Evolution of Product Lifespan and
Implications for Environmental
Assessment and Management: A
Case Study of Personal Computers
in Higher Education
C A L L I E W . B A B B I T T , *, †R A M Z Y K A H H A T ,‡
E R I C W I L L I A M S ,‡ , §A N D
G R E G O R Y A . B A B B I T T|
School of Human Evolution and Social Change, Arizona State
University, PO Box 872402, Tempe, Arizona 85287-2402,
Center for Earth Systems Engineering Management,
Department of Civil, Environmental and Sustainable
Engineering, Arizona State University, PO Box 875306, Tempe,
Arizona 85287-5306, School of Sustainability, Arizona State
University, PO Box 875502, Tempe, Arizona 85287-5502, and
Center for Evolutionary Functional Genomics, Biodesign
Institute, Arizona State University, PO Box 875001, Tempe,
Received December 16, 2008. Revised manuscript received
April 24, 2009. Accepted April 28, 2009.
Product lifespan is a fundamental variable in understanding
the environmental impacts associated with the life cycle
of products. Existing life cycle and materials flow studies of
products, almost without exception, consider lifespan to
this study provides an empirical documentation of the long-
university as a case study. Results indicate that over the
period 1985-2000, computer lifespan (purchase to “disposal”)
decreased steadily from a mean of 10.7 years in 1985 to 5.5
narrower over time. Overall, however, lifespan distribution
or materials flow forecasts of electronic waste management
for policy. We argue that these results suggest that, at least for
computers, the assumption of constant lifespan is problematic
and that it is important to work toward understanding the
dynamics of use patterns. We modify an age-structured model
of population dynamics from biology as a modeling approach
to describe product life cycles. Lastly, the purchase share and
generation of obsolete computers from the higher education
sector is estimated using different scenarios for the dynamics
of product lifespan.
The environmental management of computers and other
electronic goods is undergoing significant and increasing
attention from the public and from policy makers around
the world. Of particular concern is the management of
can contribute to growing economies (1) and narrowing the
digital divide (2, 3), but when recycled informally in
developing countries, it can potentially create significant
human health and environmental impacts (e.g., refs 4-7).
Although receiving less public and media scrutiny, envi-
ronmental impacts associated with electronic product manu-
products’ life cycle environmental impacts. Due to the
combination of relatively short lifespans and energy inten-
sive manufacturing, electronic products such as computers
are distinct from many “products with a plug” in that much
of the environmental burden over the product life cycle is
driven by manufacturing (8, 9).
When assessing environmental impacts associated with
both the upstream and downstream processes in electronic
product life cycles, product lifespan is obviously a funda-
mental variable of interest. Not only does lifespan dictate
replacement, but also the quality and operational charac-
teristics of obsolete equipment requiring end-of-life (EOL)
management. Furthermore, technological progress and the
in the environmental characteristics of manufacturing pro-
cesses, the performance characteristics of products, and in
how consumers purchase, use, and dispose of products.
Considering electronic products in use today, the chal-
lenges of understanding and managing impacts of rapid
technological innovations and change in product lifespan
are perhaps most clearly exemplified through personal
computers, which have been undergoing rapid changes in
in the United States is increasing at a rapid rate. The U.S.
Census Bureau estimated in 2005 that 62% of all households
had at least one personal computer (PC) in 2003, up from
23% in 1993 and 8% in 1984 (10). Similar trends are shown
in the commercial sector, as the number of computers in
commercial buildings increased from 30 million in 1992, to
43 million in 1995, and to 58 million in 1999 (11). However,
there has not been a clear empirical documentation of the
changes in computer lifespan over these same time periods.
Nevertheless, existing studies, policy forecasts, and life
almost without exception, consider lifespan or lifespan
distribution to be constant over time (e.g., refs 8, 12-17).
Although this oversimplifying assumption is often required
by the scarcity of publicly available and temporally variable
data, the inability to account for technological progress,
inherent to LCAs (18-21) and a significant limitation on the
and policy decisions.
computer lifespan and to identify methods by which this
aspect of technological progress could be quantified and
integrated into LCAs and other environmental assessment
studies. This approach was operationalized through a case
study of computer adoption and life cycle parameters in a
major U.S. university, Arizona State University (ASU), in
Tempe, Arizona. The primary objectives of this study were
lifespan evolution over the last 20 years and to present an
age-structured model that can be adapted to integrate
lifespan evolution in future e-waste and computer LCA
* Corresponding author phone: (480) 965-4543; fax: (480) 965-
†School of Human Evolution and Social Change.
§School of Sustainability.
|Center for Evolutionary Functional Genomics.
Environ. Sci. Technol. XXXX, xxx, 000–000
10.1021/es803568p CCC: $40.75
XXXX American Chemical Society VOL. xxx, NO. xx, XXXX / ENVIRONMENTAL SCIENCE & TECHNOLOGY9A
studies. A secondary objective was to use the case study
approach to provide the first characterization of trends in
life cycle parameters for personal computers in higher
and obsolete equipment generation. One motivation for
studying the education sector was to scope its importance
in e-waste generation and management at a national level.
The second motivation was that universities often collect
decades of data on computer purchase and disposal and
have an open perspective regarding public access to these
data, as compared to the private sector.
2. Case Study
(PCs) and related equipment purchased and used in U.S.
higher education. Colleges and universities were expected
to be major contributors to computer purchasing, use, and
disposal in the U.S., an assumption supported by recent
commercial computer usage surveys, which estimated that
the education sector (K-12 and postsecondary) accounted
for 21% of the approximately 58 million PCs in commercial
buildings in 1999 and that the overall education sector had
an ownership rate of 1335 PCs per thousand employees, as
compared to an average of 707 computers per thousand
employees for all commercial sectors in 1999 (11). Further-
more, colleges and universities are often viewed as hubs of
technological research innovation where students are in-
creasingly adept with computer and Internet applications
and instructors are more commonly using computer-based
instruction platforms. However, even with the intuitive
perception that higher education must be a significant
national trends or may possibly influence future impacts of
PC use, reuse or recycling, and disposal.
The case study focused on ownership and life cycle
parameters of institutionally owned computers at Arizona
it is one of the largest universities in the U.S. (a projected
enrollment of 66 000 students in 2008 and over 12 000
employees), and therefore, expected to be representative of
other large institutions, which may be likely to contribute
sector by the magnitude of their sizes. Furthermore, ASU
maintains an extensive database of all inventory purchased,
including computers and computer equipment, which
enabled a thorough investigation into computer life cycle
The case study was conducted by examining all data
available from the property control database related to
institutional computer purchase and use, including the
numbers of computers purchased each year, the purchase
price of each unit, the number of computers retired, or
discontinued from use at the university, each year, and the
age at which the unit was retired. No individual/student PC
ownership was included in this assessment. Data were
Calculations of annual PC stock at the university were
normalized on a per employee basis, where an employee
was defined to include any individual employed at the
university at a half-time rate or greater, which included
graduate students because they are expected to often be
student enrollment data were provided by the ASU Office of
The system boundary included all desktop and laptop
computers purchased by the university between 1985 and
2000, a 15-year window that allowed for examination of
be the most complete, as all computers purchased during
only mandated inclusion in the database of computers with
to $5000 in 2004). Although many individuals and depart-
ments continued to self-report all computers purchased,
there was no way of ensuring the completeness of data after
2000. A limited number of extrapolations were performed
There were two identified data quality issues inherent to
limited the use of more recent data. Second, during data
analyses, it was observed that there were a small number of
computers purchased each year (from 1985 onward) that
were never shown as having been retired from use, even
likely no longer be in use. To address these limitations,
of Engineering had continued reporting all computers
purchased, even after the university created the price
threshold. Therefore, data specific to this department were
as the university-wide data, to determine the similarity in
historic (1985-2000) and recent (2000-2005) trends and to
computers, those expected to have been retired because of
their age but still shown as being in active use.
3. Definition of Lifespan
Though perhaps seen as trivial, definition of lifespan,
particularly for computers, deserves special care and atten-
(1) the length of time a product is possessed by its first user,
(2) the length of time between first purchase of a product
from a manufacturer and its processing in the waste
management sector, and (3) the length of time between
obsolescence (i.e., lifespan of the “primary product” in use).
and using them interchangeably can skew results. For
Products (EuP) Directive, lifespan of a desktop computer
was assumed to be 6.6 years based on the first definition
(14). However, computers are often stored or put into
secondary use for significant periods. A 2004 survey showed
new computers every three years and that computers spend
disposal, recycle) (22). For this data set, lifespan using the
is three years. The EuP study uses the 6.6 year figure based
on the first definition to estimate total life cycle impacts of
a computer. This skews the results in two ways. First, the life
cycle impacts of operating the computer in question are
overestimated by a factor of around two since for around
the manufacturing impacts needed to provide computing
services are undercounted since the fact that the consumer
purchased a second computer halfway through the period
is ignored. This conceptual error in defining using lifespan
contributes to the conclusion drawn in the EuP study that
computer manufacturing does not significantly affect life
The reason we go into such detail in the above example
is to illustrate that it is crucial to clearly define and interpret
B 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. xxx, NO. xx, XXXX
lifespan appropriately. In our study, constraints on the data
set limit us to only studying the first definition where the
we do not and cannot describe the period a computer is
used by individuals in the organization or how long it will
take end-of-life equipment sold to the secondary market to
reach final disposition in waste management.
4. Statistical Data Analysis
be fit to a number of statistical models to describe the
distributions and calculate mean and variance parameters.
The fit of annual lifespan data to four distributional models
(normal, log-normal, gamma, and Weibull) were compared
by calculating negative log likelihoods and the Akaike
Information Criteria (AIC) (23). This method allows compet-
ing models to be ranked according to their AIC score, for
each data set, with the lowest AIC corresponding to the best
fitting model. Subsequently, best-fitting parameters were
density functions used). The maximization was performed
using the conjugate gradient method within unconstrained
solve blocks in the program MathCAD by Mathsoft (2001).
5. Age-Structured Forecasting
an empirical assessment of how dynamic PC lifetime
distributions could potentially affect the future generation
and characteristics of obsolete computer equipment. This
assessment was conducted by creating a dynamic age-
structured forecast of PC lifespans that would predict the
likelihood that computers purchased over time would be
retired at any given age. Age-structured or age-specific life
tables and models are used commonly in demography and
population ecology to investigate the effect of age on birth
and death rates or other demographic parameters of indi-
viduals in a population (24). In economics, “vintage capital
models” have been applied to predict the economic effect
of technological diffusion (25), derive optimum ICT lifespan
impact of industrial processes (27). Therefore, we sought to
In this approach, we defined a population cohort as all
of the PCs purchased in a given year. Annual inputs and
generation rate of obsolete equipment, respectively. The PC
lifespan was defined as the time between the institutional
purchase of a new computer and the ultimate sale and
removal of the computer from the university. It is expected
that this lifetime in some cases may include a storage phase
disposition. No secondary storage phase in the surplus
property operation was expected, as that unit sells all
computer inventory on a monthly basis.
Age-structured obsolescence rates were determined for
each cohort from 1985 through 2000, by quantifying the
age class. An age class i was defined as all PCs between the
“mortality,” or obsolescence, curves that demonstrated the
cumulative percentage of PCs in a cohort retired over time.
To assess how the age structure changed over time, the
obsolescence curves and the corresponding distributions of
computer age at retirement were created for four-year
averages between 1985 and 2000.
To forecast the effect of the age-structured obsolescence
curve on future generation of obsolescent equipment, a
distribution of expected lifespans was obtained using the
best fitting model, as determined in Section 4, with inputs
of the predicted future PC cohort sizes and lifespans. Future
PC purchase rates were forecast from 1985 to 2000 trends,
following the method of Tasaki et al. (28), assuming a
maximum penetration of 1.3 computers per person, con-
sistent with commercial computer use surveys (11) and an
estimated upper bound to ownership (29). Linear increases
in employee numbers were assumed to forecast the total
would affect projected obsolescence curves, three scenarios
were considered: (1) average computer lifespan would
decrease linearly, based on most recent data (1995-2000)
(most conservative estimate); (2) average lifespan would
follow an exponential decrease, with lifespans from 2000
onward below the point of inflection in the curve; and (3)
average lifespan would decrease linearly, based on trends
from 1985 to 2000 (most speculative estimate). Although
Scenario 3 is extremely unlikely, it is included to provide an
lower bound of expected lifespans. 2010 was selected as the
year for comparison among these three forecasts. MathCAD
(2001) was used to generate the forecast distributions.
Although lack of data, as described in Section 3, and the
lack of information about computers purchased recently,
and yet to be retired, prevented a full comparison of these
three scenarios with actual data, some preliminary assess-
ments were made. Approximately 200 data points from the
were compared to the right-censored distribution of actual
6. National Projections
After determining per-employee annual PC stocks and
stock and e-waste generation for the U.S. higher education
sector. It was assumed that universities would reach a
maximum penetration of 1.3 computers per person (11, 29).
Additionally, during the time required to reach that maxi-
curve, constructed based on the empirical case study data
presented herein. Obsolete equipment generation on a per
employee basis was determined for the three scenarios
described above, following the method of time series
level, estimates of U.S. total annual employees and enrolled
graduate students were obtained and projected from the
National Center for Education Statistics’ annual report
“Digest of Education Statistics” (http://nces.ed.gov/programs/
7. Historical Results for Computer Stock and Lifespan
The annual stocks and percent ownership by employees of
systems, is shown in Figure 1, for the period 1985 through
2000. This figure indicates that while ASU population
(employees and grad students) is growing, the increase of
For example, in 1985, 1990, and 2000, ownership rates were
4, 36, and 110%, respectively. For the same period, however,
the calculated log-normal mean lifespan for PC cohorts
representing standard deviation of lifespans within the
VOL. xxx, NO. xx, XXXX / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 C
of computers from the entire institution and within a
subsampled department, the Fulton School of Engineering
log-normal distribution parameters for each year in con-
sideration. As determined by Akaike Information Criteria
of data, and delta-AIC and best-fit parameters for each year
are provided in SI Table S1.
The age-structured computer obsolescence curves are
introduced in Figure 3, and show the cumulative fraction
of obsolete equipment generated in each age class of
1993-1996, and 1997-2000. Initially, the obsolescence
curve shifts rapidly to the left, showing more rapid
obsolescence over a shorter window of time. For example,
for computers purchased between 1985-1988, less than
10% had been retired by the 6th year in use. However, this
percentage increased to 35% for computers purchased
between 1989-1992 and to approximately 65% when
and 1997-2000. While the most drastic shift in the
obsolescence curve was after the first two periods, the
difference in curves for 1993-1996 and 1997-2000 was
much smaller. The major shift perhaps reflects the
decreasing use phase of PCs at the university during this
time and/or a decreasing amount of time for units to be
kept in storage before leaving the system boundaries.
Although it may be intuitively expected that shifts to
be directly correlated with decreasing lifespans, over this
period of observation, we only observed very gradual
decreases in price and no significant difference in mean
lifespans among different types of computers (SI Figure
S1). Although we observe increasing shares of notebooks,
the annual purchase cohorts are comprised by 60% or
greater of IBM-compatible desktop computers (SI Figure
S2), indicating that observed trends are not necessarily
linked to technology shifts between desktops and laptops.
Resulting distributions of PC lifespans for each of the
four-year periods also show changing trends over time
(Figure 4), specifically the decrease of computer lifespan
and a narrowing of the lifespan range. For example, for
equipment leaving the university for recycling, disposal,
or reuse might be between 6 and 12 years old. However,
for computers purchased from 1997 to 2000, it was more
likely that obsolete equipment retired from the university
would be five to seven years old. Year-by-year lifespan
distributions were also created and are provided in SI
8. Forecasting Results of Computer Stock and Lifespan
It is expected that the observed trends of decreasing
lifespan, shifting of the obsolescence curves, and a
narrowing lifespan range may continue in the future. It is
also likely that electronic equipment will continue to
To consider these potential trends, future scenarios of PC
lifespans were created with results shown in Figure 5.
from 1995 onward, extrapolated to 4.7 years in 2010, (2)
FIGURE 1. Annual computer stock and penetration rate at
Arizona State University.
FIGURE 2. Mean lifespan of PCs at ASU and subsampled in the
Fulton School of Engineering (FSE). Error bars are standard
deviation among all computers in each sampling group.
TABLE 1. Lifespan Distribution Parameters
FIGURE 3. Average obsolescence curves for computer cohorts
from four time periods.
D 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. xxx, NO. xx, XXXX
to 3.5 years in 2010, and (3) PC lifespan decreases linearly
from 1985 onward, extrapolated to 1.3 years in 2010.
To consider the effect of decrease in lifespan on
obsolescent equipment generation and to demonstrate
the utility of the age-structured model lifespan, distribu-
tions for the three scenarios were created for year 2010
(Figure 6). These scenarios indicate that there are drasti-
equipment generated, which illustrates the challenges of
recommending future e-waste management strategies for
potentially highly variable waste streams. Clearly, the
concerns and options for managing a waste stream
comprised of 1-4 year old computers with similar
be vastly different than managing a waste stream in which
computers might be anywhere between 2 and 8 years old
and would likely have different generations of hardware
and technology (Scenario 1).
Although it is impossible to determine a priori which
scenario will be most accurate for computers purchased
in the future, we can get an indication of trends using
limited data available more recently. The distribution of
cohort (n ) 196) was compared to projected 2005
distributions of the three scenarios for the same sample
size. Results (shown in SI Figure S4) indicate that this
sample distribution follows Scenario 2 very closely overall
for the years in which the retirement ages are known.
9. Scoping the Importance of Higher Education in
The challenges described herein, associated with the man-
agement of a varied and increasing flow of obsolete com-
puters from universities, would be magnified at a national
level if trends observed at ASU were consistent across the
importance of this sector at the national level, computer
ownership and obsolete equipment generation trends and
forecasts were extended to the national level (SI Figure S5).
education in the U.S. could contain a stock of 13 million
computers in 2010 and be responsible for generating ap-
proximately 2.5 million obsolete computers per year. Obvi-
lack of other institutional benchmarks, but these estimates
education and the first approximation of this nature.
evolution of PC lifespan is an important variable which
can influence forecasts, management strategies, and
as well as the energy use and emissions associated with
the use phase of computer life cycles. In addition, two
important developments of computers and their relation-
ship with society noted in this papersdecrease of lifespan
and increase in ownershipsneed to be considered more
and management of computers. Prior studies intended
for policy guidance, such as a recent assessment by the
U.S. Environmental Protection Agency (17), forecast e-
waste generation assuming a static lifespan applied to all
years considered. Without a more complete inclusion of
FIGURE 4. Average lifespan distributions for computers cohorts from four time periods.
FIGURE 5. Three projected scenarios of PC lifespan change
over time. Note: Scenario 1: linear (from 1995) decrease in
lifespan, to 4.7 years in 2010; Scenario 2: exponential decrease
in lifespan, to 3.5 years in 2010; Scenario 3: linear (from 1985)
decrease in lifespan, to 1.3 years in 2010.
VOL. xxx, NO. xx, XXXX / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 E
lifespan evolution, forecasts of e-waste generation may be
significantly skewed and/or lead to oversimplified future
characteristics of the waste flow. In addition, LCA studies
distribution of lifespans, may underestimate the natural
variability in product use phase energy consumption. As
trends in obsolete equipment storage times, user reuse
patterns, and new equipment purchase intervals become
more clear, the definition of lifespan can be continually
to be done for different computer users, such as in
trends in evolution of lifespan may be observed, implying
that sustainable use and management of electronic
products and e-waste at a national level will require a
more in-depth understanding of the dynamic nature of
lifespan and material flows for electronic products.
While this case study focused on personal computers,
lifespan dynamics could affect the life cycle assessment and
understood at present. The age-structured model and
forecasting approach presented here can contribute to
development of methodology to account for technological
progress and lifespan evolution in LCA and other types of
We gratefully acknowledge and thank Kerry Suson and
Keith Elgin of Arizona State University Capital Asset
Vadlapudi and Erin Daugherty for assistance in data
extraction and analysis. This work was supported in part
by the U.S. National Science Foundation via Grant CBET-
0731067 in the Environmental Sustainability program and
by an Industrial Ecology Fellowship from the AT&T
Supporting Information Available
Additional results and data, including disaggregated
annual lifespan distributions, comparisons of projected
scenarios with data, and the scope of computer stock for
the U.S. higher education sector. This material is available
free of charge via the Internet at http://pubs.acs.org.
Association of Electronics Recyclers (IAER): Albany, NY, 2006.
(2) Bridging the Digital Divide: Innovation Systems for ICT in
A., Muchie, M., Eds.; Adonis & Abbey Publishers Ltd.: London,
(3) Warschauer, M. Technology and Social Inclusion: Rethinking
the Digital Divide; The MIT Press: Cambridge, MA, 2004.
2008, 42 (17), 6446–6454.
(5) Exporting Harm: the High-tech Trashing of Asia; Basel Action
E-waste/technotrashfinalcomp.pdf (accessed Aug 23, 2007).
(6) E-waste Chain in Guiyu, Guangdong Province. China Envi-
ronmental Regulation Net. Qinghai Meng; 2004. http://www.
riel.whu.edu.cn/show.asp?ID)1420 (accessed March 2007).
(7) Recycling of electronic wastes in China and India: workplace
and environmental contamination, Technical note 09/2005
(Section 1); Greenpeace: Exeter, UK, 2005; http://www.
of-electronic-wastes.pdf (accessed March2007).
(8) Williams, E. Energy intensity of computer manufacturing:
hybrid analysis combining process and economic input-
output methods. Environ. Sci. Technol. 2004, 38 (22), 6166–
(9) Williams, E.; Ayres, R.; Heller, M. The 1.7 kg microchip: Energy
Sci. Technol. 2002, 36 (24), 5504–5510.
(10) Cheeseman Day, J.; Janus, A.; Davis, J. Computer and Internet
Use in the United States: 2003; U.S. Census Bureau Special
Studies: Washington, DC, 2005; pp 23-208.
(11) Computers and Photocopiers in Commercial Buildings; Energy
Information Adminstration, U.S. Department of Energy: Wash-
ington, DC, 2002; http://www.eia.doe.gov/emeu/cbecs/pc_
copier/pccopier99.html (Accessed June 2008).
(12) Choi, B.-C.; Shin, H.-S.; Lee, S.-Y.; Hur, T. Life cycle assessment
of a personal computer and its effective recycling rate. Int. J.
Life Cycle Assess. 2006, 11 (2), 122–128.
(13) Development of Environmental Performance Indicators for Ict
Products on the Example of Personal Computers; EPIC-ICT
Consortium: University of Stuttgart, Leinfelden-Echterdingen,
Germany, 2006; http://www.epic-ict.org/down/publications/
EPIC_Final_activity_report.pdf (Accessed June 2008).
Monitors, Final Report (Task 1-8) European Commission DG
Sweden, 2007; http://www.ecocomputer.org/ (Accessed July
(15) Kang, H.-K.; Schoenung, J. M. Estimation of future outflows
in California. J. Hazard. Mater. 2006, B137, 1165–1174.
FIGURE 6. Forecast computer lifespan distribution due to
changing PC lifespan, for cohorts purchased in 2010.
F 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. xxx, NO. xx, XXXX
(16) Socolof, M. L.; Overly, J. G.; Kincaid, L. E.; Geibig, J. R. Desktop Download full-text
U.S. Environmental Protection Agency: Washington DC, 2001.
(17) Electronics Waste Management in the United States, EPA530-
R-07-004a; U.S. Environmental Protection Agency, Office of
Solid Waste: Washington, DC, 2007.
distributions of emissions. J. Ind. Ecol. 2001, 4 (2), 71–91.
(19) Levine, S. H.; Gloria, T. P.; Romanoff, E. A dynamic model for
J. Ind. Ecol. 2007, 11 (4), 39–49.
(20) Mendivil, R.; Fischer, U.; Hirao, M.; Hungerbu ¨hler, K. A. New
LCA methodology of technology evolution (TE-LCA) and its
application to the production of ammonia. Int. J. Life Cycle
Assess. 2006, 11 (2), 98–105, 1950-2000.
problems in life cycle assessment, Part 2: impact assessment
and interpretation. Int. J. Life Cycle Assess. 2008, 13, 374-
in Japan and associated life cycle energy use, in 2005.
Proceedings of the IEEE International Symposium on Elec-
tronics and the Environment, Piscataway, New Jersey, 2005,
Verlag: New York, NY, 2002.
(24) Gotelli, N. J. A Primer of Ecology, 2nd ed.; Sinauer Associates,
Inc: Sunderland, MA, 1998.
(25) Chari, V. V.; Hopenhayn, H. Vintage human capital, growth,
and the diffusion of new technology. J. Pol. Econ. 1991, 99 (6),
(26) Yatsenko, Y.; Hritonenko, N. Network economics and optimal
replacement of age-structured IT capital. Math. Methods Oper.
Res. 2007, 65, 483–497.
model of the U.S. pulp and paper industry. J. Ind. Ecol 2005,
9 (3), 191–211.
(28) Tasaki, T.; Takasuga, T.; Osako, M.; Sakai, S. Substance flow
analysis of brominated flame retardants and related com-
pounds in waste TV sets in Japan. Waste Manage. 2004, 24,
(29) Yang, Y.; Williams, E. Logistic model-based forecast of sales and
Soc. Change 2009, DOI: 10.1016/j.techfore.2009.03.004.
VOL. xxx, NO. xx, XXXX / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 G