y R E S E A R C H A N D A N A LY S I S

Product Environmental
Life-Cycle Assessment Using
Input-Output Techniques
Satish Joshi
James Madison College
Michigan State University
East Lansing, MI USA

y
Keywords
automobile fuel tanks
Summary
design for environment (DfE) Life-cycle assessment (LCA) facilitates a systems view in
input-output analysis
environmental evaluation of products, materials, and pro-
ISO 14000
cesses. Life-cycle assessment attempts to quantify environ-
life-cycle assessment (LCA)
product evaluation mental burdens over the entire life-cycle of a product from
raw material extraction, manufacturing, and use to ultimate
disposal. However, current methods for LCA suffer from
problems of subjective boundar y definition, inflexibility,
high cost, data confidentiality, and aggregation.
This paper proposes alternative models to conduct
quick, cost effective, and yet comprehensive life-cycle as-
sessments. The core of the analytical model consists of the
498 sector economic input-output tables for the U.S.
economy augmented with various sector-level environ-
mental impact vectors. The environmental impacts cov-
ered include global warming, acidification, energy use,
non-renewable ores consumption, eutrophication, conven-
tional pollutant emissions and toxic releases to the envi-
ronment. Alternative models are proposed for
Address correspondence to: environmental assessment of individual products, pro-
Satish Joshi
313 Case Hall
cesses, and life-cycle stages by selective disaggregation of
James Madison College aggregate input-output data or by creation of hypothetical
Michigan State University new commodity sectors. To demonstrate the method, a
East Lansing, MI 48824 USA
[email protected] case study comparing the life-cycle environmental perfor-
http://www.msu.edu/user/satish mance of steel and plastic automobile fuel tank systems is
presented.
y
© Copyright 2000 by the Massachusetts
Institute of Technology and Yale University

Volume 3, Number 2 & 3

Journal of Industrial Ecology 95

y R E S E A R C H A N D A N A LY S I S

Introduction stone, coke, electricity, chemicals, alloys, rail-

road services, computer support , and so on,
Environmentally conscious decision making which in turn draw on resources from almost all
requires information about environmental conse- sectors of the economy, including the steel in-
quences of alternative products, processes or ac- dustry. In order to keep the analysis tractable,
tivities. Life-cycle assessment (LCA) is a most LCAs limit the scope of analysis only to the
systematic tool to analyze and assess environmen- major inputs at each stage, leading to problems
tal impacts over the entire life cycle of product. of subjective boundary definition and compara-
LCA involves tracing out the major stages and bility across studies. Moreover, data on input re-
processes involved over the life cycle of a product quirements and emissions for even such
covering: raw materials extraction, manufactur- truncated LCAs have to be collected from a large
ing, product use, recycling and final disposal, number of different suppliers leading to high
identifying and quantifying the environmental cost, time, and issues of data confidentiality and
impacts at each stage. The goal of LCA is to fa- verifiability. As a result, LCA is considered a
cilitate a systems view in product and process flawed tool that cannot deliver what it promises
evaluation (Lave et al. 1995; Miettinen and (Portney 1993; Arnold 1993; Lave et al. 1995).
Hamalainen 1997; Keoleian and Menerery 1993). To address the problem of subjective bound-
The LCA approach is widely recognized as a ary definition in conventional LCA, Lave and
useful framework and attempts are underway to colleagues proposed using economic input-out-
integrate life-cycle thinking into business deci- put analysis techniques in LCA (Lave et al.
sions. A major international initiative in this 1995). Economic input-output life-cycle assess-
direction is the series of environmental manage- ment (EIO-LCA) takes a top-down approach
ment standards (EMS) proposed by the Interna- and treats the whole economy as the boundary
tional Standards Organization, widely known as of analysis. Another strength of EIO-LCA ap-
ISO 14000. Standards being developed for in- proach is that economy-wide interdependencies
clusion under ISO 14000 include principles and in inputs are modeled as a set of linear simulta-
guidelines for conduc ting LCA for product neous equations. Although less common and
evaluation (Tibor and Feldman 1996). Similarly, limited in scope, the representation of input in-
the document “Guidance on acquisition of envi- terdependencies as systems of equations can be
ronmentally preferable products and services,” found elsewhere in the LCA literature. For ex-
prepared by the United States Environmental ample, Pento constructs input-output matrices
Protection Agency (U.S. EPA) to help imple- for the paper industry in his dynamic life-cycle
ment the President’s Executive order 12873, rec- inventory modeling (Pento 1997). Similarly, the
ommends LCA approa ch in all federal commercially available KCL-ECO LCA pro-
procurement (U.S. EPA 1995c ; Vigon et al. gram represents product life-cycle inventory as a
1993). Many eco-labeling and product take back set of linear equations of physical quantities
regulations in Europe require life-cycle environ- (Karna and Engstrom 1994).
mental analysis of products (Davis 1993). Lave and colleagues implement their eco-
Although conceptually simple and appealing, nomic input-output life-cycle assessment (EIO-
LCAs are difficult to carry out in reality. Produc- LCA) model using 498 ´ 498 commodity sector
tion of most typical products and materials re- direct requirements matrix publishe d by the
quires a large number of diverse inputs, which in U.S. Department of Commerce as a part of the
turn use many other inputs in their production. 1987 U.S. input-output tables (U.S. Commerce
Often there are interdependencies in inputs, 1994). They augment the direct requirements
which have to be modeled. Attempting to trace matrix with sector-level toxic release emission
all the direct and indirect inputs and associated factors and electricity use intensity. They ap-
environmental burdens all the way to ultimate proximate the product, whose LCA is required,
raw material extraction becomes an unwieldy by the corresponding commodity sector. For ex-
exercise. For example, steel making requires a ample, in comparing the life-cycle environmen-
large number of inputs including iron ore, lime- tal burdens of paper cups and plastic cups, they

96 Journal of Industrial Ecology

 R E S E A R C H A N D A N A LYS I S y

approximate paper cups by the industry sector ematical structure of the basic EIO-LCA. I then
“Paperboard Containers and Boxes” and plastic present alternative models for conducting LCA
cups by the industry sector “Plastic Materials of individual products; the development of vari-
and Resins.” The life-cycle toxic emissions and ous environmental impact matrices; an illustra-
electricity use from paper cup and plastic cup tive case study comparing steel and plastic
production are approximated by economy-wide automobile fuel tank systems, followed by con-
impacts to meet a given incremental final de- clusions and discussion.
mand of the corresponding commodity sectors.
EIO-LCA, as proposed, leads to a consistent
Mathematical Structure of
boundary definition. However, it is still subject
Economic Input-Output Life-
to several well-recognized limitations (Lave et
Cycle Assessment
al. 1995). First, the product of interest is ap-
proximated by its commodity sector in the na- Input-output analysis is a well-established tool
tional input-output tables with respect to input in economic analysis, where the interdependen-
requirements and environmental coefficients. cies across different sectors of the economy are
But the commodity sectors in the national in- represented by a set of linear equations. The core
put-output tables are broad aggregates that in- of the model is the inter-sectoral direct require-
clude a large number of products. As a result, ments (or technical coefficients) matrix denoted
current EIO-LCA is appropriate for comparing as a. An element aij of matrix a represents the dol-
aggregate, disparate products that are well ap- lar value of input required from sector i to produce
proximated by their commodity sectors, but not one dollar worth output of sector j ( i = 1...n, and
for comparing heterogeneous products within a j = 1...n). Let x represent the vector of total out-
commodity sector, or products that differ signifi- puts of the sectors. The exogenous change in final
cantly from representative output of the sector, demand for the output of these sectors is repre-
or completely new products. Second, EIO-LCA sented by a vector f. (For a more detailed descrip-
captures the upstream environmental burdens tion of input-output analysis, underlying
associated with raw materials acquisition and assumptions about the structure of the economy
manufacturing stages, but not those associated and limitations, refer to the work of Leontief
with product use and end-of-life options. (1966), Miller and Blair (1985), Hendrickson and
This paper extends the EIO-LCA approach to colleagues (1998); and the U.S. Department of
provide a flexible tool for comprehensive LCA of Commerce (U.S. Commerce 1994).
products. The core of the analytical model con- Because the total output of a sector is the sum
tinues to be the 498 commodity sector direct re- of final demand f and intermediate demand ax,
quirements matrix for the U.S. economy (i.e. demand as input requirement for producing
augmented with various sector-level environmen- the output of other sectors), the input-output
tal impact vectors. The environmental impacts system can be written:
covered are significantly expanded to include glo-
x – ax = f (1)
bal warming, acidification, energy use, non-re-
newable ores consumption, eutrophication, and The vector of sectoral outputs to meet a
conventional pollutant emissions. Alternative given exogenous demand f is obtained by pre-
models are proposed for environmental assess- multiplying (1) by [ I – a ]–1
ment of individual products, processes, and life-
x = [I – a]–1 f (2)
cycle stages by selective disaggregation of
aggregate input-output data or by creation of hy- The input-output technique can be extended
pothetical new commodity sectors. To demon- for environmental analysis.1 Suppose r is a k ´ n
strate the method, a case study comparing the matrix of environmental burden coefficients,
life-cycle environmental performance of steel and where rk j is environmental burden k (e.g. carbon
plastic automobile fuel tank systems is presented. monoxide emissions) per dollar output of sector
The remainder of the paper is organized as j; and e is the vector of total environmental bur-
follows: in the next section I describe the math- dens, then the economy-wide total (direct and

Joshi, Product Input-Output Life-Cycle Assessment 97

y R E S E A R C H A N D A N A LY S I S

indirect) environmental burden associated with cients. (Approximation used in Lave et al. 1995;
an exogenous demand vector f becomes Cobas-Flores 1996; Horvath 1997). Direct and
indirect effects of incremental output of a par-
e = rx = r [I – a]–1f (3)
ticular product can then be estimated by treating
The environmental burden matrix r can in- it as a change in exogenous demand for the out-
clude coefficient vectors for any environmental put of the sector. The implicit assumption here
impact of interest such as energy use, non-renew- is that input requirements and environmental
able resource use, greenhouse gas emissions, etc. burdens are proportional to product price. This
The contribution of individual industry sectors to approach is simple, quick, and does not require
the total environmental burden can be found by any additional data. It provides useful informa-
replacing each of the environmental burden co- tion, especially when comparing broad industry
efficient vectors in r, by its diagonal matrix. sectors, or typical outputs of industry sectors. It
In theory, one can develop very large techni- is also useful as an initial screening device to pri-
cal coefficient and environmental burden matri- oritize further data collection efforts.
ces representing all possible products as separate
commodity sectors, and then use equation (3) to
Model II: Product as a New Hypothetical
estimate economy-wide burdens to meet an in-
Industry Sector
cremental demand for any product. However,
this approach is not practical in view of the im- Often, it is necessary to conduct an LCA of an
mense data requirements. The largest available existing product that is not a representative out-
technical coefficient matrix, published by the put of its commodity sector, or a completely new
U.S. Department of Commerce, has 498 com- product. If information on the inputs that are re-
modity sectors (U.S. Commerce 1994). quired to produce the product, and direct envi-
ronmental burdens from the production process
are available, then the product of interest can be
Models of Economic Input-
represented as a new hypothetical sector entering
Output Life-Cycle Analysis of
into the economy. The economy-wide environ-
Products
mental burdens arising from a unit output of the
Equation (3) gives the direct and indirect new industry sector can then be estimated.
environmental burden associated with an exog- Let a = [aij ] i, j = {1, 2.....n} be the current
enous change in the demand for the output of aggregate technical coefficient matrix of the
the sector as a whole. However, a typical product economy with n sectors. Assume the product is
LCA involves estimation of environmental bur- represented as sector n+1. Let a i,n+1 be dollar
den associated with an exogenous increase in value of input required from sector i {i=1...n+1}
the demand of a specific product, which might to produce a dollar worth of the product. Further
be different from the average output of an indus- assume that the inputs used in its production are
trial sector with respect to both technical and representative output of the sectors from which
environmental discharge coefficients, or a com- they are drawn. Let [ rn+1 ] be the column vector
pletely new product/design. I present several al- of environmental burdens associated with a dol-
ternative models, in order of increasing degree lar worth output of sector n+1.
of comprehensiveness, complexity, and data in- Define a new technical coefficient matrix A,
tensity for carrying out such product LCA. where the first n row and column elements re-
main unchanged from a,
Model I: Approximating the Product é a a ù
......

by its Sector A = ê . . . . .i .n+1. . ú

ë 0 a n +1 û
This is the simplest model in which the prod-
Similarly, the new environmental burden matrix
uct of interest is assumed to be well approxi-
is reformulated as:
mated by its industry sector, with respect to both
technical and environmental burden coeffi- R = [ r1 .....rn rn+1 ]

98 Journal of Industrial Ecology

 R E S E A R C H A N D A N A LYS I S y

The economy-wide environmental burden E The first n – 1 sectors of a and A are iden-
associated with an output fn+1 of the new sector tical because the original sector n in a is a lin-
n+1 is ear aggregation of sectors n and n+1 in A ,
E = RX= R [ I – A ]–1 F (4) ain = (1 – s) Ain + s A i n+1 (6)
where F t = [0,…..0, f n+1 ] where s is the share of product n+1 in the
Alternatively, one can consider that output total output of original aggregate sector n.
of the product results in exogenous increases in Similarly, the purchases of sector j from ag-
the demand for its input requirements in the gregate sector n in a is the sum of purchases
economy, i.e. the changes in exogenous demand from disaggregated sectors n and n+1 in A, i.e.
vector f1 is equal to the input requirement vec-
an j = An j + An+1 j (7)
tor for producing fn+1 dollars worth of output of
the product. The total environmental burden Also
associated with the product is then the sum of
ann = (1 – s)(An n + A n+1 n )
environmental burden associated with the pro-
+ s(An n+1 + A n+1 n+1 ) (8)
duction of its input requirements and the direct
environmental burden associated with its pro- Equations (6) to (8) are the constraints on
duction, fn+1 ´ rn+1 . (This is the “final demand the coefficients of A such that aggregation of
approach” outlined by Miller and Blair (1985)). sectors n and n+1 in A yields a. A has a total
of 4n unknown coefficients. However, because
E = r [ I – a ]–1 f1 + fn+1 ´ r (5)
n+1
there are 2n – 1 constraints on these coeffi-
Equations (4) and (5) are equivalent. cients, data on only 2n + 1 coefficients is re-
quired. The share s of the product of our interest
in the aggregate output of the original sector n
Model III: Disaggregating an Existing
can easily be obtained from secondary sources.
Industry Sector
The technical coefficient vector [ Ai,n+1 ] for the
The underlying assumption in Model II is that product of interest can be estimated from a de-
the original technical coefficient matrix is unaf- tailed cost sheet of the product, because the ele-
fected by the introduction of the new sector. ment Ai,n+1 is dollars worth of inputs from sector
However, this may not always be the case, because i required to produce a dollar worth of the prod-
most products of interest are already included in uct n+1. Data on the sales of product n+1 to
existing commodity sectors and often used as in- different existing sectors j of the economy is re-
termediate inputs in other sectors. Suppose that quired to estimate coefficients An+1,j. The firms
the total environmental burden associated with manufacturing the product typically have infor-
exogenous changes in demand for a product, mation on the markets and market shares for its
which is already included in one of the existing products.
sectors, needs to be estimated. The sector, say sec- Similarly if r = [ r1.....rn ], the environmen-
tor n, which includes the product of interest can tal burden matrix associated with the original
be disaggregated into two sectors, one of which is matrix is available, only data on the direct en-
the product of our interest (sector n+1) and the vironm ental burden vector R n+1 associated
other sector consisting of all other products with the product n+1 is needed, to obtain the
within the sector. For example, for conducting disaggregated environmental burden matrix R,
LCA of paper cups, the sector “Paperboard Con- because:
tainers and Boxes” can be disaggregated into two
rn = (1 – s) Rn + sRn+1 (9)
sectors, “Paper Cups” and “All Other Products in
Paperboard Containers and Boxes Except Paper Once the disaggregated matrices A and R are
Cups.” It means a new n+1 ´ n+1 technical co- developed, the total environmental burden asso-
efficient matrix A , with elements Aij has to be ciated with exogenous changes in the demand
derived to represent the same economy repre- for the product (i.e. output of sector n+1) can be
sented by n ´ n matrix a (with elements aij). obtained from equation (4).

Joshi, Product Input-Output Life-Cycle Assessment 99

y R E S E A R C H A N D A N A LY S I S

Model IV: Iterative Disaggregation, When The information on input requirements and
a Limited Conventional LCA Is Available direct environmental burdens from the manufac-
turing process of intermediate inputs available in
The approach outlined in Model III can eas-
the conventional LCA is incorporated as in
ily be extended and generalized when more de-
Model III to derive the larger disaggregated ma-
tailed information on more than one stage of the
trix. The expanded technical coefficient matrix
upstream inputs is available, typically in the
and environmental burden matrices can then be
form of a conventional LCA. An expanded
used to estimate the total environmental burden
technical coefficient matrix A and environmen-
associated with an increase in exogenous de-
tal burden matrix R can be developed by itera-
mand for the product/industry sector of interest.
tive disaggregation of relevant sectors and
In essence, this approach incorporates all the
creating additional industry sectors correspond-
detailed information available, approximates miss-
ing to each product or input for which detailed
ing information, and at the same time maintains
information is available.
the whole economy as the boundary of analysis.
To continue with the paper cup example, sup-
pose a conventional LCA process model provides
detailed information on the inputs and environ- Model V: Inclusion of the Use Phase
mental burdens from the paper cup production of Product Life-Cycle Assessment
process and production of wax paper used in pa-
Alternative designs of durable goods such as
per cup manufacture. Based on this information,
automobiles, home heating devices, and washing
the sector “Paperboard Containers and Boxes”
machines may differ not only in the environmen-
can be disaggregated into “Paper Cups” and
tal burdens in the manufacturing stage, but also in
“Others Except Paper Cups” sectors, and the sec-
resource consumption and environmental impacts
tor “Paperboard and Paper Mills” can be disaggre-
in the use phase of their life-cycle. With some sim-
gated into “Wax Paper” and “All Other Paper
plifying assumptions, the above models, especially
Mill Products Except Wax Paper” sectors. The
Model II, can be used to estimate direct and indi-
other inputs for which detailed process models
rect resource use and environmental burdens from
are not available can be approximated by their
the product use phase. The use phase can be
corresponding commodity sectors (which typi-
treated as a hypothetical industry sector that draws
cally are excluded in a conventional LCA).
inputs from the existing sectors and has some asso-
To present a general case, suppose a (with el-
ciated environmental burdens. The life-cycle bur-
ements aij) is the aggregate matrix, and a disag-
dens associated with all the inputs in the use phase
gregated matrix A (with elements Aijkl ), where
can be estimated by economy-wide impacts for a
each of the aggregate sectors has been disaggre-
given output of the hypothetical industry sector.
gated into k subsectors ( k may vary for each sec-
The underlying strong assumption in this approach
tor) is needed. Let sik be the share of output of
is that the technical coefficients matrix for the
the kth subsector in the total output of aggregate
economy remains unchanged during the lifetime of
sector i. Similarly let rik be the environmental
the product and the time discount rate for envi-
burden vector associated with a dollar output of
ronmental burdens is zero. This might be a good
kth subsector of aggregate sector i.
approximation when products have short life
Along the lines of constraints (6) to (9), fol-
spans. Even when the products have longer life
lowing are the constraints on the elements of A.
spans, this approach provides a good starting point
k
å si = 1 (10) by providing comparative information under the
k
current state of the technology.
aij = å sik Aijkl (11)
k
Model VI: Inclusion of End-of-Life
aij = å Aijkl (12)
l Management Options
The product EIO-LCA approach can be ex-
ri = å sik Rik (13) tended to analyze life-cycle environmental bur-
k

100 Jour nal of Industrial Ecology

 R E S E A R C H A N D A N A LYS I S y

dens from management options at the end of the grated with EIO-LCA, thus overcoming prob-
useful life of a product, such as reuse, remanu- lems of subjective boundary definition in con-
facturing, recycling, or disposal. End-of-life (EoL) ventional LCA and aggregation problems in
management options, such as recycling, use in- EIO-LCA. Further, these models are flexible. De-
puts from existing industry sectors to process an pending on the data and cost constraints, LCAs
obsolete product and produce an output which is of different levels of complexity and accuracy can
demanded either for final consumption or as an be conducted while consistently maintaining the
intermediate input by other sectors. Typically boundary of analysis as the whole economy with
remanufactured products are used by the same all its associated interdependencies.
industry sector as substitutes for new compo-
nents, while recycled output may be demanded
Development of Comprehensive
by completely different sectors. Hence, these op-
Environmental Impact Matrices
tions can be conceptually treated as additional
hypothetical industry sectors in EIO-LCA. We augment the 498 ´ 498 commodity by
The net environmental impact from EoL commodity direct requirements (or technical
management is the sum of life-cycle burdens from coefficients) matrix of the U.S. economy for the
inputs used in EOLM, direct impacts from EOLM year 1987, with estimates of various environ-
process itself, and credits for the recycled prod- mental burdens/dollar output of each commod-
uct. For example, the net environmental impact ity sector. Only publicly available data sources
from recycling aluminum beverage cans is the are used, which have advantages of large sample
sum of: life-cycle environmental burdens from all size, transparency of methods, verifiability, and
the inputs used in collecting and processing the periodic updating by public agencies. The gen-
cans, environmental burdens from the aluminum eral impact themes covered include:
recovery process itself, and the credits for the re-
• energy use,
cycled aluminum. When a recovered product can
• non-renewable ores use,
substitute the output of an existing industry sec-
• conventional pollutant emissions,
tor, the environmental credit for recovered prod-
• toxic releases,
uct is the avoided life-cycle environmental
• hazardous solid waste generation, and
burdens from this substitution. For example, if a
• fertilizer use (as an indicator of the
ton of aluminum scrap recycling results in the re-
eutrophication potential).
covery of 0.6 MT of aluminum valued at $500,
then the appropriate credit for recovered alumi- Impacts from individual pollutant emissions are
num is the avoided economy-wide burdens from aggregated and summarized using appropriate
a reduction of $500 in the final demand for the weighting factors such as global warming poten-
output of the primary aluminum sector. When tial, acidification potential, ozone depletion po-
the recovery process results in many products, or tential, and toxicity weighting factors. Similarly,
the recovered products are likely to be used by total energy use is derived by aggregating the
many other industry sectors, new industry sectors heating values of individual fuels.2 The values of
representing the EOLM options can be added to output of the six-digit U.S. input-output com-
the input-outp ut matrix. Appropriate adjust- modity (US-IO) sectors are from the benchmark
ments can be made for the technical and envi- input-outp ut accounts for the U.S. economy,
ronmental coefficients for the existing industry published by the U.S. Department of Commerce
sectors. However, in using this approach, the (U.S. Commerce 1994). The various data
temporal effects of the useful life of the product sources are summarized in table 1. More details
are once again assumed away. on the estimation procedures, data sources, and
The above models show how the EIO-LCA limitations are available (Joshi 1998).
framework can be extended to conduct LCAs of Fuel use and energy consumption: Data on pur-
individual products and life-cycle stages. The chases of different fuels by the six-digit US-IO
models also show how more product-specific data sectors are from the work files used in preparation
available in a conventional LCA can be inte- of the 1987 benchmark input-output accounts for

Joshi, Product Input-Output Life-Cycle Assessment 101

y R E S E A R C H A N D A N A LY S I S

Table 1 Summary of environmental impacts included in EIO-LCA and data sources

Environmental impact area Data sources
Energy Consumption
Fossil Fuel Consumption US-IO work files 1987 (U.S. Commerce 1994)
Electricity Consumption U.S. Census of Manufactures 1987
(U.S. Commerce 1987)
Non-renewable Ore Consumption US-IO work files 1987
Fertilizer Consumption (Eutrophication) US-IO work files 1987
Toxic Chemical Releases U.S. Toxic Release Inventory 1993 (U.S. EPA 1993b)
Toxicity Weighting Factors Horvath et al. 1995
RCRA Hazardous Waste US-RCRA database (Horvath 1997)
Conventional Pollutant Emissions from Fuel Use US-IO work files, U.S. EPA 1995a, and IPCC 1995
Global Warming Potential US-IO work files, IPCC 1995, USTRI 93
(U.S. EPA 1995b)
Acidification Potential US-IO work files, U.S. EPA 1995b, Adriaanse 1993
Ozone Depletion Potential USTRI 93 (U.S. EPA 1995b), Horvath 1997

the U.S. economy, available from the U.S. De- centration of the chemical that cannot be ex-
partment of Commerce (henceforth referred to as ceeded during any eight-hour work shift of a forty-
US-IO work files). These work files provide data hour work week, as per the occupational health
on the value of purchases of about 7,000 com- guidelines of the American Conference of Gov-
modities, including major fuels, by various indus- ernmental Industrial Hygienists (ACGIH). The
try sectors. The quantity estimates are based on CMU-ET weighting factors are included in this
average prices of individual fuels. The data on EIO-LCA model as first-level approximations to
electricity use by the industry sectors are from the the health effects of toxic emissions. While sev-
U.S. Census of Manufactures (U.S. Commerce eral other indices covering a range of toxic effects
1987). The total energy consumption is calcu- have been proposed, data on these indices are
lated by summing the energy content of different available only for a few chemicals compared to
fossil fuels and electricity from non-fossil sources. CMU-ET (Horvath et al. 1995).
Non-renewable ores use: Data on value of di- Fertilizer use (eutrophication): Eutrophication
rect purchases of various ores by different indus- of the environment involves disturbance of eco-
try sectors are available in the US-IO work files. logical processes in water and soil due to an ex-
The average producer price data for different cessive supply of plant nutrients in the form of
ores from the Minerals Yearbook (USBM 1988) nitrogen compounds and phosphates. As a result
are used to estimate consumption intensities of of eutrophication, plant species that thrive in
ores by different industry sectors. low-nutrient environments disappear. Also, in-
Toxic releases: The earlier EIO-LCA model de- creasing nitrate levels in ground water affect
veloped by Lave and colleagues (Lave et al. 1995; drinking water quality (Adriaanse 1993). The
Cobas-Flores 1996) included data on toxic main agents responsible for eutrophication are
chemical emission coefficients at sector level for phosphates and nitrates. The main sources are
over three hundred chemicals, estimated from application of fertilizers and manure to soil in
U.S. EPA’s Toxic Release Inventory (TRI) data- agriculture, and discharge of wastewater into sur-
base (U.S. EPA 1995b). Horvath and colleagues face water. The actual eutrophication effect from
(1995) proposed a toxicity weighting scheme for fertilizer use depends on application rates, fre-
aggregating TRI chemicals called CMU-ET, quency, runoff, soil conditions, topography,
wherein the time-weighted average threshold background concentrations, and distance and
limit values of various chemicals relative to the size of the water bodies. Fertilizer purchases/dol-
threshold limit value of sulfuric acid are used as lar output by different industry sectors are in-
weights. The threshold limit value is the air con- cluded in the environmental impact matrix only

102 Jour nal of Industrial Ecology

 R E S E A R C H A N D A N A LYS I S y

as an indicator of potential eutrophication. Ag- 1995b). These have been incorporated in the en-
gregate data on value of purchases of fertilizers by vironmental impact matrix in EIO-LCA.
industry sectors are extracted from the US-IO Acid rain and total acidification potential: The
work files. The price data are from the U.S. Cen- three main contributors to acid rain are sulfur
sus of Manufactures (U.S. Commerce 1987). oxides (SOx), nitrogen oxides (NOx) and ammo-
Conventional pollutants:3 As discussed earlier, nia. We include sulfur dioxide (SO 2) and NOx
the data on consumption of various fuels by indus- emissions from fuel combustion. The data on
try sectors are extracted from US-IO work files. ammonia releases are from the U.S. Toxic Re-
Average combustion emission factors for various leases Inventory (U.S. EPA 1995b). These emis-
fuels from U.S. EPA’s compilation of emission fac- sions are aggregated using their relative
tors (U.S. EPA 1995a) and IPCC guidelines acidification potentials as weighting factors. 1/2
(IPCC 1995) were used to estimate total emis- mole (32 grams) of sulfur dioxide has the same
sions of conventional pollutants from fuel use by acidification potential as 1 mole (46 grams) of
various sectors. The pollutants covered include nitrogen dioxide and 1 mole (17 grams) of am-
carbon dioxide, carbon monoxide, sulfur oxides, monia (Adriaanse 1993). Aeq is the aggregate
nitrogen oxides, methane, and volatile organic acidification potential in SO2 equivalents.
compounds. Non-fuel use-related process emis- Stratospheric ozone depletion: Chlorofluorocar-
sions are not currently included in the model. bons (CFCs) and bromofluorocarbons (halons),
Hazardous wastes: The U.S. law governing owing to their catalytic effect, trigger off a com-
hazardous waste management, Resource Conser- plex chain reaction that results in decomposition
vation and Recovery Act (RCRA), requires all of ozone in the stratosphere. The extent to which
large-quantity generators in the U.S. to report these chemicals contribute to ozone depletion
the quantities of hazardous wastes generated, depends on the atmospheric concentrations, resi-
managed on site, received from outside sources, dence times of these chemicals in the atmo-
and shipped to off-site treatment, storage, and sphere, and their ability to break down ozone.
disposal facilities. Sectoral intensities for hazard- These effects are summarized in terms of ozone
ous waste generation, on-site management, and depletion potentials (ODP) of different chemi-
off-site shipments using the RCRA database at cals relative to CFC11. ODPs are an integral part
U.S. EPA were developed by Horvath (Horvath of national and international considerations on
1997). These have been included in the EIO- ozone protection policy, including the Montreal
LCA environmental matrix. Protocol and its amendments and the U.S. Clean
Greenhouse gas emissions and Global Warming Air Act (Wuebbles 1995). The sectoral emission
Potential: The main greenhouse gases (GHG) are factors of ozone-depleting chemicals were esti-
water vapor, carbon dioxide, methane, nitrous mated from the U.S. TRI database by Horvath
oxide, chloroflurocarbons (CFCs), and halons. (1997) and were incorporated in the environ-
The degree to which GHGs contribute to the glo- mental impact matrix. The emissions were ag-
bal warming process depends on their concentra- gregated using their ODPs as weighting factors.
tion in the troposphere and on their ability to
absorb the heat radiated by the earth. This ab-
Application of Product Input-
sorption capacity is expressed as Global Warming
Output Life-Cycle Analysis:
Potential (GWP) relative to carbon dioxide.
Steel versus Plastic Fuel Tank
(IPCC 1995; Wuebbles 1995; Adriaanse 1993).
Systems for Automobiles
GWP of different pollutants are used as weighting
factors in aggregating GHG emissions. The emis- In this section, the use of the product EIO-
sions of carbon dioxide, nitrogen oxides, and LCA model is demonstrated with a case study
methane from fuel combustion were estimated as comparing the life-cycle environmental perfor-
detailed earlier. The TRI list of chemicals in- mance of steel and plastic automobile fuel tank
cludes most of the CFCs and halons. The data on systems.
the air emissions of CFCs and halons are from the Beginning in the late seventies, there has
1993 U.S. Toxic Releases Inventory (U.S. EPA been a significant movement towards more

Joshi, Product Input-Output Life-Cycle Assessment 103

y R E S E A R C H A N D A N A LY S I S

lightweight, composite materials in automobile steel. The designs of fuel tank systems can vary
components due to increasingly stringent fuel ef- depending on the application. These two are
ficiency and weight reduction goals. As a part of picked for illustrating the method. The analysis
this trend, attempts are underway to replace tra- can easily be modified to compare other designs.
ditional steel fuel tanks in automobiles with The life cycle of a fuel tank consists of four
tanks made from lighter plastic materials. It is main stages as shown in figure 1: fuel tank manu-
forecast that by the year 2005 almost 60% of all facture, use phase on the vehicle, shredding of
passenger cars and light trucks produced in auto hulk to recover ferrous scrap, and steel mak-
North America will have fuel tanks made from ing from scrap steel in an electric arc furnace
plastic materials (OSAT 1996). Modern multi- (EAF). For this analysis, we draw a boundary
layer plastic tanks meet the functional require- around each of these stages and treat them as hy-
ments and are cheaper. However, it is not clear pothetical commodity sectors which draw inputs
if they are environmentally superior from a life- from the economy and produce some undesired
cycle perspective. Because millions of tanks are emissions. The final recycling stage yields steel
likely to be substituted, the overall environmen- which is consumed by the economy. The value of
tal impacts can be significant. inputs and environmental burdens at each stage
General Motors Corporation (GM) is replac- is estimated. All the inputs at each stage are ap-
ing traditional steel fuel tank systems on the se- proximated by their corresponding US-IO com-
lect models of the GMT600 line of vans with modity sectors. The economy-wide and hence
new blow-molded, co-extruded, multi-layer, life-cycle environmental implications are then
high-density polyethylene (HDPE) tank systems. estimated using EIO-LCA model II (equation 5).
These two tank systems were picked as examples. Because fuel tanks account for a very small frac-
A conventional LCA study comparing equiva- tion of the output of the sector “Motor Vehicle
lent designs of these two systems was conducted Bodies and Accessories,” and fuel tanks are con-
by GM and the National Pollution Prevention sumed almost exclusively by their own commod-
Center (NPPC), University of Michigan ity sector, detailed disaggregation models would
(Keoleian et al. 1997), and this paper draws on also result in very similar estimates. The aggrega-
process descriptions and data in that study. tion of results from these stages provides a com-
Each fuel tank system for the GM van con- prehensive life-cycle inventory.
sists of three major components: the tank that
holds the fuel, straps that secure the tank to the
Life-Cycle Assessment of Steel Fuel Tank
automobile frame, and a shield. The steel tank
requires a plastic shield to protect it from dam- Life-cycle environmental burdens from inputs to
age and corrosion from exposure to humidity, steel tank manufacturing: The steel tank manu-
road salt, and stones, while the plastic tank re- facturing process involves stamping, trimming,
quires a steel heat shield. For this comparative and piercing of sheet steel, welding, washing,
analysis, other components common to the two testing, and assembly operations using inputs
systems such as fuel lines, fuel filters, and send- such as steel sheet, electricity, dies, and lubri-
ing units are excluded. The fuel tank in the steel cants. The estimated value of inputs and the
tank system is made of plain carbon steel, with a U.S. commodity sectors by which these inputs
nickel-zinc coating and a paint coat. The straps are approximated are shown in the first 13 rows
are made of hot-dippe d galvanized steel with of table 2. These estimates are partly based on
painted finish. The tank shield is made of the GM-NPPC study and partly on the input
HDPE. The fuel tank in the plastic tank system purchases by the automotive stamping sector as
is a six-layer co-extruded plastic structure con- reported in the U.S. input-o utput tables
sisting of virgin and reground HDPE, adhesive (Keoleian et al. 1997; U.S. Commerce 1994).
layers, and an ethyl vinyl alcohol (EVOH) co- The economy-wide environmental burdens to
polymer permeation barrier. The straps are made meet a final demand vector equivalent of these
from hot-dipped galvanized steel coated with inputs are estimated using equation (3) and
PVC. The heat shield is made from plain carbon shown in table 3.

104 Jour nal of Industrial Ecology

 R E S E A R C H A N D A N A LYS I S y

Figure 1 Life-cycle stages of an automobile fuel tank system.

Direct environmental burdens at the manufac- Keoleian and colleagues (1997) report air emis-
turing stage: Electricity and natural gas are the sions of 0.2651 g of metals like zinc, chromium,
main sources of energy in manufacturing. The manganese, and nickel, 1.841 g of particulates,
energy associated with fuel use in electricity gen- and 0.684 g of volatile organic compounds per
eration has already been accounted for in the tank from steel tank manufacturing. Additional
input materials. Hence, the energy consumption emissions from burning of 1.29 kg (2.84 lb) of
in the manufacturing stage is 73.53MJ/tank from natural gas for process heat and other require-
natural gas use. Airborne emissions in steel tank ments are included. Waterborne emissions in
manufacture are mainly from welding opera- steel tank manufacturing are mainly from tank
tions. Based on the data from the GM plant, washing to remove the lubricant. The upper

Joshi, Product Input-Output Life-Cycle Assessment 105

y R E S E A R C H A N D A N A LY S I S

Table 2 Steel fuel tank: Input requirements at different life-cycle stages

Dollar
Input US-IO commodity sector US-IO value of input/
code used as approximation sector code tank
Inputs to Tank Manufacturing
Carbon steel sheet Blast Furnaces and Steel Mill Products 370101 23.530
HDPE shield material Plastics and Resins (HDPE Shield) 280100 1.940
Stamping, trimming dies Dies, Tools and Machine Accessories 470300 2.850
Transportation of finished tanks Motor Freight Transportation 650300 0.930
Electricity Electric Utilities 680100 0.850
Transportation of raw materials Railroad Transportation 650100 0.270
Galvanizing and coating Plating and Polishing Services 420401 0.270
Natural gas for boilers Gas Production and Distribution 680200 0.200
Packing materials Paper and Paperboard Containers 250000 0.170
Paints Paints and Allied Products 300000 0.110
Bearings and other repairs Ball and Roller Bearings 490200 0.100
Detergents for washing tanks Soaps and Detergents 290201 0.020
Lubricants and coolants Lubricants and Greases 310102 0.020

Inputs to Use Phase

Gasoline Petroleum Refining 310101 13.720

Inputs to Auto Shredding

Electricity Electric Utility Services 680100 0.055
Transportation of hulks & scrap Motor Transportation 650300 0.500
Shredder tools and repairs Dies, Tools, and Machine Accessories 470300 0.102

Inputs to EAF Steel Making*

Limestone, lime, florspar Lime and Limestone 361300 0.076
Refractories Non-clay Refractories 362100 0.060
Electrodes Carbon & Graphite Products 530700 0.208
Ferroalloys Electro-metallurgical Products 370102 0.069
Electricity Electric Utilities 680100 0.495
Natural gas Gas Distribution 680200 0.062
Maintenance of EAF supplies Industrial Electric Equipment 530800 0.131
Insurance Insurance 700500 0.033

Semi-finished Steel Recovered in EAF

Blast furnaces and steel mill products 370101 – 4.397

* These numbers are based on CMP (1987).

limit estimates of water emissions/steel tank There is no independent use phase for a fuel
based on the treatment plant data are reported tank, except as a component of an assembled
in the GM-NPPC study as 16.873 g of metals vehicle. Relevant portion of the environmental
and 170.8 g of oil and grease. These metal emis- burden associated with use phase of the com-
sions are included in the inventory of toxic plete vehicle has to be allocated to the fuel tank.
chemical releases. The typical environmental burdens during the
Environmental impacts during the use phase of a vehicle use phase cover life-cycle burdens of
steel tank: The use-phase environmental impacts various material inputs such as fuel, lubricants,
are the sum of the life-cycle burdens of inputs to coolants, tires, and repairs and maintenance ser-
the use phase and the impacts during use itself. vices over the lifetime of the vehicle, and the

106 Jour nal of Industrial Ecology

 R E S E A R C H A N D A N A LYS I S y

Table 3 Select economy-wide impacts from production of inputs for one million steel tanks
Impact area Units Quantities
Economic Impact
Value of direct inputs $ Million 31.26
Economy-wide increase in output $ Million 68.94

Non-renewable Ores Consumed

Iron ore MT* 40,942
Copper ore MT 4,464
Gold and silver ores MT 9,607

Energy Consumption
Bituminous coal MT 33,958
Natural gas MT 2,598
Motor gasoline MT 339
Light fuel oils MT 1,250
Heavy fuel oils MT 781
Electricity Mill. kWh 42.8
TOTAL ENERGY Million MJ 1,499

Fertilizers Used MT 99

Toxic Releases
TRI air releases Kilograms 17,024
TRI water releases Kilograms 3,092
TRI total environmental releases Kilograms 51,994
CMU-ET toxicity-weighted total env. releases kg H 2SO 4 eq. 131,956

Conventional pollutants
SO x MT 1,192.3
NO x MT 402.2
Methane MT 1.53
Volatile organic chemicals MT 13.89
Carbon dioxide MT 99,508
Carbon monoxide MT 77.8
RCRA hazardous wastes generated MT 2,894

Summary Indices
Global Warming Potential (GWP) MT CO2 eq. 216,313
Ozone Depletion Potential (ODP) kg CFC11 eq. 27.9
Acidification Potential (Aeq) kg SO2 eq. 404,066

* MT: metric tons

tail pipe emissions. We assume that the fuel tank lons (8.82 liters) (Keoleian et al. 1997). The
contributes to vehicle use-phase environmental value of this gasoline consumption is $13.72 in
impacts only in terms of additional fuel con- 1987 dollars. Because gasoline is a typical output
sumption due to its weight, and ignore incre- of the “Petroleum Refining” sector (US-IO Sec-
mental contribu tions to other inputs and tor Number 310101), the life-cycle environmen-
services. Keoleian and colleaugues (1997) esti- tal burde ns associated with production of
mate the contribution of the steel tank system to gasoline are approximated by the economy-wide
vehicle lifetime fuel consumption to be 23.3 gal- environmental impacts associated with an in-

Joshi, Product Input-Output Life-Cycle Assessment 107

y R E S E A R C H A N D A N A LY S I S

crease of $13.72 in the final demand for the out- scrap recovered from the fuel tank systems is as-
put of the petroleum refining sector. Even sumed to be re-melted in electric arc furnaces
though gasoline is only one of the many joint (EAF) to produce steel mill products. Each steel
products from petroleum refining, this procedure fuel tank yields 21.26 kg of processed scrap,
allocates the environmental burdens of refining which can be processed into 20.24 kg of semi-
to different refinery products on the basis of finished steel mill products valued at $4.397.
their relative market value. The various inputs required for EAF recycling
The life-cycle energy associated with the use and the US-IO sectors by which they are ap-
phase is 3.37 GJ/steel tank, (3.19 ´ 10 6 BTU/ proximated are shown in table 2. Estimates of
tank) consisting of heating value of 23.3 gallons input requirements are based on a report on
(88.2 liters) of gasoline, and the life-cycle energy electric steel making by the Center for Metals
of 316 MJ associated with production of gasoline. Production (CMP 1987). The separately esti-
The use-phase emissions are the vehicle mated direct emissions and energy use from EAF
emissions allocated to the fuel tank based on its steel making are included in table 4. This output
contribution to lifetime fuel consum ption. of steel mill products is assumed to reduce the
Keoleian and colleagues (1997) estimate that final demand for the output of the industry sec-
the 1996 G-van emits 1,969 kg of carbon mon- tor “Blast Furnaces and Steel Mills (US-IO
oxide, 101.2 kg of hydrocarbons and 220 kg of Number 370101)” by a corresponding amount.
nitrogen oxides over a lifetime travel of 110,000 The environmental credits for recycled steel are
miles (177,028 kilometers). These emissions are the avoided economy-wide, environmental bur-
allocated to the steel tank in proportion to its dens from $4.397 worth of steel mill products
contribution to the lifetime vehicle fuel con- substituted by the recycled steel.
sumption of 6,707 gallons. Select environmental burdens over different
Environmental burdens of steel tank scrap pro- life-cycle stages of the steel tank are summarized
cessing: The prompt scrap produced during steel in table 4.
tank manufacturing and the steel recovered from
the auto hulk at the end of vehicle life are used
Life-Cycle Assessment of Plastic Fuel Tank
as scrap input into steel production. A total of
21.26 kg of steel scrap per tank enters the shred- The environmental burdens of the plastic
ding process and we assume that 100% of its fuel tank system over its entire life cycle, cover-
weight is recovered as processed scrap. Sterdis ing production of input materials, manufactur-
(1997) estimates that a ton of auto hulk process- ing, use phase, and end-of-life management, are
ing consumes about 53.6 kWh (0.193 GJ) of analyzed using similar steps, as in the case of the
electricity, and $5.90 worth of tools and mainte- steel tank. Estimates of inputs and direct envi-
nance service. Motor transportation costs per ronmental burdens at each life-cycle stage were
steel tank are estimated to be $0.614 (Keoleian developed. The economy-wide environmental
et al. 1997). These inputs to recycling are ap- burdens from these inputs were calculated using
proximated of their industry sectors. The esti- EIO-LCA model II (equation 5). The input re-
mated (1987) dollar values and corresponding quirements at various life-cycle stages of the
US-IO sectors are shown in table 2. The life- plastic tank and corresponding commodity sec-
cycle environmental burdens of these inputs are tors are shown in table 5 .
estimated using equation (3). Direct environ- The plastic tank manufacturing process begins
mental burdens from the shredding operation in- with the mixing of resin with appropriate addi-
clude energy consumption in shredding (1.13 tives in separate mixing vessels. These mixtures
kWh per tank ) and automobile shredder residue correspond to the composition of the six layers of
(ASR), which is landfilled. It is assumed that the tank, namely virgin HDPE with carbon
the HDPE shield will be a part of ASR and black, reground HDPE, an adhesive layer, ethyl
landfilled as hazardous waste. vinyl alcohol (EVOH) co-polymer permeation
Environmental burdens of steel recovery from barrier, another adhesive layer, and virgin HDPE.
scrap in an electric arc furnace: All the ferrous These mixtures are fed through six individual ex-

108 Jour nal of Industrial Ecology

 R E S E A R C H A N D A N A LYS I S y

truders to produce layers just before entering the economy-wide environmental burdens are esti-
blow molder. The polymer layers are simulta- mated along the same lines as the steel tank.
neously blow molded to the required shape of the As in the case of the steel tank, all the ferrous
tank. The multi-layer tank is then sent to the scrap from the plastic fuel tank system is as-
piercing and machining station. Components sumed to be processed into semi-finished steel
such as rollover valves and clips are welded onto mill products by re-melting scrap in an electric
the tank, followed by assembly of the sending arc furnace: 3.41 kg/tank of steel scrap entering
unit, straps, shield, and fuel lines. All finished the EAF process yields 3.25 kg of semi-finished
tanks are then packed, placed on shipping racks, steel mill products valued at $0.70. The credit
and sent to the vehicle assembly plant by truck. for recycled steel is estimated similarly.
The initial part of table 5 shows the estimated A summary of the environmental burdens
dollar value of inputs from different commodity over the entire life cycle of the plastic fuel tank
sectors for plastic tank manufacture. Their life- is shown in table 6.
cycle burdens are summarized in table 6.
Direct environmental burdens during manufac- Comparison of Steel and Plastic Fuel Tank
turing of plastic tanks: The major source of emis- Life-cycle Environmental Impacts
sions in plastic tank manufacturing is the A summary comparison of the total life-cycle
extrusion–blow molding process. Emission fac- environmental impacts of the two fuel tank sys-
tors for HDPE blow molding process from Barlow tems is presented in table 7. Comparative de-
(Barlow et al. 1996) were used to estimate pro- composition by life-cycle stages of selected
cess emissions during plastic tank manufactur- impacts is presented in the form of a series of
ing. In addition, emissions from burning natural graphs in figures 2–7.
gas for process heat and other requirements, es- From table 7, it is apparent that the total life-
timated at 1.0 gram of carbon monoxide, 3.25 kg cycle environmental burdens from the steel fuel
of CO 2, 3.91 grams of NOx and 0.08 grams of tank are larger than the burdens from the plastic
methane per tank, are included. fuel tank, on all the dimensions except RCRA
Environmental burdens in the use phase of the plas- hazardous generation and total ozone depletion
tic tank: As in the case of steel tanks, the life-cycle potential (ODP) of air emissions. The environ-
impacts of gasoline production and exhaust emis- mental impacts from the steel tank as a percent-
sions during vehicle use are allocated to the fuel age of impacts from the plastic tank range from
tank. The gasoline use allocated to the plastic fuel 125% for total TRI toxic chemical releases to
tank is 14.95 gallons valued at $8.80. The GM- the environment, to 3,842% for net iron ore use.
NPPC study estimates that the G-Van, fitted with From tables 4 and 6, and figures 2–7, it can be
a plastic tank, consumes 6,699 gallons of gasoline seen that the environmental burdens from the
over its lifetime (110,000 miles), and emits 1,967 steel fuel tank are larger than the burdens from
kg of carbon monoxide, 101 kg of hydrocarbons, the plastic fuel tank in most of the individual
219.7 kg of NOx, and 58,733 kg of carbon dioxide. life-cycle stages also.
These use-phase emissions are allocated to the In conclusion, the EIO-LCA suggests that if
plastic fuel tank based on its weight. all other performance parameters were equal,
Environmental burdens from recovery of steel the plastic tank would be preferable in this ap-
from the plastic fuel tank: For this analysis, it is plication, on the basis of its relatively better en-
assumed that the whole tank system is shredded, vironmental performance.
i.e., there is no disassembly of components prior
to shredding. That is, the tank system weighing Comparison with Results from the GM-
14.6 kg enters the shredding operation, from NPCC Study
which 3.41 kg of steel corresponding to the As mentioned, General Motors and National
shield, straps, and prompt scrap will be recov- Pollutio n Prevention Center, University of
ered and the balance, 11.2 kg, will be landfilled Michigan (GM-NPCC) carried out a conven-
as automobile shredder residue. The input re- tional life-cycle inventory of these two fuel tank
quirements for the shredding operation and the systems. The boundaries of analysis, data sources,

Joshi, Product Input-Output Life-Cycle Assessment 109

 Table 4 Summary of life-cycle environmental burdens of the steel tank system (per tank)
y

110
Environmental Inputs to Manu- Inputs to Inputs to EAF steel Credit for
impact manufacturing facturing use phase Use phase shredding Shredding making EAF steel TOTAL
Energy
Coal (kg) 34.00 0.00 1.42 0.00 0.190 0.00 1.58 –5.73 31.46
Nat. gas and LNG (kg) 2.80 1.29 3.26 0.00 0.018 0.00 0.18 –0.38 7.168
Petroleum fuels (kg) 2.70 0.00 1.80 65.00 0.193 0.00 0.15 –0.34 69.503
Electricity (kWh) 42.80 17.7 4.80 0.00 0.10 1.13 1.12 –7.07 60.58
Total energy (MJ) 1499 73.50 316 3055 10.20 4.10 61 –191 4827.8
Toxic Releases (g)
R E S E A R C H A N D A N A LY S I S

Air 17.02 0.27 4.39 0.00 0.06 0.00 0.76 –2.33 20.17
Water 3.09 0.02 0.33 0.00 0.01 0.00 0.21 –0.48 3.18

Jour nal of Industrial Ecology

Land & underground 31.88 0.00 1.83 0.00 0.03 0.00 0.71 –5.14 29.31
Total env. releases 51.99 0.29 6.56 0.00 0.11 0.00 1.68 –7.96 52.67
Releases + transfers 258.3 0.29 58.10 0.00 0.39 0.00 32.79 –43.74 306.13
Toxicity-Weighted Releases H2SO4 eq g
Air 10.72 0.17 0.42 0.00 0.02 0.00 0.55 –1.87 10.01
Water 2.14 0.02 0.16 0.00 0.01 0.00 0.05 –0.34 2.04
Land & underground 119.10 0.00 2.44 0.00 0.12 0.00 4.06 –20.60 105.12
Total env. releases 131.95 0.19 3.02 0.00 0.15 0.00 4.66 –22.81 1117.16
Releases + transfers 834.60 0.19 61.22 0.00 1.01 0.00 115.70 –148.6 864.08
Conventional Air Pollutants (g)
SOx 1192 0.00 51.95 0.00 7.07 0.00 56.87 –200.20 1107.69
NO x 402 3.91 31.20 760 3.86 0.00 19.0 –66.42 1153.55
Methane 1.54 0.08 0.41 41.6 0.04 0.00 0.08 –0.23 43.52
VOCs 13.89 0.68 3.55 350 0.63 0.00 0.47 –1.90 367.32
CO 77.84 1.0 17.27 6840 5.05 0.00 2.60 –10.42 6933.34
CO 2 (kg) 99.51 3.25 17.11 206.2 1.11 0.00 4.85 –16.20 315.83
RCRA Waste (kg)
Generated 2.89 0.00 9.40 0.00 0.03 3.34 0.34 –0.36 15.64
Managed 1.94 0.00 8.90 0.00 0.03 0.00 0.03 –0.23 10.67
Nonrenewable Ores
Iron ore (kg) 40.94 0.00 0.12 0.00 0.015 0.00 0.01 –7.58 33.505
Copper ore (kg) 4.52 0.00 0.23 0.00 0.024 0.00 0.57 –0.68 4.664
Gold, silver ores (kg) 9.61 0.00 0.10 0.00 0.014 0.00 0.09 –1.71 8.104
Summary Indices
GWP (kg CO2 eq) 216.30 4.38 26.17 427.1 2.23 0.00 10.37 –35.47 651.08
ODP (g CFC11 eq) 0.03 0.00 0.002 0.00 0.00 0.00 0.001 –0.002 0.031
Aeq (g SO2 eq) 404.0 2.72 27.88 529.0 3.40 0.00 19.29 –67.08 919.21
 R E S E A R C H A N D A N A LYS I S y

Table 5 Plastic fuel tank: Input requirements at different life-cycle stages

Commodity sector EIO
used as an sector Dollar value
Input approximation code of input/tank
Inputs to Plastic Tank Manufacture
HDPE, PVC, EVOH Plastics and Resins 280100 7.60
Steel straps and shield Automotive Stampings 410201 3.52
Electricity Electric Utility Services 680100 1.38
Glycol and other supplies Industrial Org. & Inorg. Chemicals 270100 0.85
Natural gas Gas Production 680200 0.20
Packing materials Paperboard and Containers 250000 0.69
Molder spare parts Special Industry Machinery Parts 480600 0.24
Carbon black Carbon Black 270405 0.10
Adhesive layer material Adhesives and Sealants 270403 0.15

Inputs to Use Phase

Gasoline Petroleum Refining 310101 8.80

Inputs to Auto Shredding

Electricity Electric Utility Services 680100 0.038
Transportation of hulks & scrap Motor Transportation 650300 0.343
Shredder tools and repairs Dies, Tools and Machine Accessories 470300 0.070

Inputs to EAF Steel Making

Limestone, lime, florspar Lime and Limestone 361300 0.01175
Refractories Non-clay Refractories 362100 0.00969
Electrodes Carbon & Graphite Products 530700 0.00334
Ferroalloys Electro-metallurgical Products 370102 0.00112
Electricity Electric Utilities 680100 0.07940
Natural gas Gas Distribution 680200 0.00995
Maintenance of EAF supplies Industrial Electric Equipment 530800 0.02101
Insurance Insurance 700500 0.00533

Semi-finished Steel Recovered in EAF

Blast furnaces and steel mill products 370101 – 0.700

coverage, and classification of environmental Franklin Associates for the life-cycle burdens
burdens in the GM-NPPC study are different from gasoline production. The GM-NPCC study
from EIO-LCA. In the analysis of materials pro- reports the following environmental burdens:
duction phase, the GM-NPCC study considers energy consumption, solid waste generation, air
only steel, HDPE, PVC, and EVOH. The envi- emissions of CO2, CO, NOx, hydrocarbons and
ronmental data on steel are from a European da- particulate matter, and water effluents covering
tabase on packing materials (FOEFL 1991). The suspended solids, dissolved solids, oil and grease,
data on PVC and HDPE are from other Euro- and metals. These are different from those cov-
pean databases (Boustead 1993, 1994). EVOH is ered in EIO-LCA. As a result, only a limited di-
approximated by HDPE. In EIO-LCA, steel is rect comparison is possible between the results
approximated by the U.S. steel sector, and all from these two studies. The common elements
plastics are approximated by the U.S. plastics are summarized in table 8.
materials and resins sector. Similarly, GM- Two things are noticeable about the com-
NPCC use proprietary information from parative results. One, despite using very different

Joshi, Product Input-Output Life-Cycle Assessment 111

y R E S E A R C H A N D A N A LY S I S

Figure 2 Total energy.

Figure 3 Global Warming Potential.

112 Jour nal of Industrial Ecology

 Table 6 Summary of life-cycle environmental burdens of the plastic tank system (per tank)
Environmental Inputs to Manu- Inputs to Inputs to EAF steel Credit for
impact manufacturing facturing use phase Use phase shredding Shredding making EAF steel TOTAL
Energy
Coal (kg) 7.43 0 0.91 0.00 0.13 0.00 0.26 –0.92 7.81
Nat. gas and LNG (kg) 1.69 1.29 2.10 0.00 0.01 0.00 0.03 –0.06 5.06
Petroleum fuels (kg) 1.10 0 1.16 41.71 0.13 0.00 0.03 –0.05 44.08
Electricity (kWh) 16.1 23 3.08 0.00 0.07 0.78 0.18 –1.14 42.07
Total energy (MJ) * 1246 156.8 202.8 1960 7.00 2.82 10.0 –31 3555
Toxic Releases (g)
Air 17.22 0.45 2.82 0.00 0.04 0.00 0.12 –0.37 20.28
Water 2.23 0 0.21 0.00 0.01 0.00 0.03 –0.08 2.4
Land & underground 18.99 0 1.17 0.00 0.02 0.00 0.12 –0.83 19.47
Total env. releases 38.38 0.45 4.21 0.00 0.08 0.00 0.27 –1.28 42.11
Releases + transfers 89.38 0.45 37.28 0.00 0.27 0.00 5.26 –7.02 125.62
Toxicity Weighted Releases H2SO 4eq g
Air 2.26 0.16 0.27 0.00 0.02 0.00 0.08 –0.30 2.49
Water 1.44 0 0.10 0.00 0.01 0.00 0.01 –0.06 1.5
Land & underground 34.44 0 1.57 0.00 0.08 0.00 0.65 –3.30 33.44
Total env. releases 38.14 0.16 1.94 0.00 0.10 0.00 0.75 –3.66 37.43
Releases + transfers 110.84 0.16 39.28 0.00 0.69 0.00 18.56 –23.84 145.69
Conventional Air Pollutants (g)
SOx 266.89 0.00 33.33 0.00 4.86 0.00 9.12 –32.11 282.09
NO x 92.24 3.91 20.02 488 2.65 0.00 3.05 –10.65 599.22
Methane 0.54 0.08 0.26 27.2 0.03 0.00 0.01 –0.04 28.08
VOCs 5.56 0.45 2.28 225 0.43 0.00 0.07 –0.30 233.49
CO 26.21 1.00 11.07 4390 3.47 0.00 0.42 –1.67 4430.5
CO 2 (kg) 26.03 3.25 10.97 132.3 0.76 0.00 0.78 –2.60 171.49
RCRA Waste (kg)
Generated 2.34 0 6.03 0.00 0.02 11.20 0.05 –0.06 19.58
Managed 2.06 0 5.71 0.00 0.02 0.00 0.01 –0.04 7.76
Nonrenewable Ores
Iron ore (kg) 2.00 0 0.08 0.00 0.01 0.00 0.002 –1.22 0.872
Copper ore (kg) 1.18 0 0.15 0.00 0.02 0.00 0.09 –0.11 1.33
Gold & Silver ore (kg) 0.78 0 0.06 0.00 0.01 0.00 0.02 –0.27 0.60

Joshi, Product Input-Output Life-Cycle Assessment

Summary Indices
GWP (kg CO2 eq) 54.22 4.38 16.78 274.1 1.53 0.00 1.66 –5.69 346.98
ODP (g CFC11 eq) 0.05 0.00 0.001 0 0.00 0.00 0.00 0.00 0.051
R E S E A R C H A N D A N A LYS I S

95.7 2.72 17.88 339.6 2.34 0.00 3.09 –10.76 450.57

y

Aeq (g SO2 eq)

113
*Energy including the heating values of HDPE, PVC & EVOH (78.3, 31.5 & 52 MJ/kg respectively).
y R E S E A R C H A N D A N A LY S I S

Table 7 Comparison of total life-cycle environmental impacts of steel and plastic fuel tanks
Environmental Plastic tank Steel tank Steel tank impacts as %
impact impacts impacts of plastic tank impacts
Energy Use
Coal (kg) 7.81 31.46 403
Nat. gas and LNG (kg) 5.06 7.168 142
Petroleum fuels (kg) 44.08 69.503 158
Electricity (kWh) 42.07 60.58 144
Total energy (MJ) * 3555 4827.8 136

Toxic Releases (g)

Air 20.28 20.17 99
Water 2.4 3.18 133
Land & underground 19.47 29.31 151
Total env. releases 42.11 52.67 125
Releases + transfers 125.62 306.13 244

Toxicity-Weighted Releases H2SO4eq g

Air 2.49 10.01 402
Water 1.5 2.04 136
Land & underground 33.44 105.12 314
Total env. releases 37.43 117.16 313
Releases + transfers 145.69 864.08 593

Conventional Air Pollutants (g)

SO x 282.09 1107.69 393
NOx 599.22 1153.55 193
Methane 28.08 43.52 155
VOCs 233.49 367.32 157
CO 4430.5 6933.34 156
CO 2 (kg) 171.49 315.83 184

RCRA Waste (kg)

Generated 19.58 15.64 80
Managed 7.76 10.67 138

Nonrenewable Ores
Iron ore (kg) 0.872 33.505 3842
Copper ore (kg) 1.33 4.664 351
Gold, silver ores (kg) 0.6 8.104 1351

Summary Indices
GWP (kg CO 2 eq) 346.98 651.08 188
ODP (g CFC11 eq) 0.051 0.031 61
Aeq (g SO2 eq) 450.57 919.21 204

114 Jour nal of Industrial Ecology

 R E S E A R C H A N D A N A LYS I S y

Figure 4 Total toxic releases to the environment.

Figure 5 Toxicity-weighted environmental releases.

data sources, methodologies, and levels of aggre- pollutant emissions. Second, contrary to the ex-
gation, there is a high degree of correspondence pectation that EIO-LCA would report higher
between the comparable results from these two impacts due to greater coverage, the absolute
studies. The qualitative conclusions from the amount of impacts reported in EIO-LCA here
two studies are the same: that the plastic tank are generally lower than those reported in the
appears to perform better on most of the envi- GM-NPPC study (except solid waste genera-
ronmental dimensions except solid waste gen- tion). The energy consumption in the materials
eration; and use-phase impacts dominate total production stage is higher in the case of EIO-
life-cycle energy use and life-cycle conventional LCA as expected. Though both studies use the

Joshi, Product Input-Output Life-Cycle Assessment 115

y R E S E A R C H A N D A N A LY S I S

Figure 6 Acidification potential of air emissions.

Figure 7 RCRA hazardous waste generation.

same quantities of gasoline to estimate use phase to differences in allocations between co-prod-
energy, unit energies are different. While the ucts in petroleum refining. EIO-LCA also pro-
GM-NPPC uses total energy (precombustion + vides energy credits for recycled steel. The
combustion) of 42.03 MJ/liter of gasoline, the GM-NPPC study uses heating value of 81 MJ/kg
estimated precombustion + combustion energy of HDPE while EIO-LCA uses 78.3 MJ/kg of
content of gasoline in EIO-LCA is 38.2 MJ/liter. HDPE. If these estimates were used in EIO-
Because the GM-NPPC estimate is from a pro- LCA, the total life-cycle energy in EIO-LCA
prietary source, we can not pinpoint the source would be higher than those reported in the GM-
of the difference. However, it is most likely due NPPC study.

116 Jour nal of Industrial Ecology

 R E S E A R C H A N D A N A LYS I S y

Table 8 Comparison of results from EIO-LCA and GM-NPCC LCA

Burden Steel tank Plastic Tank
Category GM-NPPC EIO-LCA GM-NPPC EIO-LCA
Life-cycle energy MJ/tank 4860 4828 3631 3555
Energy in material production MJ/tank 1020 1499 1020 1246
Energy in use phase MJ/tank 3710 3371 2360 2163
Energy in EOM MJ/tank 14 –126 13 –21
Life-cycle CO2 emissions kg/tank 295 316 191 172
Life-cycle CO emissions g/tank 7014 6933 4524 4431
Life-cycle NOx emissions g/tank 1284 1154 942 600
Life-cycle VOC emissions g/tank 1381 367 894 234
Life-cycle solid waste kg/tank 13.4 15.64(haz) 14.5 19.6(haz)

A major limitation of the current EIO-LCA emissions, fertilizer use, and non-renewable ore
is that it reports conventional pollutant emis- use from publicly available sources. The pro-
sions from fuel combustion only and does not posed methodology is demonstrated by applying
include non-fuel process emissions. Process it to a real-life decision problem of evaluating
emissions are included in the GM-NPPC study. environmental performance of alternative fuel
HDPE and PVC manufacturing involve signifi- tank systems.
cant process emissions of NOx and VOCs. Ac- The proposed EIO-LCA method, however,
cording to Boustead (1993), these total to 21 g has several potential limitations that have to be
of VOCs and 10 g of NO x per kg of HDPE, and considered while applying the method. These
19 g of NOx and 26 g of VOCs per kg of PVC. limitations and caveats arise both from the na-
Inclusion of these process emissions in EIO- ture of input-output analysis and from the spe-
LCA is necessary for comparability of conven- cific character of this method. EIO-LCA shares
tional pollutant emissions. If we add only these the fundamental limitations of input-ou tput
direct NOx and VOC emissions to the reported analysis, such as: linear approximation in tech-
EIO-LCA emissions, the resultant emission nical coefficients, static analysis, and omission
numbers will be larger than the emissions re- of capital services. Static analysis and linearity
ported in the GM-NPPC study. provide good approximations for relatively small
changes, typical in product LCA. Because capi-
tal services are consumed over a large number of
Discussion and Conclusions
individu al units, the environmental impacts
This paper extends and refines the EIO-LCA from capital services are likely to be small.
model to make it a practical and flexible tool for Expressing all the environmental impact co-
comprehensive product life-cycle assessment efficients in terms of environmental burden/dol-
that can be used by companies, regulators, and lar value of sector output instead of familiar
consumer groups to conduct quick, cost-effec- physical units such as tons, liters, service hours,
tive, and comprehensive life-cycle assessments. and so on makes modeling simpler by avoiding
The extension enables life-cycle assessment of the difficulty of incorporating multiple physical
individual products, comparison of products that units and appropriate conversion factors in the
may belong to the same sector or completely new model. Under constant price conditions, these
products, and incorporation of environmental impact coefficients can easily be translated to
impacts from product use phase and end-of-life physical unit bases. However, geographical, tem-
management (reuse, remanufacturing, recycling) poral and tax-related variations in prices can in-
stages in EIO-LCA. The extension combines to- troduce errors. Care should be taken to use
gether comprehensive environmental impact co- appropriate price indices to adjust for these
efficient matrices for the U.S. economy, covering variations to reflect national average producer
fuel and energy use, conventional pollutant prices in the base year.

Joshi, Product Input-Output Life-Cycle Assessment 117

y R E S E A R C H A N D A N A LY S I S

Co-prod uct allocation is a much-debated porting. Similarly, only facilities employing at

topic in both construction of national input- least ten people or using at least 1,000 lb/year of
output tables and LCA literature (Miller and the toxic chemicals are required to report. A study
Blair 1985, Curran 1996). In the construction of by the U.S. General Accounting Office concluded
US-IO tables, a commodity technology assump- that reported TRI releases might be as low as 5%
tion is used in estimating technical coefficients of the actual toxic releases (U.S. GAO 1991). As
for co-products (refer to U.S. Commerce 1994 a result, the TRI release data in the model are
for more details). EIO-LCA allocates environ- likely to be an underestimate. The conventional
mental burdens on the basis of market value. pollutant emissions are based on generic emission
Other bases of allocation used in the LCA litera- factors for the primary use of different fuels. Varia-
ture include mass, dry mass, energy content, in- tions in emissions due to non-linearity over the
cremental processing energy, stoichiometry, and operating range and differences in fuel quality,
heat of reaction (Curran 1996, SETAC 1993, combustion technology, and efficiencies have
Boustead 1993). For some specific systems, one been overlooked in using these generic emission
of these allocation methods might be more ap- factors. Non-fuel combustion-related emissions
pealing than other methods. Science does not are not currently included in the model due to the
dictate the method of allocation. Market value unavailability of data. Attempts are underway to
can be a reasonable basis when higher prices re- include data on non-fuel process emissions esti-
flect increased processing and emissions. mated from the AIRS (Aerometric Information
In deriving the U.S. input-output tables, the Retrieval System) database of U.S. EPA into EIO-
technical coefficients for comparable imports LCA. However, the major strength of EIO-LCA is
are assumed to be similar to the corresponding that the national averages (and derived estimates
domestic U.S. industry sectors (U.S. Commerce for disaggregated sectors) are mainly used as ap-
1994). The EIO-LCA models assume that the proximations for missing data and all the available
environmental burden coefficients of imports more accurate data can be included in the model.
are also well-approximated by the corresponding More detailed models will be less affected by errors
U.S. domestic industry sectors. This may intro- in aggregate data. Results from the cruder models
duce errors in analyses of products with high im- should be interpreted more as indicators of rela-
port content and very different foreign tive performance in comparing products than as
production technologies. Conceptually, creation absolute performance indicators.
of appropriate hypothetical industry sectors In conclusion, EIO-LCA is a flexible and pow-
(Model II) representing the technology of such erful addition to the toolkit of LCA practitioners.
imports can address the issue. In its simpler forms it serves as an initial screening
Because data from many different sources and device for scoping and prioritizing data collection.
time periods have been collected, normalized, More advanced models can be used to validate re-
aggregated, and averaged to arrive at sector-level sults at different stages of a conventional LCA to
environmental indices, there are significant un- check if boundary definitions and consideration of
certainties in the estimates.4 First, there are uncer- economy-wide interdependencies affect results.
tainties in the national technical coefficient Efforts are underway to offer EIO-LCA as a web-
matrix arising from aggregation, sampling errors, based software in the public domain.
imputation procedures for missing data, co-prod-
uct allocation procedures, and linearity assump-
Acknowledgments
tions. The data on dollar value of purchases of
fuels, fertilizers, and ores are from the US-IO work This article is based on my doctoral disserta-
files. The quantities were derived using an average tion. I thank Lester Lave, Chris Hendrickson,
producer price, which overlooks regional and Francis McMichael, Lowell Taylor, and other
quality-related price variations and inventory faculty members at Carnegie Mellon University,
changes. Large uncertainties are associated with Pittsburgh. I thank Ron Williams and Robert
the U.S. TRI data. Many industries such as mining Stephens from General Motors for their help
and utilities until recently were excluded from re- and comments. I am grateful to the editor and

118 Jour nal of Industrial Ecology

 R E S E A R C H A N D A N A LYS I S y

two anonymous referees for their input, which Boustead, I. 1994. Eco-profiles of European plastics in-
strengthened the article considerably. Financial dustry—polyvinyl chloride: Report 6. Brussels: The
support for the research was provided by U.S. European Center for Plastics in the Environment.
EPA, National Science Foundation, and mem- CMP (Center for Metals Production). 1987.
Technoeconomic assessment of electric steelmaking
bers of the Green Design Consortium, Carnegie
through the year 2000 (Report EPRI EM-5445).
Mellon University.
Pittsburgh, PA: CMP.
Cobas-Flores, E. 1996. Life-cycle assessment using in-
Notes put-output analysis. Ph.D. dissertation, Carnegie
Mellon University.
1. There is a vast theoretical and empirical litera- Cumberland, J. H. and B. Stram. 1976. Empirical ap-
ture on extension of input-output techniques for plications of input-output models to environ-
environmental analysis (for example, see Ayres mental protection. In Advances in input-output
and Kneese 1969; Cumberland and Stram 1976; analysis: Proceedings of the sixth international con-
Forsund 1985; Duchin et al.1990; Duchin 1994; ference on input-output techniques, Vienna, April
Lutz 1993; UN 1993). However, most of them are 22–26, 1974, edited by K. R. Polenske and J. V.
aimed at broad national-level policy analysis. In Skolka, pp. 365–382. Cambridge: Ballinger.
that respect, the current application for micro- Curran, M. A. 1996. Environmental life-cycle assess-
level product and process evaluation is novel. ment. New York: McGraw Hill.
2. I do not attempt to quantify or monetize the fi- Davis, G. A. 1993. The use of life-cycle assessment in
nal health and ecological impacts. These in- environmental labeling programs , EPA/742-R-93-
volve many other difficul t concept ual and 003. Washington, DC: U.S. Environmental Pro-
empirical issues. tection Agency.
3. In U.S. regulatory jargon, conventional pollutants Duchin, F., G. M. Lange, and T. Johnsen. 1990. Strategies
refer to carbon monoxide, sulfur oxides, nitrogen for environmentally sound development: Progress: Re-
oxides, methane, and volatile organic compounds. port 5. New York: Institute of Economic Analysis.
In the analysis described in this article, carbon di- Duchin, Faye. 1994. Industrial input/output analysis:
oxide is included in this category as well. Implications for industrial ecology. In The green-
4. Data quality is a major concern of LCA practi- ing of industri al ecosyste ms, edited by B. R.
tioners and many formal quantitative and quali- Allenby and D. R. Richards, pp. 61–68. Wash-
tative indicators of data quality in LCA have ington, DC: National Academy of Engineering.
been proposed (see Vigon et al. 1993 and refer- Forsund, F. R. 1985. Input-output models, national
ences therein). However, only qualitative de- economic models and the environment. In
scriptions of data quality are provided here due Handbook of natural resource and energy economics ,
to time and data constraints. vol. I, edited by A. V. Kneese and J. L. Sweeny,
pp. 325–394. Amsterdam: North Holland.
FOEFL (Federal Office of Environment, Forests and
References Landscape, Switzerland). 1991. Eco-balance of
Adriaanse, A. 1993. Environmental policy performance packaging materials: State of 1990. Berne, Switzer-
indicators . The Hague: Sdu Publishers. land: FOEFL.
Arnold, F. 1993. Life-cycle design doesn’t work. Envi- Hendrickson, C., A. Horvath, S. Joshi, and L. Lave.
ronmental Forum. Sept/Oct: 19–23. 1998. Economic input-output models for life-
Ayres, R. U. and A. V. Kneese. 1969. Production , cycle assessm ent. Environme ntal Science and
consumption and externaliti es. The American Technology : 13(4): 184A–191A.
Economic Review, LIX(June): 282–297. Horvath, A., C. Hendrickson, L. Lave, F. C.
Barlow, A., D. Contos, M. Holdren, P. Garrison, L. McMichael, and T. S. Wu. 1995. Toxic emission
Harris, and B. Janke. 1996. Developmen t of indices for green design and inventory. Environ-
emission factors for polyethyle ne processing . mental Science and Technology 29(2): 86–90.
Journal of the Air and Waste Management Associa- Horvath, A. 1997. Estimation of environmental implica-
tion 46: 569–580. tions of construction materials and designs using life-
Boustead, I. 1993. Eco-profiles of European plastics in- cycle assessment techniques . Ph.D. dissertation,
dustry—polyethylene and polypropylene: Report 3. Carnegie Mellon University.
Brussels: The European Center for Plastics in the IPCC (Interg overn mental Panel on Climate
Environment. Change). 1995. IPCC guidelin es for national

Joshi, Product Input-Output Life-Cycle Assessment 119

y R E S E A R C H A N D A N A LY S I S

greenhou se gas inventor ies, vol. 1–3. UNEP, neering and Public Policy, Carnegie Mellon
OECD and IPCC. University.
Joshi, S. V. 1998. Comprehensive product life-cycl e Tibor, T. and I. Feldman. 1996. ISO 14000: A guide to
analysis using input-output techniques . Ph.D. dis- the new environm ental managem ent standard s.
sertation, Carnegie Mellon University. Chicago: Irwin Publishing.
Karna, A. and J. Engstrom. 1994. Life-cycle analysis UN (United Nations). 1993. Integrated environmental
of newsprint: European scenarios. Paperi ja Puu and economic accounting. New York: United Na-
76(4): 232–237. tions Department of Economic and Social Infor-
Keoleian, G. A. and D. Menerey. 1993. Life-cycle de- mation and Policy Analysis, Statistics Division.
sign manual (EPA 600/R-92/226). Washington, USBM (United States Bureau of Mines). 1988. Min-
DC: U.S. Environmental Protection Agency. erals yearbook 1988.
Keoleian, G. A., S. Spatari, and R. Beal. 1997. Life- U.S. Commerce (United States Department of Com-
cycle design of a fuel tank system (draft report). merce, Economics and Statistics Administra-
Ann Arbor, MI: National Pollution Prevention tion ). 1987. Census of manufac tures 1987.
Center, University of Michigan. Washington DC: U.S. Bureau of Census.
Lave, L. B., E. Cobas-Flores, C. T. Hendrickson, and U.S. Commerce (United States Department of Com-
F. C. McMichael. 1995. Using input-output merce, Inter-industry Economics Division ).
analysis to estimate economy-wide discharges. 1994. Input-output accounts of the U.S. economy,
Environ mental Science and Technology 29(9): 1987 benchmark (diskettes).
420A–426A. U.S. EPA (United States Environmental Protection
Leontief, W. 1966. Input-output economics. New York: Agency). 1995a. Air Chief (CD-ROM). Wash-
Oxford University Press. ington, DC: U.S. EPA.
Lutz, E., ed. 1993. Toward improved accounting for the en- U.S. EPA (United States Environmental Protection
vironment. Washington, DC: World Bank. Agency). 1995b. 1987–1993 toxics release inven-
Miettinen, P. and R. Hamalainen. 1997. How to ben- tory. (CD-ROM, EPA 749/C-95-004). Washing-
efit from decision analysis in environmental life- ton, DC: U.S. EPA, Office of Pollution
cycle analysis. European Journal of Operations Prevention and Toxics, June.
Research 102: 279–294. U.S. EPA (United States Environmental Protection
Miller, R. A. and P. D. Blair. 1985. Input-output analy- Agency). 1995c. Guidance on acquisition of envi-
sis: Foundations and extensions . Englewood Cliffs, ronmentally preferable products and services. Fed-
NJ: Prentice Hall. eral Register: 60(189): 50721–50735.
OSAT (Office for the Study of Automotive Transpor- U.S. GAO (United States General Accounting Of-
tation). 1996. Delphi VIII: Forecast and analysis of fice). 1991.Toxic chemicals— EPA’s Toxic Release
the North American automotive industry: vol. 3, Inventory is useful but can be improved (GAO/
materials. Ann Arbor, MI: Office for the Study of RCED-91-121). Washington, DC: U.S. GAO.
Autom otive Transporta tion, University of Vigon, B. W., D. A. Tolle, B. W. Cornaby, H. C.
Michigan Transportation Research Institute. Latham, C. L. Harrison ,T. L. Boguski, R. G.
Pento, T. 1997. Dynamic life-cycle inventory model- Hunt, and J. D. Sellers. 1993. Life-cycle assess-
ing with input-outpu t tables and joined time ment: Inventory guidelines and principles (EPA/
projections. Seventh Annual Meeting of Society of 600/R-92/245). Cincinnati, OH: Risk Reduc-
Environmental Toxicology and Chemistry—Eu- tion Engineering Laboratory, U.S. EPA.
rope, 6–10 April. Amsterdam: SETAC. Vigon, B. W. 1997. Life-cycle inventory: Data quality
Portney, P. R. 1993. The price is right: Making use of issues. Proceedings of the 1997 total life-cycle con-
life-cycle analysis. Issues in Science and Technol- ference— life-cycle management and assessment .
ogy X(2): 69–75. P310 (971162): 47–54. Warrendale: Society of
SETAC (Society of Environmental Toxicology and Automotive Engineers, Inc.
Chemistry). 1993. Guidelines for life-cycle assess- Wuebbles, D. J. 1995. Weighing functions for ozone
ment: A code of practice . Workshop report. depletion and green house gas effects on cli-
Pensacola, FL: SETAC. mate. Annual Review of Energy and Environment
Sterdis, A. 1997. The future of automobile materials re- 20: 45–70.
covery. Unpublished paper. Department of Engi-

120 Jour nal of Industrial Ecology