Home > Input-Output Based Life Cycle Analysis for New Economy Models
Input-Output
Based Life Cycle Analysis for New Economy Models
H. Scott Matthews*
Department of Civil and Environmental Engineering
Department of Engineering and Public Policy
Carnegie Mellon University
Porter Hall 119
Pittsburgh, PA 15213 USA
hsm@cmu.edu
Chris T. Hendrickson
Head, Department of Civil and Environmental Engineering
Carnegie Mellon University
Porter Hall 119
Pittsburgh, PA 15213 USA
cth@cmu.edu
Using the 1992 U.S. Benchmark
and 1997 Annual Input-Output tables, we have created a freely available
Internet tool to facilitate I-O analysis at www.eiolca.net. Beyond
economic transaction data, the site has sector-level data on fuel and
electricity use, conventional air pollutant and greenhouse gas emissions,
toxic and hazardous waste emissions, employment, worker safety and fatalities,
ores and fertilizers. Our primary purpose for providing this information
is to facilitate Life Cycle Assessment of products and processes, an
environmental analysis tool to track all resource inputs and environmental
outputs from extraction of raw materials through product disposition.
Extensions of the model allow
users to enter multiple sectoral demands, graph results, and show side-by-side
comparisons. The architecture of the model is open, and input-output
data from other countries can be accommodated (and is being solicited)
via the same web-based portal. German and Japanese I-O data will be
added in 2002.
We demonstrate this model by comparing the energy use and greenhouse gas emissions for buying books from online vendors versus traditional retail bookstores in the United States. In addition, we compare our results with results from a non-input-output based estimate for book purchasing in Japan.
The Economic Input-Output Life
Cycle Assessment (EIO-LCA) model was originally developed in 1995 to
describe connections between economic activity, energy and resource
use, and environmental emissions using the 1987 Benchmark Input-Output
Table developed by the U.S. Department of Commerce Bureau of Economic
Analysis [Lave 1995]. The EIO-LCA model was originally developed
as a proprietary Microsoft Windows application. However, a grant
from the U.S. National Science Foundation in 1998 allowed development
of a free, Internet-based version of the model and an update to the
1992 Benchmark I-O table [Lawson 1997, Hendrickson 1998]. Since
2000, nearly 100,000 analytical comparisons have been made using this
tool, available at http://www.eiolca.net/. Amongst the many features available
are direct and total supply chain estimations from a given input change
in multiple sectors of demand, analysis of aggregate and disaggregate
models, corresponding estimates of non-economic effects, and bar graphs
of individual and total sector contributions.
The EIO-LCA model was created
in a modular way so that new input-output data (as well as corresponding
non-economic vectors) could be easily added and updated. Benchmark
I-O tables for the US economy are available every five years (e.g. 1987
and 1992) and are released 4-5 years subsequently. In 2001, we
updated to the 1997 Annual Input-Output table from the U.S. Department
of Commerce [Kuhbach 2001]. In 2002, we will make available the
German Statistical Office 1993 German input-output table as well as
information on conventional pollutant and greenhouse gas emissions for
a roughly 60-sector model. In late 2002, we will update to the
1997 U.S. Benchmark and present a similar model for the Japanese economy.
Our motivation for providing
such comprehensive supplemental non-economic data support in conjunction
with the I-O tables is to facilitate environmentally conscious decision
making. One approach to support environmentally conscious decision-making
is life cycle assessment (LCA). In a U.S. Environmental Protection Agency
(EPA) document, Vigon [1993] defines life cycle assessment as “A concept
and a methodology to evaluate the environmental effects of a product
or activity holistically, by analyzing the entire life cycle of a particular
product, process, or activity. The life cycle concept has often been
referred to as “cradle to grave” assessment. A life cycle view of
a product is intended to yield environmental improvement by revealing
the complete environmental picture of a product, rather than just the
emissions generated in the usual course of production by the manufacturer.
LCA is useful beyond the scope of a manufacturer as well. Service providers,
government agencies, and other interested parties can use these methods
to consider the total impact of their global business activities.
The Society for Environmental
Toxicology and Chemistry (SETAC) life-cycle assessment technical framework
workshop report published in January 1991 summarized the status of the
field at that time and was one of the initial documents that outlined
a basis for life cycle studies [SETAC 1991]. The United States Environmental
Protection Agency (EPA) [Vigon 1993, Curran 1996] accepted and built
on the SETAC framework. The research by EPA and SETAC led to a four-part
approach to LCA that is accepted today: scooping and definition,
inventory, impact assessment, improvement analysis. The
major obstacles in performing life cycle assessments are in dealing
with boundaries and circularities within the system of study and in
collecting the necessary data. In defining the "boundary"
of the analysis, e.g., in the LCA of a paper cup by the manufacturer,
the practitioner may decide to consider only the inventory of effects
arising from the 10 highest cost items in the production process to
save time and effort. The boundary assumption is an important one, as
it draws the line around what will be excluded from consideration in
the inventory, and inevitably, from the overall assessment. The necessary
data collection and interpretation is contingent on a proper understanding
of where each stage of a life cycle begins and ends. Any effects that
lie outside of the boundary are ignored. This boundary assumption can
potentially lead to significant under-estimation of the inventory of
effects of a product across its life cycle. Circularities
arise when boundaries are drawn broadly, and ‘feedback loops’ exist,
e.g. when considering the LCA of steel there is some trucking of materials
to factories needed, and the manufacture of trucks requires steel.
Quantifying the inputs and
outputs (materials and energy use, and environmental discharges) associated
with each stage of the life cycle is an exhaustive task. This is typically
accomplished by either initiating new research to estimate the inventory
data, or consulting existing databases of inventory information. Data
collection is driven by the study’s goal. For example, existing estimates
on the environmental impact of electricity generation can be used, but
such estimates need to be relevant and specific to the particular application
to be of use in such a setting. To yield accurate results, the inventory
of electricity effects for a particular processing plant need to be
relevant to the local mix of electricity purchased. Thus, estimates
need to be available that reflect the use of renewable or non-renewable
resources used to make the plant's power. Getting such data could be
difficult, as it will rely on fuel and technology assumptions (e.g.
data on a power plant burning low-sulfur coal or using flue-gas desulfurization
for control of air emissions). In the end, the success or failure of
any LCA will depend greatly on the boundary assumptions, data quality,
and the level of economic resources available. In short, the data requirements
for even a small system can be tremendous.
Life Cycle Impact Assessment
(LCIA) further extends the analysis and interprets the results of the
inventory in order to assess the impacts of the product or project on
human health and the environment. EPA [2000] reports “Impact indicators
are used to measure the potential for the impact to occur rather than
directly quantifying actual impacts. This approach works well to simplify
the LCA process making it a more useful tool. A variety of environmental
impact indicators and associated indicators have been developed and
more continue to be used as LCA method evolves. The categories for indicators
range from a global level, such as contribution to global warming and
ozone depletion, to local impacts, such as photochemical smog formation.
As an example, a recent study conducted for the US EPA defines eight
impact categories and indicators for: global climate change, stratospheric
ozone depletion, acidification, photochemical smog, eutrophication,
human toxicity, ecological toxicity, and resource depletion.
Finally, the improvement analysis
is a systematic evaluation of the needs and opportunities to reduce
the environmental impacts, energy use, and materials use during the
lifecycle. This analysis may include both quantitative and qualitative
measures of improvements. Introductory material on life cycle analysis
is further documented in Barnthouse [1997], Curran [1996], Graedel [1995],
and Vigon [1993].
As stated above, a life cycle includes all the steps, from extracting the resources to product disposal. Each of these life cycle steps has impacts on the environment. For example, raw materials extraction results in depletion of nonrenewable resources like petroleum and ores. In the case of ores, mining machinery requires large quantities of energy, generally from burning fossil fuels that release carbon monoxide, nitrogen oxides, and particulate matter into the air. Manufacturing and disposal of a product also require energy and result in discharges. In short, every step of a product's life cycle has both inputs from and discharge to the environment. Over the life cycle, the sum of these inputs and releases can be substantial.
Many SETAC-based LCA software
tools exist to aid decision-makers in performing life cycle inventories.
Most consist of a graphical user interface front-end to a database of
existing product or process-specific inventory data. The inventory data
may be proprietary, from public data sources, or both. If boundaries
are appropriate and care is taken in selecting data sources, meaningful
and relevant results are possible. However the accuracy costs time and
money. It is not uncommon for detailed LCA studies to require hundreds
of thousands of dollars and 6 months to complete. If time is a primary
driver, results may not be available until the next production cycle,
which is often too late to make improvements.
The concern over the cost and
time required for LCA has resulted in researchers investigating methods
to simply the analysis while still retaining the information needed
to satisfy the study goals. One concept that has received attention
is Streamlining LCA. EPA 2000 describes this concept as follows; “A
continuing concern over the cost and time required for LCA encouraged
some practitioners to investigate the possibility of “streamlining”
or simplifying LCA to make it more feasible and more immediately relevant
without losing the key features of a life-cycle approach. When the concept
of streamlining was first introduced, many LCA practitioners were skeptical,
stating that LCA could not be streamlined. Over time, however, there
has been growing recognition that “full-scale” LCA and streamlined
LCA are not two separate approaches but are, instead, points on a continuum.
As a result, streamlining an LCA becomes part of the scope and goal
definition process. The key is to ensure that the streamlining steps
are consistent with the study goals and anticipated uses, and that the
information produced will meet the users’ needs. From this perspective,
the scope and goal definition process involves determination of what
needs to be included in the study to support the anticipated application
and decision”. The concept of Streamlined LCA is developed in Graedel
[1998]; Bennett [2000] presents an example using the approach.
Despite these advances, this
LCA approach still requires setting tight boundaries around the problem
to make it tractable. As we show below, the parts of the supply chain
that are outside the boundaries are generally important, leading to
significant changes in the use of resources and environmental discharges.
Thus, while this LCA approach has become easier and cheaper to apply,
it still has the inherent difficulty of excluding a significant part
of the life cycle.
The limitations of traditional
SETAC LCA methods are largely addressed in an input-output model based
LCA system. For example, production of goods and services can
be completely traced through the supply chain with EIO-LCA (i.e. the
boundary is all production in the economy). Circularity is overcome
by use of the Leontief equation. By considering additional ‘use-phase’
purchases (e.g. electricity use), the entire life cycle of implications
can be considered. For example, given appropriate information
on currency differences, a user could see the implications of producing
$1 million US dollars’ worth of steel in several countries at once.
When coupled with environmental or energy data, users could quickly
see the energy efficiency differences present in each nation.
Below we detail more of the available analytical tools available in the EIO-LCA model, discuss data sources and availability, solicit additional help in enlarging the data sets, and show an example of EIO-LCA analysis for online versus traditional book publishing in the new economy.
As noted above, the EIO-LCA
model is organized in a modular way to allow any input-output table
to be used as a basis for a Leontief type modeling exercise. It
has been designed to be extended as data is updated and other country
IO data becomes available. Currently the EIO-LCA model has the
following input-output data sources:
The U.S. models are supplemented
by publicly available sector-level data on the following non-economic
effects (with government sources noted):
As consistent with I-O modeling, any vector of effects that is available for all sectors on a per-unit-of-output basis can be included in EIO-LCA. Further, any additional national IO tables can be easily added.
It is tempting to assume that
the sale of products on the Internet is beneficial to the environment.
For example, emissions from vehicles driven to shopping malls can be
avoided, retail space can be reduced, and inventories and waste can
be reduced. However, a product ordered online may be shipped partially
by airfreight across the country and require local truck delivery. Also,
the product is likely to be packaged individually, and the packaging
may not be reused. For urban dwellers relying on public transportation,
delivery by courier service probably implies increased fossil fuel use.
Residential energy consumption will increase somewhat with additional
time spent at home shopping on-line. The adverse impacts on the environment
can be significant, and the net effect of different logistics systems
is not obvious.
Books are regularly purchased
online as well as in retail stores. This work compares environmental
and economic performance of traditional retailing and e-commerce logistic
networks for the case of books. Traditional retailing involves
a retail outlet to which books are shipped from the publisher through
distributors and warehouses. The customer then purchases the book at
a retail store and brings it home. The e-commerce model ships the book
from the publisher to a warehouse then to a courier's regional hub where
it is sent by delivery truck to the customer.
The high number of remainders
(unsold books) suggests an additional factor favoring online retailing.
After sales have peaked, these remainders are either discarded, recycled
or sold to a discount bookstore. E-commerce allows for lower inventories
and fewer remainders at the sales end of the supply chain (since there
is only one inventory point), thus possibly reaping environmental benefits
due to avoided warehousing and paper production. Romm and collaborators
suggest that the reduced building space requirements of the online retailing
imply lower energy consumption [Romm 1999]. Matthews [2000] outlined
factors contributing to environmental performance.
This work initiates the task
of quantative systems analysis comparing the two logistics systems.
We present two life cycle assessments of online versus traditional retailing,
one for the case of the US and a subsequent one for Japan. In terms
of methodology, the former uses an extended version of economic input-output
Life Cycle Assessment (EIO-LCA) [CMU GDI 2002], which includes total
supply chain effects, while the latter is based on a traditional LCA.
Both studies consider the energy consumed in distribution, packaging
and personal transport, but beyond these basic factors the focus issues
are different. The US study has a larger system boundary, and highlights
switching of truck-rail-air modes and inventory reductions associated
with online retailing, while the Japan work focuses on the effect of
population density, mode of consumer transport, and changes in residental
energy consumption.
Case Study of Energy Requirements
from Book Retailing in the United States1
The traditional method of retail,
where books are sold at retail stores, can be modeled as a series of
transport links among facilities. The books are printed, transported
to a national warehouse, and then shipped again to a regional warehouse.
From the regional warehouse, the books are transported to a retail store,
where a customer buys a book and takes it home. In addition there is
a return link for unsold copies as roughly 35% of best sellers are unsold
[Publishers Weekly 1997]. We assume all transportation is carried out
by truck, and the distance between all destinations (e.g. warehouses
and stores) is separated into segments of 805 km (500 miles). The average
consumer lives 16 km (10 miles) away from a bookstore [Brynjolfsson
2000] but consumers tend to buy more than one item at a bookstore (or
as part of a shopping trip); thus, only a round-trip distance of 5 miles
(8.3 km) was used for the round-trip to the bookstore. This model
builds upon Matthews [2001].
We assume that the 35% remainder
rate for books in traditional retail inherently causes the production
of 35% more books than sold (or a total of 386,000 books). All of these
books are transported in boxes of 10 to bookstores. Assuming that each
box is 51 x 41 x 41 cm (20 x 16 x 16 inches) and weighs 910 grams (2
pounds), the cost of each box is $1.33 [ULINE 2001].
The environmental effects of
automotive trips made by consumers to bookstores to purchase books must
also be taken into account. Assuming the fuel economy of a passenger
car to be 9.6 km/l (22.5 mpg), and the fuel economy of a light truck
to be 6.5 km/liter (15.3 mpg) [US EPA 1997], we can calculate that the
energy required per mile for a passenger car is 3.6 MJ/km (5.8MJ/mi)
and for a light truck is 5.3 MJ/km (8.6 MJ/mi). Thus, the 5 mile (8.3
km) round trips by passenger vehicle would require 29 and 43 MJ, respectively.
Assuming the fleet is 65% cars [NHTSA 1998] and 286,000 trips are made,
9.7 TJ of energy would be required for all trips. Energy is also required
to produce the fuel consumed in these trips, as seen below.
Returns of unsold books from
retailers in the traditional model are an important issue. Shipping
of returns involves an additional truck leg, which we again assume to
be 805 km (500 miles). We ignore returns from customers after purchases;
we assume they would involve similar personal trips for both traditional
and e-commerce retailing.
In the e-commerce method of
selling a book, we assume that the books are bulk shipped 805 km (500
miles) from the printer to the company’s major distribution warehouse
via truck. We also assume that this warehouse is located near or at
an air hub of a major logistics carrier (e.g. UPS or FedEx) so transfer
from warehouse to the carrier is negligible. When an order is received,
the books are then air freighted or sent by truck to a regional courier
center (again assuming a distance of 805 km) from which the books are
delivered by local courier truck to the customer’s residence.
The air shipping scenario represents a 'worst case' scenario for analysis.
The packaging used in e-commerce tends to be corrugated cardboard boxes. Using an actual amazon.com shipping box for a single book, we assume a box size of 30 x 23 x 11 cm (12 x 9 x 4.5 inches), a weight of 317 g (0.7 pounds), the cost of each box to be $0.41 [ULINE 2001] and that the books are packaged individually, we can calculate the cost in individually packaging the total shipment of $1 million worth of books as $117,000. We assume no remainders or returns in this model. However, the cost of the bulk packaging of 286,000 books also needs to be included, $38,000, for a total of $155,000.
Selling $1 million of books in the traditional model with remainders requires 386,000 to be produced and shipped given the 35% remainder rate. The total weight of shipments in the traditional model is 455 Mg (501 short tons)– including 420 Mg (463 short tons) of books and 34.5 Mg (38 short tons) of bulk packaging. A base production of $1 million of bestseller books in the e-commerce (no remainders) model requires only 286,000 books to be shipped. The e-commerce model ships a total of 338 Mg (371 short tons) in bulk (including 343 short tons of books and 29 short tons of packaging) and a total of 403 Mg (443 tons) individually. A comparison of these costs is shown in Table 1. We present estimates for two traditional models - with and without remainders. We use the 35% remainder rate to scale up costs where appropriate.
| Item | Traditional Retailing | E-Commerce by Air Retailing | |||
| Without
Returns |
W/ 35% Return | ||||
| Calculation Notes | Cost
($1000) |
Cost
($1000) |
Calculation Notes | Cost
($1000) |
|
| Packaging | $1.33 x 286,000 books/10 books per box | 38 | 51* | ($1.33 x 286,000 books/ 10 books/box) + ($0.41 x 286,000 books) | 155 |
| Bulk Truck Shipments to Warehouse | 3 trips*805 km/ trip (500 mi) of 338 metric tons (371 short tons) at $0.18/mt-km ($0.26/ ton-mile) (US DOT 1999) | 144 | 195* | Only one 805 km (500 mile) shipment of 338 metric tons (371 short tons) | 48 |
| Air Freight | None | 0 | 0 | One 805 km (500 miles) trip - 403 tons (443 short tons) at $0.55/mt-km ($0.80/ ton-mile) (US DOT 1999) | 177 |
| Local
Delivery/ Pickup |
8 km (5 miles) at $0.21/ km ($0.33/mile) for 286,000 pickups. | 472 | 472 | Local delivery charge of $1.50 for 286,000 books | 440 |
| Retailing Overhead Cost | 12% revenue with average $15/book (Meeker 1997) for 286,000 books | 515 | 695* | 4% of revenue with average $15/book (Meeker 1997) for 286,000 books | 172 |
| Return Shipping from Retailer | 805 km (500 mi) of returns (100,000 books *10 books/box * $0.26/mi) | 0 | 17 | ||
| Return Production Cost | $3.5 of 100,000 books | 0 | 350 | ||
| Total without private auto | 697 | 1,308 | 992 | ||
| Total with private auto | 1,169 | 1,780 | 992 | ||
Table 1
Estimated Costs of Logistics and Returns for Traditional versus E-Commerce
Book Retailing
With a zero return rate, the
traditional system has a slightly higher overall cost than e-commerce
but can provide immediate service to customers. But generally, a certain
proportion of the books published will remain unsold, and will be either
returned to the publisher to be recycled or sold to discount stores.
Assuming the average return rate for bestsellers of 35%, our estimate
of e-commerce retailing costs is far lower than the traditional system.
Our estimates do not include any costs associated with stock-outs in
the traditional system; the e-commerce model places books not immediately
available on back-order for eventual delivery. Purchasing a book via
e-commerce or at a bookseller are not entirely equivalent “goods”.
The recreational aspect of visiting a bookshop, for instance, is an
important factor. Also, on-line booksellers may have a wider selection
than conventional bookstores.
E-commerce logistics systems
involve more reliance upon airfreight service than truck or rail modes.
Airfreight requires much higher energy and fuel usage, with corresponding
large air pollution emissions. Figure 1 shows comparative supply chain
energy effects from trucking, airfreight and rail (CMU GDI 2002). Table
2 shows the use of energy for the trucking, air freight, packaging,
fuel production and book production for the traditional and e-commerce
retail models.
Figure 1:
Comparison of Freight Mode Energy Requirements
In order to quantify the energy
impacts associated with the production of the fuel used in passenger
vehicles, a producer price of $0.90/gallon was assumed for our calculations.
This figure was then combined with the values for the fuel efficiencies
of passenger cars and light trucks above to arrive at the dollar amount
of fuel used for passenger trips to the bookstores. Using a fleet composition
of 35% light trucks and 65% passenger cars [NHTSA 1998], the dollar
cost of fuel for one round trip to the bookstore is $0.225. For 286,000
trips to the bookstore, the monetary cost of the fuel used is $64,400.
Our results indicate significant
differences between the retail fulfillment modes, with e-commerce having
comparable impacts when compared to the traditional model with returns.
The comparison between e-commerce and the traditional model without
returns is less clear. Overall, energy from passenger vehicle trips
(including fuel production) contributes significantly. By eliminating
these trips, energy use is significantly reduced in the e-commerce model.
However, the increased air freight and packaging of the e-commerce system
outweighs much of the benefits from reduced passenger trips.
Energy use in the production
sectors was estimated using the Economic Input-Output Life Cycle Assessment
(EIO-LCA) model [CMU GDI 2002]. The economic costs from Table
1 were used as inputs to the model for the relevant economic sectors.
As seen in Table 2, the e-commerce impacts lie roughly between the zero
and 35% return traditional models. In the case of the worst-case e-commerce
air scenario, it is only slightly more energy intensive than the traditional
system. Results in Table 2 show total results (and direct only effects
in parentheses). Direct effects do not include supply chain energy
use, as found in EIO-LCA. Due to data and model uncertainties, none
of the four models show significantly different results. It is
clear from Table 2 that the increased energy used for delivery logistics
trades off with the passenger pickup component. Passenger trips
are much less efficient than freight logistics networks, thus any energy
savings would need to come from reducing the passenger transportation
component of e-commerce retailing. While not detailed in Table
2, total energy use for shipping two books per order are 57 and 46 MJ
for traditional and 48 and 44 MJ for e-commerce by air and truck.
Estimated emissions of greenhouse gases range from 4.5-5.3 kg per book
in traditional retailing and 5.1-5.7 kg for online. Greenhouse
gases increase relatively in the online method due to increased use
of air transportation.
| Trad.
(35% returns) |
Trad.
(no returns) |
E-commerce Air | E-commerce Truck | |
| Trucking (w/ returns) | 18 (15) | 12 (10) | 4 (3) | 9 (8) |
| Air | N/A | N/A | 13 (11) | N/A |
| Courier Deliveries | N/A | N/A | 37 (31) | 37 (31) |
| Logistics Subtotal | 18 (15) | 12 (10) | 54 (45) | 46 (39) |
| Book Production | 18 (2) | 13 (1) | 13 (1) | 13 (1) |
| Packaging | 3 (<1) | 2 (<1) | 9 (1) | 9 (1) |
| Passenger Trips | 34 (34) | 34 (34) | N/A | N/A |
| Passenger Fuel Production | 5 (3) | 5 (3) | N/A | N/A |
| Passenger Subtotal | 39 (37) | 39 (37) | N/A | N/A |
| Total | 78 (55) | 67 (49) | 77 (48) | 68 (41) |
Table 2
Comparative Total (and Direct) Energy Effects for Book Retailing, per
book sold, all units in MJ
US
Case Study Conclusions
We analyzed a generic scenario
for traditional versus E-Commerce retailing of a single commodity, best-selling
books. Our analytical approach can be adjusted for different assumptions
about shipping distances, return rates or shopping purchase allocations.
By altering these critical parameters, E-Commerce can be found to be
more or less costly or energy efficient than the traditional system.
It is noted that different assumptions about population density (and
thus, distances to retail stores) and order sizes significantly change
the results. Examples of potential spillover effects from electronic
commerce that were not analyzed include structural changes to the economy,
substitution of manual or physical processes to digital systems, et
cetera [Romm 1999]. Nevertheless, our base analysis case suggests
that E-Commerce sales have a cost advantage and comparable energy use
and greenhouse gas emissions. Given the assumption of air delivery of
e-commerce purchases, and underlying data uncertainties, we conclude
that e-commerce is neither more nor less energy intensive than traditional
methods.
Two conditions that should
substantially affect the environmental performance of b2c e-commerce
are population density, which is related to consumer transport modes
and distances, and the amount of goods purchased per order. Japan makes
an excellent geographical choice for studying the former factor as it
contains both densely populated urban regions with extensive public
transport and rural regions where personal automobiles are the main
form of transport.
We review the Japanese book
sector in order to set the stage for analysis. The industry functions
under a fixed price system in which discounting below the retail price
is not permitted (for domestic books) and the profit margins for bookstores
and distributors and fixed at 20% and 10% respectively.2
The central players in the book industry are the distribution companies,
who decide the inventories of bookstores and orders made to publishing
houses. Of the 7 large distribution firms, the two major ones, Nippan
and Tohan, control 80% of the book market. The publishing and bookstore
industry is by contrast quite diffuse, with a few large firms but no
clear dominating group.
79% of publishing houses are located in the Tokyo area or outskirts of Tokyo. Transport of books between publisher, distributor, and bookstore is generally contracted out to trucking firms. Shipping from the e-commerce company to consumer is handled by courier services, which are dominated by 3 large firms. Trucking is the favoured mode of transport for book shipments, with little use of rail and essentially no shipments via air. This contrasts with the US case in which air shipments are not
uncommon.
Bulk shipment (lg.)
Personal transport
Courier
service
Bulk shipment (med.)
Bulk shipment (lg.)
Printer
Distributor
Book Store
Reader’s household
e-commerce
firm
Reader’s household
Paper
&
Ink
System boundary
Figure 2: Structure of Japanese distribution systems
This section presents a physical
model of energy use for the distribution of books in Japan. The
two systems for e-commerce and traditional retail distribution are depicted
pictorially in Figure 2. The analysis of energy includes the phases
falling within the system boundary marked in the figure. In contrast
to the US study, the production phase is neglected here. The book supply
chain in Japan is managed by a few mega-distributors who practice sophisticated
supply chain management. The existence of a difference in remainders
between online and traditional retailers is unclear in this context.
The four factors included in
the simulation of energy use are the fuel used in transport by shipping
and courier services, fuel used by the consumer in travel to and from
bookstore, energy to produce packaging, and electricity and fuel consumed
at the sales point, either at the bookstore or by the consumer at home
making an e-commerce purchase.
In order to evaluate the dependence
on population density, distribution and sales in three different regions
in Japan were considered: Tokyo, Tochigi and Hokkaido. Tokyo is taken
as representative of a densely populated urban region (5,600 inhabitants/km2),
Tochigi as a suburban area (310 inhabitants/km2), and Hokkaido
as a rural area (68 inhabitants/km2).
Fuel used in transport by shipping and courier services
The two main pieces of information
needed to estimate the consumption of fuel by shipping and courier vehicles
are the typical distances between nodes in the distribution network
and the type of truck used between each node. This data was gained through
a set of interviews with seven major firms involved in publishing, distribution,
sales, and transport of books. The basic formula used was:
Energy per book [M] = distance [km] 1/truck fuel efficiency [l/km]
energy content
of fuel [MJ/l]
volume share in truck [%],
where the sum is over legs
of the distribution path. In the e-commerce case, shipments between
distributor and e-commerce firms are carried in large (10-ton) trucks.
At the e-commerce firm’s distribution centre, individual customer
orders are placed into small parcels and shipped via courier service.
The courier service uses large trucks to transport packages to regional
distribution centres, after which small (2-ton) trucks carry to individual’s
residences. For conventional retail, the leg connecting distributor
and bookstore is handled by large trucks (10-ton) as far as nodal distribution
centres, after which shipments were divided and placed in small (2-ton)
trucks for delivery to local bookstores. Fuel efficiencies were assumed
to be 3 km/l and 5 km/l for large and small trucks respectively. The
amount of energy allotted to transport of a book shipment was calculated
according to the volume (not weight) fraction of the parcel of compared
to the capacity of the truck. In the courier service leg of the e-commerce
case, the volume was calculated assuming 2 books are shipped in box
of volume 7,290 cm3 (measured from actual boxes used by Amazon.co.jp).
Fuel used by the consumer in travel to and from bookstore
The mode of transport used
and distance travelled by the consumer was estimated based on information
on the number of bookstores located in different regions [Regional databases,
1999]. Typical distances between bookstores were calculated assuming
stores are on average uniformly distributed in a prefecture. The resulting
store-to-store distances are 1 km, 5.2 km and 13.4 km for Tokyo, Tochigi
and Hokkaido respectively. The consumer typically travels half this
distance to reach the nearest store. We assumed that the total
travel distance was that for a round trip to and from the nearest bookstore,
solely for the purpose of purchasing books (i.e. not part of a shopping
trip). This is admittedly a crude estimation, only intended identify
the order of magnitude of energy use. In general consumers may travel
further than the closest store and shopping trips are often combined.
Given the above assumptions
on typical distances, the short distance to the nearest bookstore in
Tokyo suggests that walking or bicycle is the preferred mode of transport,
while personal automobiles are used in Tochigi (suburban) and Hokkaido
(rural) cases. The typical fuel efficiency of automobiles was taken
as 13 km/l [EDMC 2001].
Energy to produce packaging
Packaging for shipping from
distributor to e-commerce warehouse or bookstore is apparently the same:
distributor firms report that medium sized cardboard boxes holding about
40 books are used. Bookstores generally put books in light paper bags
or wrap in a paper cover for the customer. E-commerce firms ship
books in small cardboard boxes. We obtained representative samples of
actual packaging used and measured their weight and volume. The energy
used to produce the packaging was calculated according to life cycle
process data from the European BUWAL database [BUWAL 1996]. BUWAL data
reports that energy to produce 1 kg of cardboard is 25 MJ, and 45 MJ
for paper. Physical characteristics of packaging and estimated
energy investment are summarized in Table 3.
| Type | Use | Weight (g) | Energy (MJ) |
| Medium cardboard box (40 books) | distributor to bookstore or e-commerce firm | 734 | 18 |
| Small cardboard box (1-3 books) | e-commerce to home | 329 | 8.1 |
| Small paper bag | bookstore to consumer | 9 | .41 |
| Book cover (paper) | bookstore to consumer | 7 | .32 |
Table 3: Packaging
characteristics and energy investment
Sales point energy
Energy is also consumed at
the sales point, either at the bookstore or the home. The bookstore
case was calculated combining macro statistics on energy consumption
for utilities per unit area in retail stores [EDMC 2001], the total
area of bookstore space in Japan (Publishers 2000), and data on number
of books sold [Shuppan 2000]. The result of this calculation is that
0.68 MJ of energy for lighting, heating, and air-conditioning was consumed
per book.
Estimation of point-of-sale
energy for e-commerce is more difficult as there is little existing
data on the differential increase in residential energy consumption
due to additional time spent at home. We approached the problem by constructing
a simple model of residential energy consumption. The purchase of one
book via e-commerce was assumed to take 20 minutes, with an additional
10 minutes per additional book purchased. Residential consumption includes
electricity used by computers and lighting as well as energy for heating
or cooling. Central heating is rare in Japanese homes, climate is generally
controllable room-by room on-demand. It is assumed that making an e-commerce
purchase involves heating or cooling one room, one third of an average
residence. The energy consumption for computer use is taken at 65.5
W, which averages consumption of desktop and laptop models (Miyamoto/Tekawa/Inaba
1998). We assumed that lighting of one room typically takes 200W of
power. Annual per household use of energy for climate control is 14,250
MJ [EDMC 2001]; this is converted to wattage/room by assuming 12-hour/day
usage 365 days/year. The resulting average power consumption for climate
control is 303 W per room. Combining the above results yields
that home energy consumption associated with purchase of one book via
e-commerce is 0.7 MJ.
The estimations indicate that
point-of-sale energy consumption at the bookstore and home are very
similar in scale. This underlines the importance of including increased
residential energy consumption in analysis of ICT modes, a factor often
neglected. For instance, Romm and collaborators cite a comparison of
business energy expenditure of e-commerce and traditional booksellers
as evidence that e-commerce is energy efficient [Romm, 1999, 26]. We
suggest that the energy savings at the firm level is lost due to increased
residential consumption.
Japan Case Study
Results and Conclusions
Calculations of the energy
use associated with each of the four factors for the case of one book
are presented in Table 4. The case of two books per sale is shown in
Table 5. The results suggest a crossover in performance as population
density changes. The conventional system uses less energy in dense urban
regions due to additional packaging and courier fuel use for e-commerce.
As the population density decreases, e-commerce saves energy because
courier services are apparently more efficient than “shipping” via
personal automobiles. Not surprisingly, the energy efficiency per book
improves substantially as the number of books in an order increases.
| Unit: MJ | Shipping, Courier | Personal transport | Package | Point-of-
sale |
Total | |
| Tokyo | E-commerce | 0.15 | 0 | 8.5 | 0.7 | 9.3 |
| Traditional | 0.013 | 0 | 0.85 | 0.68 | 1.6 | |
| Tochigi | E-commerce | 0.66 | 0 | 8.5 | 0.7 | 9.9 |
| Traditional | 0.10 | 6.8 | 0.85 | 0.68 | 8.5 | |
| Hokkaido | E-commerce | 3.1 | 0 | 8.5 | 0.7 | 12 |
| Traditional | 0.37 | 14 | 0.85 | 0.68 | 16 |
Table 4. Per book energy use (1 book per purchase)
| Unit: MJ | Shipping,
Courier |
Personal transport | Package | Point-of-sale | Total | |
| Tokyo | E-commerce | 0.075 | 0 | 4.5 | 0.5 | 5 |
| Traditional | 0.013 | 0 | 0.65 | 0.68 | 1.35 | |
| Tochigi | E-commerce | 0.33 | 0 | 4.5 | 0.5 | 5.5 |
| Traditional | 0.10 | 3.4 | 0.65 | 0.68 | 5 | |
| Hokkaido | E-commerce | 1.5 | 0 | 4.5 | 0.5 | 6.5 |
| Traditional | 0.37 | 7 | 0.65 | 0.68 | 9 |
Table 5. Per book energy use
(2 books per purchase)
Although the study addresses a single commodity and a particular region, the results may reflect some general lessons on the conditions under which b2c e-commerce becomes an environmentally friendly transport technology. Packaging is evidently a key issue, both in terms of the energy investment to produce it and for courier services to realize efficient load factors. Minimization of packaging should thus be a priority for e-commerce firms. Population density and number of books per order are also important factors. Both are addressable to some extent through consumer choice, although it is unclear to what extent behaviour can be changed. This is the long-standing challenge of green consumerism. Policies and prices for shipping give e-commerce firms some influence over the size of an order.
Model Comparisons and Conclusions
In this section we compare
and integrate the results of the two case studies. The EIO-LCA and conventional
LCA methods yield qualitatively similar results - i.e. that 'crossover
points' exist between traditional and e-commerce enabled book retailing.
This implies that there is no definitive answer as to which sales and
distribution method is more energy efficient, rather it depends on the
conditions of implementation, in particular the transport modes used
by shippers and consumers. E-commerce is most favoured when shipping
modes are unaltered (i.e. no added air freight) and consumer automobile
travel is substituted by courier services.
The quantitative results of
the two case studies differ greatly. Excluding the production of books,
total energy for sales and logistics in the US ranges from 53-63 MJ
for one book versus 2-16 MJ for Japan. Some portion of this gap arises
from the difference in methodologies and factors considered. The
US study utilises input-output models, which consider the effects across
the supply chain from production (e.g. estimates of transportation energy
include the fuel cycle). A comparable Japanese input-output model
was not available. Traditional LCA models only focus on effects
from processes included in the analysis. Thus input-output models
will generally lead to higher estimates. However, the main contribution
to the gap between results for the US and Japan lies in geographical
and social differences between the two countries. Since Japan is more
densely populated and smaller than the U.S., distances for shipping
and passenger trips to bookstores are significantly smaller. Also, Japan,
with its extensive public transportation networks and train “culture”,
has a much lower reliance on the automobile. This implies a much larger
fraction of shoppers travel by train, bus, bicycle, or foot.
The differences in composition
of the distribution energies in the US and Japan suggest a distinct
set of priorities for energy savings. The logistics energy per book
in the US compares with or even exceeds the production energy, thus
remainder rates affected the result less than might be expected. Avoiding
use of air travel and shortening/aggregating travel distances becomes
a top priority. For Japan, the lower energy use in transport gives a
greater weight to packaging, remainder rates and building energy use.
As seen above, the US and Japan
models differ in data availability, sources, and method. However,
they show similar views of the energy requirements for retail book distribution.
In addition, the sensitivities of the models to assumptions like shopping
trips make potentially large differences. The end result suggests
that while the actual energy requirements for purchasing books will
differ greatly by customer, the overall estimate is that the methods
are comparable. However, estimation to a higher order faces substantial
methodological and data obstacles.
These case studies lead to some additional observations regarding future work. First, as the relative impacts of other ICT-enabled business/lifestyle models (e.g. telework and telecommuting) are considered, similar 'crossover' effects should be expected. Second, in cases where the model results are comparable, data uncertainties alone might account for the differences. Finally, since few business processes are explicitly designed to be energy efficient, studies like this are most useful as a comparative benchmark to see whether new business models can be energy efficient. Customers are unlikely to stop using a particular method merely because of energy issues; however, policy shifts may tend to discourage energy inefficient processes.
Bibliography
Ed. Barnthouse, L. et. al. Life-Cycle Impact Assessment: The State-of-the-Art, 2nd edition. The Society of Environmental Toxicology and Chemistry. Pensacola, FL. 1997.
Bennett, E.B., Graedel, T.E. “Conditioned Air”: Evaluating an Environmentally Preferable Service. Environmental Science & Technology. 34(4). 2000. Pp. 541-545.
Brynjolfsson E., Smith MD. (2000): Frictionless Commerce? A Comparison of Internet and Conventional Retailers. Mgmt. Sci. 2000;46:563-585.
BUWAL Economic Inventory of Packaging (also known as BUWAL 250) Database, Swiss Federal Office of Environment, Forests and Landscape and the Swiss Packaging Institute. 1996.
Carnegie Mellon University Green Design Initiative. Economic Input-Output Life Cycle Assessment software. http://www.eiolca.net. Accessed March 1, 2001.
Curran, M. A., Life Cycle Analysis; Island Press: New York, 1996.
EDMC Handbook of Energy and Economic Statistics (Enerugii keizai toukei youran). The Energy Conservation Center, Tokyo, Japan (in Japanese). 2001.
White Book of Personal Computers (Pasokon hakusho 1999-2000), Japan Electronics and Electronic Industry Development Association (JEIDA): ComputerAge.Co., Ltd., Tokyo, Japan (in Japanese). 1999.
Graedel, T.E. and B.R. Allenby, Industrial Ecology, Prentice Hall, NJ, 1995.
Kuhbach, Peter D., and Mark A. Planting, “Annual Input-Output Accounts of the U.S. Economy, 1997”, Survey of Current Business, January 2001.
Lester Lave, Elisa Cobas, Chris Hendrickson and Francis McMichael "Using Input/Output Analysis to Estimate Economy-Wide Discharges", Environmental Science & Technology, Vol. 29, pp. 420A-426A, September 1995.
Chris T. Hendrickson, Arpad Horvath, Satish Joshi and Lester B. Lave, "Use of Economic Input-Output Models for Environmental Life Cycle Assessment," Environmental Science & Technology, April 1998.
Lawson, Ann M., “Benchmark Input-Output Accounts for the U.S. Economy, 1992: Requirements
Tables”, Survey of Current Business, December 1997.
Matthews HS, Hendrickson CT, Lave L. Harry Potter and the health of the environment. IEEE Spectrum. 2000; November, pp. 20-22.
Matthews HS, Hendrickson CT, Soh DL. Environmental and Economic Effects of E-Commerce: A Case Study of Book Publishing and Retail Logistics. Trans. Res. Rec. 2001; 1763: 6-12.
Meeker M, Pearson S. The Internet Retailing Report, Morgan Stanley U.S. Investment Research, 1997, online http://www.morganstanley.com/techresearch/inetretail/, last accessed July 3, 2001.
Miyamoto S, Tekawa M. Inaba A. Life Cycle Assessment of Personal Computers for the Purpose of Design for Environment. Energy and Resources. 1998; 19 :75-80 (in Japanese)
United States Department of Transportation. National Highway Traffic Safety Administration National Center for Statistics and Analysis, Traffic Safety Facts 1998. Data implied from Table 3.
Editor. They shall return. Publishers Weekly. Apr 7, 1997.
Publishing Research Institute. Publishing Indicators Annual (Shuppan shihyou nenkan), by Publishing Research Institute, National Publishing Industry Association (in Japanese). 2000.
Regional databases Report on Industry Statistical Surveys (shougyou toukei chousa houkoku) , Tokyo, Tochigi, and Hokkaido (in Japanese), 1999.
Romm, J. The Internet Economy and Global Warming. Center for Energy and Climate Solutions, Technical Report, 1999.
The Publishing Yearbook 2000 (Shuppan nenkan 2000), Shuppan News Co., Ltd., Tokyo, Japan .
Introduction to the Distribution System for Books (Yoku wakaru shuppanryuutsuu no shikumi), Tohan co., Ltd., Tokyo, Japan (in Japanese), 2001.
SETAC and SETAC Foundation for Environmental Education, Inc., "A Technical Framework for Life-Cycle Assessment", Fava J, Denison R, Jones B, Curran MA, Vigon B, Selke S and Barnum J. Washington, January 1991
ULINE Shipping Supply Specialists http://www.uline.com. Accessed Dec. 10, 2001.
United States Department of Transportation. National Transportation Statistics 1999.
United States Environmental Protection Agency National Vehicle and Fuel Emissions Laboratory, EPA420-F-97-037, Annual Emissions and Fuel Consumption, 1997. http://www.epa.gov/otaq/ann-emit.htm. Accessed March 25, 2001.
Vigon, B. W.; Tolle, D. A.; Cornaby, B. W.; Latham, H. C.; Harrison, C. L.; Boguski, T. L.; Hunt, R. G.; Sellers, J. D., Life Cycle Assessment: Inventory Guidelines and Principles, U.S. Environmental Protection Agency: Washington, DC, February 1993; EPA 600/R-92/245.
WRI 1998-99 World Resources - A Guide to the Global Environment, Oxford University Press.
1 Cost calculations in this case study are based on the comparative costs involved in selling a million dollars worth (at production) of 'best seller' books, or roughly 286,000 books at an assumed production cost of $3.50 each. We assume each book is 23 x 6 x 16 cm in size (9 x 2.25 x 6.25 inches) and weighs 1.1 kg (2.4 pounds).
2 This is in contrast to the United States where online retailers aggressively discount prices to gain market share, brand recognition, and loyalty.
All Rights Reserved Powered by Free Document Search and Download
Copyright © 2011