciriacy-Wantrup, S.V. 1952. Resource Conservation: Economics and Policies. Berkeley University of
California Press.
Ciriacy-Wantrup, S.V. and William E. Phillips. 1970. "Conservation of Calfornia Tule Elk: a
socioeconomic study of a survival problem." Biological Conservation. 3(1): 23-32
Cline, W.R, 1992. The Economics of Global Warming. Institute for International Economics,
Washington, D.C.
Fisher, Eric O'N. 1992. "Sustained growth in the model of overlapping generations," Journal of Economic
Theory. 56:77-92.
Hartwick, John M. 1977. "Intergenerational equity and the investing of rents from exhaustible resources"
American Economic Review 67(5)972-74.
Hartwick John M., and Nancy D. Olewiler. 1986. The Economics of Natural Resource Use. New York:
Harper and Row, Publishers.
Howarth Richard B. 1991. "Intertemporal equilibria and exhaustible resources: an overlapping
generations approach." Ecological Economics 4:237-252.
Howarth Richard B., and Richard B. Norgaard. 1990. "Intergenerational resource rights, efficiency, and
social optimality," Land Economics. 66(1): 1-11.
National Academy of Sciences, National Academy of Engineering, and Institute of Medicine. 1992. Policy
Implications of Greenhouse Warming: Mitigation. Adaptation and the Science Base. National
Academy Press, Washington D. C..

Nordhaus, William D. 1993. "Reflections on the economics of climate change" Journal of Economic
Perspectives 7(4): 11-25.
Norgaard Richard B. 1991. "Sustainability as intergenerational equity: the challenge to economic thought
and practice," Working Paper, Office of the Chief Economist of the Asia Region, The World Bank.
Page, Talbot. 1977. Conservation and Economic Efficiency An Approach to Materials Policy.
Baltimore: Johns Hopkins University Press
Perrings, Charles. 1971. "Reserved rationality and the precautionary principle: technological change,
time and uncertainty in environmental decision making." In Robert Costanza, ed. Ecological
Economics. New York. Columbia University Press.
Randall, Alan. 1991. "Thinking about the value of biodiversity," Unpublished manuscript, Department of
Agricultural Economics and Rural Sociology, Ohio State University.
Randall, Alan and Michael C. Farmer. 1995. "Benefits, costs, and the safe minimum standard of
conservation" In Daniel W. Bromley, ed. Handbook of Environmental Economics. Cambridge,
Mass., Blackwell.
Randall, Alan. Forthcoming. "Thinking about the value of biodiversity," Unpublished manuscript,
Department of Agricultural Economics and Rural Sociology, Ohio State University.
Randall Alan and Michael Thomas. 1991. "The role of limits of economics in decision-making regarding
non-indigenous species." Draft prepared for the Office of Technology Assessment U.S.
Shackle, G.L.S. 1952. Expectations in Economics. Cambridge. England, The University Press.
United Nations, Department for Economic and Social Information and Policy Analysis. 1993. Integrated
Environmental and Economic Accounting: Interim Version. United Nations, New York.
Woodward, Richard T. and Bishop, Richard C. forthcoming. "Efficiency, sustainability and global
warming." Ecological Economics.

Jim Hrubovcak
Michael LeBlanc
Kelly Eakin

Detailed information derived from the national income and product accounts provide the
basis for economic interpretations of changes in the nation's income and wealth. Our intent in this
paper is to more accurately measure agriculture's contribution to national income. We develop a
theoretically consistent framework for incorporating natural capital and environmental goods into
the existing income accounts. Next, we apply the framework and adjust agricultural income and
national income to reflect the depletion of natural capital (land and water) caused by agricultural
production and the non-market effects of agricultural production on output in other sectors of the
economy and consumers. Specifically, the effects of soil erosion on agricultural productivity and
income, the economic effects of decreased water quality, and the depletion of water stocks are
presented as examples of the potential scope of accounting adjustments needed in the agricultural
sector. Estimated adjustments to net agricultural income are in the range of $4 billion and have
declined as a percentage of net farm income since 1982. Our estimates suggest that agriculture's
contribution to social welfare far exceeds the environmental damages and deterioration of the stock
of natural capital resulting from the production of food.

National income accounting is one of the most important economic policy making tools
developed in the last 50 years. Detailed information derived from the accounts provide the basis
for economic interpretations of changes in the nation's income and wealth. These national income
and product accounts (NIPA) through their measures on Gross Domestic Product (GDP) and Net
National Product (NNP) often provide the only meaningful indicator of the effects of public policy
interventions, Nearly from the inception of national income accounting, however, economists have
criticized the NIPA by identifying inconsistencies with the underlying theory and the empirical
application of the theory.
Early criticism of the NIPA centered around the treatment of capital, leisure, and
government expenditures. Recent critiques, with historical roots in the early 1970s, question the
use of estimates of NNP as a measure of social welfare because it does not account for the value
of changes in the stock of natural resources nor does it include the value of environmental goods
and services. Critics question the credibility of the accounts because natural and reproducible
capital are treated asymmetrically and the value of non-marketed environmental goods and services
is not captured (Prince and Gordon, 1994).1 NNP, it is argued, is not a useful measure of long-
term sustainable growth partly because natural resource depletion and environmental goods are not
considered. The failure to explicitly consider the environment in the accounts misrepresents the
current estimate of well-being, distorts the representation of the economy's production and
substitution possibilities, and fails to inform policy-makers on important issues related to economic
growth and the environment.
Several attempts to adjust income measures to account for the environment exist, it is
most common for these studies to focus on accounting for natural resource depletion (Repetto,
1992; Smith, 1992; Nestor and Pasurka, 1993; U.S. Department of Commerce, 1994).2
1	Our definition of non-marketed goods includes environmental amenities and disamenities.
2	Smith (1992) suggests his work should be characterized as environmental costing rather than
environmental accounting.

Theoretical and empirical problems persist, however, particularly when the level of environmental
services and damages are estimated. For example, no consistent approach for the treatment of
"defensive expenditures" in response to or in anticipation of environmental injury has emerged from
the literature (Ahmad, El Serafy, and Lutz, 1989).
Our intent in this paper is to more accurately measure economic well-being. Improving the
measure of current economic activity requires incorporating non-market final goods and bads into
the existing accounts. Economic well-being, however, extends beyond current economic activity
and must also reflect future production possibilities. We begin by developing a theoretically
consistent framework for incorporating natural capital and environmental goods into the existing
income accounts, Next, we empirically apply the framework and adjust agricultural income and
national income to reflect the depletion of agricultural natural capital (land and water) and the non-
market effects of agricultural production on output in other sectors and consumer utility.
The theoretical framework developed for this study is grounded on the work of Arrow and
Kurz (1970), Weitzman (1976), Solow (1986), Hartwick (1990), and Maler (1991). Weitzman has
shown that the current-value Hamiltonian in a neoclassical growth model of the aggregate economy
can be interpreted as NNP.3 Solow incorporated exhaustible resources as distinct capital assets
into Weitzman's treatment of NNP. Hartwick and Maler extended Solow's approach to capture
renewable resources and environmental capital (pollution abatement). In our analysis, the
Hartwick-Solow-Weitzman framework is extended to include three production sectors (agriculture,
non-agriculture, and household production). This extension allows us to adjust both agricultural
and national income. Rather than viewing non-market environmental goods as externalities, we
follow the prescription of Solow (1992) and cast the environment as a set of natural capital assets
providing flows of goods and services to the economy. Economic use of natural capital results in
feedback effects: depletion of stock of natural capital reduce future flows of goods and services
from the environment, degradation due to the disposal of residuals results in costs imposed on third
3 This interpretation requires a re-normalization of the current value Hamiltonian.

parties. In addition, firms and households are allowed to make expenditures for pollution
abatement and control.
Results from a dynamic optimization model are utilized to adjust NNP and net farm product
(NFP) for the use of natural capital assets, in addition, NNP reflects the value of net changes in
capital goods (net investment) and the value of net changes in the stock of natural capital.
Optimizing the current value Hamiltonian yields scarcity values for all capital stocks including
natural capital. The optimization process, therefore, generates relationships for adjusting current
NNP to account for the current value of the loss of natural capital stocks from using exhaustible
resources and depleting and degrading renewable and environmental resources.
Theoretical results from our model mirror Hartwick's results. That is, GDP includes priced
resource input flows and these flows from capital stocks should be off-set by deductions from GDP
to incorporate the value of changes in natural resource capital stocks to arrive at NNP.4 Our
empirical application suggests only minor changes are necessary when agricultural natural resource
effects are incorporated into the national income accounts. Adjustments to the national accounts
are minor because agricultural production is a small component of GDP (less than 2 percent) and
most extra-agricultural effects are currently captured in GDP. Larger changes are warranted,
however, in the adjustment of net agricultural income, Most effects represent income transfers
between agriculture and other sectors.
Agricultural income is adjusted to reflect the value of changes in the stocks of "effective"
farmland, water quality, and the stock of ground-water. These natural capital stocks may change
due to damages associated with agricultural production. Specifically, the effects of soil erosion on
agricultural productivity and income, the economic effects of decreased surface-water quality, and
the depletion of ground-water stocks are presented as examples of the potential scope of
accounting adjustments needed in the agricultural sector. We adjust income for changes in the
stock of ground-water because in some regions there has been a sustained withdrawal of ground-
4 Possible increases in the value of natural or environmental capital are not excluded.

water stocks in some regions of the United States. Our estimated adjustments would require net
agricultural income to be revised downward by $4 billion (6 percent). These estimates of
adjustments to net farm income are consistent with a view of U.S. agriculture where environmental
problems exist and the resource base is depreciating, but also suggest that agriculture's
contribution to social welfare far exceeds the environmental damages and deterioration of the stock
of natural capital resulting from the production of food.
National Income Accounting
The national income and product accounts (NIPA) were developed primarily to monitor the
macroeconomic performance of the economy. The most widely used measure or statistic of
economic activity is gross domestic product (GDP). GDP is highly correlated with employment and
capacity utilization and therefore central to how business cycles are defined and tracked.
A simple circular flow diagram is a powerful model to illustrate the flow of final goods and
services from the business sector to the household sector and the" concurrent flow of factor
services from households to firms (Figure 1). In a monetized economy, goods and services
exchange for consumer expenditures while primary factors of production (endowments of capital,
Figure 1. Circular Flow Model

labor, and land) exchange for wages and salaries, rent, interest, and profit. The circular flow model
suggests two methods for measuring the monetary value of current GDP: flow-of-output and flow-
of-income. In a flow-of-output approach, all expenditures on final goods and services are added
together. This measure captures the transactions from the "upper loop" of the circular flow model
and includes the value of new capital (gross investment), government purchases of goods and
services, and net exports. The flow-of-income alternative yields an equivalent measure of GDP and
is computed by summing payments to the primary factors of production. Because GDP is a
measure of final goods and services, purchases of intermediate goods must be excluded. The
failure to exclude intermediate goods and services from national income results in "double-
counting" and an over-statement of the level of economic activity.
Table 1 provides a summary of the NIPA for 1992. The table illustrates the flow-of-income
and flow-of-output approaches. Though arrived at in different ways, the calculation of national
income and GDP are equal in either case ($6 trillion). The flow of income approach include
compensation of employees ($3.6 trillion), proprietors income ($0.4 trillion), corporate profits ($0.4
trillion), net interest ($0.4 trillion), and rental income. The flow of output approach includes
expenditures on final goods and services by households ($4.1 trillion), the government ($1.1
trillion), and gross investment by firms ($0.8 trillion).6
Net of taxes, the largest single item differentiating GDP from national income is the
consumption of fixed capital or depreciation. For 1992, U.S. GDP exceeded $6 trillion while
national income approached $5 trillion. Depreciation of the U.S. capital stock was estimated at
$657.9 billion or about 11 percent of GDP. The concept of capital stock depreciation is particularly
important when we turn our attention to natural capital and environmental assets.
Table 2 summarizes the calculation of farm income for 1992 using a flow-of-income
approach. Gross farm income in 1992 was $84.4 billion or about 1.4 percent of U.S. GDP. While
6 However, the current NIPA system attributes household and government investment to current

wage income (compensation to employees) is by far the largest income category at the national
level (74 percent of U.S. national income), proprietors income (65 percent) and net interest (15
percent) are the largest components of net farm income. Consumption of fixed capital in
agriculture is 26 percent of gross farm income, over twice as large as the aggregate national rate.
Income accounts are subject to mismeasurement either by improperly including or excluding
items. Including the exchange of intermediate goods and services in the measure of national
income is an example of improper inclusion. Similarly, counting transfer payments or non-
productive redistributions such as social security payments, welfare payments, and agricultural
deficiency payments as gross income is inconsistent with the received definition of national
Improper exclusion occurs when the value of a final good or service is not included in the
accounts. This occurs when a good or service is traded in informal markets commonly referred to
as the "underground economy." Often these transactions in the form of "cash-only" arrangements
are undertaken to avoid taxes. "Non-market" goods and services are also often excluded from the
income accounts because they are difficult to measure. Examples include unpaid housework and
child-care and environmental goods and services. In some cases, market values have been imputed
for "non-market" goods and the income accounts adjusted accordingly. The value of housing
services received from owner-occupied houses is the best example.
The treatment of several elements in the accounts remain controversial and unclear,
Leisure, for example, has properties associated with a normal economic good. Yet, whether and
how to include the consumption of leisure in the national income accounts is unresolved. Another
example is criminal activity. Criminal activity is typically viewed as reducing not enhancing social
welfare and therefore not included in GDP. Legal gambling services in Nevada and New Jersey are,
however, included. Excluding criminal transactions reflects a moral judgement about the
desirability of illegal goods and services as indicators of social well-being. The cost of this moral
judgement is to reduce the accounts usefulness as a measure of economic activity.

Government expenditures on military defense, police, and environmental clean-up add to
the conventionally measured income accounts. Nordhaus and Tobin (1972) argue, however,
increases in these expenditures reflect the increasing "disamenities of urban life" that decrease
social well-being. Similarly, increases in household "defensive" expenditures on items like mace
and bottled water may signal a decrease in social welfare.
Environmental Accounting
Environmental accounting addresses the improper exclusion of the services provided by
environmental goods and the asymmetric treatment of natural capital and reproducible capital
within the existing accounts. Including the provision of environmental goods and services greatly
increases the complexity of properly adjusting the income accounts. Environmental goods and
services rarely have observed market prices or easily measurable market quantities. The absence
or incompleteness of these markets can have distorting effects on the good for which markets
exist. Thus, even if environmental goods and services are not included in the accounts, their
existence may cause distortions in the relative prices in traditionally measured sectors. If so, the
view of measured NNP as the current consumption value of a dynamically optimal resource
allocation is flawed.
Income accounting in the U.S. does not correct for price distortions. In developing
countries, however, significant effort is made to correct income accounts for market distortions
when the correction may be important for deciding among competing investment projects. The
implicit rationale for not adjusting market prices in developed countries is markets are well
developed and distortions, to the extent they exist, are small. However, price distortions with
respect to environmental and agricultural goods may be relatively large.
Changes in environmental quality have multiple effects across sectors and consumers.
Producers are affected because changes in environmental quality can affect the productivity of
other resources. Consumer utility is affected directly through changes in consumption and

indirectly through effects in option or existence value. Environmental effects are, therefore, a
mixture of private good, public good, and quasi-public good effects.
The income accounts can be extended using the flow-of-output approach to value
environmental goods and services produced. To avoid double-counting it is important to capture
only the value of the final environmental goods and services, Accounting for intermediate external
effects is needed only to compute sectoral income. If, however, an accurate measure of national
income alone is sought, then intermediate external effects can be ignored. In many cases
externalities are intermediate goods whose value is imbedded in the bundle of final goods and
services. Including the intermediate good in the income accounts is double-counting. A similar
argument holds for the flow-of-income approach. Economic rents generated by a non-market
externality are captured in payments to factors of production.
Accounting for non-market goods requires adjusting GDP for environmental goods and
services and transactions from the informal or underground economy. If changes to income consist
largely of accounting for environmental effects, then adjusted aggregate income might be termed
"green GDP". Adjusting GDP requires deriving a shadow price and physical measure for each final
non-market good. No information is necessary on intermediate goods.
There is considerable agreement that national accounts, although flawed, are useful
measures of economic performance and these accounts can be modified or extended to improve
the measure economic activity. Some economists have argued for developing alternative
accounting systems. Satellite accounts, a related but separate set of environmental accounts, may
be a preferred alternative to further diluting the quality of the market-based data with imputed
transactions. Critics of integrating the accounts argue that although flawed, the current income
accounts reasonably represent the market economy. Satellite environmental accounts would
include current market environmental expenditures as well as shadow accounts for non-market
environmental goods. A complete system of satellite environmental accounts would allow the
analyst to calculate the non-market adjustments and trace productivity effects across sectors.

The United Nations System for Integrated Environmental and Economic Accounting (SEEA)
is a set of satellite environmental accounts supplementing the current System of National Accounts
(SNA).6 The intent is to develop an environmental accounting framework consistent with the
concepts and principles underlying conventional income. Harrison (1989) presents criteria for
guaranteeing the satellite accounts are complementary to rather than a substitute for the current
accounts. A primary requirement is the parallel treatment of "natural capital" (natural resources)
and physical capital in the national accounts.
Although there have been other attempts to capture environmental effects in national
accounts (Nordhaus and Tobin, 1972), Nestor and Pasurka (1993) is the most ambitious. Nestor
and Pasurka disaggregate the U.S. input-output tables into environmental and non-environmental
components, Adopting the framework of Schafer and Stahmer (1989), Nestor and Pasurka divide
the environmental account into three categories. The "internal environmental protection sector"
captures intermediate goods and services produced and used within the environmental protection
industry. The "external environmental protection sector" captures the purchase of intermediate
inputs from outside the sector. Examples include waste disposal, sewage treatment, and
environmental construction activities. The "final demand sector" for environmental protection
includes fixed capital formation for environmental protection, direct pollution abatement activities
by governments and households and net exports of environmental protection goods.
The Nestor and Pasurka approach is consistent with the proposed system for environmental
and economic accounts (United Nations, 1993) and indicates the importance of environmental
protection activities in GDP. Through disaggregation, they estimate the 1982 total value-added for
environmental protection to be 0.3 percent of GDP. This is less than 20 percent of the $80.6
billion (1.7 percent of real GDP) estimate of real pollution and abatement control expenditures for
1991 (Rutledge and Leonard, 1993). While the Nestor and Pasurka approach provides more
information on the contribution of market expenditures on environmental protection, it does not
6 See United Nations (1992) and Bartelmus, Stahmer, and von Tongeren (1991).

change the overall measure of GDP because it does not include non-market activities.
NNP and Welfare
NNP is the premier indicator of current market-based economic activity. NNP has also been
promoted and, more importantly, interpreted as an indicator of social welfare. Samuelson (1961)
rejected all current income concepts as meaningful welfare measures and argued instead for a
"wealth-like magnitude" such as the present discounted value of future consumption. Weitzman
(1976) bridged the gap between Samuelson's argument for a wealth-based indicator of welfare and
current measures of income by demonstrating NNP captures both current consumption and the
present value of future consumption. A current income concept and a wealth-like magnitude, he
argues, "are merely different sides of the same coin." Weitzman's results are illustrated in figure 2.



\ a
! \\ slope-p
| \\^
dK*/dt B dK/dt
Figure 2. National Income

In figure 2, the production possibilities frontier B'B represents the economy's technical ability to
transform investment goods into consumption goods. The budget constraint C'C represents
society's willingness to trade-off future consumption for current consumption which depends on
the rate at which society discounts future consumption. The economy is located at point A on the
production possibilities frontier B'B. Optimal consumption and net investment are given by C* and
dk*/dt. Real NNP, is geometrically represented as OC'. The only point where measured income is
supported by production is at A. OC' is a strictly hypothetical consumption level at the present
time, because the largest permanent consumption level obtainable is OB'. Production and income
are equivalent only at A unless the transformation of investment goods into consumption goods
does not exhibit diminishing marginal returns. That is, if the production possibilities frontier is
linear, OB' is income, where income is interpreted as the maximum consumption possible. The
correct measure of "income" or NNP at the dynamic optimum is indicated by A. The level of
constant consumption OC' gives the same present value of welfare as the discounted maximum
welfare received along the optimal consumption path. Thus, Weitzman calls OC' the stationary
equivalent of future consumption.
Weitzman argues, income accounts, properly measured, provide a measure of the welfare
of society and give concrete economic form to the concept of sustainability. The current income
accounts do not adequately measure welfare or sustainable income because they fail to consider
non-market environmental goods and services and the degradation or depletion of non-renewable
If natural capital has a market, but is excluded from the accounts, then the accounts fail to
accurately measure true NNP. The only correction needed is to adjust the national accounts is to
deduct the value of the natural capital consumption (resource depletion). If natural capital does not
have a market, however, or the market price is distorted, then adjusting the accounts for natural
capital consumption is not as straightforward. Difficulties arise because there is a non-optimal level
of resource depletion and the shadow-price of resource depletion, an endogenous value, differs

from the socially optimal price. Similarly, if natural capital is substitutable for reproducible capital,
properly measured NNP also represents the maximum level of sustainable income for society.
However, if natural capital cannot substitute for reproducible capital, the link between aggregate
NNP and sustainable income is more problematic.
Application Framework
In this analysis, the environment and natural resources are treated as natural capital assets
generating a flow of services. Such a treatment allows for substitution between natural and
reproducible capital and is consistent with notions of weak sustainability. By adjusting national
income for changes in environmental quality and natural resource stocks, the national accounts
provide a more accurate economic interpretation of changes in the nation's assets. This approach
implies information about stocks on their own is not a sufficient statistic for well-being.
The model developed for this analysis draws significantly on Hartwick (1990) and Maler
(1991). Our work differs from previous work in that our model includes three production sectors
(agriculture, non-agriculture, and household production), three roles for land, and equations
describing the change in "effective" productivity of farmland, surface water quality, and the stock
of ground-water over time. Land, surface-water quality, and the stock of ground-water are treated
as natural capital.
Land in its natural state contributes directly to social welfare but is not used in any
production sector. Land is used in the agricultural sector and also contributes directly to social
welfare by providing rural landscape. We distinguish between the productivity of farmland and its
role in providing rural landscape because efforts to increase productivity are not likely to provide
added rural landscape. The third use of land is as an input in the production of non-agricultural
goods. This land makes no direct contribution to social welfare, but influences welfare by
contributing to the production of non-agricultural goods and services.
Water quality directly contributes to social welfare and is also an input into the production

of non-agricultural goods. Agricultural production, however, adversely affects water quality as a
result of soil erosion and chemical run-off. We adjust income for changes ground-water stocks
because in some regions there has been sustained withdrawal ground-water over time.
Each of our natural capital assets are regenerative or renewable but may be exhausted from
over-use if the rate of use exceeds the natural and managed regenerative rate of the asset. The
net rate of regeneration is the rate at which the stock of the asset changes over time.7 For land,
surface-water quality, and the stock of ground-water, the net rate of regeneration depends on the
intensity of use, the natural rate of regeneration, and the effectiveness of management to offset
the intensity of use of the asset. Land, for example, is usable until the productivity of soil for
producing agricultural goods approaches zero. The loss in soil productivity is offset by the soil's
natural capability to regenerate itself. The productivity of soil to produce agricultural goods is also
enhanced (managed) by applying labor, intermediate inputs (fertilizer), and capital to improve soil
Surface-water quality is characterized in a similar fashion. Natural regenerative processes
offset surface-water quality deterioration. The net rate of regeneration is a function of water
quality damage from agricultural production, the natural rate of regeneration, and the effectiveness
of management to offset degradation. The treatment of ground-water is potentially more
problematic because there may be resource degradation associated with the water stock's quantity
and quality. Treatment of ground-water in this analysis does not consider changes in ground-water
While agriculture's share of NNP includes a deduction for the consumption of physical
capital, a similar deduction is not made for other types of capital including farmland or natural
resource stocks such as water quality or water quantity, In addition, NNP is not adjusted for
externalities associated with agricultural production. For example, agriculture's contribution to NNP
is not reduced by offsite damages to water quality associated with soil erosion.
7 The net rate of regeneration defines the equation of motion for each asset.

In this analysis, farm income is adjusted to reflect changes in the effective level of farmland
in agriculture over time and the damages associated with soil erosion on surface-water quality. We
also correct farm income for the sector's contribution to the overall decline in the stock of ground-
water. Because data is limited, the value of scenic preservation of farmland and the value to
society of land in its natural state are not addressed. We also do not correct for the value of
leisure, We do not correct GDP or NNP for the value of leisure or the production of household
For the interested reader, the theoretical model is developed in detail Appendix A. The
work of Weitzman and originally Arrow and Kurz (1970) provide the necessary connection between
the current value Hamiltonian and NNP. In their work and our model, net welfare is expressed as
the linearized version of the current value Hamiltonian, NNP is reduced to the sum of the social
value of an economy's consumption and the social value of the changes in its capital stocks. By
capital stocks we mean manufactured or reproducible capital as well as natural capital stocks.
Net welfare measure in terms of final goods and services is:
NWM -	^ + Ja-wu
aglan, ^ 3/c, dz^ ar, 1 aw, 1J
dXi dn2 * dkP BY dLi
- -=-l* + *4 + *s + * + 4 + 4 + 4 + 4 + 4 + 41
+ ^y
+ E^+P3^1 +PaY+p6W,	">
The first line of equation (1) represents expenditures on final goods and services produced

by the agricultural sector as the sum of the value of the marginal contributions of each input used
in producing the agricultural good. That is, the expenditures on final agricultural goods is the sum
of the value of labor (n,), capital (k,), an environmental input (Z,), effective farmland (T,), and the
stock of ground-water (W, ) that is used to produce the agricultural good. The inputs used to
produce the agricultural good are valued in terms of the marginal contribution of the agricultural
good to the utility of society (dll/dq). The second line in equation (1) represents expenditures on
total goods and services produced in the non-agricultural sector. Expenditure on these goods is a
function of the value of labor (n2), capital (k2) water quality (Y), and land (L2) used to produce non-
agricultural goods, valued in terms of the marginal contribution of these goods to the utility of
society (3U/3x2). The third line in equation (1) represents expenditures on intermediate inputs used
to produce the agricultural and non-agricultural goods and services. Intermediate expenditures are
excluded from NNP to avoid double counting.
Deleterious environmental effects from agricultural production increase the cost of
production and require devoting additional productive resources to improve damaged water quality.
These additional intermediate inputs in the production of non-agricultural output are reflected in
lower current measured output in final consumer goods. The long-term effects on the production
of final consumption goods caused by environmental damages from agricultural production are not
included in conventionally measured NNP.
The fourth line in equation (1) represents the value (3U/3Y) of the stock of clean water (Y)
to consumers. This value is also not captured in conventionally measured NNP. The final line in
equation (1) reflects the addition of the value of net investment in both reproducible capital (kj) and
natural capital: effective farmland (T), water quality (Y), and ground-water quantity (W). Current
period production is valued in terms of its marginal contribution to the utility of society today. Net
investment in both reproducible and natural capital are valued by their marginal contributions to the
utility of society today and their marginal contribution to the utility of society in the future. 8
8 The conditions for optimality are presented in Appendix B.

The last two lines in equation (1) represent our adjustment to NNP. We suggest that the
conventional measure of NNP be corrected to reflect environmental impacts of agricultural
production on the stock of clean water as well as the future environmental impacts of agricultural
production on the stocks of effective farmland (T), water quality (Y), and ground-water quantity
Effective Farmland/Soil Productivity
The link between agricultural production practices, erosion, and farmland's ability to
produce output has been studied extensively (Crosson, 1986). In 1989, as part of the Second
Resources Conservation Act (RCA) Appraisal, the USDA estimated a 3 percent loss in productivity
over the next 100 years if farming/management practices remained as they were in 1982 (Table 3).
Similarly, Alt, Osborn, and Colacicco (1989) found that the net present value of both the crop yield
losses and the additional fertilizer and lime expenses associated with agricultural production totaled
$28 billion, Both studies employ a crop production model, Erosion Productivity Impact Calculator
(EPIC), which link production practices, erosion rates, and productivity, to provide estimates for
physical depreciation rates of land." Linking physical depreciation rates with crop prices can
provide an estimate of economic losses attributable to soil erosion over time. However, a
productivity loss of 3 percent over 100 years will not change NNP significantly.
While our theoretical model for adjusting NNP for the impact of erosion on loss of soil
productivity is straightforward, it is more difficult to assess a more comprehensive view of land
quality over time (National Academy of Sciences, 1993). For example, the RCA report also
concluded that less than 50 percent of all agricultural land was "adequately" protected.
Adequately protected soil was defined as soil within acceptable limits with respect to soil erosion
and other factors limiting sustained use. Soil scientist have developed "soil loss tolerance" or "T-
values" which vary by type of soil. A general rule of thumb is that erosion rates less than 5 tons
9 EPIC is a physical-process model that simulates interaction of the soil-climate-plant management
processes in agricultural production. EPIC was developed by USDA/ARS scientists and has been used
extensively in the RCA and elsewhere (e.g. Faeth, 1993).

per acre per year (T) do not result in damage to crop yields. Although results from the RCA seem
to indicate soil erosions effect on productivity are economically unimportant, the report also
indicates about 40 percent of cropland was eroding at rates greater than T.
Water Quality
More important than the productivity impacts of agricultural production on effective
farmland are the impacts of erosion on water quality and therefore on recreation, commercial
fishing, navigation, water storage, drinking supplies, industrial supplies, and irrigation. Ribaudo
(1989) estimated the average annual offsite erosion costs for the U.S. at $1.78 per ton ($ 1986).
Even if productivity effects are negligible, soil erosion associate with an acre of land causes, on
average, $9 in offsite damages.
Because data is limited on wind erosion our estimates focus on the offsite effects
associated with sheet and rill erosion, We link sheet and rill erosion and the adsorption of nutrients
to soil particles to estimate the effects of agricultural production on siltation, stream sedimentation,
and water pollution. Table 4 presents estimates of sheet and rill erosion for cropland and
pastureland for 1982, 1987, and 1992 from the National Resources Inventory (USDA).
It is possible for agents to mitigate the effects of pollution through defensive expenditures
of capital, labor, and other intermediate inputs. For example, increased siltation diminishes the
usefulness of a reservoir for producing electricity. The effects of siltation can be offset by
dredging. The attempt to offset the effects of soil erosion may result in additional costs
(expenditures) in electricity generation. In this case, part of the costs of agricultural production are
shifted to electricity generation. Similar arguments can be made for other industries. Economy-
wide NNP, therefore, should not be increased or decreased to reflect the transfer of costs from one
industry to another because aggregate NNP is correct. There is, however, a misallocation of
income among sectors. Conventionally measured farm income is higher if the costs of repairing the
reservoir are included as an intermediate expense of the affected industries rather than as an
intermediate expense of agricultural production.

Soil erosion also affects consumer utility. An increase in sedimentation in a reservoir can
reduce recreational activities. Because many recreational activities are unpriced and therefore are
not included in conventionally measured NNP, the diminished value of the resource does not
directly affect the income accounts although decreases in expenditures on complementary goods
will appear. In the inter-industry example there was a misallocation of income but economy-wide
NNP was accurate. In the second case, conventionally measured NNP fails to fully reflect the loss
of welfare due to the loss of the recreational resource. Therefore, the off-site damages to
consumers caused by agricultural production should be counted as an overall decline in NNP.
Similarly, the noncommercial loss of fish and waterfowl populations associated with
increased sedimentation are not fully represented in NNP. In addition to the impacts on recreation,
there may be an "existence" value component for the health of these riparian ecosystems. Such a
value is also excluded from the national accounts as currently measured.
We do measure the stock of water quality (Y) or the marginal utility of water quality
(3 U/3Y). Because no comprehensive measure exists, we use Ribaudo's (1989) estimate of the off-
site damages to water quality from soil erosion. The off-site damages in dollars per ton of soil
erosion ( converted to $1982) are listed in table 5. The estimates reflect the off-site effects of soil
erosion on freshwater and marine recreation, water storage, navigation, flooding, roadside ditches,
irrigation ditches, freshwater and marine commercial fishing, municipal water treatment, municipal
and industrial uses, and steam power cooling. We reorganize the damages into those affecting
industry (water storage, navigation, flooding, roadside ditches, irrigation ditches, freshwater and
marine commercial fishing, municipal water treatment, municipal and industrial uses, and steam
power cooling) and those directly affecting consumers (freshwater and marine recreation). The
industry and consumer damages per ton of soil erosion are highest in the Northeast.
The value of total damages presented in tables 6 and 7 are calculated by applying
Ribaudo's per ton estimates to the total level of sheet and rill erosion for cropland and pasture by

region.10 Total damages are $4.4 billion 1992, with $3.0 billion associated with industry affects,
Interestingly, while the dollar per ton effects are highest in the Northeast, the total industrial
damages are greatest in the Southeast ($390 million).
The effects of sheet and rill erosion on consumers totaled $1.1 billion in 1982, $1.2 billion
in 1987, and $1.3 billion in 1992. In addition to reducing farm income, these adjustments reflect a
decline in NNP and overall welfare. The effects on other industries were about twice as large as
the consumer impacts. Estimated industry effects are $2.4 billion in 1982, $2.7 billion in 1987,
and $2.7 billion in 1992. While these adjustment lower agricultural income, they do not reflect a
decline in NNP and overall welfare. They are treated as a transfer from one production sector of
the economy to another.
Ground-Water Quantity
Our final adjustment to the national and agricultural sector accounts is an adjustment for
the value of the change in the stock of ground-water over time. In the long-run, an equilibrium is
generally reached in terms of recharges (precipitation, imports from other regions) and discharges
(natural evapotranspiration, exports to other regions, consumptive use, and natural outflow) from
any ground-water system. However, in five water resource regions, the rate of discharge has
consistently been greater than the rate of recharge and has lead to a continued decline in the stock
of ground-water (U.S. Department of the Interior). Those five regions are: the Missouri Basin
(Montana, Wyoming, North Dakota, South Dakota, Nebraska, and parts of Colorado and Kansas),
the Arkansas-White-Red (southern Kansas, Oklahoma, north Texas, and western Arkansas), the
Texas-Gulf (most of Texas), the Lower Colorado (Arizona), and California. While it is difficult to
assess agriculture's contribution to the overall change in the stock of ground-water in those
regions, the sector accounted for 79 percent to 88 percent of total ground-water withdrawals in
the U.S. (Table 8).
10 Ribaudo's 1982 estimates are inflated to 1987 and 1992 by the change in the gross domestic
product implicit price deflator.

Because the most recent estimate of the change in the stock of ground-water for the U.S.
is for 1980 (U.S. Department of the Interior) and because the data are not specified by sector of
use, we adopt the following four step procedure. First, we employ the 1980 water resource
budgets and use agriculture's share of total ground-water withdrawals (Solley, et.al.) to allocate
the change in the stock of ground-water for each of the five water resource regions exhibiting
declines in the stock of ground-water in 1980. For example, in 1980 agriculture accounted for
about 86 percent of ground-water withdrawals in the California water resource region. Therefore,
we assume that agriculture accounted for 86 percent (1.2 billion gallons per day (BGD)) of the total
decline in the stock of ground-water in the California water resource region (1.4 BGD) for 1980.
Second, because water use data is collected every five years, we use the change in total
ground-water withdrawals to update the total change in the stock of ground-water for each of the
water resource regions. For example, from 1980 to 1985, the total (both agriculture and non-
agricultural) withdrawals of ground-water for the California region fell by about 30 percent from
21.0 to 14.8 BGD. Therefore, we assume that the rate of ground-water depletion in the region fell
by about 30 percent from 1.4 BGD to 1.0 BGD.
Third, we again use agriculture's share of total ground-water withdrawals to allocate the
change in the stock of ground-water. Continuing with our California example, in addition to the
decline in overall ground-water withdrawals, the share of withdrawals attributed to agriculture fell
from 86 percent to about 70 percent. Therefore, the rate at which agriculture contributed to the
decline in the overall stock of ground-water in the California water resource region fell from 1.2
BGD in 1980 to 0.7 BGD in 1985 (Table 9).
This process leads to some interesting comparisons over time. The change in overall
ground-water withdrawals coupled with changes in agricultural uses indicates that by 1990,
agriculture's contribution to overall decline in the stock of ground-water declined since 1980 and
remained stable since 1985. Regionally, however, there are some differences. For the Lower
Colorado and California water resource regions both total ground-water withdrawals and the share

of ground-water withdrawals attributed to agriculture has fallen significantly. In both regions, the
share of ground-water withdrawals attributed to agriculture has fallen from close to 90 percent in
1980 to about 75 percent by 1990. Much of this decline in ground-water withdrawals can be
attributed to the decline in irrigated acres in the Pacific coast over that period. 11 However, for the
Missouri Basin, Arkansas-Red-White, and Texas-Gulf, agriculture's share of total withdrawals of
ground-water has remained fairly constant since 1980.
Finally, we need to associate values with the estimated changes in the rate of ground-water
depletion, We estimate the value of ground-water based on the ratio of energy expenses for on-
farm pumping of irrigation water to the estimated amount of water applied to farms from wells.
The data on energy expenses and water application is from Farm and Ranch Irrigation Surveys
(U.S. Department of Commerce, Census of Agriculture). 12 The values range by water resource
region and for 1992 range from $0.10 to $0.12 per 1,000 gallons in California, the Lower
Colorado, and the Texas Gulf to $0.07 to $0.09 in the Missouri Basin and the Arkansas-Red-White
region. While there is considerable uncertainty regarding the appropriate value of water, the
estimates used in this analysis are similar to those used by Grambsch and Michaels (1994).
Grambsch and Michaels estimate, based on water price data for the 120 largest metropolitan areas
and government capital and operating expenses, was $0.09 per 1,000 gallons. The adjustment to
farm income presented in table 9 combines the value of ground-water with the rate of ground-
water depletion associated with agriculture. Total damages range from $212 million in 1987 to
$291 million 1992.
Impacts on Income
Agriculture affects both production in other sectors of the economy and consumer utility
through its use of environmental and natural resource assets. Production in other sectors of the
11	Irrigated acres in the Pacific coast fell from 12 million to 10.5 million from 1978 to 1992 (USDA,
12	The data in the Census of Agriculture are for 1979, 1984, and 1988. The GDP implicit price
deflator is used to match census years with the dates used in this analysis.

economy are affected because changes in environmental assets affect the productivity of other
inputs and therefore the cost of producing non-agricultural goods and services. Consumer utility is
affected directly through changes in consumption and indirectly through changes in option or
existence value.
The approach here is to extend the existing flow-of-output accounts to value environmental
goods and services. Double-counting is avoided by recognizing that the inter-industry externalities
caused by agricultural production are captured in the existing accounting framework as
intermediate expenses in non-agricultural production. Accounting for intermediate external effects
is needed only to compute sectoral income. If, however, an accurate measure of national income
alone is sought, then intermediate external effects can be ignored. The production externality is an
intermediate good whose value is imbedded in the bundle of final goods and services. Agriculture's
contribution to the decline in surface-water quality cause a transfer of accounting income from the
agricultural sector to the non-agricultural sector of the economy in 1982 of $2.4 billion, in 1987 of
$2.7 billion, and in 1992 of $2.7 billion. These adjustments reduce agricultural income and
increase income in other sectors of the economy but do not reduce economy-wide NNP. Including
intermediate goods in the income accounts is double-counting. Similarly, economic rents generated
by a non-market externality are captured in payments to factors of production in the flow-of-
income approach. This is not the case, however, when consumer utility is affected directly through
changes in consumption and indirectly through changes in option or existence value.
Our estimates suggest only minor adjustments to NNP are made necessary by the effects of
agricultural production on the environment and natural resource base. This result follows partly
from agriculture's small share (less than 2 percent) of GDP. Even large changes in net farm income
have only modest effects on NNP. Adjustments to total farm income and economy-wide NNP for
1982, 1987, and 1992 are displayed in table 10. In each year, total farm income is reduced by
about $4 billion when adjustments are made for agriculture's contribution to the decline in surface-
water quality and stock of ground-water. Overall, agriculture's contribution to economy-wide NNP

falls by $1.3 billion in 1982, $1.4 in 1987, and $1.6 in 1992 when adjustments are made for
agriculture's contribution to the decline in surface water quality and stock of ground-water. About
85 percent of the adjustment is caused by agriculture's contribution to the decline in surface-water
The relative effects on net farm product are significantly greater. Adjustments to net farm
product range from 6 to 8 percent. The relative share of environmental adjustments to
conventional net farm product, however, decreased from 1987 to 1992, Measured agricultural
environmental costs per dollar of farm income are declining. This suggests estimated
environmental costs flowing from agriculture are not growing as fast as farm income. One possible
explanation is policies and programs for controlling soil erosion were effective during this period. In
particular, highly erodible acreage enrolled in the Conservation Reserve Program increased from
13.7 to 35.4 million acres from 1987 to 1992. Removing nearly 22 million acres of highly erodible
land from production contributed to a nearly 21 percent decrease in estimated soil erosion on
cropland during this period even though planted acreage for grains increased by 6 percent.
Conservation compliance requirements promulgated under the 1985 farm legislation have provided
additional incentives for reducing erosion.
The estimates are consistent with Smith's (1992) work on environmental costing. Smith
aggregates the effects of off-site soil erosion, wetland conversion, and ground-water contamination
and estimates environmental costs relative to the value of crops produced in 1984. His estimates
range from 0.08 to 7.5 percent in the Mountain region to 3.5 to 40 percent in the Northeast. Corn
Belt estimates range from 6 to 7 percent. 13
Our estimated adjustments represent average costs of environmental damages and resource
use. Marginal costs are likely to be higher. It is possible that the distortionary effect of commodity
13 Smith suggests the work on Viscusi and Magat (1991) on energy implies that the environmental
costs of agriculture are comparable to those estimated from several energy sources. Both the Smith
and Viscusi and Magat work differ from Nestor and Pasurka's estimates of total value-added for
environmental protection of 0.3 percent.

programs is alone sufficient to lead to marginal decreases in social welfare. Accounting for natural
resource deterioration and environmental injury, in such a case, would lead to further reductions in
social welfare. In addition, our national estimates may be masking significant regional or local
problems. Estimated costs of erosion in terms of lost productivity, for example, is not a significant
national problem, but may be a significant regional or state problem. Faeth (1993) shows negative
net economic value per acre after accounting for soil depreciation and off-site costs for
Pennsylvania's best corn-soybean rotation over 5 years. The work demonstrates there may be
significant regional variation in resource depreciation and off-site costs of agricultural production.
Growing interest in the environment has raised questions about the adequacy of current
measures of national income particularly when these measures are used as social welfare
indicators. The intent of this paper is to more accurately measure agriculture's contribution to
national income. Improving the measure of current economic activity requires incorporating non-
market final goods and bads into the existing accounts. We focus attention on treating natural
capital assets used or affected by agricultural production parallel to how reproducible capital is
treated in the national accounts. Net national income and agricultural income are adjusted to
reflect the value of changes in the stock of effective farmland, surface-water quality, and ground-
We first develop a theoretically consistent framework for incorporating natural capital and
environmental goods into the existing income accounts. Next, we apply the framework and adjust
agricultural income and national income to reflect the value of the depletion of agricultural natural
capital (land and water) and the non-market effects of agricultural production on output in other
sectors of the economy and consumers. Specifically, the effects of soil erosion on agricultural
productivity and income, the economic effects of decreased surface-water quality, and the
depletion of ground-water stocks are presented as examples of the potential scope of accounting

adjustments needed in the agricultural sector. Our estimates suggests only minor adjustments to
NNP are made necessary by the effects of agricultural production on the environment and the
natural capital base. This result follows from agriculture's small share of GDP and because the
environmental effects considered in this paper are largely captured in the existing accounts.
Adjustments to net farm income are relatively greater and fall in the range of 6 to 8 percent.
Our estimates of "green" adjustments to net farm income are consistent with a view of
U.S. agriculture where environmental problems exist and the resource base is depreciating, but the
extent of the effects is in the range that can adequately be addressed by thoughtful policy. Our
estimates suggest that agriculture's contribution to social welfare far exceeds the environmental
damages and deterioration of the stock of natural capital resulting from the production of food.
Estimates of adjusted or "green" income presented here are incomplete. Because the
objective of our analysis is to illustrate some of the adjustments necessary to improve NNP and
NFP as measures of social welfare, we restrict our scope to consider a few key agricultural effects.
Other adjustments, including additional environmental damages and valuing environmental services,
are necessary before a credible measure of welfare or sustainability can emerge. We have not, for
example, estimated the cost of farm chemical volatilization on air quality, or valued the benefits of
landscape preservation or increasing wildlife habitat. In addition, on the cost side, we have not
examined how soil quality characteristics, other than erodibility, affect productivity or wildlife
habitat. Valuation of farm program benefits warrant further exploration. Program payments are
currently treated as income transfers, included in net farm income but excluded from gross farm
income. An alternative approach views the Government purchasing environmental benefits like
scenic value or wildlife habitat.

Ahmad, Yusuf, Salah El Serafy, and Ernst Lutz. eds. (1989V Environmental Accounting for
Sustainable Development. Washington, DC: World Bank.
Alt, Klaus, Osborn, C. Tim, and Dan Colacicco (1989). "Soil Erosion: What Effect on Agricultural
Productivity." Agricultural Information Bulletin No. 556, U.S. Department of Agriculture, Economic
Research Service.
Arrow, Kenneth and Mordecai Kurz (1970). Public Investment, The Rate of Return and Optimal
Fiscal Policy. Baltimore, MD, Johns Hopkins University Press.
Bartelmus, P., Stahmer, C. and Van Tongeren, J. (1991). "Integrating Environmental and Economic
Accounting: Framework for a SNA Satellite System. " The Review of Income and Wealth. Vol. 37.
pp. 111-47.
Crosson, Pierre (1986). "Soil Erosion and Policy Issues." in Agriculture and the Environment (T.
Phipps, P. Crosson, and K. Price, eds.) Resources for the Future. Washington, DC.
Dasgupta, Partha and Karl-Goran Maler (1990). "The Environment and Emerging Development
Issues." Proceedings of the World Bank Annual Conference on Development Economics, 1990.
The World Bank. pp. 101-31.
Eisner, Robert (1988). "Extended Accounts for National Income and Product. " Journal of
Economic Literature. December, pp. 1611-84,
El Serafy, Salah (1991). "The Environment as Capital." in Ecological Economics: The Science and
Management of Sustainabilitv. (R. Costanza, cd.) Columbia University Press, New York.
Faeth, Paul (1992). "National Economic Implications of a Transition to Sustainable Agriculture in
the United States: Implications for Policy Reform. " Project Summary, World Resources Institute,
Washington D. C., November.
Faeth, Paul (1993). "Evaluating agricultural Policy and the Sustainability of Production Systems:
An Economic Framework. " Journal of Soil and Water Conservation. March/April pp. 94-99.
Faeth, Paul, Robert Repetto, Kim Kroll, Qi Dai, and Glenn Helmers (1991). "Paying the Farm Bill:
U.S. Agricultural Policy and the Transition to Sustainable Agriculture." World Resources Institute,
Washington D.C.
Grambsch, Anne and R. Gregory Michaels (1994). "The United Nations Integrated Environmental
and Economic Accounting System: Is it Right for the U.S." Paper presented to the 1994 AERE
Workshop Integrating the Environment and the Economy: Sustainable Development and
Economic/Ecological Modelling. June.
Hartwick, John (1990). "Natural Resources, National Accounting and Economic Depreciation."
Journal of Public Economics. Vol. 43, No.3. pp. 291-304.
Hicks, John (1939). Value and Capital. Oxford University press.
Hung, N.M. (1993). "Natural Resource, National Accounting, and Economic Depreciation: Stock
Effects." Journal of Public Economics. Vol. 51, No.3. pp. 379-89.

Maler, Karl-Goran (1991). "National Accounts and Environmental Resources." Environmental and
Resource Economics. Vol. 1, No, 1. pp. 1-15.
National Academy of Sciences. National Research Council (1993). Soil and Water Quality: An
Agenda for Agriculture. National Academy Press. Washington. DC
Nestor, Deborah Vaughn and Carl A. Pasurka (1994). "Environment-Economic Accounting and
Indicators of the Importance of Environmental Protection Activities." Paper presented at the 1993
Annual Meeting of the Southern Economic Association. March.
Nordhaus, William and James Tobin (1972). "Economic Growth." National Bureau of Economic
Prince, Raymond and Patrice Gordon (1994), "Greening the National Accounts." CBO Papers.
Congressional Budget Office. March.
Repetto, Robert (1992). "Accounting for Environmental Assets." Scientific American, pp. 94-100,
Repetto, Robert (1992). "Earth in the Balance Sheet: Incorporating Natural Resources into
National Income Accounts." Environment pp. 43-45, September.
Ribaudo, Marc (1986). "Reducing Soil Erosion: Offsite Benefits. " Agricultural Economics Report
No. 561, U.S. Department of Agriculture, Economic Research Service.
Ribaudo, Marc (1 989). "Water Quality Benefits from the Conservation Reserve Program."
Agricultural Economics Report No. 606, U.S. Department of Agriculture, Economic Research
Ribaudo, Marc and Daniel Hellerstein (1992). "Estimating Water Quality Benefits: Theoretical and
Methodological Issues." Technical Bulletin No. 1808, U.S. Department of Agriculture, Economic
Research Service.
Rutledge, Gary, L. and Mary L. Leonard (1993). "Pollution Abatement and Control Expenditure,
1987-91." Survey of Current Business, U.S. Department of Commerce, Bureau of Economic
Analysis, Vol. 73, No. 5, pp. 55-62, May.
Smith, V. Kerry (1992). "Environmental Costing for Agriculture: Will It Be Standard Fare in the
Farm Bill of 2000?" American Journal of Agricultural Economics. Vol. 74, No. 5, pp. 1076-1088.
Solley. W. B., Chase, E. B., and W.B. Mann, IV (1983). "Estimated Use of Water in the United
States in 1980." U.S. Geological Survey Circular 1001.
Solley. W. B., Pierce, R. R., and H.A. Perlman (1993). "Estimated Use of Water in the United States
in 1990." U.S. Department of Interior Circular 1081.
Solow, Robert. (1986). "On the Intertemporal Allocation of Resources." Scandinavian Journal of
Economics. Vol.88, No.2, pp. 141-56.
Solow, Robert. (1992). "An Almost Practical Step Towards Sustainability." An Invited Lecture on
the Fortieth Anniversary of Resources for the Future. Mimeo.

United Nations (1993). "Integrated Environmental and Economic Accounting." Interim Version.
Studies in Methods. Series F. No. 61, Handbook on National Accounting. Department for
Economic and Social Information Policy Analysis. United Nations: New York.
U.S. Department of Agriculture (1989). "The Second RCA Appraisal: Soil, Water, and Related
Resources on Nonfederal Land in the United States, Analysis of Conditions and Trends." June.
U.S. Department of Agriculture (1994). Soil Conservation Service. "Summary Report 1992
National Resources Inventory." July.
U.S. Department of Agriculture (1994). Economic Research Service. "Agricultural Resource and
Environmental Indicators." Agricultural Handbook No, 705. December.
U.S. Department of Commerce (1982). Bureau of the Census. 1978 Census of Agriculture.
"1979 Farm and Ranch Irrigation Survey." AC78-SR-8. November.
U.S. Department of Commerce (1986). Bureau of the Census. "1984 Farm and Ranch Irrigation
Survey." AG84-SR-1. June.
U.S. Department of Commerce (1990). Bureau of the Census. 1987 Census of Agriculture.
"Farm and Ranch Irrigation Survey." AC87-RS-1. May.
U.S. Department of Commerce (1993). Bureau of Economic Analysis. "Survey of Current
Business." August.
U.S. Department of Commerce (1994). Bureau of Economic Analysis. "Accounting for Mineral
Resources: Issues and BEA's Initial Estimates." April.
U.S. Department of the Interior (1984). U.S. Geological Survey. "National Water Summary 1983-
Hydrological Events and Issues." Water Supply Paper 2250, p.26.
Viscusi, W.K. and W.A. Magat (1991). "Interim draft report on efficient energy pricing." Report to
U.S. Environmental Protection Agency.
Weitzman, Martin (1976). "On the Significance of Net National Product in a Dynamic Economy."
Quarterly Journal of Economics. Vol. 90, No. 1, pp. 156-162.

Table 1. Overview of the Existing NIPA Accounts, 1992, $Billion
Flow of Income
Flow of Output
Compensation of Employees
Personal Consumption Expenditures
Proprietors Income
Gross Domestic Investment
Corporate Profits
Government Purchases
Net Interest
Net Exports
Rental Income

Natkinaf tm&nt*
Gro«* Domestic Product

Consumption of Fixed Capital
Business Transfer Payments
Rest of World Net Factor Income
Individual Tax and Nontax Liability
Statistical Discrepancy
Subsidies Less Government Surplus
Business Transfer Payments
Consumption of Fixed Capital
Individual Tax and Nontax Liability
Gross National Income
Subsidies Less Government Surplus
Statistical Discrepancy

Gross National Income

Rest of World Net Factor Income

dro&e Domestic Product
Nattonat Income

Source: Survey of Current Business, 1993.
Table 2. Summary of Farm Income, 1982, 1987, 1992 $Billion
Flow of Income Components
Compensation of Employees
Proprietors Income
Corporate Profits
Net Interest Income
Nfet Farm income'
54 3

Indirect Tax and Nontax Liability
Subsidies Less Current Government
Consumption of Fixed Capital
Gross Farm Product

Source: Suivey of Current Business, various years.

Table 3. Productivity Impacts on Cropland Associated with Soil Erosion, 1982
Sheet and Rill

Lake States
Corn Belt
Northern Plains
Southern Plains
* = less than 0.01%
Source: U.S. Department of Agriculture, The Second RCA Appraisal.
Table 4. Gross Annual Sheet and Rill Erosion (Cropland and and Pasture/Range)


Million Tons
Million Tons
Million Tons

Lake States
Corn Belt
Northern Plains
Southern Plains

Source: U.S. Department ot Agriculture. Soil Conservation Service.

Table 5. Off-Site Damages Associated with Soil Erosion, 1982

I ndustry

T otal








Lake States


Corn Belt




Northern Plains


Southern Plains








Source: Ribaudo, 1989.

Table 6, Estimated
Annual Soil
Erosion Damages

1 992

Million $
Million $
Lake States
Corn Belt
Northern Plains
Southern Plains
* 2,424,0

Table 7. Estimated Annual Consumer Soil Erosion Damages

Million $
Million $
Lake States
Corn Belt
Northern Plains
1 0
Southern Plains
1,254,7 !
Table 8. Ground-Water Withdrawals by Water Resource Region (Billion Gallons per Day)
1 980
1 990

T otal
\gricultu re
T otal
Missouri Basin
Lower Colorado

Source: Solley, et. al.

Table 9. The Effects of Agricultural Production on Ground-water Storage (Billion Gallons per Day,
and Million $ per Year).

Million $
Million $
Million $
Missouri Basin
1 08
Lower Colorado
Source: 1980 data from U.S. Department of the Interior, U.S. Geological Survey.

Table 10. Summary of Adjusted National Income and Product Accounts 1982, 1987 and 1992

1 982
Income Components

$ Billions

Traditional Farm Income
Water Quality

Industry Transfer
Consumer Effects
Water Quantity
Green farm facortte
Traditional Non-Farm Income
Water Quality

Industry Transfer
+ 2.4
+ 2.7
+ 2.7
Consumer Effects

Water Quantity

<3**eeri Non-Farm income
Traditional National Income
Water Quality

Industry Transfer
Consumer Effects
Water Quantity
<3r0en National tncotfia

Appendix A: Theoretical Model
Output of the agricultural sector (q) is given by the production function:
q - tfn,, 2,. r„ aW,)	(A.1)
n,: agricultural labor,
k,: agricultural capital,
Z,: environmental input,
T,: "effective" stock of land used in agricultural production,
e: ground-water extraction rate, and
W,: stock of ground-water.
Output of the non-agricultural sector (x) is given by the production function:
* = /fe. V, LJ	(A 2)
x: non-agricultural good,
n2: non-agricultural labor,
k2: non-agricultural capital,
Y: water quality effect on non-agricultural production (3x/3Y > O), and
L2: land used in non-agricultural production.
Household or non-market production (h) is given by:
h - h(/j6, Xf, kj	(A.3)
na: household labor,
xe: intermediate inputs used in household production,
ke: household capital.
The household production function includes non-marketed activities beyond those related to the
The equation of motion for the effective productivity of farmland is
t, ¦ Y<£ ^ - dL,	(A 4)
where land can be managed (improved) by adding labor, intermediate inputs (fertilizer), and capital

according to a management function
, . ,<£ A, *)	(A.5)
K: is a rate of appreciation,
n3: labor used in managing land,
x3: intermediate inputs used in managing land,
k3: capital used in managing land, and
d: soil erosion rate.
The management function yi ) is assumed linearly homogeneous in its arguments (n3/L,, x3/L1f
k3/L,) and in n3, X3, and k3.
The equation of motion for water quality is
Y =[a - D(Z) + ti(n4, kj)Y	(A.6)
where the impact of agricultural production on water quality is represented by:
D - HZ)	(A.7)
Water quality can be managed (improved) by adding labor, intermediate inputs, and capital:
11 = ti(a?4, Xa, /Q)	(A.8)
n4: labor used in managing water quality,
x4: intermediate inputs used in water quality,
k4: capital used in managing water quality, and
a: natural repair of water quality.
The damage function D^) and the repair function /7O are also assumed linearly homogeneous in
their respective arguments.
Our equation of motion for the stock of ground-water is
- [~ " *ns, Xs, kJ]W,	(A.9)
where the extraction of ground-water for use in agriculture is represented by:

e = e(ns, kj
(A. 10)
n6: labor used in extracting ground-water,
x6: intermediate inputs used in extracting ground-water, and
ksi capital used in extracting ground-water, and
i/j: the rate ground-water is replenished.14
As discussed in the text, each natural capital asset is regenerative or renewable but could
be exhausted from over-use. The net rate of regeneration, as captured by the equations of motion
is a function of the intensity of use, the effectiveness of management to offset the intensity of use
of an asset, the level of the stock of the resource itself, and the natural rate of regeneration.
Social welfare (U) is defined as a function of final goods and services (q,x2), household
production (h), an index of water quality (Y), land in its natural state "-o)> land used in agriculture
(L,), and leisure (n7). The social planner's goal is to maximize:
q : agricultural output (final good),
x2: non-agricultural (final) goods and services,
h : household production,
Y : index of water quality, (3U/3Y > 0) 16
Lg: unused land (natural state),
L,: land used in agriculture,
n7: leisure, and
r: social discount rate
subject to the equations of motion for the stock of effective land, surface-water quality, and the
14 This is a simplified representation. The ground-water replenishment rate tfi is a function of
precipitation, inflows and outflows, and the return flow of water extracted for agricultural uses.
16 Corner solutions are problematic. For perfect water quality human efforts at improvement have
no impact. With no water quality agriculture creates no added damages. We assume these situations
are unique so that our results are not affected.
The Model

stock of ground-water:
% =Y(^.^. ^ - ^	(A-12)
V-=[a-0(2.) + T,K.>i, U]Y	(A. 13)
W, = |* - fl(/7s. A* kJW	(A. 14)
In addition to natural capital, there are equations of motion for each of our six types of
reproducible capital:
kt = 1/ - 8/f/ fori' 1	6	(A.15)
where: 'i represents gross investment in the ith type of reproducible capital and represents the
depreciation rate for each type of reproducible capital.
A materials balance equation and constraints for labor and land complete the model:
*(/%, Aj. Y, ij) - *2 + *3 + *4 + Xs * % + A + l2 + *3 + U + k + U	(A. 16)
N=Y,n,	(A.17)
/. 1
L= Y: L,	(A.18)
/- 0
The materials balance equation accounts for the output of the non-agricultural sector, x, in the
economy. For example, some non-agricultural output goes to final non-agricultural consumption
goods and services x2. Non-agricultural output is also used as investment goods inputs that go
into managing the stock of effective farmland x3, water quality x4, and the stock of ground-water
x6; and as inputs in the household production function x8.

The current value Hamiltonian in flow of output terms is:
H = U [q(nv kf, 2;, Tv eWJ.
¦*(n2> k2, Y,	- x4 - x$ - Xg - /, - l2 - /j - /4 - /5 - /#i
^j)i K ^o» i-ii /^J
* P3lY(Tf' Li' 7^ " dL*]
+ P4[a - 0(3) + r\(n4, x^, kj\Y
+ Psfo - e(n5> /^)]W,
+ £ l*!'/ ~ 6A1
- »l£ n, - N[
i' 1
-Q[£ *•/-*¦!	(A-19)
/» 0
where pif //„ wi( and Q, are co-state variables.
The Measurement of Net Welfare
The Hamiltonian along the optimal trajectory is the national welfare measure in utility terms
(Maler, 1991; Hung, 1993). The linear approximation of the Hamiltonian along the optimal path is
the exact correspondence to the net national welfare measure. It measures the current utility of
consumption (of goods and services and environmental services) and the present value of the
future utility stream from current stock changes. This follows because stock prices measure the
present value of the future contribution to welfare from a marginal increase in the stocks.

Net welfare is measured as
+ i^[J* ^ ~ -%v +
dXjTdr^ * dkt* dY dL
J^l*3 + *4 + *s + * + 'i + k + + U + 7s + 'J
~	+ «. *Lq + ^ + + Wy+
a/»la^ % dxf9 d^ dLo^ az^H ay an,*
* E eft - »AJ
i' 1
P,M^. ^)i, - <*ij
+ P
final consumption of the non-agricultural good.
The fourth line of equation (A.20) captures implied expenditures on the household product,
natural-state land, aesthetic farm landscape, water quality, and leisure. The fourth line that
contains most of the extensions to the traditional GDP accounts. However, some of these
expenditures may already be included in the GDP accounts. For example, government expenditures
to improve water quality and explicit expenditures by environmental groups to save natural-state
land such as old growth forests already show up in the accounts. The fifth line of equation (A.20)
captures net investment in each of the six types of physical capital, while the last three lines report
net investment in the three types of natural capital. The gross investment components of these
last three lines are also extensions of the GDP accounts.
The first three lines of equation (A.20) and the gross investment components of line 5 sum
to the traditional measure of GDP. Adding line 4 and the gross investment components of lines 6,
7, and 8 gives the extended GDP measure. Lines 1, 2, 3, and 5 sum to the traditional NNP
measure. The entire expression given by equation (A.20) represents the extended NNP measure.
Two final observations stemming from equation (A.20) are worth noting. First, concern for
sustainability and properly valuing natural resource depletion leads to extending the accounts by
including lines 6, 7, and 8 of equation (A.20). Second, concern with including "non-market" goods
(e.g. housework, land in its natural state, rural landscape, water quality, and leisure) in the
accounts leads to expanding the accounts by including line 4.

Appendix B: The Optimality Conditions
The optimality conditions are obtained by partially differentiating the Hamiltonian (equation
A. 19) with respect to the control and state variables. The control variables are the seven uses of
labor, the uses of the manufactured output x, gross investment in the six type of reproducible
capital, the three uses of land, and the level of water pollution, Z,. For labor, the optimality
conditions are:
rn = w Ja. - „ . o	(B.i)
dq d/i|
dH dU dx	n	/D
	 =		 o) = 0	(B.2)
dn2 dXj, dn^
—	= p3-^- - u = 0	(B.3)
—	= P4— Y - o = 0	(B.4)
dn4 dnA
*£L =	W -	- u = 0	(B.5)
3% dqdedng 1 sdns 1
M =	(B.6)
dnt dh drtg
dnj drtj
Equations (B.1), (B.2), and (B.6) indicate the value of the marginal product of labor is
equalized across the three production sectors. This value at, the shadow wage rate, is also the
marginal value of leisure, equation (B.7), and the marginal value of labor in enhancing land,
equation (B.3), repairing water quality, equation (B.4), and depleting ground-water stocks, equation
The manufactured good x can be directly consumed (x2), used as intermediate input or for

investment. The optimality conditions for x as an intermediate input for improving land, water
quality, and depleting ground-water stocks are:
HL = jwiajtew _ du _ p = o	(B 10)
dXg dqdedX;	8% dXs
These conditions show that the value of the marginal product of the manufactured good in
each of its intermediate uses must equal 3U/3x2, the opportunity cost of direct consumption.
The optimality conditions for x as investment in reproducible capital are:
¦^ = ~ + ^/ = °	(/=1	6).	(B.11)
Off	0*2
As with intermediate goods, the marginal value of investment in each type of capital (#) must equal
the marginal value of the consumption good x2 (3U/3x2).
Partially differentiating with respect to each land type determines the distribution of land
across sectors:
where A = rVL,, B = x3/Llf and C = k3/L,. Because y is assumed homogeneous of degree 1 in A,
B, and C, equation (B. 13) reduces to:

The remaining use of land, L2, is chosen so that
M. = dU dx - Q = 0	(B.15)
di^ 3% dL?
Recall the unique character of each type of land. Land in its natural state, Lo» has only a
direct welfare effect and no productivity effect. Land used in non-agricultural production, L2,
affects welfare indirectly as an input in production. Farmland, Lv however, has both a productivity
effect in agriculture and a direct welfare effect in utility in terms of providing rural landscape.
The shadow value Q gives the price of land in its natural state. This price exceeds the
direct marginal contribution of farmland to welfare because some farmland erodes, while pristine
land and non-agricultural land are assumed not to erode. This price Q also equals the value of the
marginal product of land in the non-agricultural sector.
An additional control variable to consider is Z,, the environmental input to agricultural
production. The optimality condition for this variable is:
Here the choice of Z, can be interpreted as the optimal use of an environmental input,
water quality. Equation (B.16) indicates that the value of the marginal product of water pollution in
agricultural production is equal to the marginal change in welfare from increasing water quality.
The optimality conditions associated with the state variables describe the choice of stock
levels for the six types of physical capital and the three types of natural capital. For the physical
capital variables, the optimality conditions are

ft ¦" *	(B-17)

P3-|jj7 = (' + *3)^3 " 1*3	(B 19)
p4|v=h8>4-1*	(b-2°)
dUd$de_w _ dew m (r + ^ _ £	(B 21)
dqdedks 1 K58A^ 1 v * 5
= (r + « \,i _ ,1	(B.22)
a/? 3A, v 6/ 6 8
These conditions demonstrate that the value of the marginal product of reproducible capital
in each activity (including land enhancement, water quality repair, and diminishing ground-water
stocks) is equal to a rental price of capital. Because the investment good is treated as the
undifferentiated intermediate good, =^2 =^3 =Ub =Va- However, the rental prices may differ
because of different economic depreciation rates.
The final optimality conditions involve our natural capital stocks: effective farmland, water
quality, and ground-water stocks. These conditions are:
*•	 * *>' 
Unlike the conditions for physical capital stocks, equation (B.23) does not have a depreciation rate.
Soil erosion, which is similar to a physical depreciation rate, is already captured in equation (B.23).
The optimality condition for the stock of water quality is also a rental rate similar to those for
physical capital. However, given the form of equation (B.24), this rental rate is adjusted for water
quality appreciation rather than depreciation.
Finally, it is interesting to compare the shadow values for reproducible capital to natural
capital. For example, a unit of reproducible capital that is used to in the agricultural sector has a
In other words, the value of a unit of reproducible capital in time t is equal to the discounted value
of the future services it will provide in terms of agricultural output. An increase in the discount rate
(r) or the rate of depreciation Wi) will reduce the value of capital.
Our shadow value of natural capital has similar characteristics. For example, a unit of
water quality has a shadow value:
dqdk^ + ti,
* ' (r+ «,) + (r * 5,)
dU + dUdx
dY 3*2 3V	p4
P4 =	+ [a - 0(3) + H(n4, k4, *<)]

For natural capital, an increase in the natural rate of regeneration or an increase in human
attempt to improve the quality of water reduces the discount rate and increases the shadow value
associated with water quality. In addition, unlike reproducible capital, the shadow value captures
the discounted value of water quality to both consumers (3 U/3Y) and producers of the
manufactured good [(3 U/3x2)(3x/3Y).

Some Issues Related to Ecological and Economic Modeling of Ecosystem
Nancy Bockstael and Jackie Geoghegan
Discussion paper for 1994 AERE Workshop Participants
In this discussion paper are outlined some of the issues and problems we are
encountering in a multidisciplinary (ecological and economics) research endeavor
sponsored by EPA. The "vignettes" that follow correspond to sections of our
presentation and are supplied here in hopes of stimulating discussion and some good ideas.
After a brief description of the project, we address each of the following topics in turn:
•	the general structure of an ecological-economic model
•	the treatment of spatial data in economic analysis
•	modeling landscape reconfiguration.
Overview of Project
This work is sponsored by EPA's OPPE (Mary Jo Kealy and Michael Brody,
project officers.) The researchers include ecologists from the Center for Estuarine and
Environmental Studies, U. of Maryland (R. Costanza, W. Boynton L. Wainger) and
economists from the Department of Agricultural and Resource Economics (N. Bockstael,
I. Strand, J. Geoghegan K. Bell).
The immediate goal of the project is to model the spatial configuration and
dynamic evolution of an ecological landscape by capturing ecological fictions, human
behavior, and their interaction. This will provide a means of describing the evolving
landscape under different policy scenarios on land use controls, non-point source pollution
regulations, etc. Also, the effort may ultimately provide some insights into the valuation
of ecosystems - and even the much debated issue of sustainability, although neither of
these topics will be given much attention here.
The watershed chosen for the case study is the Patuxent watershed in southern
Maryland, one of the nine river basins of the Chesapeake Watershed and covering about
1,000 square miles. This includes parts of seven counties, ranging from the Washington
DC suburbs and the state capital to predominantly rural counties at varying stages of
development. Significant portions of land within the area are dedicated to each of the
major land uses - commercial, high/medium/low density residential, agriculture (mainly
cropland and pasture, with few orchards), forests (both deciduous and coniferous), and
wetlands. There are a few industrial centers and some military establishments. It is worth
noting that agriculture comprises 32% of the watershed's land and forests comprise 46%.

We begin with a very cursory description of a generic ecological landscape model
developed by Costanza and Maxwell (1991), because it serves as the starting point for the
research effort. The term "landscape" model which we use throughout has come to mean a
spatially-articulated dynamic model of an area of land. Traditional ecological studies,
similar to traditional economic studies, assumed that systems and actors were spatially
homogeneous Landscape ecology is an outgrowth which analyzes and interprets landscape
heterogeneity and spatially explicit ecological processes (see Turner and Gardner, 1991).
In our subsequent discussions we will be focusing on the economic issues, but it is the
landscape nature of the ecological model that has led us in the particular direction we are
taking. In fact, we propose a development parallel to that from conventional to landscape
ecology for economic modeling.
Serendipitously, the approach taken in the landscape ecology model of Costanza
and associates' is particularly close in spirit to one which seems appropriate for economic
land use problems. Analogous to the generic ecosystem model that predicts expected
changes in habitat conditions, with inter-cell flows of hydrological information linked with
physical and chemical parameters, we are interested in predicting expected changes in land
use, with inter-cell flows of economic information of spatial and aspatial variables.
A compelling feature of this model is that it is designed to simulate a variety of
ecosystem types with a fixed model structure. While the structure is general, however,
different sets of ecosystem functions are activated for any site in the landscape, depending
on its location and ecosystem type. Additionally, parameters of these functions are
specific to the ecosystem type and site and are derived from field data. The underlying
model structure is more complex than any particular application is likely to need, but
allows for selection among functions and aggregation over levels of detail where
applicable. The generic approach is appealing because it is an efficient way to construct
models of this sort. Recalibration for a particular ecosystem is time consuming, but not so
costly as reinventing the entire model. Additionally a sort of comparability and uniformity
across applications becomes possible. Differences in results can be attributed to differing
ecological conditions rather than modeling idiosyncrasies.
An important feature of the generic model is its spatial disaggregation. In broad
terms, the model operates by dividing the landscape into cells and modeling the ecological
functions within each cell and the vertical fluxes of mass above and below sediment. The
horizontal mass fluxes of water, soil and nutrients between cells are then simulated over
time using a spatial dynamic simulation program. The model is driven largely by
hydrological algorithms (varying depending on the ecosystem type) and focuses
predominantly on the responses of macro- and microphytes to nutrient availability, light,
temperature, water availability, etc. Approximately 14 sectors (including a number of state
variables) are incorporated, such as the inorganic sediments sector, dissolved phosphorus
sector, hydrologic sector, macrophyte sector, etc. (see Table 2). The Patuxent application
of the model focuses on nutrient and sediment loading in the watershed and predicts such
things as changes in water quantity and quality, vegetation and amount and quality of
wildlife habitat, all at a spatially disaggregated level.

The ecosystem functions and the parameters of those functions that are simulated
for any given cell in the landscape are dictated by the cell's "land use" or "habitat"
designation at the beginning of any simulation round. Then conditioned on that land use
and the stocks of the state variables at that point in time in the cell, the processes and
fluxes are calculated. Conceptually, there are two "levels" at which human behavior
could be expected to affect this simulation. One is in the land use designation of a cell; the
other is in the nature of ecological processes that occur within a cell conditioned on its
land use.
Understandably, the ecosystem model without economic input, imposes rather
than models this human behavior. Consider the land use designation. The ecological
model calculates land use designation through a "habitat switching" model which
determines when through natural succession or weather-driven ecological catastrophe
(e.g. flood, forest fire), the habitat shifts from one type to another. Human instigated land
use changes must be imposed exogenously and hypothetically. Perhaps the most
important contribution of the economists will be to model this human land use conversion
and how it is related to both the ecological and economic features of the landscape.
Human interactions with the environment conditioned on land use, are similarly
imposed in the current ecological model, which uses something akin to a fixed coefficient
technology to capture these. For example, if a cell is designated as being in cropland, then
a given set of processes and parameters are assumed to operate, conditioned on ecological
features such as slope/soil type. Variation across individuals or responses to external
stimulae, like changing prices, are ignored. In order to assess the effects of some non-
point source policy, the model must impose an assumed change in these processes and
parameters, ignoring human response to the change in the regulatory environment. The
second type of contribution that the economists can make is in modeling these conditional
human interactions. Our first endeavor of this sort involves modeling farmer's behavior,
both in crop choice and best management practices adoption, as functions of ecological
and economic forces. We anticipate that a transportation sector or a residential sector
might follow.
The General Structure of the Integrated Model
The shortcomings of an ecological model with no "moving economic parts" are
obvious to economists. The shortcomings of our own treatment of ecosystem-related
problems should be equally obvious. While we are not primarily interested in valuation
here, how economists have treated ecosystem valuation is relevant to the discussion. With
the exception of a few who have written largely in conceptual terms, most economists
have been forced to consider only those services of ecosystems that are well-defined, are
easily measurable using conventional market or non-market valuation methods, and have
immediate consequences for humans. Piecemeal valuation of this sort ignores the more
subtle, long range contributions of the ecosystem to human welfare; and it ignores the

importance of the mnfiguration of the ecosystem landscape in determining its value,
Where things are matters. Analysis that ignores spatial location and spatial arrangement
misses important dimensions of the problem. One way of thinking about ecosystem
valuation might be: how do we value the reconfiguration of the landscape in its various
states as it evolves over time?
The appeal of a joint modeling effort that looks at the interaction of
ecological processes and economic behavior in a spatially disaggregate framework as it
plays out over time seems self-evident. The pressing question is how to structure such a
modeling effort. Simply put the purpose of having an integrated model is to capture how the
distribution of human activities (farming, electric power generation, commercial and residential
development, recreation wastewater treatment highway construction fishing) affect the
ecosystem as well as to capture the effect of the ecosystem landscape on the quality and value
of goods and services (e.g., recreation, wildlife enjoyment water quantity and quality, housing
environmental aesthetics, etc.) and, therefore on human decisions. The model needs also to
capture how human activity and its impact on the ecosystem may differ under different
regulatory regimes.
But an integrated model need not be a "black box". At this point we do not intend to
meld both ecological and economic models into one "super-model." Instead, we plan for the
two types of models to exist in parallel but to exchange information on ecological and .
economic elements generated by the other. This approach preserves the integrity and intuition
of both models. It also allows the appropriate choice of time step, geographical scale, and
level of aggregation which might differ between the ecological and economic models. The
inconsistencies that are likely to arise in these dimensions are worth discussing because they
pose problems in information exchange, no matter how the integration is structured.
By design the ecological model establishes physical boundaries. From the start the
ecologists wanted agreement on these physical boundaries since these determine how many
cells and of what types must be covered in their model. For them, the area of interest ranges
from the tops of the trees to the depths of the groundwater and extends up to the limits of the
drainage basin. The economists had no particular problems with the vertical boundaries
(except to the extent that air quality issues of certain types maybe omitted from consideration).
However, there is no reason to expect that the drainage basin of the Patuxent is a meaningful
economic boundary. The question from our perspective has to do with the extent of relevant
markets, and the relevant markets will belabor and land markets and possibly markets for
products of the area-principally agricultural, forest and recreational products.
Market boundaries are largely undefinable and sometimes are not related to
space at all. However the markets we are interested in - land, labor, recreation and products
that have high transportation costs or are perishable-are likely to peter out or dissipate with
distance. Regional economic modeling deals with these artificial boundaries of markets, and
generally does sousing political boundaries. The attached map shows the difference between
the county boundaries and the watershed boundaries, but both are clearly arbitrary delineations

Patuxent Watershed Counties
~	County Boundaries
~	Watershed Boundary


from an economist's perspective. The boundary definition affects what explanatory variables
are considered endogenous to the system (and thus must be predicted internally by the model
for any scenario simulation) and what variables are considered exogenous to the system (and
thus must be predicted by some model of the regional economy). The problem is more
complicated than our usual economic intuition would suggest because we are not dealing with
aggregates within and outside our region, but with spatially disaggregated decisions that are
probably serially correlated over space (more about this in the next section.) Decisions made
at point x in the landscape may be affected by characteristics of the landscape within y miles of
the spot. Thus if we are interested in simulating activity in the watershed we may need to
know about activity that extends some y miles beyond the watershed boundaries -or up to
some natural geographic barrier such as an ocean major river, etc. Because our GIS data is
available by county, we currently have information that extends up to the political boundaries in
the attached map. This poses no problem for information exchange between the two models,
since the ecological model will use only that part of our information which it needs. But the
"sliding" boundary problem remains an issue for the internal workings of the economics model.
Organizational Complexity
The ecological model is structured around a desired level of ecological complexity and
resolution. It simulates the activities of thirteen sectors and tracks twenty-five state variables
all of which are fisted in Table 2. Modeling with greater levels of disaggregation is possible,
but costly, and requires a significant amount of additional data collection and programming.
However, many of the state variables of importance for tracking ecological processes are not of
direct interest to the modeling of human behavior (or for assessing the value of the landscape
Unfortunately, the ecological model cannot afford much detail in the state variables
that are most visible and important to humans. The animal kingdom is represented by one state
variable "consumers" and the macro-plant kingdom by the state variable, macrophytes levels
of aggregation decidedly unacceptable to the economists. A proposed solution is to introduce
the details exogenously by determining in side calculations the subgroups of macrophytes and
consumers that are likely to be found in a given cell depending on its habitat/land use type,
surrounding land uses and distances to critical ecological features (streams, etc.) and human
disruptions (highways, etc.) The model will include markers and detailed rules for species loss,
treating habitat evaluation as a side calculation. A function will be developed that will indicate
the likelihood of game species and other forms of wildlife in particular habitat types. This
approach allows for the provision of additional complementary information without increasing
the level of disaggregation of the generic ecosystem model.
Time Scale
The ecosystem model operates on a time scale of a day or less. Yet, given available
data the economic models will be estimated on an annual basis. This means that the timing of
the exchange of information and the time-dependent nature of that information must be
carefully thought out. The economic decisions can be modeled on an annual basis but then
distributed over the year according to rules or separate side calculations. For example,
intra-year timing of the agricultural decisions will be easy to predicts, since these are governed

by growing seasons. The timing of those decisions that are prompted by weather can be driven
by the ecological simulation that incorporates weather pattern simulations as well.
Land conversion decisions cause more trouble. Their timing is important because
construction can have different immediate ecosystem impacts depending on the season. One
solution is to use independent data on building starts and construction durations to forecast the
seasonal impacts of construction. We may also add another habitat/land use type- land in
transition, since this state can cause more sediment loss than almost any other.
Geographical Scale
The geographical scale of resolution between the two models will also differ. The
ecosystem model divides the study area into cells each covering approximately 0.364 km2 or
90 acres. But for land use conversion decisions in states like Maryland, 90 acres is far too large
relative to the decision unit. We have a choice (described in a later section) of using actual
ownership parcels or of dividing the landscape into calls but the cells would likely be smaller
than 90 acres and more closely matched the size of areas that are observed to convert in a
given year.
In any event, the economic model will need to use observational units smaller than the
cells in the ecological model and this poses one of the more serious modeling conflicts for the
project. The economic model of land use conversion will generate predictions at a higher level
of spatial resolution than the ecological model will need. However, since the former is likely to
take the form of a discrete choice model it will produce predicted probabilities that can be
interpreted as proportions. The ecological model can accommodate heterogeneity in the form
of shares) within a cell if it is not necessary to preserve information on the specific locations of
the heterogeneous factors within the cell. Devising weights to monitor what is happening in
cells, thresholds can be set so that cells could go from homogeneous to heterogeneous units
and vice versa. This additional detail will also allow the model to make inferences on a wider
variety of land use restrictions (i.e., agricultural policies, zoning policies, and environmental
protection policies).
Disaggregating the output from the ecological model for use in the economics
model may be more difficult. The exact locations of some features of the landscape that
are important to economic decisions will be known independently and will be unchanging
(e.g. location of streams) but information about others (e.g. quantity and quality of stream
flow) will be available only as output from the ecological model. We do not yet know how
serious a problem this will turn out to be.
Time Horizons
Some ecosystem effects of human actions take a long time to play out, and as a
consequence the ecologists are interested in scenarios of at least 20 years. A time horizon
of this length makes economists nervous - so much that affects human behavior
(technology, changing preferences, . . .) is impossible to predict very far in advance.
Nonetheless, we have agreed to 20 year scenarios, but there will be an array of these
subject to a host of different assumptions.

Spatial Data/Issues in Economics
The emphasis in the discussion so far has been on "space" and "time". If
"ecological-economic" modeling has any meaning at all, it must have something to do with
the interactions between humans and natural systems over space and time. Economists
have excelled at modeling time dynamics, but spatial issues have received much less
attention. Perhaps this is because the markets for most goods are not spatially driven.
Land, while not the only exception, is certainly the most obvious one.
What happens to land, not just aggregate land but the spatial arrangement of land,
is a topic of increasing interest to multiple disciplines. Land use is inextricably tied to
public infrastructure demands that are more or less costly depending on their spatial
distribution. Land use is almost synonymous with locational externalities - visual, noise,
etc. And land use has environmental consequences that differ markedly depending on the
pattern of remaining habitat and the size and proximity of disturbances to ecologically
sensitive areas. The configuration of land is one of the major contributors to the quality of
Yet, traditional fields of economics have reduced the complexity of spatial
relationships, almost to the point of making spatial issues non-issues. Either aggregate
relationships have been specified or the spatial components in a model have been reduced
to uni-dimensional variables, e.g. the distance between economic activities in a location
model, the wage differential in a migration model, cost of access in a transportation mode
choice model. The concept of a landscape mosaic of natural and human-managed
patches is foreign to economists.
Data drives analyses. In the absence of spatially articulated data, there has been no
impetus to develop broadly adopted methods for analyzing two dimensional space. But
now that GIS data is becoming more readily available, economists are reconsidering their
analytical tools. Along with others at a similar stage of thinking, we are looking for away
to take full advantage of this new type of data. Is there some way of explicitly thinking
about spatial interactions and their impacts on decision making beyond including location
specific amenities and distances to features of importance? Can we model these spatial
issues using higher dimensions in order to increase the predictive power of our model? If
not a totally new approach, can we use these new data to better describe the aspects of
space that matter?
While most economists know that GIS is a technology that can store, analyze, and
display spatial and descriptive data, not so many economists have had the opportunity to
work with such data. A GIS technology takes information from existing maps or aerial
photographs and digitizes it, keying points, lines or polygons in one way or another to
map coordinates. GIS software is used to manage the database system to store and
retrieve data to analyze data and to report analyses and display maps.

Our GIS data includes mappings of land uses at four points in time for the counties
of interest. We also have, or will soon have, access to digitized maps of ecological
features, such as slopes, soil types, elevations, and hydrology (streams, rivers, etc.), as
well as the output of the ecological model simulations that will provide values for state
variables in a GIS format. We expect to obtain GIS data on zoning and land use controls
in our counties, as well as likely scenarios for future land use management. Additionally,
our data base includes transportation networks, business districts, street addresses, etc.
The latter allows us to match information from other sources (including a tax assessment
data base and a survey of farmers) to map coordinates. GIS software provides a means of
obtaining a variety of measures, including calculating distances, registering contiguous
attributes, and measuring percentages of areas of various shapes and sizes made up of
different attributes.
While we are still searching for the most valuable way to use these data there are
some spatial attributes that are clearly of importance to the value of land in different uses,
For example, the value of a parcel in residential use will be affected by access to
employment centers (given by transportation networks and proximity to business districts)
and private and public infrastructure (shopping, schools, recreational facilities), etc. But it
will also be affected by the spatial arrangement of ecological features and man-made
structures making different parcels equi-distant from employment centers of differing
value because of these spatially oriented amenities/disamenities. Additionally, the ability
to convert land to a developed use will be circumscribed by regulatory mechanisms and
incentives: zoning, land use controls, taxation patterns, best management practice
incentives, etc. The value to society of land in an undeveloped state will also depend on
attributes of the land and its spatial arrangement. For example, the suitability of a patch
for wildlife habitat will depend on its water and vegetative features, its size, shape and
habitat edges and its proximity to human disturbances and human access.
Spatial Measures in Modeling
The disciplines of landscape ecology and geography, as well as a sub-field of
econometrics called "spatial econometrics" (see, for example, Anselin, 1988) offer
interesting alternatives to conventional measurements of space. Here we discuss two
types of measures of spatial pattern that have emerged in some of this literature: measures
which capture in a two-dimensional way relationships among cells in the landscape and
measures that capture the complexity of spatial pattern. We consider their application to
economic models that attempt to describe what goes on at any given location in the
The original motivation for the following measures were driven by regional
economic development issues. Therefore, all the following measures were derived in
order to use spatial aggregate data, on large spatial units such as counties or census
bureau tracts. However, our model will use disaggregate data on much smaller spatial

units, such as land parcels, so the following measures will have to be modified to use on
disaggregate data. Given this caveat, we now describe some of these measures.
Spatial contiguity matrices describe spatial relationships between all pairs of spatial
units in the landscape. The simplest is based on binary contiguity between spatial units,
where each cell is represented by a row and a column in a matrix. For any i,j combination
of cells a 1 appears as the i,j element of the matrix if the cells are contiguous and a 0
otherwise. This requires dividing the landscape into units, and for regular structures
mathematical properties are well defined. It is also possible to define higher order
measures of contiguity.
Spatial weight matrixes are extensions of spatial contiguity matrices that add
weights to the contiguity measure. Matrices with terms such as the following are
commonly employed:
w. = b a B ¦
ij VIJ l
wv = a exp(-c/(> 10)
bv = binary contiguity factor
a, = the share of area i in the entire spatial system
Bij = the proportion of the interior boundary of unit i in contact with unit j
d„ — distance between unit i and unit j
and a, f3= parameters.
In the first two expressions above, the matrix contains non-zero information only
for contiguous cells, although as the third example suggests, weight matrices can easily be
defined that allowed relationships with more distant cells. In our disaggregate model,
which has much smaller spatial units, these measures can be modified in a number of ways.
For example, the' first two measures, which are based on binary contiguity, can be
extended to allow for higher levels of contiguity. In this way, land use parcels can be
affected by other spatial attributes that are not directly contiguous, but yet are of interest
for their potential impact on the land area in question. Measures in the spirit of the third
example above already permit impacts from noncontiguous units, so can easily be used to
create matrices that incorporate influences from a further distant.
An obvious way to use these matrices in econometric modeling is to add structure
to the pattern of correlation among errors. Spatial data introduces the likelihood of spatial
autocorrelation. One can also imagine using these weights to discount location-specific
explanatory variables with distance. In this context, one might think of these weights as
spatial lag operators.

Landscape ecologists have also developed indices of complexity of spatial pattern,
derived from information theory and fractal geometry. These indices have been used
principally to compare spatial heterogeneity across landscapes of considerable size, but
seem adaptable to our type of problem (O'Neill et al 1988). A well-known and
commonly used measure of diversity (or conversely dominance) from information theory is
applicable here: Within any given sized sub-area of landscape, diversity of land uses could
be measured by:
H = -f.Pt\nP,
k=l .
where Pic is the percent of the sub-area in land use k and m is the total number of land
uses. H ranges from 0 when all land in the sub-area is of the same land use to In m, the
value of H when all land uses are represented equally. Consequently, a measure of
dominance is given by:
D = \nm~ H
and ranges from 0 to In m, at maximum dominance.
A second and less well-known information theory measure is a measure of
"contagion". This index is concerned with edges and contiguity, and reflects the extent to
which land uses are clumped.
C = 2/ilnrt + ££&,ln£^
where Qij is the proportion of cells of type i adjacent to cells of type j and n is the total
number of cells in the sub-area. Note that 2n in n is the maximum value of the second
term. At high values of C, land uses are highly concentrated; at low values the landscape
is heavily dissected.
Finally a measure adopted from fractal geometry is frequently used to capture the
complexity of the sub-area. The fractal dimension is twice the slope of the regression line
found by regressing the log of one-quarter of the perimeter on the log of the area. The
fractal dimension ranges from 1.0 if all patches are simple square shapes to 2.0, which
represents an patch with the same area, but with a very complex shape.
In the next section, we explore how we might actually do some economic
modeling in space and how we might use some of the above concepts to add richness to
our modeling effort.

Modeling Land Conversion
Recognizing that the ecological effects of human activity are driven by the specific
uses man chooses to make of the stock of natural capital, one of the major contributions
we can make to the ecologists' landscape model is an understanding of how the land use
decisions are made by individuals. This is critical for the integrated modeling effort, since
the simulation of each geographically designated cell's ecological functions are driven by
land use designation.
More specifically, the purpose of this phase of the project is to develop the ability
to predict future land use of a parcel or unit of land, given information on its history,
relevant zoning and other land use restrictions, the general level of regional economic
activity, and the variety of often spatially related economic and ecological variables that
affect the value of the parcel in different uses. Given this information, we intend to
predict the probabilities that a parcel of land with certain characteristics will stay in its
present land use or convert to alternative uses.
There have been numerous attempts by economists to model land use conversion
(see, for example, models of urban fringe development by Dunford, Marti, and
Mittlehammer, 1985; Alig and Healy, 1987; Barnard and Butcher, 1989; McMillan, 1989)
but they have been hampered by limited data. The data we have available, while not
perfect, offer the potential for a richer and more spatially disaggregate model than has
previously been possible. But as the previous section explains, we are still uncertain as to
how to take full advantage of this spatial data.
We have two interesting data sources that contain information on land use
conversion. The first consists of snapshots, at four points in time, of land uses in the
seven Patuxent watershed counties prepared by the Maryland State Office of Planning.
These are GIS data, and in this format different land uses are recorded as polygons on a
digitized map. A polygon of a minimum of 10 acres will appear on the map with a
separate land use designation. This GIS database allows us to see three periods of land
use changes - from 1973 to 1981, from 1981 to 1985, and from 1985 to 1990. The land
use designation categories are reported in Table 1, but can be summarized as types of
agricultural land, types of forests, types of residential, industrial, institutional or
commercial development, barren land, wetlands, etc.
Tax assessment files comprise the second data source of interest and were acquired
from Maryland's Department of Assessments and Taxation. These files include
observations on each individually or publicly owned parcel of land in the seven relevant
counties as of 1993. The database includes fields for a wide range of interesting
characteristics of the land parcel and the owner. Not all fields are filled in for all
observations or all counties, but there remains considerable information on each parcel.
Variables include size, location, zoning, land use designation property factors (e.g. sewer,
water, historic, etc.), structure description, market value, tax assessment, building value,
land value, etc. Of particular interest to us are the variables that report property transfer

information, and year built, if a structure exists on the property. Because the data base
includes addresses, we can, at least in theory, map the locations of these parcels onto our
GIS database using Census Bureau TIGER files that supply GIS coordinates for street and
road addresses. This process is underway and we are currently attempting to resolve the
matching problems that invariably occur with such data sets.
These two data sources together provide important information and can be merged
in the GIS database, but they are different in a number of important dimensions. The State
Office of Planning land use maps give us a good picture of land use change over time.
Land use changes are recorded in terms of polygons switching from one land use
designation to another, rather than individual parcel owners' decisions. This maybe a
useful format if we choose to employ a grid or cell type approach in defining our units of
observation. In that case we could model the proportion of each cell in a given land use at
a point in time. However, from these maps changes can be observed only in
approximately 5 year intends. This obscures observation of the sequencing of changes
and lengthens the time unit of measurement even further relative to the ecological model.
In contrast, the tax assessment database is extremely detailed and includes data by
parcel of ownership, should we choose to use that as the unit of observation. However,
because it records information as of the current period, it must be used creatively to
extract information about past changes. These changes must be deduced from information
on time and conditions of property transfer, date at which property was converted from
one tax category to another, and year structure was built. Additionally, if we have
difficulty mapping all parcels, we may encounter selection biases in our sample of
Despite the shortcomings in both data sets, merging them will provide far better
information than has been available to analyze land use conversion in the past. At this
point we expect that our observational units will be cells rather than parcels, in part
because even in the tax assessment data base the observations are based on parcel
ownership only in the current time period not at the time the decision was made. Thus,
from these data it will be impossible to determine whether several new housing units came
from one or more conversion decisions. Economists are more comfortable using
observations on decision makers than on units of the commodity, but somewhat related
problems arise in some types of surveys. For example, on site surveys in recreation yield
samples of trips rather than samples of recreationists.
Also included in the GIS data base is information on transportation networks and
central business districts, hydrology (streams, rivers, etc.), land slopes, soil types and
elevations. (We are currently attempting to acquire "hiatn-ri^Ql transportation information.)
This is all in addition to the GIS level data supplied by the ecological model. These data
allow us to calculate, for any arbitrarily small cell in the landscape, such things as distances
from roads and highways, towns and employment centers, and natural ecological features
of interest - like shoreline or recreational facilities. It also provides a means of calculating

variables that reflect what is going on around a particular point on the landscape and what
may be happening to the quality of the environment.
Two other external sources of information are worth mentioning at this point. The
state of the regional economy is likely to be an important factor in determining land use
conversion in the Patuxent watershed. In order to simulate future scenarios of land
conversion, we need a forecasting model of the economic activity in the region. The most
likely candidate for this is a well recognized regional model of Maryland developed and
marketed by Mahlon Strazheim of the Economics Department of the University of
Another source of externally supplied information will come from the Patuxent
Demonstration Project. This is an inter-governmental research group that has assembled
the current zoning and land use restrictions for the Patuxent watershed counties in a
detailed GIS format. They have also developed a set of potential land use management
scenarios that could conceivably evolve over the next two decades in this area. These
include zoning based on comprehensive plans and sewer/water service plans; forest
conservation and agricultural best management practices programs; and clustering
requirements together with urban best management practice programs.
While we have a host of data related problems to overcome, including some
potentially serious sample selection problems, our ultimate data set is likely to consist of
discrete panel data: time series observations on land use of individual parcels or of equi-
sized cells in the landscape. We plan to use models of discrete panel data (see Heckman,
1983) either to predict the probability of any parcel of land, or the proportion of any equi-
sized geographic cell, being in a given land use at time t.
Heckman's treatment of panel data incorporates intertemporal connections among
decisions and the resulting increased complexity in error structures. Adapting Heckman's
general model, we consider that the continuous latent random variable (i our problem
reflecting utility or returns from putting parcel i in land use m at time t) can be given by a
systematic function of exogenous variables and variables capturing the dynamic nature of
the decision (i.e. functions of past decisions and values of past latent variables) as well as
an error term.
Heckman frames his problem in a dichotomous choice context, but we will have
either polychotomous choices or nested dichotomous choices. In a general model, we
would be interested in predicting the probability that a parcel, conditioned on current land
use, will end up in any of m land uses in the next period, where m might be as many as 5
or more land uses depending on aggregation over categories in Table 1. For our particular
study area, however, most conversions take place from some relatively undeveloped use
(e.g. forest or agriculture) to a developed use (some type of residential or, far less often,
commercial). Rather than having a majority of zero cells in a conversion matrix, we might
alternatively consider a series of nested dichotomous choices:
1. develop or not

2.	if develop residential or commercial
3.	if residential: low or high density
Both polychotomous choice and nested dichotomous choice problems are most easily
framed in the context of multinomial logit. However, the complicated error structures that
are suggested below are not possible in a logit framework unless they can be captured by
fixed effect terms.
Dynamic Issues
There are a variety of types of dynamic relationships possible in discrete panel
data. First and foremost, the parcel's state in the previous period will be expected to have
an effect on the decision. This type of term would appear in a simple Markov chain model.
In our problem it is clear that the land use in time t-1 will have an important effect on land
use choice in time t, because of inertia and varying costs of transition (none of which are
likely to be fully captured with explanatory variables). It is not obvious that choices in
time periods t-j, j> 1 can be expected to have a separate effect. The Markov effects may
not be stationary, however. They maybe changing overtime because of (otherwise
unmeasurable) changes in land use policies, for example.
Additionally the cumulative history of the parcel might matter. For example, the
valuation of a parcel in a particular use maybe affected by how long the parcel has been in
its current state. Accumulation and depreciation of natural, human, and structural capital,
as well as other forms of time dependency, can be reflected this way. "Renewal" terms of
this sort may have interesting interpretations in land conversion models related to soil
depletion, timber cycles, or depreciation of man-made capital, but whether our data will
support such subtleties remains a question.
We might also expect a lagged adjustment to past valuations of alternative states.
Given the near irreversibility of some land use conversion decisions, responses to a
persistent economic signal are more likely than sudden responses to a one-time change in
economic conditions. Additionally, given the time it takes to plan, obtain permits, etc.
there will likely be a lag between conversion decision and observable action.
Exogenous Variables and Spatial Issues
The model needs to capture those factors that dictate the value of a parcel of land
in different uses. These maybe ecological features of the landscape, such as soil type,
slopes, water availability, scenic amenities. They will also include man-made features of
the landscape, such as access to employment centers (given by transportation networks
and proximity to business districts), and access to both private and public infrastructure
(shopping, schools, recreational facilities), etc. The ability to convert land and its ultimate
value in alternative uses will be circumscribed by regulatory mechanisms and incentives:
zoning, land use controls, taxation patterns, best management practice incentives, etc. But
even this relatively straightforward consideration has locational spill-over effects, since the
zoning of the land next door has an effect on the value of a particular parcel. Finally, the
value of a parcel in a given land use is very much affected by the land uses of surrounding
land, not just specific features with point locations.

There are a few ways we could imagine incorporating the spatial measures
mentioned earlier into the systematic part of our land use conversion model. A weighting
scheme based on distance and contiguity might be used as a spatial lag operator on
exogenous variables. For example, instead of using as an explanatory variable equal to the
distance to the nearest employment center, we might include all relevant employment
centers measured by their size and weight them by the spatial weights (i.e. discount them
by distance and/or contiguity.)
We could also apply spatial lag operators to variables that reflected land use, the
dependent variable, for surrounding cells. The probability that a particular undeveloped
parcel will be developed during time period t will be affected by the land use configuration
surrounding the parcel at the beginning of t. We might measure the proportion of land
within any concentric circle, for example, in a particular land use, and then weight these
measures by the distance of the concentric circle from the parcel and/or by contiguity
These measures are promising but might not fully capture the aesthetic,
congestion, access, etc. aspects of land configuration that make location so important in
land values. Models of land value or land conversion may make particularly good use of
the measures of spatial heterogeneity and complexity that have, up till now, been used
principally for description. By choosing an appropriate size for a sub-area, we can
calculate measures of diversity, contagion, etc. for circles or squares centered on each cell.
By doing so, we encounter the "sliding" neighborhood phenomenon making knowledge
of areas within an ever expanding boundary necessary for simulation. This argues all the
more strongly for modeling an area bounded by geographical "walls" such as bays, rivers,
etc. rather than the boundaries of the watershed.
Error Structure
The error structure in our problem poses particular problems. In his general
model, Heckman assumes that the errors are distributed with mean vector zero and
covariance, Z, which is a TxT positive definite matrix. This specification allows non-
stationary and serially correlated errors. Although more general than previous models,
Heckman's specification assumes that e(i) is independent of e(j), j*i, because he is
concerned with panel data in which the cross section observations are taken over
randomly selected individuals.
In our case, the observations will be overland parcels and the spatial relationship
among parcels is likely to dictate a pattern in the error structure in the cross-section
dimension as well as the time series dimension. Perhaps the most obvious use of the
spatial contiguity or spatial weight matrices described in the last section is to provide
structure for the covariance matrix of the errors. Clearly our model will not capture all
relevant factors and the omitted ones will certainly be correlated over space because of the
immense importance of locational spill-over effects. The types of weight matrices

discussed above can provide structure - dependent on distance and contiguity factors - for
the covariance matrix of the errors.
The complexity of the error structure, together with the polychotomous or nested
nature of the choice problem, poses estimation difficulties. If we assume a generalized
extreme value distribution for the 6's, thus generating a multinomial logit specification
then only a fixed effect model is practicable. However, if we wish to represent the likely
error structure, we would need to assume a normal distribution (as does Heckman). We
have not yet resolved this modeling problem and any ideas will be gratefully received.

Table 1
— The Habitat or Land Use Designation Types
Developed/Urban Land Uses:
Low density residential
Medium density residential
High density residential
Open urban land
Row/garden crops
Deciduous forest
Evergreen forest
Mixed forest
Wetlands (by State Land Use definition; not Section 404 definition)
Bare ground

Table 2
— State Variables in the Generic Ecosystem Model
Hydrology Sector:
Surface water
Unsaturated water
Saturated water
Hydrodynamic Sector:
Horizontal Flows (rivers, waves)
Vertical Flows (snow, rain)
Inorganic Sediments Sector:
Deposited inorganic sediments
Suspended inorganic sediments
Pore space
Salt (NaCl) Sector (Conductivity):
Salt crystals
Salt in surface water
Salt in sediment water
Dissolved Phosphorus Sector:
Phosphate in surface water
Phosphate in sediment water
Dissolved Nitrogen Sector:
Dissolved inorganic nitrogen in surface water
Dissolved inorganic nitrogen in sediment water
Dissolved Oxygen Sector:
Dissolved oxygen in surface water
Non-Macrophyte Sector:
Algae (phytoplankton and/or periphytons)
Macrophyte Sector:
Macrophyte photosynthetic biomass
Macrophyte non-photosynthetic biomass
Above Sediment Organic Matter and Detritus Sector:
Suspended organic matter
Standing detritus
Organic Sediments/Soil Sector:
Deposited organic matter
Consumer Sector:
Consumer biomass (all fauna except microscopic decomposes)
Fire Sector:
Fire Igniters
Fire Propagates

Alig, Ralph J. and Robert G. Healy. "Urban and Built-Up Land Area Changes in the
United States: An Empirical Investigation of Determinants." Land Economics 63 (3
1987): 216-226.
Anselin, Luc. Spatial Econometrics: Methods and Models. Dordrecht, The Netherlands:
Kluwer Academic Publishers, 1988.
Barnard, Charles H. and Walter R. Butcher. "Landowner Characteristics: A Basis for
Locational Decision in the Urban Fringe." American Journal of Agricultural Economics 71
(3 1989): 679-684.
Costanza, Robert and T. Maxwell. "Spatial Ecosystem Modelling using Parallel
Processors." Ecological Modeling 58 (1991): 159-183.
Dunford, Richard W., Carol E. Marti and Ronald C. Mittlehammer. "A Case Study of
Rural Land Prices at the Urban Fringe Including Subjective Buyer Expectations."Land
Economics 61 (1 1985): 10-16.
Heckman, James J. "Statistical Models for Discrete Panal Data." In Structural Analysis of
Discrete Data with Econometric Applications, ed. Charles F. Manski and Daniel
McFadden. Cambridge, MA: The MIT Press, 1983.
McMillen, Daniel P. "An Empirical Model of Urban Fringe Land Use." Land Economics
65 (2 1989): 138-145.
O'Neill, R.V., J.R. Krummel, RH. Gardner, G. Sugihara, B. Jackson, D.L. DeAngelis,
B.T. Milne, M.G. Turner, B. Zygmunt, S.W. Christensen V.H. Dale, and R.L. Graham.
"Indices of Landscape Pattern." Landscape Ecology 1 (3 1988): 153-162.
Turner, Monica G. and Robert H. Gardner. "Quantitative Methods in Landscape Ecology:
An Introduction." In Quantitative Methods in Landscape Ecology, ed. Monica G. and
Robert H. Gardner Turner. New York: Springer-Verlag, 1991.

Landscapes, Ecosystem Value, and Sustainability
Robert Gottfried, David Wear and Robert Lee
This paper offers an ecologically-based view of land and land value, building upon
the concepts of ecosystems as multiproduct assets and of landscape ecology.*
Having briefly reviewed landscape ecology, the paper questions the ability of
markets to create optimal landscapes, even when traditional methods of
internalizing externalities are applied. The paper concludes that attempting a
complete valuation of ecosystems appears to be a rather quixotic enterprise.
Managing natural systems to optimize production of certain valued outputs,
perhaps subject to certain sustainability provisions, may represent a more practical
goal. Achieving sustainable landscapes, however, requires both sufficient
ecological knowledge and institutions capable of bringing about this result
inasmuch as the unaided market cannot do so. The paper argues that landscape
modeling may help provide needed information, and examines forms of public and
private ownership to assess how well particular institutional conditions might
facilitate ecological adaptation. Flexibility and creativity will be needed in
designing institutions that can deal effectively with landscape-scale management.
* This paper is the outgrowth of a series of discussions by the authors as part of the US Man &
the Biosphere Temperate Zone Directorate Project "Land Use Patterns in the Olympic and
Southern Appalachian Biosphere Reserves: Implications for Long Term Sustainable
Development and Environmental Vitality."
Copies of the manuscript may be obtained from:
Dr. David Wear
U.S.D.A. Forestry
3041 Cornwallis Road
Research Triangle Park,
North Carolina 27709

People and Parks: Economic Management of Khao Yai National Park, Thailand
Heidi J. Albers
Food Research Institute
Stanford University
October 1994
Following the policy literature on people-park conflicts, this paper provides an economic
analysis of the efficiency of park management decisions and their impact on rural incomes in
developing countries. Using Khao Yai National Park (KYNP) in Thailand as a case study,
analysis of an economic model reveals the importance of spatial and intertemporal characteristics
of land use in and around a park area for establishing management schemes that meet both
preservation and rural development goals. Sensitivity analysis of the model reveals the role of
discount rates, the importance of habitat size and spatial externalities, and the impact of the
perspective of the manager-local, national, international-on optimal land use. The spatial
analysis suggests that current management of KYNP fails to consider the impact of the park on
economic development in surrounding and, in so doing, allocates too much land to a pure
preservation use. A buffer zone policy paired with rights for extractive good collection within the
park would increase the social benefits created by KYNP.
Resented at the Association of Environmental and Resource Economists Workshop, "Integrating the
Environment and the Economy: Sustainable Development and Economic/Ecological Modeling." Boulder,
Colorado, May 5-6, 1994.
Copies of this manuscript may be obtained from:
Professor Heidi Albers
Food Research Institute
Stanford University
Stanford, CA 94305

R David Simpson
Roger A. Sedjo
John W. Reid
Resources for the Future
1616 P Street NW
Washington, DC 20036
Presented to the
1994 AERE Workshop
Integrating the Environment and the Economy:
Sustainable Development and Economic/Ecological Modelling
6 June 1994

There has been considerable recent interest in "genetic prospecting" among wild
plants and animals for novel chemical compounds. Such prospecting might uncover new
pharmaceutical products and provide a mechanism for saving endangered ecosystems. It
is unclear what values may arise from such activities, however. Evidence from observed
transactions is incomplete. Existing theoretical investigations are flawed in their treatment
of the probability of discovery of novel chemical compounds. In this paper we develop a
simple model in which the "marginal species" maybe redundant with respect to its
potential as a source of new chemical leads. By optimizing the value of the marginal
species with respect to the probability with which it yields a commercially successful
product we are able to place an upper bound on its value. This upper bound may itself be
relatively modest. Slight modifications in assumptions lead to drastic reductions relative
to this upper bound. We also extend our findings from the value of the marginal species
to that of the marginal hectare of habitat by combining our results with a common model
of the species-area relationship. We find that the incentives for habitat conservation
generated by pharmaceutical research are also, at best, very modest, and are more likely to
be negligible.

There has been considerable recent interest in "genetic prospecting." Genetic
prospecting is the search for chemicals produced by wild organisms. In nature, these
compounds are employed to escape predators, capture prey, increase reproduction, and
fight infection. These chemical compounds might be of considerable commercial value if
adapted to industrial, agricultural, and, particularly, pharmaceutical applications.
Genetic prospecting has also been touted as a tool for the conservation of
biodiversity. It has been argued that incentives for the preservation of areas m which
genetic diversity is greatest, particularly tropical rain forests, might be increased if
landholders could be compensated for the values generated by endangered organisms used
in new product research (this argument has been made, with varying degrees of
enthusiasm, by, among others, Farnsworth and Soejarto, 1985; Principe 1989: Wilson,
1992; Reid et al, 1993; and Rubin and Fish, 1994).
In order to determine the strength of such conservation incentives, we would need
to know the value of the "marginal Species"' in genetic prospecting. A number of studies,
including those of Farnsworth and Soejarto [1985], Principe [1989] McAllister [1991],
Harvard Business School [1992], Pearce and Puroshothamon [1992], and Aylward
[1993]2 have adopted, with differing degrees of sophistication, a straightforward
approach to valuing biodiversity for pharmaceutical search. In each of these
contributions, the authors have multiplied an estimate of the probability of discovering a
commercially valuable substance by the value of such a discovery. There is considerable
disagreement among the studies as to the magnitude of estimation of the
latter quantity, although the sober estimates offered by the more recent studies seem the
1 We will argue in Section VI that the "marginal species" is in fact a meaningful concept. To anticipate
that discussion many biologists-and even many describing apocalyptic scenarios-model the loss of
species as a continuous function of the conversion of habitat rather than as a catastrophe discontinuity.
* An excellent summary of all these studies may be found in Aylward, 1993.

more probable. The results of these exercises vary widely, ranging from as little as $44
per untested species in situ [Aylward, 1993] to as much as $23.7 million (Principe, 1989].3
The studies in which the value of indigenous genetic resources in pharmaceutical
research have been more thoughtfully derived are useful in that they incorporate detailed
treatments of the nature of the benefits to be derived from new Product discovery.We
believe the method underlying all these studies to be flawed, however. It is curious that
this existing work on economic valuation of genetic resources takes little account of
scarcity. Redundant, resources are not scarce, and hence are not of great value on the
margin. By multiplying the probability with which an organism sampled at random
contains some chemical compound of commercial value--whether unique to that organism
or not-by the expected value of a successful commercial product earlier researchers have
failed to recognize the possibility of redundancy among natural Compounds.4 Thus
potential values may be overstated in even the more carefully conducted work.
Our approach is more closely related to that of Brown and Goldstein [1984]: we
value the marginal species on the basis of its incremental contribution to the probability of
making a commercial discovery. Our work is also related to that of Polasky and Solow
[1993], Solow, Polasky, and Broadus [1993], and Weitzman [1992, 1993]. In these
papers the authors measure biological diversity in terms of the genetic "distances"5 between
related species; in fact, Polasky and Solow [1993], and Weitzman [1992] show how their
proposed measures of diversity can be related to the incremental probability of discovering
' There is also some confusion in many of these studies between the average and the margin value of
biodiversity. The total value maybe truly astronomical, and hence the average value substantial. We
show that the value of the marginal species is likely to be negligible, however.
4 Note that we emphasize the possibility of redundancy, rather than assert its existence Our findings do
not rest on the existence of redundant compounds,but rather on the fact that if the marginal material from
which sampling may occur is so rare as not to be redundant the probability of its discovery is small. This
point is made more formally below, but should be borne in mind through the entire discussion .
® See Weitzman [1992] for an explanation of how distance may be measrured by matching DNA.

commercially valuable compounds. In each of these papers, however, the authors are
attempting to describe a measure of biodiversity; that is, a ranking by which one collection
of organisms may be said to be more or less diverse than another.* In our work, we
accept current taxonomiC7 practice as the appropriate measure; we suppose that all species
within a particular taxon are "equally different." We then ask by how much is value
augmented by increasing the number of species that maybe tested in new drug research.
Valuation methods based on the work of these other authors will prove more
valuable as greater information concerning the genetic constitutions of species-and even
individuals-becomes available. Our simpler approach is closer to practical application,
however. Biologists estimate there to be between ten and one hundred million living
species. Of these, only about 1.4 million have been described [Wilson, 1992] and a far
smaller number have been subjected to chemical or genetic analysis [Farnsworth, 1988].
The types of measures suggested by Weitzman and Polasky, Solow, and Broadus simply
cannot be performed on a broad scale with existing data and computational limitations. In
our work we will treat each new species to be evaluated as an independent Bernoulli trial
with an equal probability of yielding the commercial product for which it is being tested.
Since much of the literature on biodiversity preservation emphasizes the importance of
saving as yet unknown species as genetic insurance against as yet unidentified diseases,
our approach seems appropriate.
The reader may find it curious that the roundabout methods we describe for
determining values are necessary. One might suppose that our questions could be
® A more recent paper by Polasky and Solow [1994] does deal explicitly, and in a relatively sophisticated
manner, with valuation issues. The Polasky and Solow paper does not address values on the margin,
however, and it does not incorporate any costs ofprospecting-hence, there is no "stopping rule" to
determine when additional search is justified. Finally, it would appear that the recent Polasky and Solow
paper wee written in part to address omissions in an earlier version of this paper.
7 We will use "taxonomy," "taxon" and its plural, "taxa" often in this paper. A taxon is a collection of
species, or a collection of collections of species, etc.; e. g., a genus, class or order. Taxonomy is the
science of categorizing species according to the successfully narrower taxa to which they belong.

answered merely by observing market transactions. We discuss the reasons for which this
is not feasible in the next section. Following that, we provide a very brief overview of the
natural products pharmaceutical research process. We then turn to a discussion of
possible sources of redundancy in genetic prospecting. Our main results are presented in
the fourth through sixth sections of the paper. We present a simple model in which
discoveries may prove redundant. We are able to derive an upper bound on the value of
the marginal species--and, by extension, on the marginal unit of habitat on which it exists.
We demonstrate that this upper bound will be substantial only under very optimistic
assumptions, and that the value of the marginal species falls off very rapidly if the
probability of discovery differs from that which maximizes the marginal value.
Any model that purports to measure something as speculative as the value of a
species for its pharmaceutical research potential must be built on a number of simplifying,
assumptions. We discuss these assumptions and their implications in a seventh section,
but we can summarize hereby saying that we do not believe that a more realistic treatment
would change our results much.
We state our conclusions in a final section but we should emphasize one point
now. This paper is concerned solely with pharmaceutical reseachers' willingness to pay
for indigenous genetic resources as inputs into commercial products. Biodiversity may
have important values over and above those as inputs into pharmaceutical research. Our
point is not that biodiversity has little value at the margin; it may give rise to a great
number of other ecological, moral, and esthetic values that are not captured in market
transactions. To the extent that the incipient markets for genetic resources will not
generate revenues adequate to support the preservation of endangered habitats, it is all the
more important that alternative means for financing conservation be developed.

I. The Value of Genetic Resources in Observed Transactions
One reason for which there is little evidence concerning the prices at which genetic
resources have traded is that they are non-rival goods and property rights in them have
typically not been well established [see Sedjo, 1992; see also Chichilinsky, 1993; and
Vogel, 1993]. The seminal contributions of Coase [1960] and Demsetz [1976; see also
Barzel, 1988] suggest that property rights will come to be established either de facto in the
form of contracts between parties or de jure when the benefits of their definition exceed
the costs of their enforcement. The legal and institutional treatment of indigenous genetic
resources is, in fact, changing. The Biodiversity Convention [UNEP, 1992] prepared for
the 1992 UNCED meetings in Rio de Janeiro and recently signed by the United States
guarantees states sovereignty over their genetic resources and forbids their appropriation
without prior informed consent. Organizations in many countries are now entering into
commercial agreements with foreign pharmaceutical researchers. The most noted of these
is probably that signed between Merck and Company, a large U.S. pharmaceutical firm,
and Costa Rica's Instituto Nacional de Biodiversidad (INBio). This agreement calls for a
fixed payment of some one million dollars and premises of substantial royalties in the
event of new product discovery [Sittenfeld, 1993].
While institutional developments are indicative of a new enthusiasm and optimism
concerning the value of indigenous genetic resources, they provide little evidence
concerning the value of unimproved genetic resources in situ. "Markets" for transactions
in indigenous genetic resources are just beginning to emerge. While payments of between
$50 and $200 per kilogram for samples have been reported [Laird, 1993], the
interpretation of fixed payments for samples as a measure of the value of resources in situ
is suspect for at least two reasons. The first is suggested by our dicussion above: it is not
entirely clear that the collector has (or should have) legal title to the samples she sells. For
this reason, observed "prices" might be misleadingly low. The second reason is that sample
collection is typically a much more difficult process than it may appear at first. Payments

made for samples may reflect compensation for collection and processing labor and
taxonomic expertise rather than rents for the materials themselves.®
Compensation for access to samples is often not made in the form of simple cash
transactions, however. Many agreements specify royalty provisions rather than up-front
payments. Inasmuch as the terms of these provisions are generally secret, and the parties'
estimation of both the probability of discovery and the payoff in the event that a valuable
discovery is made are unknown, little can be inferred about the value of resources in situ
from public information concerning these contracts. For these reasons, most existing
attempts to estimate the value of indigenous genetic resources for pharmaceutical research
have been based on inferences from indicators other than observed transactions.
II. The Use of Indigenous Genetic Resources in Pharmaceutical Research
Indigenous genetic resources are the genetic codes containing the "recipes" for
chemical compounds of potential value m pharmaceutical products. These recipes can be
exploited for commercial purposes by acquiring a breeding stock of the organism that
produces the desired compound transplanting genes, or using the naturally occurring
compound as a model for the synthesis of the same or related compounds. Pharmaceutical
research on natural products is more often intended to develop "leads" than to identify
natural products that can be used in an essentially unmodified form. Leads are promising
molecules: blueprints of compounds that may show promise in their naturally occurring
form, but must be modified to increase efficacy or reduce side-effects.
Part of the reason for the increased recent interest in natural products research is a
renewed appreciation of the importance of natural leads. While considerable efforts at
"rational design" of drugs from inorganic materials continue, researchers have also come
8 The Merck-INBio agreement illustrates this point. Of the million-dolloar up-front payment, less than ten
percent was designated for conservation activities. The remainder went for equipment purchases and to
defray INBio's expenses [Sittenfeld and Gamez 1993].

to recognize that nature has perfected chemicals that synthetic chemists might never dream
up [Reid et al., 1993]. Wild plants and animals have evolved elaborate chemical means to
enhance reproductive success, deter predators, and resist infection. These chemicals may
have great promise in pharmaceutical applications.
The development of new drugs from indigenous genetic resources proceeds in
many steps and may take ten or more years from the time a promising lead is discovered
to the first commercial sales of derived compounds. The process begins with field
collection. It is important that collection be undertaken by trained taxonomists;
appearance and location must be carefully recorded so that finds will be replicable.
Samples are next dried and ground. While these processes may sound straightforward,
they must also be performed to tight tolerances. The next step is typically to extract active
compounds with a chemical solvent. Extracts are then tested to determine activity for
certain purposes. These tests, or assays, are today typically performed in vitro in a matter
of minutes, and are intended to determine if a certain chemical reaction occurs.
Once products with promising properties are identified, their active compounds
must be isolated. These isolated active compounds may then be "optimized" that is,
chemically modified to increase efficacy or reduce side-effects. Experimental drugs are
subjected to several rounds of clinical trials, which may, of course, be terminated at any
point if it is determined that the research is unlikely to be successful. Production planning,
patent application, and pursuit of regulatory approval maybe conducted concurrently with
other activities. Finally, if tests have beta successful and regulatory hurdles cleared,
commercial sales may begin.
III. Value and Redundancy in Indigenous Genetic Resources
In this paper we seek to determine the value of indigenous genetic resources in situ
for pharmaceutical research, and, by extension, the incentives that might be created by

pharmaceutical research for the preservation of undisturbed habitat; we derive a demand
curve for indigenous genetic resources and their habitat. We then determine from this
demand curve the willingness to pay for the "marginal SpCCifiS."'
In deriving this demand curve we must consider not only the likelihood that useful
products will be found in one sample, but that they will be duplicated by other finds. The
marginal value of genetic information for medicinal purposes is measured by its
contribution to the improvement of available health care. For example, the value of a new
cancer treatment is determined by its capacity to improve remission rates, reduce side
effects, lower costs, and so forth. A new drug that maybe effective but is identical or
inferior to an existing treatment is of little value. While the discovery of a novel
compound may not often prove completely superfluous it is often the case that one
product will largely duplicate another, or that discovery of one effective compound will
reduce the urgency, or even eliminate the need to continue research on others. '®
The essence of the argument we will make more formally below is that regardless
of the probability with which the discovery of a commercially useful compound may be
made, if the set of organisms that may be sampled is large, the value of the marginal
species may be very small. At any given time, researchers will be searching for
compounds effective in particular application. If the probability that a species chosen at
random will yield an effective compound is high, the probability that two or more species
® We will, for want of a better index, treat "species" as the basic units of genetic differentiation. It would
be inaccurate to suppose that the all species are seperated by the same degree of genetic variation. It is
common, however, to consider the species both as the basic unit of biological diversity [Wilson, 1992] and
of economic value.
I® This point is illustrated by taxol, a drug derived from the bark of the pacific yew tree that is used to
combat ovarian cancer. Though perhaps the most important anti-cancer find in recent years, the drug
provides only an incremental improvement in our ability to treat the disease. Comparing it to the most
effective alternative treatment, while taxol has in some tests shrunken tumors in a higher proportion of
women for a few more months, is has more severe side effects, costs three times as much, and has not
conclusively extended lives. "The Aura of Miracle Fades from a Cancer Drug," Gina Kolata, New York
Times, November 7, 1993.

will be found to do so is also high. To the extent that additional species from which to
sample are likely to be redundant, their marginal value will be low. Conversely, if
potentially valuable compounds are so rare as to make their discovery in two or more
species highly unlikely, the probability of their discovery in any species will be unlikely.
We will treat these issues more formally below; we note in passing, however, that
there are several reasons for which redundancy of genetic resources may be relatively
common. First, individuals of the same species maybe redundant. The same species may
be found over a wide range. If all representatives of a species produce a particular
compound, individuals in excess of the number needed to maintain a viable population are
redundant. Second, there are numerous instances in which identical drugs, or drugs with
similar clinical properties, have been isolated from different species [Farmsworth, 1988]. "
To give a recent example, the discovery of the anti—cancer drug taxol m the Pacific Yew of
Western North America has set pharmaceutical researchers looking for similar compounds
in its old-world relatives.^ Given the numerous examples of parallel morphological
development in the evolution literature, it should not be surprising to find that different
organisms that have evolved in similar ecological nicks have developed similar chemicals.
Finally, there is a dimension of what we might label clinical or medicinal,
redundancy. Very different compounds, perhaps even drugs working through different
mechanisms, may be effective in treating the same set of symptoms. Moreover, while the
inventiveness of nature in developing useful compounds is much extolled as a factor in the
increased demand for natural products for pharmacological research [Findeison and Laird,
1991], it is possible that synthesis from non-organic sources would yield substitutes for
natural product leads.
'' It may also be the case that there are a host of other sources of common compounds that remain
undiscovered because current sources are adequate.
See, e. g., "A New Cancer Drug May Extend Lives - at Cost of Rare Treees," Marilyn Chase, Wall
Street Journal, April 9,1991.

IV. A Simple Model
In this section we derive a simplified demand function for indigenous genetic
resources in pharmaceutical prospecting, determine the maximum willingness to pay for
the "marginal species," and consider the sensitivity of the value of the marginal species to
the probability of discovery and assumptions concerning overall profitability. We begin
with a very simple model. Suppose that medical researchers have identified a need for a
new product. A new product, if successfully developed, will earn net revenues of R. R is
assumed to be net of production, advertising, and marketing costs, but gross of any costs
of product research and development (i. e., costs of determining whether or not a natural
material will in fact lead to a commercially successful product). These costs of research
and development will be denoted by c.
Suppose that there are n species of organisms that may be sampled in the search
for the new product. Suppose further that p is the probability with which any species
sampled at random yields a successful commercial product. We treat each new sampling
as an independent Bernoulli trial with equal probability of success. Testing for a particular
application ends with the first success: once a successful product is found, further
discoveries would be redundant. Thus, the value of the entire collection of n samples is
V(n) = p/?-c+(l-p)[pi?-c] + (l-p)2[/7/?-c]+ ... +(l-p)"~I[pl?- c]
. i^£[l-(!-„)¦].	(1)
That is, with probability p, the first organism tested yields a commercially successful
product and the search ends. With probability 1 -p, the first organism tested does not yield
a successful product and the second organism is tested, and so on. If none of the n
organisms tested yields a commercially successful product, search ceases.
What is the value of the "marginal species?" In other words, how much does total
expected value increase with the addition--or decrease with the loss--of a species that

could be tested? The increase in total value to be realized by the preservation of an
additional species is
V(n + 1) - V(n) =	_ ££z£[i_(1_p)»]
=	(2)
we will abbreviate this expression for the value of marginal species as v(n) in what follows.
Note the straightforward intuition underlying expression (2): the value of the marginal
species is the expected payoff in the event it is sampled pR - c, times the probability with
which search is unsuccessful in the set of n other species, (1 - py-
Obviously, the buyer must believe that pR-c > 0 if any sampling is deemed
worthwhile; on the other hand, as p becomes larger the magnitude of (1-pY1 declines more
quickly than than of pR - c increases. In what follows, we describe how the value of the
marginal species varies with the probability of success m any given trial. We derive two
main results in this section. First, one must make optimistic assumptions in order to
believe that the value of the marginal species is very large even if the probability of success
in each trial were that which maximizes the value of the marginal species. Second, the
function relating the value of the marginal species to the probability of success in any
given trial is sharply peaked. With large numbers of organisms from which to sample, not
only is the maximum value of the marginal species low, but the value also falls off
steeply if the probability of success differs even slightly from the maximizing probability.
Differentiate (2) with respect to p to find that
^ - n(pR ¦ c)(l-p)- +	*(1 -/>)•
= [ R- c - (n+l)(p/?-c)](l-p)"'1 = 0	(3)
when p is chosen to maximize v(n).
The second-order condition for a maximum requires that

d2 v
^-r = — (« —1)[/? - c - (n + l)(/7^-c)](l-/7)""2 - (n + l)/?(l-p)"_l £ o.
As the satisfaction of the first-order condition requires that the expression in square
brackets is zero at the maximum the second-order condition is satisfied. It is also easy to
see that there is only one extreme point on the interval [0, 1], so the probability that
maximizes the value of the marginal species is unique.
The first-order condition may now be expressed as
* D	R ~ C
P R ~ c = ——,
. R + nc 1 n c
D* = 	 a 	 +		(A)
{n+l)R /i+l n+lR
The restrictions that p*R - c>0 and p* c."
Using (4), we can derive the maximum possible value of v, which we will call v*:
R—c( R—t
v* * v(n)
p• H+u
" T - i
'-c n V
R «+lJ
The approximation I 	• I "» — (where e is the base of the natural logarithm,
\ /i+l/ e
approximately 2.718) is very accurate for values of n on the order of those we are
considering for wild species. Incorporating this approximation, we have
R-c (R-cY
v* ' 7—rH —I •	(6)
Expression (6) still involves a number of variables concerning whose magnitudes
and relative magnitudes we have not yet said anything. At this point we can see, however,
that it is entirely possible that the maximum possible value of the marginal species could
be insubstantial. As n grows large, v* will be small for even relatively small values of c.
This is true for two reasons. The first is the re + 1 in the denominator of (6). The second
1 3 Of course, we would expect R » c; the value of a proves discovery substantially exceeds the cost of

is that 	 is raised to the nth power in (6); for large values of n, this expression will
become quite small for even moderate values of c relative to R.
It is also revealing to express (6) in another way. From (1), we can define the
expected revenues of a program searching for a particular product as
n —	— (l — p) j, and the total expected costs as K = —£l - (1—¦/?) j. We can
then rewrite
R-c 1 pK
R	n'
Using (4) to evaluate this expression at p*, we find
^ (n + l)n — nK
V *
For large n, we have approximately
R-t "

and the maximum value of the marginal species is approximately
v*(n) » 7^T7:«n~*	(7)
AS K approaches II, v*(n) again approaches zero. In short, the value of the
marginal species can only be high if the expected aggregate profitability of the research
venture is high. In Figure 1 we illustrate this relationship.
It also bears mentioning both that the marginal species takes on its maximum value
at a probability relatively close to that at which prospecting "breaks even" and that the
value of the marginal species declines relatively rapidly with respect to probability after
having reached a maximum. Recall that prospecting is only profitable in expectation if
14 The curve in Figure 1 quickly approaches a linear relationship; recall from (7) that
For R»K, the exponential term is almost constant, so the linear term in R - c dominates.

pR - c > 0, i. e., p > R/c. Our statements about relative closeness maybe made more
concise if we define a basic unit
, R 1 R-c
" 7 = 7m—-	(8)
Note that U. is necessarily less than	.
If we now consider v, the value of the marginal species, as a function of p, the
probability of success in any given trial (fixing n), it follows that v(p* - |i ) = 0. More
/ *. \ / .^R — c( n — m R — cY
v(p' + m(l) = (m+l,—
For large n, the approximation
( * , \ R - c m+if R - cX
is very accurate. Thus, to a very close approximation
v(p* + m\L) » v(p*).	(9)
The shape of this function is illustrated in Figure 2; it is, of course, the same as the
graph of (pR-c)(\-pY- Note the extreme concentration at the function's peak. Recall that
1	10
< 	; thus, on an interval of length less than , vfn) varies from 0 to its
n + 1	n+1
maximum value to 10r» = 0.0012 times its maximum value, p* itself is greater than
	. If, as seems likely, a researcher cannot predict the probability with which she
n + 1
anticipates success in any given sample evaluation within an order of magnitude ex ante,
her expectation of the value of the marginal species is likely to be very low.
V. Some Specific Examples
It is impossible to estimate the value of marginal species with any precision. Even
deriving an estimate for its maximum possible value is a highly speculative exercise. We
can, however, get some idea as to the magnitudes involved by using some data from the

pharmaceutical industry. While our estimates are little more than back-of-the-envelope
calculations, a more careful treatment might well yield still lower numbers.
In order to relate our model to real-world data, we must aggregate over all
possible discoveries. Some of what we believe to be the excessive enthusiasm for the
potential of genetic prospecting as a conservation strategy stems from an unrealistic view
of the number of products" to be generated from prospecting activities. " One rarely finds
things for which one does not look. Genetic prospectors subject samples to a limited
series of tests at any given time. While the history of science records many serendipitous
discoveries, they are the exceptions. It would be difficult to come up with a figure for the
number of applications for which species are tested, whatever that number, however, we
do have statistics on the numbers of new products developed. We should require as a
reality check that the probability of discovery times the number of applications for which
tests are performed not vastly exceed current numbers of new products developed. ^
We will suppose that there exist a series of "potential products" that might be
derived from genetic resources. Potential products might be regarded as cures for
diseases. The demand for them may arise as new infectious diseases become
widespread," as demographic characteristics change and the health needs of certain
groups become more important,19 or as new technologies are developed.20 We label these
13 We do not treat agricultural and industrial applications here. Casual empiricism and conversations
with researchers suggest that the value of the marginal species for these purposes may be much lower still,
as a still greater number of substitute research opportunities may be available (in agricultural research, for
example, pest-resistant strains can often be developed from the large number of very close-often of the
same species-relatives of cultivated varieties).
16 Conversations with researchers suggest that on the order of one hundred tests or less are done on
species for their pharmaceutical potential.
If more thorough genetic prospecting activities did in fact yield a deluge of new products we would
have to wonder again if the marginal new product were of any appreciable value.
" For example, the AIDS virus was not identified until the 1980s.
" The aging of the population and the increased need for geriatric care are good examples here.

as potential products, as there is no assurance that solutions to newly identified needs can
actually be found. It is not unreasonable to suppose that new potential products are
generated by a Poisson process with parameter X. Then, in expectation, X potential new
products will be identified every year. We will suppose that X remains constant overtime:
potential new products are identified at a more-or-less constant rate.
We might suppose that each new potential product j identified at time I would
have a stream of revenues net of research and development costs denoted by Rp
Similarly, we could say that the cost of evaluating the potential of the ith species for its
use in deriving the /(h potential product at time t is a random variable *fr It is not
unreasonable to assume, at this level of detail, that all the R's and c's are statistically
independent and denote the expectation of each as R and c, respectively. If future returns
are discounted at a constant rate r, the expected value of the marginal species is simply
£X(1+r)'(pR - c)(l-p)m = —(pR - c)(l-p)\	(10)
r-0	r
As was noted above, if we are considering extremely large numbers of species, the
value of any one species must be negligible. While biologists are unable to specify the
number of living species to within even an order of magnitude, a reasonable lower bound
would be ten million specks. The "base case" estimate we report below would have been
reduced by forty-one orders magnitudes if we had assumed that all of ten million
species were equally likely to yield a successful product.
Let us, therefore, narrow the range of species over which we consider searching.
Some have argued that phytochemicals-compounds produced by higher plants-have
exceptional pharmaceutical potential [see, e. g., Joffe and Thomas 1989]. These
compounds may be unlikely to be produced by other types of organisms, and may have
substantial pharmaceutical value. Aspirin, quinine, and the anti-cancer drugs vincristine,
For example, the demand for immunosuppresant drugs has increased greatly as a result of the progress
that has been made in organ transplant surgery.

vinblastine, and taxol are all derived from higher plants. There are estimated to be at least
250,000 living species of higher plants [Myers, 1988; Wilson, 1992]
We will consider the value of the marginal species of higher plant assuming that p
is chosen so as to maximize that value. Regrettably, there are no reliable estimates of the
parameters X, R, or c each might be inferred indirectly from knowledge of aggregate
industry success rates, revenues, and costs, however. We will ask what the values of the
parameters we seek would be if observed data were generated by the probability of
success that maximizes the value of the marginal species.
Between 1981 and 1993 the U. S. Food and Drug Administration approved an
average of 23.8 new drugs per year [PMA, 1982-1994]. This rate was relatively stable
(see Table 1), varying between 14 in 1983 and 30 in 1985 and 1991. There is no
discernible trend in the data. As new drug applications include both compounds first
approved in the U.S. and subsequently sold to the rest of the world, as well as drugs
already sold elsewhere but just being approved in the U. S., we take these figures to be
representative of world discovery rates.
About one third of all prescription drugs are derived from higher plants
[Chichilnisky, 1993]; we will assume that ten new drugs per year are expected to be
discovered from investigating higher plants. The expected number of new products
developed per year is the expected number of new potential products identified, X., times
the probability with which a successful commercial product is developed, i- (1 -PY-
Di Masi, et al. [1991] estimate pharmaceutical research and development
expenditures per successfully derived product to be $231 million. A recent report
suggests that "a reasonable upper bound" on the figure is $359 million [OTA 1993]. We
Farmsworth [1988] places the number at between 250,000 and 750,000, so our estimates of the value of
the marginal species should again be biased upward.

will assume a value of $300 million for our calculations. In our notation the R&D cost
c	K
per successful product developed would be expressed as — as	. .
P 1 — (1 -P)
We summarize some data relating net revenues to R&D costs for major
pharmaceutical companies in Table 2. We assume that marketing and administrative costs
vary in proportion to the number of products marketed, so we define net revenues as sales
less production costs and marketing and administrative costs.
This data cannot be applied directly, however. In our model we have assumed that
samples arc evaluated, costs are incurred, and revenues received instantaneously. In the
real world of course, these things occur overtime. Let us consider, then, a
pharmaceutical company that earns a stream of revenues from products of various
vintages. For simplicity, suppose that products differ only by their dates of discovery;
each product of the same age earns the same net revenues (in expectation) regardless of
when it reaches that age. Hi (l-PY\ is the number of products expected to be
developed in any given period and let be the expected net revenue received by a product
of age t. Then the total expected net revenues of a firm of age T will be
If we assume that net revenues of older products eventually decay and the firm is
sufficiently old, the firm's total expected net revenues should be constant over time under
our assumptions.
The expected present value of the net revenues of products developed in period T
will be less than the value of its current receipts, however, as these revenues will not be
received immediately. That is,
*[.-(!-„)¦]* -
A reasonable specification of the ^t's might be to suppose a stylized model of patent
protection. Suppose that new products are the exclusive property of their inventors for T

periods, during which constant expected net revenues of  are received. After the
expiration of the patent we will suppose that all profits are competed away. Under these
assumptions we would find that
It is clear that	 is less than one. To give some idea of general magnitudes, if
r = 0.10 and T = 17--values that might be assumed in consideration of pharmaceutical
, tt o 1 ""
company discount rates and patent law m the U.b.	would be about 0.49.
We might also do a similar correction for the timing of research expenditures; even
in a steady state, a firm's current R&D expenditures overstate the expected present value
of its expenditures on products under development, as the latter will be incurred in the
future, and hence, discounted. The most favorable assumption that we could make on
costs would be that they are all incurred at the last possible moment, however. All R&D
costs are, by definition, incurred before a product is marketed, so revenues are not
received until all costs are incurred. Thus, if we discounted from the time at which
research begins until costs are incurred, we would also want to discount from the time at
which research begins until revenues begin to be received. These would be offsetting
corrections, however (we care about the ratio of total expected costs to total expected
Combining all these considerations it seems generous to suppose that an
investment in pharmaceutical R&D pays a fifty percent return. If the cost per successful
product developed is $300 million, then, we will suppose that the net revenue is R = $450
million. Finally, we will suppose that pharmaceutical firms discount future returns at ten
percent per year.
The results of an exercise based on expression (6) and these assumptions are
summarized in Table 3. Our assumptions imply that the probability of hitting on any given
species for any given potential product that maximizes the value of the marginal species

would be about twelve in a million. Over an entire collection of 250,000 species from
which to sample the probability of making a hit is slightly over ninety-five percent. The
expected cost of evaluating a sample is around $3,600. The maximum possible value of
the marginal species is slightly less than $10,000.
We must emphasize that these estimates are extremely sensitive to changes in
assumptions, however. Recall that we have evaluated the marginal species at that
probability of success that maximizes its value. The results reported in Table 3 indicate
that p*= 0.000012. If we continue to assume that c = $3600 and R = $450,000,000, but
allow p to vary, we may get very different results. We must have p <£ 0.000008 in order to
have the expected value of conducting any test be positive. From that level, however, the
value of the marginal species quickly increases to the peak at $9,431. If p were to
increase further, to 0.000040, the value of the marginal species declines to only about $67.
If p were an order of magnitude greater than p*--but still only on the order of 10"*~the
value of the marginal species would plummet to less than $0.0000005!
The second assumption that can make a great deal of difference in our results
concerns the relative magnitude of net revenues and costs. In our base case scenario we
assumed that expected net revenues exceed expected research costs per successful new
product derived by fifty percent. If we assumed instead that expected net revenues
exceed expected costs per successful product by twenty-five percent, the value of the
marginal species would be only $1,017.53; if expected net revenues exceed expected costs
per successful product by ten percent the value of the marginal species would be $2.20.22
We will seem the next section that even numbers on the magnitude of $10,000
may translate into very limited incentives for the preservation of threatened habitats. It is
22 Of come, if we assumed that net revenues exceed expected costs per product developed by a wider
margin, we would obtain greater values for the marginal species. At a certain point however, these
results become implausible for other reasons; we should not expect the overall profitability of the industry
to reach unlikely levels.

worth emphasizing again, however, that we have generated values of that magnitude only
under what we regard as generous assumptions. We do not claim to have proved that the
marginal species is necessarily of negligible value; extremely fortuitous circumstances may
combine to create greater values. Our results do suggest, however, that only very
optimistic researchers might demonstrate a substantial willingness to pay.
VI. Incentives for the Conservation of Endangered Habitat
We have concentrated to this point on efforts to evaluate the worth of the
"marginal species." We are, perhaps, past due in Mining this concept and justifying its
importance. Economists should be familiar with the notion of valuing resources on the
margin but maybe uncomfortable with applying marginal analysis man ecological
context. How can one identify the marginal element of a large and complex ecosystem?
We will elaborate on our assumptions in this context in a moment; it suffices to say for
now that we will assume that the number of species in an ecosystem declines as a
continuous function of habitat loss.
It is important to note, however, that we are addressing explicitly only questions
concerning the value of the marginal hectare of land on which the marginal species grows.
That is, we are concerned only with matters of land conversion. Other human impacts
may be more widely felt. The introduction of exotic species, the release of pervasive
pollutant, or the effects of global climate change may have devastating impacts on
biological diversity. A marginal analysis maybe inappropriate for the consideration of
such phenomena. In the event of apocalyptic ecosystem collapse, however, the lost
potential for pharmaceutical research might well be the least important of our worries.
Much of the current concern with respect to the extinction of species arises from
the destruction of habitat. There is an extensive literature on the relationship between
habitat area and the richness of species. We will employ a widely used model in the
ecological literature, advanced by Preston [1960; 1962] and incorporated by McArthur

and Wilson [1967] in their influential theory of island biogeography. While this model has
been widely criticized by ecologists [See for example, Simberloff and Abele, 1982;
Boeklen and Gotelli, 1984; and Zimmerman and Bierregaard, 1986] for its inability to
predict the viability of individual populations and its resultant lack of utility in refuge
design, its predictions are likely to bias the estimate of the value of the marginal hectare
Upward,23 and for this reason we will employ it. We might also note in passing that it is
generally species-areas relationships that are employed to generate even the more
apocalyptic estimates of impending biodiversity losses.
The theory of island biogeography predicts that the number of species, Ilj, in a
particular taxon found in an area of size Aj is given by
n, = Mf,	(li)
where Oh is a constant that measures the species richness potential of an area and Z a
constant whose value is approximately 0.25 [see e.g., McArthur and Wilson, 1967;
Preston, 1962; Wilson, 1988].
To infer the maximum possible value for the marginal hectare of land for genetic
prospecting, then, we can differentiate V[n(A)] with respect to A to find that
dv = dvdn
dA dn dA
dtij/dAf can be found by differentiating (11) with respect to A:
~ = Za,A?~l =	= ZDf,	(12)
oA	A,
where Dj is the species density, i.e., the number of species per unit
23 Island biogeography, as the name suggests, is based on the distribution of species in physically isolated
habitats—islands in mid-ocean, labs in large land masses, isolated mountaintops, and the like. The
degree to which habitat conversion by, for example, felling forests for agriculture, actually isolates
populations is much disputed [ace, for example, Lugo, Parrotta and Brown, 1993].

We can combine expression (12) with our earlier results presented in Table 3 to
estimate the conservation incentives that would arise in particular threatened habitats. If
we accept the figure of $9,431 for the value of the marginal species of higher plant, we
can translate this number into a figure for a pharmaceutical company's maximum
willingness to pay to conserve a marginal hectare. In Table 4 we have entered data on
Norman Myers's [1988; 1990] eighteen biodiversity "hot spots." We find that the greatest
willingness to pay might be on the order of $20 per hectare in Western Ecuador. In other
areas with less genetic diversity the willingness to pay would be considerably lower, on the
order of a dollar per hectare or less. Again, it should be emphasized that even these very
low estimates arise under optimistic assumptions concerning the probability of discovery
and expectations of profitability. Equally plausible conjectures concerning these
parameters would yield radically lower values.
VII. Caveats and Extensions
The simple model we have developed above and on which we based the numerical
exercises we have reported is unrealistic in several respects. In this section we consider
two ways in which it might be improved and how our findings might differ if a more
realistic-if less tractable-model had been specified. We then discuss how other sources
of uncertainty might affect our results. We conclude this section with some reasons for
which we believe the model presented in Section III nevertheless provides useful insights.
Sequential Testing
In the simple model specified above we treat the cost of testing each individual
species as a random variable drawn independently from the same distribution. In the real
world, of course, testing is a complicated and extensive process. The first test may be
very simple (e.g., the "test" may consist of determining whether or not a given species

belongs to a taxon considered likely to contain the desired compound), the next test
somewhat more complicated and expensive, and so forth.
Consider a simple example. Suppose that two tests are required to determine if a
sample contains the desired product. Suppose that the (expected) cost of the first test is
C| and that of the second CDenote by Pi the probability that a sample chosen at random
"passes" the first test and by Pi the probability that it "passes" the second. As before, let R
be the (expected) net revenues earned by a successful product--i.e., one that passes both
tests. Then the value of the marginal sample is the expected value of evaluating a sample
at random, net of expected testing coats, times the probability with which no successful
product is identified among the first n species sampled. That is,
v(«) - [piPj*-(c,+ACi)][l-(l-PiPa)"]-
Differentiating with respect to both Pi and Pj yields two first-order conditions:
[PiPiR-ici+PiC^np^l-PtPt)"1 + (p2*-c2)[l-(l0
[pxPi* - (ci +A*i)]»ViU ~ PiPi+ Pi1^} - U ~ P1P2)"] - 0
Suppose that both of these conditions hold. Multiply the first by p2 and the second by px.
As both expressions are equal to zero, we must then have
{PxPiR - ACj)[l - (1 - PxPlY] - PiPiR[l - (l - PiPi)"\or
pxc 2 = 0.
Obviously, Pj cannot be zero if the species is to have any value. If CjWeiC zero
we would have the problem we have already solved above, with p replaced by P{P2 and no
meaningful basis for regarding the probability as being separate. Thus, for C2 >0 we
conclude that the value of the marginal species is maximized ifp, = 1; that is, the
assumption that the first-order conditions are simultaneously satisfied is contradicted It is
easy to demonstrate that this result generalizes to any finite number of required sequential
tests. We conclude, then that the assumption that all sequential tests are compressed into

a single number denoting the expected cost of all testing does not bias our estimate of the
value of the marginal species downward.
Continued Search
Another way in which our simple model has not been realistic is in its treatment of
search following initial sampling successes. We have assumed that search stops after the
first success. As we have noted above, however, practice differs from this abstraction.
The identification of compounds of potential value in one species may lead to a continued
search for similar but more effective compounds in others. Let us consider how this
consideration might be incorporated in a more realistic model, what might be gained in
detail, and what might be lost in tractability.
A more realistic treatment might specify the payoff to a particular sample taken at
random as a random variable 0; Assume again that the cost of evaluating a sample--of
determining the realization of 0—is c. We can generalize the model we have presented
above by noting that, under reasonable distributional assumptions, once a realization of 0
in excess of some certain value, call it 0*. is encountered search will cease. That is, let
m be the distribution of 0 and (0, ©) its support (it is convenient-and realistic--to set
the lower bound of the support of 0 equal to zero: the pharmaceutical researcher cannot
be obliged to develop products of negative value). Suppose also that the 0's are
independently and identically distributed across species.
The expected gain to be realized from evaluating an additional sample given that
one of value x has already been identified is
|(0-x)/(0)d0 - c.	(13)
Denote by 0* that value of x for which (13) is exactly ZCfO.^*
^ Obviously, such a 0T i 11 exist if 0 is finite. More generally, we must require that there not be too
much mass in the right tail of the distribution of 0. It seems entirely reasonable to suppose that such a 0*
exists in our context.

Suppose that 9{n) is the greatest value of 9 encountered in a collection of size n
(i. e., 8(n) is the greatest order statistic in a collection of size n). Now we cart denote the
expected value of a collection of n species with respect to a particular potential product as
V(n) = ([l-F(0*)]£(0l0£0*)-c) + F(0*)([l-F(0*)]£(0l0 2>0*)-c) +
F(0*)2([l-F(0*)]E(0l0£0*)-c) + ...+ F(0*)"~l([l-F(0*)]E(0I0 k 9*)-c) +
F(9*)" e(0(/i)I 9(n) < 0*).	(14)
This expression is relatively straightforward--and similar to (1). Its mth (m n)
term consists of the probability with which the mth species yields a product so successful
as to obviate the need for further search, times the expected value of the product given
that it is sufficiently valuable that search is suspended less the cost of sample evaluation,
all times the probability that a product so successful as to motivate the suspension of
search is not discovered m the previous m - 1 species sampled. The final term is the
product of the probability that no species sampled yields a product sufficiently valuable as
to motivate the end of search and the expected value of the most valuable product found
in searching over all n species, conditional on none yielding a value greater than 0*.
Note that
F(0*)"e(0(/i)I0(/i)<0*) a J9nf(9)F(9)"~'d9,
as ni(9)F(9y-lb the probability density of the greatest order statistic in a sample of size n.
It is now straightforward to show that
v(/t) - V(n + 1) - V(») a ([l - F(0*)]£(0I0 2:0*) - c)F(0*)"
•• ¦
+ Jy(0)F(0)"'l[(«+l)F(0)-n]d0.	(15)
The term on the first line to the right of the equal sign is familiar from (2); it is (2), with p
replaced by 1- F(0*) and R replaced with E(0I0 ^ 0*). It is obvious that (2) and (15)
coincide when the distribution of 0 is sharply bimodal: if all "failures" are without
commercial value and the value of all "successes" are tightly clustered.

The question is, then, whether the value of successes are clustered. We believe
that they are likely to be. Continued search for pharmaceutically active compounds for a
particular purpose after one "successful" compound has been discovered is likely to be
geared toward finding other species in which the same or similar compounds are produced
more plentifully. In other words, continued search may be undertaken in order to lower
costs of production. Production costs are a relatively unimportant component of
pharmaceutical industry profits. Thus, large increments in value maybe unlikely to result
from subsequent discoveries.
Moreover, it must be remembered that we are asking what the expected value of
an untested species is at the margin and ex ante. Some additional testing maybe done
because conditional expectations of value are high enough to justify it. While variations in
chemical properties among related species may motivate continued search, the lion's share
of the value may be realized by finding an organism that serves to identify the taxon to be
the subject of further search. All organisms in the taxon may be fairly close substitutes for
this purpose. All organisms not in the identified taxon have a conditional value of zero.
Two Additional Sources of Uncertainty
While we have mentioned that R and c my be regarded as the expectations of
random variables, we have not dealt explicitly with* underlying stochastic expressions.
If we replace each by the corresponding random variable, it can be shown that the
maximum value of the marginal species in our simple model-expression (6)--is convex in
both. If we sum overall anticipated future potential products and evaluate the resulting
expression at the expectations of R and c, our estimate of the maximum possible
marginal value will be biased downward. This consideration does not greatly concern us,
however. As shown in figure 1, and explained in footnote 14, expression (6) is nearly
linear when profit margins are appreciable. The function is sharply curved only when
marginal values are negligible anyway.

Another source of unmodeled uncertainty may be more problematic. The
extinction of a species is the example par excellence of an irreversible (dis-)investment.25
It is well known [see, e. g., Pindyck 1991] that such investments should be made only
when their expected benefits exceed their costs by a positive differential. The size of this
differential is determined by the parameters of the stochastic process by which benefits
(and, in a fuller treatment, costs) are assumed to be generated. In particular, greater
uncertainty in the process induces a greater differential. This "option value" argument is
also often emphasized in the ecological and environmental literature on the value of
endangered resources for pharmaceutical research.
We do not propose to suggest a figure by which our earlier numerical examples
might be inflated in order to correct for this uncertainty. We will suggest, however, that
overall uncertainty may not be great. It is true that spectacular new medical needs are
identified from time to time. The sum of marginal values with respect to the various
potential products for which testing may take place might evolve considerably more
smoothly, however.
Other Extensions
We have just noted two ways in which our treatment of uncertainty may result in
estimates of the maximum possible value of the marginal species that are too low. It is
also likely that the sharply peaked shape of the value of the marginal species that are too low. It is
of the probability with which any species sampled at random yields a "hit" is an artifact of
our assumption that all "hits" are equivalent--although, inasmuch as we think this
assumption is approximately true, we regard our results as being highly suggestive as well.
There are some technological optimists who maintain that the premise of Jurassic Park is not far from
being realizable, but more sober estimates suggest that retreating extinct species will remain the stuff of
science fiction for the foreseeable future.

Other omissions and simplifications in our model have likely led us to overestimate the
value of the marginal species, however.
One of these omissions concerns timing and discounting. We have assumed that
different species are sampled sequentially, but that each is evaluated instantaneously. To
have inserted discounting in our simple model would not have complicated matters much;
it could be accommodated by multiplying our expression for the value of the marginal
species by a discount factor. If, as seems likely, it could take years before the marginal--or
"last"--species would even be evaluated, values would be considerably lower.
Of course, research does not proceed by evaluating all samples sequentially. In
practice, firms also decide in how much capacity they ought to invest. Firms with greater
research capacity can evaluate different species simultaneously. To evaluate a large
number of species simultaneously is to increase the probability with which redundant
expenses are incurred however.
Redundant expenses are one of the reasons for which a more realistic treatment of
market structure might also result in lower estimates of the willingness to pay for the
marginal species. Over and above the fear of being beaten to a promising lead by a
competitor, rivals may also dissipate values by overinvesting in research and development.
There are a number of models in the industrial economics literature [see, e. g., Loury,
1979; Brander and Spencer, 1984] in which firms innovate too fast--incurring too great an
expense--in an effort to finish first.
More importantly, our numerical example does not recognize the abundance of
potential sources of new pharmaceutical products. In constructing our numerical example
we have supposed that all the world's species--and more generally all possible research
opportunities-can be separated into those that might possibly yield a product and those
that definitely do not. We suspect that restricting our attention to higher plants is very
unrealistic. Major pharmaceutical products have been developed from a microorganism
first found in the soil of a Japanese golf course and from a spore that happened to float

through the window of a laboratory in New Jersey and contaminated an ongoing
experiment. Synthetic chemistry and other inorganic sources provide other alternatives.
The number of available substitutes maybe much higher than we have supposed.
Finally, we have not included Bayesian updating m our analysis. We have
supposed that researchers' beliefs concerning the probability that any organic source could
contain the product sought do not decline regardless of lack of success. To suppose that
downward revisions in expectations would not occur after an unbroken string of failures
would imply either a very optimistic investigator or one with a very pessimistic prior; if the
latter, one would have to wonder if search would have been undertaken in the first place.
VIII. Conclusions
We have developed a simple model of the demand for indigenous genetic resources
for use in pharmaceutical research. We have demonstrated that the upper bound on the
value of the marginal species-and by extension of the "marginal hectare" of threatened
habitat--may be fairly small under even relatively favorable assumptions. Moreover, the
value of the marginal species may be a very sharply peaked function of the probability with
which any species chosen at random yields a commercially valuable discovery. Finally, we
have argued that our model, even though it is very simple, may yet offer some important
insights into the real values that biodiversity prospecting might generate for conservation.
Even if the reader rejects all of our other assertions, we would argue that the
development of a model of the demand for genetic resources is an important contribution
in and of itself. The valuation of genetic resources for pharmaceutical prospecting is an
important issue in conservation policy. Despite numerous contributions from ecologists,
environmental advocates, and, recently, economists, there has not yet been any adequate
treatment of this subject. Whatever else the drawbacks of our study maybe, we have
modeled values with an eye to the importance of scarcity. In addition, several recent

papers have advanced economic theories of the measurement of diversity. In none of
these instances were these concepts reduced to monetary values, however.
We would also argue that our numerical examples merit serious consideration. It
is true that, by making very generous estimates of the profitability of the industry and
supposing very fortuitous realizations of the probability of discovery, one might generate
moderate estimates for the conservation incentives provided by genetic prospecting. One
would have to take a very rosy view to suppose that the probabilities of discovery happen
to be precisely those that generate the maximum possible value for the marginal species.
If one takes the more reasonable perspective that researchers have some subjective
probability distribution over the probability with which individual species sampled will
yield commercial products, it seems quite likely that the perceived value of the marginal
species will be miniscule. This view seems to be consistent with information concerning
observed transactions. This subject should be studied further, and the extensions we have
discussed above pursued, but we would not expect a reversal of the conclusion of our
analysis, however the value of the marginal species for use in pharmaceutical research,
and, by extension, the incentive to conserve the marginal hectare of threatened habitat, is
We should emphasize again in closing that none of our conclusions imply that we
should not be concerned with the problems of declining biodiversity.Our point is,
rather, that if the international community values biological diversity, it should be actively
seeking other alternatives for financing its conservation.
^ We should note in passing that the social value of the marginal species for pharmaceutical research
may be higher than the private, as a successful researcher cannot appropriate the entire surplus for new
drug discovery. This does not detract from our conclusion that private incentive to conserve endangered
habitats for Pharmaceutical research will not be great. We doubt, however, that even the social incentives
for this purpose would be large.

Aylward, Bruce A., 1993, "A Case Study of Pharmaceutical Prospecting," Part III of
Bruce A. Aylward, Jaime Echeverria, Liza Fendt, and Edward B. Barbier, The
Economic Value of Species Information and its Role in Biodiversity
Conservation: Case Studies of Costa Rica's National Biodiversity Institute and
Pharmaceutical Prospecting, Report to the Swedish International Development
Bard, Yorem, 1989, The Economic Analysis of Property Rights (Cambridge:
Cambridge University Press).
Boeklen, William J. and Nicholas J. Gotelli, 1984, "Island Biogeographic Theory and
Conservation Practice: Species-Area or Specious-Area Relationship?" Biological
Conservation 29,63-80.
Brander, James A., and Barbara J. Spencer, 1983, "Strategic Commitment with R&D:
The Symmetric Case," Bell Journal of Economics 14, 225-35.
Brown Jr., Gardner and Jon H. Goldstein, 1984, "A Model for Valuing Endangered
Species; Journal of Environmental Economics and Management, 11, 303-309.
Chichilnisky, Graciela, 1993, Property Rights and Biodiversity and the pharmaceutical
Industry: A Case Study," working paper, Graduate School of Business, Columbia
Cease, Ronald H., 1960, "The Problem of Social Cost," Journal of Law and Economics 3,
Demsetz, Harold, 1967, "Toward a Theory of Property Rights," American Economic
Review 57 (Papers and Proceedings), 347-59.
DiMasi, Joseph A., Ronald W. Hansen, Henry G. Grabowski, and Louis Lasagna, 1991,
"Cost of Innovation in the Pharmaceutical Industry," Journal of Health Economics
10, 107-42.
Farnsworth, Norman R., 1988. "Screening Plants for New Medicines." In Biodiversity,
E.O. Wilson and Frances M. Peter (editors), Washington, D.C. , National
Academy Press.
Farnsworth, Norman R., and D. Soejarto, 1985, "Potential Consequences of Plant
Extinction in the United States on the Current and Future Availability of
Prescription Drugs," Economic Botany 39, 231-40.

Findeison, Christina, and Sarah Laird, 1991, ""Natural Products Research and the
Potential Role of the Pharmaceutical Industry in Tropical Forest Conservation,"
report prepared by the Periwinkle Project of the Rainforest Alliance.
Harvard Business School, 1992, "The INBio/Merck Agreement: Pioneers in Sustainable
Development," NI-593-015.
Joffe, Steen, and Robert Thomas, 1989, "Phytochemicals: A Renewable Global
Resource," AgBiotech News and Information 1, 697-700.
Laird, Sarah A., 1993, "Contracts for Biodiversity Prospecting," Chapter Four in Reid et
al., 99-130.
Loury, Glenn C., 1979, "Market Structure and Innovation," Quarterly Journal of
Economics 93, 395-410.
Lugo, Ariel E., John A. Parrotta and Sandra Brown, 1993, "Loss in Species Caused by
Tropical Deforestation and Their Recovery Through Management," Ambio, 22,
MacArthur, R. and E.O. Wilson, 1967. The Theory Island Biogeography (Princeton:
Princeton University Press).
McAllister, D. E., 1991, "Estimating the Pharmaceutical Value of Forests, Canadian and
Tropical," Canadian Biodiversity 1, 16-25.
Myers, Norman, 1988, "Threatened Biotas: 'Hot Spots' in Tropical Forests," The
Environments list 8(3), 187-208.
Myers, Norman, 1990, "Biodiversity Challenge Expanded Hot-Spots Analysis:" The
Environmentalist, 10(4), 243-256.
Office of Technology Assessment (OTA), 1993, Pharmaceutical R&D: Costs, Risks, and
Rewards, OTA-H-523 (Washington U.S. Government Printing Office).
Pearce, David and Seema Puroshothamon, 1992, "Preserving Biological Diversity: The
Economic Value of Pharmaceutical Plants," Discussion Paper 92-27, CSERGE
Pharmaceutical Manufacturing Association (PMA), 1982-94, "New Drug Approvals,"
Annual report published by MPA, Washington, D.C.
Polasky, Steven, and Andrew Solow, 1993, "Measuring Biological Diversity," working

Polasky, Steven, and Andrew Solow, 1994, "On the Value of a Collection of Species,"
working paper.
Pindyck, Robert, 1991, "Irreversibility, Uncertainty, and Investment," Journal of
Economic Literature XXIX, 1110-1148.
Preston, F.W., 1960. "Time and space and the variation of species," Ecology, 41, 611-
Preston, F.W., 1962. "The canonical distribution of commonness and rarity," Ecology,
43, 185-215,410-432.
Principe, Peter, 1989, The Economic Value of Biodiversity Among Medicinal Plants
(Paris: OECD).
Reid, Walter V., Sarah A. Laird, Carrie A. Meyer, Rodrigo Gamez, Ana Sittenfeld, Daniel
H. Janzen, Michael A. Gollin, and Calestous Juma, "A New Lease on Life,"
Chapter 1 in Walter V. Reid, Sarah A. Laird, Carrie A. Meyer, Rodrigo Gamez,
Ana Sittenfeld, Daniel H. Janzen, Michael A. Gollin, and Calestous Juma,
(editors), 1993, Biodiversity Prospecting: Using Genetic Resources for
Sustainable Development (Washington World Resources Institute).
Rubin, Steven M., and Stanwood C. Fish, 1994, "Biodiversity Prospecting: Using
Innovative Contractual Provisions to Foster Ethnobotanical Knowledge,
Technology, and Conservation," Colorado Journal of International
Environmental Law and Policy 5, 23-58.
Sedjo, Roger A., 1992, "Property Rights, Genetic Resources, and Biotechnological
Change; Journal of Law and Economics 35, 199-215.
Simberloff, Daniel S. and Lawrence G. Abele, 1982, "Refuge Design and Island
Biogeographic Theory: Effects of Fragmentation," American Naturalist, 120, 41-
Sittenfeld, Ana, 1992, "Tropical Medicinal Plant Conservation and Development Projects:
The Case of the Costa Rican National Institute of Biodiversity (INBio)," Paper
presented at the Rainforest Alliance Tropical Forest Medical Resources and the
Conservation of Biodiversity symposium New York, N.Y
Sittenfeld, Ana, and Rodrigo Gamez, 1993, "Biodiversity Prospecting by INBio," Chapter
3 of Reid et al. (editors), op. cit.
Solow, Andrew, Stephen Polasky, and James Broadus, 1993, "On the Measurement of
Biological Diversity," Journal of Environmental Economics and Management 24,

United Nations Environment Program (UNEP), 1992, "Convention on Biological
Vogel, Joseph H., 1993, Privatization as a Conservation Policy (Melbourne: Centre for
International Research on Communication and Information Technologies).
Weitzman, Martin L., 1992, "On Diversity," Quarterly Journal of Economics CVII, 363-
Weitzman, Martin L., 1993, "What to Preserve? An Application of Diversity Theory to
Crane Conservation," Quarterly Journal of Economics CVIII, 157-84.
Wilson, Edward O., 1988, "The Current State of Biological Diversity," in Biodiversity,
E.O. Wilson and Frances M. Peter (editors) (Washington: National Academy
Wilson, Edward O., 1992, The Diversity of Life (Cambridge Belknap).
Zimmerman, B.L., and R.O. Bierregaard, 1986, "Relevance of the Theory of Island
Biogeography and Species-Area Relations to Conservation with a Case from
Amazonia," Journal of Biogeography, 13, 133-143.

Figure 1
1.3	1.4	1.5	1.6	1.7
Rollo of expected net revenues to expected costs

Figure 2
p* 0.000016	0.000024
Probability of success in any one Mai

Table 1
New Drug Approvals
Source: U.S. Food and
Drug Administration

Table 2
NAt revenues and R&D expenses of three major pharmaceutical companies


Materials & production costs
Marketing & administrative expenses

Research & development expenses

Materials & production costs
Marketing & administrative expenses

Research & development expenses
Bristol-Myers Squibb

Materials & production costs
Marketing & administrative expenses

Research & development expenses
Source: 1992 annual reports of companies

Table 3
Base Case Scenario
Number of species
Expected number of new products
Cost of developing a new product
Net revenue-to-cost ratio
Net revenue
Discount rate
Probability of a hit
Value of the marginal species

Table 4
Maximum wiMngness to pay to presetve a hectare of land in 18 biodiversity "hot spots"



of Plant


of Plant
Endemic to
Species per
(1000 HA)
To Pay
Western Ecuador
Southwestern Sri Lanka
New Caledonia
Western Ghats of IncHa
Atlantic Coast Brazil
Uplands of Western Amazonia
Cape Fkxlstic Province of South Africa
Peninsular Malaysia
Southwestern Aushala
Ivory Coast
Northern Borneo
Eastern Himalayas
Colombian Choco
Central Chile
California Floilstic Province
Source: My era (1988; 1990) and authors' calculations.