Journal of Environmental Management 90 (2009) 3322–3331

Contents lists available at ScienceDirect

Journal of Environmental Management

journal homepage: www.elsevier.com/locate/jenvman

How much e-waste is there in US basements and attics? Results from

a national survey
Jean-Daniel M. Saphores a, *, Hilary Nixon b,1, Oladele A. Ogunseitan c, 2, Andrew A. Shapiro d, 3
a
Civil and Environmental Engineering, Economics, and Planning, Policy, & Design Departments, University of California, Irvine, CA 92697 USA
b
Department of Urban and Regional Planning, San Jose State University, San Jose, CA 95192, USA
c
Department of Population Health & Disease Prevention, University of California, Irvine, CA 92697, USA
d
Department of Electrical Engineering & Computer Science, University of California, Irvine, CA 92697, USA

a r t i c l e i n f o a b s t r a c t

Article history: The fate of used electronic products (e-waste) is of increasing concern because of their toxicity and the
Received 28 July 2008 growing volume of e-waste. Addressing these concerns requires developing the recycling infrastructure,
Received in revised form but good estimates of the volume of e-waste stored by US households are still unavailable. In this context,
29 March 2009
we make two contributions based on a national random survey of 2136 US households. First, we explain
Accepted 3 May 2009
Available online 4 June 2009
how much e-waste is stored by US households using count models. Significant explanatory variables
include age, marital and employment status, ethnicity, household size, previous e-waste recycling
behavior, and to some extent education, home ownership, and understanding the consequences of
Keywords:
E-waste recycling, but neither income nor knowledge of e-waste recycling laws. Second, we estimate that on
Poisson regression average, each US household has 4.1 small (10 pounds) and 2.4 large e-waste items in storage. Although
Negative binomial regression these numbers are likely lower bounds, they are higher than recent US Environmental Protection Agency
Inflated counts (EPA) estimates (based on narrower product categories). This suggests that the backlog of e-waste in the
US is likely larger than generally believed; it calls for developing the recycling infrastructure but also for
targeted recycling campaigns.
Ó 2009 Elsevier Ltd. All rights reserved.

1. Introduction improper handling causes widespread damage to human health

and to the environment (Grossman, 2006).
The fate of used electronic products (e-waste) is of increasing The difficulty of managing e-waste is further compounded by the
concern because of the toxicity of e-waste and its growing volume. increasing popularity of CEDs linked to innovation and to falling costs.
Indeed, consumer electronics devices (CEDs) contain numerous From 1990 to 2007, the average number of functioning CEDs in
toxic materials, such as mercury, lead, zinc, and cadmium, which a typical household increased from 10 to 25 according to the
pose a threat to public and environmental health if CEDs are Consumer Electronics Association (Consumer Electronics Association,
improperly discarded. As one of the fastest-growing segments of 2008). As a result, there is an ever-expanding pile of used and obsolete
the municipal solid waste stream, the US Environmental Protection CEDs ready to enter the municipal waste stream. In addition, the
Agency (US EPA) estimates that approximately 2.3 million tons of e- Federal Communications Commission-mandated switch from analog
waste are sent to landfills and incinerators each year (US EPA, to digital broadcasting by June 12, 2009 (Federal Communications
2006). Although some states, such as California, mandate domestic Commission, 2009) creates an additional challenge for e-waste
recycling for e-waste, anecdotal evidence suggests that a significant managers and environmental policymakers. The 20% of US house-
percentage of e-waste ends up in developing countries where its holds who do not own a digital television and who receive only free,
over-the-air programming (US Government Accountability Office,
2007) will be unable to view any television broadcast after this date
* Corresponding author. Tel.: þ1 949 824 7334; Fax: þ1 949 824 8566. unless they purchase a digital television or an analog-to-digital
E-mail addresses: saphores@uci.edu (J.-D.M. Saphores), hilary.nixon@sjsu.edu converter box. This may trigger an ‘‘e-waste tsunami’’ as millions of
(H. Nixon), oladele.ogunseitan@uci.edu (O.A. Ogunseitan), aashapir@uci.edu (A.A. analog televisions become obsolete (Swann, 2007).
Shapiro).
1
Tel.: þ1 408 924 5852.
E-waste is by no means only a US problem, as illustrated by the
2
Tel.: þ1 949 824 6350. situation in the European Union (EU) and Japan. Even though the
3
Tel.: þ1 818 393 7311. EU has passed Waste Electric and Electronic Equipment directive

0301-4797/$ – see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jenvman.2009.05.008
 J.-D.M. Saphores et al. / Journal of Environmental Management 90 (2009) 3322–3331 3323

(that prescribes a collection rate of 4 kg per capita), only 40% of Toxic Characteristic Leaching Procedure (TCLP) for lead, which is
larger appliances are collected for salvage and recycling; the used by the US to examine if a solid waste is hazardous, although
collection rate is 25% for medium sized household items and close actual samples of landfill leachate do not present toxic levels above
to zero for most small appliances (Science Daily, 2007). In Japan, the legal limits. Lincoln et al. (2007) focus on cell phones; they show
Law for the Recycling of Specific Home Appliances, which covers that different tests give conflicting information about toxicity and
televisions, refrigerators, washing machines and air conditioners, find that cell phones fail the TCLP for lead and California’s total
has created a big incentive to export obsolete and unusable appli- threshold lLimit concentration (TTLC) test for copper, nickel, lead,
ances to China and South-East Asia by requiring consumers to pay antimony, and zinc. Finally, in their study of computers, Li et al.
for transportation and recovery costs unless these appliances are (2006) report that motherboards are of most concern for lead
sold for ‘‘reuse’’ (Chung and Murakami-Suzuki, 2008). A similar leaching.
problem exists for used personal computers owned by businesses A handful of studies examine the economic and financial aspects
(regulated by the Law for the Promotion of Effective Utilization of of recycling e-waste. Macauley et al. (2003) analyze the costs and
Resources). the environmental benefits associated with computer monitor
Both public demands to curb e-waste exports to developing recycling under different policy options. More recently, in an
countries and concerns about e-waste toxicity are putting pressure attempt to better understand consumer willingness to pay for
on policymakers and on recyclers to expand the recycling infra- ‘‘green’’ electronics and environmentally-responsible end-of-life
structure. To do this effectively, however, requires good estimates of management, Saphores et al. (2007) find that Californians are
the large volume of e-waste already stored by US households. willing to pay a 1% premium, while Nixon and Saphores (2007)
Unfortunately there is currently no reliable, survey-based estimate of report that Californians will support a 1% advanced recycling fee on
this volume. The handful of published studies on this topic relies on electronics to fund recycling programs.
industry sales data, estimated product lifespan, and assumptions A major concern regarding e-waste is shipments from devel-
regarding consumer behavior (see, e.g. Carnegie-Mellon University, oped to developing countries. Wong et al. (2007) conduct an
1991; Matthews et al., 1997; National Safety Council, 1999). extensive soil, air, and water analysis in the city of Guiyu, China,
In this context, we make two contributions based on a national which is a notorious treatment center for imported e-waste. They
survey of 2136 US households. First, we estimate count models such find high concentrations of heavy metal contaminants from elec-
as Poisson, negative binomial regression and the corresponding tronics throughout the community. Although developing countries
inflated models to explain how many small and large e-waste items receive large volumes of e-waste from developed countries, they
are stored by US households using a wide range of demographic, are also becoming a major source of domestic e-waste, particularly
socio-economic and attitudinal variables. Second, we generate in India and China. Hicks et al. (2005) analyze the legislative and
simpler count models to predict the number of small and large e- market responses to e-waste recycling and disposal in China; they
waste items held in storage by US households using weights that find that the existence of an extensive informal sector, combined
correct for some potential biases in our sample. with a lack of environmental awareness, are complicating the
This paper is organized as follows. In the next section, we development of a financially and environmentally sound recycling
present a brief review of the relevant literature. We then introduce and disposal system. Liu et al. (2006) study changes in e-waste
our survey instrument and the data we collected before motivating generation and management in Beijing prompted by a new Chinese
the models we estimated. In Section 4, we summarize our results law that mandates local governments to take responsibility for end-
before discussing our conclusions and some of the policy implica- of-life management of e-waste. Recently attention has also turned
tions of our research in Section 5. to e-waste in Africa. Osibanjo and Nnorom (2007) and Schmidt
(2006) focus on Nigeria, where most e-waste brought by interna-
2. Literature review tional trade is junk. Currently, Africa has neither the appropriate
waste management infrastructure to deal with problem-waste nor
In recent years, the number of published studies dedicated to e- the institutional capacity to enforce e-waste bans.
waste has increased as e-waste has become of increasing concern Finally, in order to design effective e-waste recycling programs,
for NGOs, the public, and policymakers. This growing literature it is essential to understand consumer preferences for recycling.
covers a wide range of topics from general e-waste issues; leaching Saphores et al. (2006) find that previous experience with recycling
studies; the economics of recycling; e-waste management in (e.g. conventional recyclables such as newspapers), in addition to
developing countries; and e-waste recycling preferences. We convenience, is directly related to willingness to recycle electronics.
briefly review this body of literature before focusing on two aspects It is therefore advisable to establish recycling programs first in
that are of particular interest for our study: the development of the communities with conventional recycling. In a recent paper, Nixon
e-waste recycling infrastructure and estimates of the quantity of et al. (2009) examine the impact of environmental attitudes and
stockpiled e-waste by households. beliefs on willingness to pay for increasing the convenience of
Electronic waste is a relative newcomer on the list of environ- recycling electronic waste; their paper provides a useful review of
mental concerns. Widmer et al. (2005) provide a comprehensive the recent research linking environmental attitudes and beliefs
analysis of salient issues, review key relevant legislative initiatives, with recycling behavior, which we do not have space to review
and compare e-waste recycling practices in Beijing, Delhi, and herein.
Johannesburg. Similarly, Babu et al. (2007), Herat (2007), and (Selin
and VanDeveer, 2006) give excellent overviews of the e-waste 2.1. E-waste recycling infrastructure
problem; a key strength of these articles is their analysis of
management policies (e.g. extended producer responsibility) as it With the passage of numerous state regulations banning e-waste
relates to consumer electronics. from household trash and with increasing concerns about exporting
A major concern about e-waste is the toxic materials it contains e-waste to developing countries for processing, it is even more
and their potential for contaminating groundwater and soils if they pressing now to develop the US recycling industry. As of May 2008,
are improperly discarded. Several studies have examined the 11 states (AK, CA, CT, MA, MA, MN, NC, NH, NJ, OR, RI) have passed
leaching potential of e-waste in landfills. Jang and Townsend bans on landfilling CRTs, and in some cases, other electronic items. At
(2003) find that printed wiring boards from computers fail the the federal level, households and small generators (<220 lb) are
3324 J.-D.M. Saphores et al. / Journal of Environmental Management 90 (2009) 3322–3331

exempt from regulation under the Resource Conservation and research, which analyzes a range of consumer electronic products,
Recovery Act. deals primarily with the recycling and reuse industries in the US.
Although some states require manufacturers to take responsi- In two 2007 reports, the US Environmental Protection Agency
bility for the end-of-life management of regulated electronic (US EPA) tries to develop a better understanding of the current
devices, in most cases, this responsibility falls to local municipali- status of the electronics industry and to generate new estimates on
ties. In fact, numerous local laws also regulate e-waste disposal. In the quantity of stockpiled material nationwide. In the first report
its 2004 report on cathode ray tube (CRT) management, the Cal- (US EPA, 2007a), market research sales data are combined with
ifornia Integrated Waste Management Board (2004) reports that estimated product lifespans to estimate the quantity of e-waste
financial issues top the list of local governments’ concerns, followed available for end-of-life management annually. In 2005, it estimates
by end markets. Indeed, the lack of a reasonable assessment of the that 121.9 million CRTs were in storage nationwide. The second
quantity of e-waste awaiting recycling and uncertainly about end report (US EPA, 2007b) relies on government sales data instead, but
markets makes it difficult to establish an economically-strong otherwise follows a similar approach. It finds that, in 2005, more
recycling industry. These concerns are also emphasized in a 2006 than 77 million pounds of lead from e-waste were landfilled, 2.2
report published by the International Association of Electronics million pounds were incinerated, and only 15.4 million pounds
Recyclers (IAER, 2006). were recycled.
A few studies specifically focus on modeling the waste flow of A couple of studies adopted a different approach, however. A
consumer electronics as it relates to the recycling industry, 2001 California Integrated Waste Management Board study (Cal-
including Linton and Yeomans, (2003) and Kang and Schoenung ifornia Integrated Waste Management Board, 2001) analyzes the
(2005, 2006). A common theme across these papers is the impor- capacity of the recycling industry to process CRTs. Upon extrapo-
tance of knowing how much e-waste has been stockpiled in order lating results of a telephone survey of 1003 California households to
to predict the demand for recycling. the whole state, it estimates that 6.1 million CRTs were stockpiled in
(Linton and Yeomans, 2003 examine the role of industrial California homes in 2001. Similarly, a 2003 study of 800 Florida
ecology in a sustainable economy using televisions as an example. households asked residents to indicate the number of working,
They develop a forecast through 2050 of television disposal in the non-working, and broken televisions and computer monitors in
US based on sales data, estimated product lifespan, and end-of-life their home (Florida Department of Environmental Protection,
management behavior. To deal with technological change, which is 2003); it calculates that Floridian households have approximately
a key source of uncertainty in their model, and to examine the 3.35 million CRTs in storage.
sensitivity of their results to various assumptions, they rely on Based on our review of the relevant literature and on govern-
Monte Carlo simulation. In addition to recommendations about end ment publications, it appears that there are no recent nationwide
markets for CRT glass, they underscore the importance of under- survey data of households to estimate the quantity of stockpiled e-
standing waste flows for developing a stable recycling industry. waste. Other studies rely on sales data and assumptions regarding
Kang and Schoenung (2005) are concerned with CRT glass, product lifespan and consumer EOL management behavior. Our
plastics, and metal recycling. They review the capabilities of the US study therefore fills this gap and provides an opportunity to
electronics recycling industry and discuss recycling technologies compare estimates obtained via different approaches.
before arguing that a steady supply of inputs and a stable demand
for outputs is required for the electronics recycling industry to 3. Methods
mature. In a subsequent study, Kang and Schoenung (2006)
conduct a time-series materials flow analysis to estimate the 3.1. Survey instrument and data
number of central processing units (CPUs) and monitors available
for recycling through 2013. Like (Linton and Yeomans, 2003, they Our unique dataset was collected via an Internet-based survey
explore how technological change will influence the material of a random subset of Knowledge Networks’ (KN) online research
composition for end-of-life management. They also emphasize the panel, which currently has approximately 43,000 members. KN,
need to know both ‘‘when’’ and ‘‘how much’’ e-waste will be which was founded in 1998 by two Stanford University professors,
available to develop the necessary recycling infrastructure. is a fast growing market research firm that provides survey-based
information for commercial, academic, and policy research appli-
cations. Unlike typical Internet surveys that rely on self-selected
2.2. E-waste inventory respondents, KN has built a panel representative of the US pop-
ulation based on probability sampling. Prospective panel members
As electronic products have gained in popularity, many house- are randomly recruited by telephone using random digit dialing
holds have stockpiled e-waste because they do not know how to (RDD) sampling techniques that cover the country’s entire resi-
safely and conveniently dispose of it. Various estimates suggest that dential population with telephone access. KN’s approach meets the
this backlog is huge, but its size is still quite uncertain. To date, only Federal Government’s quality standards for RDD surveys.
a few published studies explore this issue, and almost all of them Once a household accepts to join KN’s panel, its members are
rely on a combination of sales data, estimated product lifespan, and asked to provide socio-economic and demographic information.
assumptions about consumer behavior. This core information, which is updated every year, is available for
The pioneering study on the accumulation of e-waste must be subsequent surveys. If needed, household members are provided
credited to Carnegie-Mellon University (1991). It examines the with a free WebTV appliance and Internet access in exchange for
waste stream generation of three products (personal computers, taking part in online surveys. Households who already have
refrigerators, and telephones) to develop design strategies that a computer and Internet access are asked to answer surveys using
would foster the reuse, remanufacturing, or recycling of these their own equipment but they receive points redeemable for cash
items. A follow up study by Matthews et al. (1997) focuses on upon completing surveys. Panel members need to complete at least
personal computers; it revises the assumptions underlying the one of every six surveys they are assigned to remain on KN’s panel;
1991 study to better reflect current market conditions. We should however, they are never asked to complete more than four surveys
also mention the 1999 National Safety Council study, which is one in a month. Panel members are notified by email or conventional
of the most widely cited studies on e-waste generation. This mail when they have been assigned an Internet survey. Email
 J.-D.M. Saphores et al. / Journal of Environmental Management 90 (2009) 3322–3331 3325

reminders are sent to non-respondents, followed by a phone ones (see Fig. 1). As expected, households tend to have more small
reminder after at least 3 days. For more details about KN’s meth- than large e-waste items (3.99 on average compared to 2.32), but
odology, see (KN, 2009a); for a recent bibliography of academic we observe a lot of variability in the data: the standard deviations
papers that analyze data collected using KN’s panel, see (KN for the counts of small and large items are 5.07 and 3.83 respec-
,2009b). tively. For modeling purposes, we also note that a large number of
After getting feedback on a preliminary version of our survey respondents report having no e-waste in storage.
from students and colleagues from the University of California, For generalizing our results, it is important to explore the
Irvine (UCI), we asked KN to conduct a pilot study of our survey. representativeness of our sample. Compared to 2000 US Census
Between February 15, 2006 and February 20, 2006, 167 panelists data, we find that our respondents’ demographic characteristics are
were solicited and 110 completed our survey. Based on their generally similar to those of the US population (see Table 1). Our
answers and comments, we updated our survey to clarify some respondents are slightly older, more likely to be married, and more
questions and reduce the median completion time. The revised likely to live in and own a single-family home. In addition, our
survey was then fielded from March 28, 2006 to April 7, 2006; 2136 sample is less ethnically diverse than the population (Hispanics and
out of 3048 solicited panelists answered our questions, for African Americans are under-represented); a lower percentage of
a completion rate of 70.1%, which is typical of KN’s online surveys. our respondents are without a high school degree; and households
Our survey had four parts. The first part asked some general with an annual income over $100,000 are under-represented.
questions about environmental beliefs and involvement with However, these differences are relatively minor so we believe our
voluntary organizations, which have been found to play an analysis will yield useful insights regarding the quantity of e-waste
important role in explaining recycling behavior (e.g., see the stored in US households.
references in Nixon et al., 2009). The second part gathered infor-
mation about household trash and recycling activities; a few 3.2. Modeling strategy
questions also dealt with recycling household hazardous waste,
which was the focus of the third part. Finally, the last part asked For modeling e-waste counts for both small and large items, we
additional demographic questions to complement the data proceed in two steps. First, we seek to explain how many e-waste
provided by KN. A copy of our survey instrument is available upon items are stored by households using the best available informa-
request. tion, including data on environmental beliefs and past recycling
Analyses presented in this paper concentrate on the data behavior. Unfortunately, not all respondents provided this type of
collected in Part 3, where we asked our respondents to inventory information, so in a second step, we rely only on socio-demo-
small and large e-waste items. For the former, we explicitly asked to graphic information provided by Knowledge Networks to build
account for small TVs (21 inches or less), small computer monitors simplified models which we use to predict counts for small and
(15 inches or less), laptop computers, conventional phones and cell large items for all of our respondents. We then combine these
phones, lightweight printers, and ‘‘other items’’ such as handheld predicted counts using the weighting scheme provided by KN to
electronics, keyboards, or digital cameras. For large e-waste items, obtain averages for the entire US population.
we explicitly inquired about TVs over 21 inches, computer monitors
over 15 inches, desktop computers, as well as large consumer 3.2.1. Principal components analysis
electronics products weighing more than 10 lb each, such as large To limit our long list of explanatory variables, we condense
printers, stereos, old-fashioned radios, large fax machines or eight survey questions (see Table 2) on recycling beliefs by prin-
microwave ovens. A total of 1630 respondents provided the cipal components analysis (PCA; Kline, 1994); our goal here is to
required information for small e-waste items and 1648 for large create meaningful indexes to help interpret our results. PCA is an

Fig. 1. Reported number of small and large e-waste items. For small e-waste items, we explicitly asked our respondents to account for small TVs (21 inches or less), small computer
monitors (15 inches or less), laptop computers, conventional and cell phones, lightweight printers, and ‘‘other items’’ such as handheld electronics, keyboards, or digital cameras. For
large e-waste items, we explicitly inquired about TVs larger than 21 inches, computer monitors over 15 inches, desktop computers, as well as large consumer electronics weighing
more than 10 lb each, such as larger printers, stereos, old-fashioned radios, large fax machines or microwave ovens. A total of 1630 respondents provided the required information
for small e-waste items and 1648 for large ones.
3326 J.-D.M. Saphores et al. / Journal of Environmental Management 90 (2009) 3322–3331

Table 1 considered three questions dealing with moral values related to

Selected demographic characteristics of respondents in our sample. recycling, as moral considerations have been found important to
Characteristic Category Respondents US Census explain recycling behavior (e.g., see Schwartz, 1970). However,
in our sample for 2000 (%) these questions (‘‘Households like mine should not be blamed for
(%, n ¼ 2136) environmental problems caused by excessive trash generation’’;
Age (years) 18–29 19.4 22.2 ‘‘My responsibility is to provide only for my family and myself’’;
30–44 26.2 31.4
and ‘‘Given increasing environmental problems with solid waste, I
45–59 28.0 24.5
60 26.3 21.9 feel a moral obligation to recycle’’) were not amenable to factor-
Gender Male 48.3 49.0 ization. To assess the adequacy of our principal components, we
Female 52.0 51.0 rely on common statistical tests, which are summarized below
Marital status Married 58.0 54.4 Table 2.
Divorced 12.5 9.7
Widowed 4.8 6.6
Separated 1.8 2.2 3.2.2. Count models
Single (never married) 22.9 27.1 The Poisson regression model is a natural first choice for
Ethnicity White 77.7 69.1 modeling count data, as it is well known (e.g., see Long and Freese,
Hispanic 7.3 12.6
2006) that the linear regression model is biased and inconsistent
African American 8.5 12.1
Two or more races 2.8 1.6 for such data. In our context, let Ni denote the random variable that
Other 3.7 4.6 indicates the number of e-waste items stored by household i. Then,
Education Less than high school 11.4 19.6 if Ni has a Poisson distribution, the probability that Ni ¼ n  0 is
High school 33.3 28.6 given by:
Some college 29.5 27.3
Bachelors degree or higher 25.8 24.4 n
Household income ($K) <10 8.1 9.6 expðli Þli
Pri ðnÞ ¼ (1)
10–24 18.5 19.1 n!
25–49 32.8 29.3
50–74 19.8 19.5 where li is the Poisson parameter for household i. To fit sample
75–99 11.6 10.2 data, we assume that the natural logarithm of li is a linear
100 9.7 12.3 combination of various explanatory variables (such as household
Home ownership Own 69.9 66.2 size, income, or recycling beliefs), denoted by Xi, and a vector b of
Rent 24.4 33.8
Other 5.7 NA
unknown coefficients:
Type of dwelling Single-family detached 67.2 61.4
t
Duplex 5.5 5.6; lnðli Þ ¼ b Xi (2)
Mobile home 7.9 7.0
Apartment/condominium 17.6 25.9 However, the Poisson regression model implies that the variance
Other 1.8 0.1 of the count random variable Ni equals its mean (¼ li). This
Household size 1 20.7 21.4 assumption does not seem to hold here as Fig. 1 suggests an
2 38.8 32.5
excess of 0 counts for the Poisson model. As a result, the b coef-
3 17.2 16.5
4 14.0 14.3 ficients estimated with the Poisson model are likely to be biased
>4 9.3 11.0 and their standard errors underestimated (see Cameron and
Percentages may not add exactly to 100% due to rounding. Trivedi, 1998).
To overcome this limitation, we consider two approaches. First,
orthogonal linear transformation that reduces a set of correlated we estimate a negative binomial model, which is obtained from the
variables into a smaller number of uncorrelated variables called Poisson model by adding to Eq. (2) an error term, which is inde-
principal components (PC) such that each successive PC accounts pendently gamma distributed with mean 1 and variance a. This
for as much of the remaining data variability as possible. We also yields a conditional probability that can be integrated to give

Table 2
Principal components analysis of recycling beliefs.

Survey items and principal components (PC) Index weights Principal component statistics
I1. Perceived barriers to recycling
1. ‘‘Finding room to store recyclable materials is a problem’’ 0.06548 a ¼ 0.684; KMO ¼ 0.684;
2. ‘‘Storing recyclable materials at home is unsanitary’’ 0.07224 Bartlett: p < 0.001
3. ‘‘Recycling drop-off centers are too far away’’ 0.05044
4. ‘‘Storing recyclable materials at home is safe’’ 0.06184

I2. Awareness of the consequences of recycling

1. ‘‘Recycling substantially reduces the use of landfills’’ 0.06244 a ¼ 0.693; KMO ¼ 0.734;
2. ‘‘Recycling conserves natural resources’’ 0.06761 Bartlett: p < 0.001
3. ‘‘Recycling won’t make much difference in the quality of the 0.05999
environment’’
4. ‘‘Recycling creates jobs’’ 0.05996

For questions in both indexes, the scale is as follows: 0, strongly agree; 1, mildly agree; 2, unsure; 3, mildly disagree; and 4, strongly disagree, except for question 4 of I1 and
question 3 of I2, for which this scale is reversed. To find the value of one of the indexes for a respondent, simply calculate the weighted sum of the answers using the weights
above. Note that the index weights of each principal component (I1 and I2) were normalized to sum to 0.25 each so both I1 and I2 are comprised between 0 and 1. A lower
value of I1 indicates higher perceived barriers to recycling. Conversely, a lower value of I2 indicates higher perceived benefits from recycling for environmental quality.
Cronbach’s alpha indicates how well a set of variables measures a single underlying construct; it is high when inter-item correlations are high, so in a way it measures the
reliability of our indexes. A value of at least 0.6 is desirable (its maximum is 1). The Kaiser–Meyer–Olkin (KMO) statistic detects excessive correlations, which would lead to
multicollinearity; a satisfactory KMO should exceed 0.6 (KMO ranges between 0 and 1). Bartlett’s test of sphericity checks whether the correlation matrix of the variables
differs significantly from the identity matrix; if not, the principal component model is inappropriate. For PCA to work well, the Bartlett test should reject the null hypothesis
that the correlation matrix is the identity matrix.
 J.-D.M. Saphores et al. / Journal of Environmental Management 90 (2009) 3322–3331 3327

  !a1  n employee, not a homemaker). We are therefore confident that self-

G a1 þ n a1 l selection is not a major issue in our study.
Pri ðnÞ ¼   (3)
G a1 n! a1 þ l a1 þ l
4.2. Principal components analysis
This formulation allows the mean to differ from the variance as
follows:
Results from our principal components analysis are presented in
VarðNi Þ ¼ EðNi Þ½1 þ aEðNi Þ (4) Table 2. We developed two indexes to reflect recycling beliefs; each
is based on four questions. The first index (I1) summarizes
where a is a measure of dispersion. The b coefficients and the perceived barriers to recycling and the second one (I2) reflects
a parameter can be estimated via maximum likelihood. A likeli- awareness of the consequences of recycling. To facilitate their
hood ratio test of a ¼ 0 can be used to assess if the negative bino- interpretation, we normalized I1 and I2 so they are comprised
mial is preferred to a Poisson model. between 0 and 1. To assess their validity, we performed some
Another approach is to estimate zero-inflated models, which common statistical tests, which all turned out to be satisfactory;
postulate the existence of two latent (unobserved) groups: a first they are explained in the notes below Table 2.
group where the count of each household is always zero, and
a second group with a non-zero probability of a positive count. To 4.3. Explaining how much e-waste households store at home
justify these models in our context, we follow (Cameron and Trivedi
,1998, p. 126) and assume that people in the always zero group We then estimated the four types of count regression models
incorrectly misreported the number of their e-waste items as zero. presented above for both small and large e-waste items on the 1562
If ji denotes the probability that respondent ‘‘i’’ is in the first respondents who gave us complete information. Simple statistics of
group, and if Pri(n) denotes the probability that respondent ‘‘i’’ has the explanatory variables considered in these models are summa-
a count equal to n  0 given that he/she is in the second group, then rized in Table 3, and results from our count regression models are
a zero-inflated model can be summarized by: presented in Table 4 for small e-waste items (results for large items
 are omitted for brevity). While there are many similarities between
PrðNi ¼ 0Þ ¼ ji þ ð1  ji ÞPri ð0Þ these models, some interesting differences are worth highlighting.
(5)
PrðNi ¼ n > 0Þ ¼ ð1  ji ÞPri ðnÞ The base set of independent variables for explanatory count
where group membership (i.e., ji) can be conveniently modeled via models is identical. These variables include the two principal
a logit model that depends on explanatory variables characterizing components (I1 and I2) dealing with the influence of recycling
each household and a set of unknown coefficients g. Using the beliefs; responses to three survey questions on moral values,
Poisson model (Eq. (1)) to define Pri(n) gives the zero-inflated recycling and the environment; knowledge of relevant e-waste
Poisson (ZIP) model; conversely, if Pri(n) is given by a negative laws; previous e-waste recycling behavior; as well as demographic
binomial regression model (Eq. 3), we obtain the zero-inflated and socio-economic characteristics such as age, marital status,
negative binomial (ZINB) model. The unknown parameters b, g, and ethnicity, primary language, household size, education, income,
a (for ZINB) can again be estimated by maximum likelihood. home ownership, and employment status.
To select the most appropriate count model, we relied on the
likelihood ratio test for alpha (to discriminate between the negative
4. Results binomial and the Poisson regression models), on the Vuong test (to
compare non-inflated and inflated models), and on our models’
Our results were estimated with Stata 9.1. After considering self- ability to predict observed e-waste counts. Based on this informa-
selection, we summarize results from our principal component tion, we identified the zero-inflated negative binomial regression
analysis and discuss models developed to explain and predict the model as the best fit for small e-waste items and the negative
number of small and large e-waste items stored by households. binomial regression model as best for large e-waste items (e.g., see
details in the notes below Table 4). Note that variables that are
strongly statistically significant often have the same sign and
4.1. A look at self-selection magnitude across different models, which suggests that our results
are fairly robust. Let us now summarize our more salient results,
Out of 2136 respondents, 1562 provided all of the information starting with models for small e-waste items.
necessary for explaining the number of items stored by US
households. Most respondents omitted just a few questions but 4.3.1. Explaining the number of small e-waste items stored by
some probably found that giving us an inventory of their e-waste households
items was too demanding so they skipped those questions entirely. Let us start with the binary equation for the zero-inflated
Self-selection of respondents may therefore be a problem here, negative binomial regression model, which indicates whether
although it is likely to be much less of an issue than in labor or in respondents belong to the category of people with no small e-waste
health economics, for example, as we anticipate that our non- items (likely people who under-reported). As expected, we see that
respondents do not share systematic characteristics (except for respondents who perceive lower barriers for recycling (higher
their dislike of long surveys). To examine the issue of self-selection values of I1) are more likely to belong to the category of people with
in a very simple way, we estimated logit models where the no e-waste. No other variable is statistically significant in the binary
dependent variable equals 1 for respondents who reported their model. However, when the least significant variables are sequen-
estimated number of small (large) e-waste items, and 0 otherwise; tially removed (results not shown), I2, ‘‘Paid/unpaid employee’’,
explanatory variables were all the socio-economic and demo- ‘‘Single, never married’’, ‘‘Some college education’’, and ‘‘College
graphic variables systematically collected by KN. Results are degree’’ become significant (p  10%). The first three of these
omitted for brevity; only three variables were found to be variables increase the likelihood of having no e-waste, whereas the
statistically significant: two decreased slightly the probability of last three have the opposite effect.
non-reporting information (being single and belonging to a dual- Let us now examine the count equation. Recycling beliefs and
income household) while the third one increased it (being an perceived barriers to recycling play a minor role and moral values
3328 J.-D.M. Saphores et al. / Journal of Environmental Management 90 (2009) 3322–3331

Table 3 have less available free time to use consumer electronics and thus
Descriptive statistics for key variables. have fewer obsolete items.
Variable Mean SD In addition, we find that ethnicity is statistically significant in
Dependent variables our models. Both African Americans and Hispanics tend to have
Number of small e-waste items in household (range 0–39) 3.96 4.50 more obsolete small e-waste items compared to Whites (our
Number of large e-waste items in household (range 0–37) 2.31 3.59 reference category), which may indicate less opportunities to
Independent variables (yes [ 1, unless otherwise specified) recycle e-waste or insufficient information about the potential
Principal components dangers of e-waste.
I1: Perceived barriers to recycling (normalized to range 0.55 0.23 Finally, larger households and homeowners (as opposed to
from 0 to 1)
I2: Awareness of consequences of recycling (normalized 0.23 0.17
renters) have significantly more small e-waste items. This result is
to range from 0 to 1) not unexpected. Larger households typically have more electronic
Moral considerations products of many kinds (for communication and entertainment for
Strongly agree/agree with statement: ‘‘Given increasing 0.70 0.46 example). Since homeowners tend to be less transient than renters,
environmental problems with solid waste, I feel a moral
they are more likely to accumulate e-waste items waiting for an
obligation to recycle’’
Strongly agree/agree with statement: ‘‘Households like 0.35 0.48 opportunity to reuse or recycle them.
mine should not be blamed for environmental problems On the other hand, neither income nor education seem to have
causes by excessive trash generation’’ an impact on the number of small e-waste items stored by US
Strongly agree/agree with statement: ‘‘My responsibility 0.24 0.43 households, possibly because small CEDs are now ubiquitous.
is to provide only for my family and myself’’
Socio-economic variables
Likewise, the knowledge of e-waste laws is not statistically signif-
Age 30–44 years 0.26 0.44 icant, which suggests that information about these laws does not
Age 45–59 years 0.27 0.45 include where to dispose of e-waste.
Age 60 years and greater 0.26 0.44
Single, never married 0.24 0.43
4.3.2. Explaining the number of large e-waste items stored by
Divorced 0.12 0.32
Widowed 0.04 0.20 households
Separated 0.02 0.14 Results for our negative binomial model for large e-waste items
African American, non-Hispanic 0.08 0.27 show similarities with a few differences; results are omitted to save
Other race, non-Hispanic 0.03 0.18 space. First, we find that people who express a high degree of
Hispanic 0.07 0.25
Two or more races, non-Hispanic 0.03 0.17
awareness about the consequences of recycling tend to have more
Primary language is not English 0.03 0.16 of the large e-waste items at home; we speculate that they may
Interaction: female  head of household (range 0–2) 1.25 0.75 keep these items until an opportunity to recycle arises. Likewise,
Logarithm of household size (range 0–2.3) 0.83 0.52 respondents who previously recycled e-waste tend to have more of
High school education only 0.34 0.47
the large e-waste items; one possibility is that they keep a number
Some college education 0.25 0.43
College degree 0.25 0.43 of these until they have another opportunity to recycle.
Logarithm of household income, in thousands 3.71 0.84 Our demographic results are fairly similar. As before, both African
(range 0.92–5.19) Americans and Hispanics tend to have more of the large e-waste
Dual-income household 0.56 0.50 items stored in their homes; the same holds for multiple-race
Homeowner 0.69 0.46
respondents, although this result is only marginally significant. As
Paid/unpaid employee, not homemaker 0.57 0.50
High-speed Internet access 0.60 0.49 before, larger households tend to have more of the large e-waste
Knowledge of recycling items, but so do singles (they may need more entertainment). One
Not aware of e-waste laws (local or state) 0.67 0.47 difference, however, is that respondents between the ages of 30 and
Previously recycled e-waste 0.35 0.48
44 years tend to have more large e-waste items than other age
Sample size: n ¼ 1562. All variables are binary (0/1) unless otherwise specified. groups, including our reference category, adults from 18 to 29 years.
While we expect older adults to have fewer CEDs than younger ones,
the difference between small and large e-waste items here is inter-
regarding recycling and the environment are not statistically esting. It is likely that younger people (<30 years) have more small
significant. Respondents with higher scores for I2 (awareness of e-waste items given the popularity of products such as MP3 players,
consequences of recycling) tend to have more small, obsolete CEDs cell phones, and other mobile devices, while adults in the 30–44 year
stored in their homes, holding all other variables constant; age group are more likely to have acquired larger consumer elec-
a possible explanation is that people who are more aware of the tronics such as larger televisions and monitors.
consequences of recycling are more likely to recycle their e-waste Unlike for small e-waste items, education is now statistically
when a good opportunity arises. significant as respondents with only a high school degree have
Not unexpectedly, older adults tend to have fewer small e-waste fewer large e-waste items. Income is again not a factor, but
items in their homes. In general, younger people tend to be more surprisingly neither is owning a home (and therefore having more
comfortable with information and communication technologies. storage space).
They are therefore more likely to own a wider range of consumer
electronics (see, e.g. Horrigan, 2007) and, by extension, accumulate 4.4. Estimating how much e-waste items are stored at home by US
more e-waste. households
Marital status also plays a significant role in our models. Respon-
dents who are widowed tend to have more of the small e-waste items As mentioned above, only 1562 respondents provided us with
in their homes. Time–activity pattern could be a factor here: they may all the information needed to explore links between the number of
have more electronic items in general because they may have more e-waste items stored by households in their homes and various
free time. Similarly, people with access to high-speed Internet have demographic, socio-economic, and belief variables.
more small e-waste items. The same reason may also explain why To estimate how many e-waste items on average are stored by
female heads of household and dual-income households have fewer US households, we estimate reduced models that rely only on the
small e-waste items stored in their homes: these households may relevant demographic and socio-economic variables KN requests
 J.-D.M. Saphores et al. / Journal of Environmental Management 90 (2009) 3322–3331 3329

Table 4
Explaining the number of small e-waste items stored by US households.

Independent variable Poisson Negative Zero-inflated Zero-inflated negative

binomial Poisson (ZIP) binomial (ZINB)

Count Binary Count Binary

Principal components
I1. Perceived barriers to recycling 0.201 (0.141) 0.335** (0.143) 0.013 (0.126) 1.075*** (0.304) 0.106 (0.150) 2.047** (0.815)
I2. Awareness of consequences of recycling 0.389 (0.238) 0.370* (0.211) 0.417* (0.215) 0.314 (0.438) 0.558** (0.235) 1.642 (1.388)

Moral considerations
Strongly agrees/agrees with: ‘‘Given increasing 0.090 (0.075) 0.090 (0.071) 0.039 (0.067) 0.201 (0.154) 0.069 (0.072) 0.231 (0.298)
environmental problems with solid waste,
I feel a moral obligation to recycle’’
Strongly agrees/agrees with: ‘‘Households like 0.088 (0.069) 0.067 (0.066) 0.088 (0.062) 0.072 (0.130) 0.105 (0.076) 0.272 (0.390)
mine should not be blamed for environmental
problems causes by excessive trash generation’’
Strongly agrees/agrees with: ‘‘My responsibility 0.139** (0.070) 0.135* (0.070) 0.111* (0.062) 0.133 (0.151) 0.148 (0.107) 0.118 (0.736)
is to provide only for my family and myself’’

Socio-economic variables
Age 30–44 years 0.092 (0.090) 0.095 (0.089) 0.088 (0.079) 0.015 (0.209) 0.136 (0.102) 0.285 (0.863)
Age 45–59 years 0.219** (0.107) 0.174* (0.102) 0.153 (0.095) 0.255 (0.216) 0.090 (0.122) 0.669 (0.951)
Age 60 years and greater 0.457*** (0.122) 0.393*** (0.117) 0.320*** (0.108) 0.400 (0.252) 0.331** (0.140) 0.514 (0.986)
Single, never married 0.089 (0.089) 0.167* (0.086) 0.024 (0.080) 0.378* (0.206) 0.074 (0.102) 0.727 (0.687)
Divorced 0.062 (0.115) 0.054 (0.105) 0.047 (0.105) 0.100 (0.219) 0.026 (0.116) 0.109 (0.634)
Widowed 0.340** (0.144) 0.378*** (0.154) 0.238* (0.133) 0.394 (0.337) 0.339* (0.184) 0.287 (1.097)
Separated 0.030 (0.228) 0.072 (0.231) 0.049 (0.197) 0.212 (0.421) 0.144 (0.238) 0.620 (1.020)
African American, non-Hispanic 0.497*** (0.101) 0.481*** (0.101) 0.384*** (0.092) 0.451* (0.245) 0.446*** (0.108) 0.325 (0.621)
Other race, non-Hispanic 0.190 (0.186) 0.074 (0.167) 0.181 (0.168) 0.061 (0.345) 0.074 (0.203) 0.143 (1.145)
Hispanic 0.386*** (0.102) 0.409*** (0.105) 0.223** (0.090) 0.759*** (0.280) 0.277** (0.116) 1.254 (0.940)
Two or more races, non-Hispanic 0.292 (0.180) 0.262 (0.165) 0.210 (0.172) 0.397 (0.413) 0.227 (0.216) 0.590 (1.967)
Primary language is not English 0.395** (0.181) 0.379** (0.168) 0.270* (0.147) 0.556 (0.373) 0.226 (0.161) 1.265 (0.773)
Interaction: female  head of household 0.078* (0.042) 0.076* (0.043) 0.069* (0.037) 0.060 (0.091) 0.090* (0.051) 0.071 (0.310)
Logarithm of household size 0.225*** (0.066) 0.275*** (0.066) 0.241*** (0.058) 0.006 (0.144) 0.272*** (0.068) 0.094 (0.305)
High school education only 0.245** (0.106) 0.249** (0.103) 0.161* (0.094) 0.257 (0.211) 0.171 (0.116) 0.380 (0.626)
Some college education 0.089 (0.111) 0.144 (0.108) 0.036 (0.100) 0.293 (0.230) 0.062 (0.126) 0.729 (1.066)
College degree 0.024 (0.121) 0.073 (0.116) 0.058 (0.108) 0.478* (0.248) 0.038 (0.127) 1.001 (0.850)
Logarithm of household income in $1000 0.007 (0.046) 0.016 (0.041) 0.008 (0.041) 0.062 (0.089) 0.028 (0.054) 0.048 (0.423)
Dual-income household 0.167** (0.071) 0.177** (0.071) 0.160** (0.064) 0.054 (0.148) 0.189** (0.076) 0.038 (0.347)
Homeowner 0.177*** (0.071) 0.182** (0.072) 0.120* (0.064) 0.240 (0.160) 0.144* (0.084) 0.289 (0.469)
Paid/unpaid employee, not homemaker 0.179*** (0.067) 0.142** (0.068) 0.067 (0.060) 0.417*** (0.147) 0.066 (0.074) 0.742 (0.475)
High-speed Internet access 0.238*** (0.069) 0.232*** (0.067) 0.169*** (0.063) 0.291** (0.136) 0.180** (0.083) 0.418 (0.471)

Knowledge of recycling
Not aware of e-waste laws (local or state) 0.069 (0.070) 0.066 (0.068) 0.083 (0.063) 0.004 (0.135) 0.055 (0.080) 0.005 (0.450)
Previously recycled e-waste 0.121* (0.068) 0.081 (0.066) 0.144** (0.062) 0.166 (0.131) 0.124 (0.078) 0.347 (0.354)
Constant 1.097*** (0.253) 1.025*** (0.229) 1.363*** (0.225) 1.320*** (0.504) 0.988*** (0.270) 3.709 (2.282)
a (dispersion coefficient) – 1.130*** (0.060) – – 0.776*** (0.098)
D 0.609 0.118 0.365 0.058

Robust standard errors in parentheses. Sample size: n ¼ 1562. *p  0.10, **p  0.05, and ***p  0.01, respectively. D is the sum of absolute mean differences between predicted
and actual counts of small e-waste items for counts between 0 and 20. The likelihood ratio test of H0: a ¼ 0 strongly rejects the null hypothesis so the dispersion coefficient is
highly significant and the negative binomial model is preferred to the Poisson model. Likewise, the zero-inflated negative binomial model is preferred to the zero-inflated
Poisson model. The Vuong test comparing the Poisson and the zero-inflated Poisson (ZIP) models equals 12.98 (p < 0.001), so the ZIP model is preferred to the Poisson model.
The Vuong test comparing the negative binomial and the zero-inflated negative binomial (ZINB) equals 4.09 (p < 0.001), so the ZINB model is preferred to the negative
binomial model. Overall, the zero-inflated negative binomial model is preferred; it also predicts best observed counts for small e-waste items.

from its panel member as explanatory variables. This approach has items is a zero-inflated negative binomial regression model,
two advantages: first, it allowed us to consider all of our respon- whereas for large e-waste items, a simpler negative binomial
dents; and second, it enabled us to apply a weighting scheme that regression model is preferred. Coefficients for these models are
corrects for non-response and for several small defects in KN’s typically close to their counterparts in the explanatory models,
panel. These defects result in deviations from the equal probability except in the binary model for small e-waste items. Overall, these
of selection methods; they include, for example, the fact that RDD models do a good job at explaining observed counts of e-waste
sampling rates are proportional to the number of phone lines in items compared to the other models we estimated; as shown at the
a household or minor oversampling of Chicago and Los Angeles by bottom of Table 5, the sum of the absolute mean differences
KN due to early pilot surveys in those two cities. Weights were between predicted and actual counts (between 0 and 20 items) of
created by KN using iterative proportional fitting over age, gender, e-waste items is respectively 0.068 for small and 0.057 for large
census region, education, metropolitan status, and Internet access e-waste items.
to make the weighted sample cells match those of KN’s panel. After With these models, we can predict the number of e-waste items
review, the weights were trimmed at the tails of their distribution in storage for each of the 2136 respondents in our sample, form
(0.9% and 99.55%) and scaled to sum to 2136 (our total sample size). a weighted sum of these predictions using the weights discussed
Final estimates of our ‘‘best’’ predictive models are presented in above, and divide the result by the number of respondents (which
Table 5; only significant variables are shown for conciseness. As is also the sum of the weights). We obtain 4.1 small and 2.4 large
with our explanatory models, the best model for small e-waste e-waste items in storage on average for US households. These
3330 J.-D.M. Saphores et al. / Journal of Environmental Management 90 (2009) 3322–3331

Table 5
Predicting the number of e-waste items stored by US households using only socio-demographic data from KN.

Independent variable Small e-waste items: zero-inflated Large e-waste items:

negative binomial (ZINB)a negative binomialb

Countc Binaryd
Age 45–59 years 0.264*** (0.075) – –
Age 60 years and greater 0.263*** (0.090) 0.727** (0.366) –
Single, never married – – 0.233*** (0.088)
African American, non-Hispanic 0.435*** (0.096) – 0.452*** (0.119)
Hispanic 0.319*** (0.916) – 0.338** (0.138)
Two or more races, non-Hispanic 0.279* (0.158) – 0.382* (0.220)
Interaction: female  head of household 0.119*** (0.041) – –
Logarithm of household size 0.284*** (0.065) 0.531* (0.311) 0.330*** (0.088)
High school education only 0.200*** (0.071) – 0.333*** (0.081)
Some college education 0.137* (0.074) – –
Dual-income household 0.212*** (0.069) – 0.157* (0.083)
Homeowner 0.162** (0.066) – –
High-speed Internet access 0.209*** (0.067) – –
Paid/unpaid employee, not homemaker – 0.911*** (0.369) –
D 0.068 0.057
Weighted average of predicted e-waste 4.11 2.42
items per household

Robust standard errors are in parentheses. *p  0.10, **p  0.05, and ***p  0.01, respectively.
D is the sum of absolute mean differences between predicted and actual counts of small e-waste items for counts between 0 and 20.
a
Sample size, 1630; non-zero observations, 1193; zero observations, 437.
b
Sample size, 1648; non-zero observations, 948; zero observations, 700.
c
The count equation presents results for the group of households with e-waste. Coefficients are interpreted as for Poisson or negative binomial regression models.
d
The binary equation presents results for the group of households in the ‘‘Always Zero’’ group, i.e., those households who have no e-waste. Coefficients are interpreted as for
a binary logit model.

numbers are slightly larger but close to those provided by our 1562 In light of these findings, a variety of policies could be
respondents used to estimate our explanatory models (see the top considered. In the short run, it is necessary to deal with the
of Table 3). backlog of accumulated e-waste by developing the recycling
Our averages enable us to calculate an estimate of the average infrastructure, publicizing information about how to recycle
number of e-waste items stored by a US household. The US Census e-waste, and organizing staggered recycling campaigns for
Bureau (2007) estimates that there were 114,384,000 households selected types of e-waste products (such as printers, for example)
in 2006, so there are at least 470 million small and 277 million large to avoid recycling overloads. These measures could be financed by
e-waste items stored by US households, for a total exceeding 747 a temporary sales tax on electronic products. In the long run, laws
million items. should be adopted to foster the comprehensive recycling of
electronic products. Deposit-refund systems could be imple-
mented to give consumers an incentive to return obsolete
5. Conclusions electronic products to certified collectors, just as for lead acid
car batteries. In addition, extended producer responsibility (EPR)
This paper presents count models to explain and predict the (e.g., see Sinha Khetriwal et al., 2009) could be mandated for CED
number of small and large e-waste items stored by US house- firms to make them responsible for the sound environmental
holds. Statistically significant explanatory variables include age, management of the end of life of their products; this would give
marital and employment status, ethnicity, household size, them an incentive to design products that are easier to recycle and
previous e-waste recycling behavior, and to some extent educa- maybe even ‘‘green’’ electronics products that do not need special
tion, home ownership, and awareness of the consequences of treatment at the end of their useful life. Ideally, these measures
recycling, but neither income nor knowledge of e-waste recycling should be adopted at the federal level to create a coherent system;
laws. this would make it simpler for manufacturers to adapt, and it
Our simplified prediction models based only on standard socio- would avoid loopholes created by a patchwork of inconsistent
economic and demographic data collected by KN from all its state laws.
panelists suggest that each US household has, on average, 4.1 small Future research could consider the potential impacts on the
and 2.4 large e-waste items in storage. These numbers are also volume of obsolete TVs of the upcoming switch from analog to
almost certainly under-estimates of the true e-waste count for at digital broadcasting; it should also explore ways of greening elec-
least two reasons: first, it is time consuming for a household to tronic products and of facilitating recycling both for households
precisely inventory of all its obsolete e-waste items; and second, we and for e-waste recyclers. Finally, it would be of interest to
relied on a narrow definition of e-waste that excludes large appli- understand how many large, older appliances such a refrigerators,
ances such as refrigerators and washing machines. Extrapolating freezers, and air conditioners are in storage, not only because of
our results to the whole country indicates that there are at least 470 their potential contribution to e-waste but also because they likely
million small and 277 million large e-waste items stored by US use a lot more power than recent models.
households, for a total exceeding 747 million items. This is
considerably more than the EPA estimate of 180.3 million items in
storage, although their estimate is based on a smaller range of Acknowledgements
products (printers, desktop and laptop computers, monitors,
peripherals, as well as televisions). This suggests that the backlog of This research was supported in part by grants from the National
e-waste in the US may be much larger than generally believed. Science Foundation (DMI-0223894 and CMS-0524903) and from
 J.-D.M. Saphores et al. / Journal of Environmental Management 90 (2009) 3322–3331 3331

the Green Materials Lead Campus Component of the University of Li, Y., Richardson, J.B., Walker, A.K., Yuan, P.-C., 2006. TCLP heavy metal leaching of
personal computer components. Journal of Environmental Engineering 132 (4),
California Toxic Substances Research & Teaching Program (UC-
497–504.
44157). Additional support was provided by the Program in Lincoln, J.D., Ogunseitan, O.A., Shapiro, A.A., Saphores, J.-D., 2007. Leaching
Industrial Ecology at UC-Irvine and by the office of the Executive assessments of hazardous materials in cellular telephones. Environmental
Vice Chancellor at UCI; their support is gratefully acknowledged. Science & Technology 41, 2572–2578.
Linton, J.D., Yeomans, J.S., 2003. The role of forecasting in sustainability. Techno-
We also thank Amy Jerominek for valuable research assistance. logical Forecasting and Social Change 70 (1), 21–38.
Finally, we thank the editor and three anonymous referees for very Liu, X.L., Tanaka, M., Matsui, Y., 2006. Generation amount prediction and material
helpful comments. flow analysis of electronics waste: a case study in Beijing, China. Waste
Management & Research 24, 434–445.
Long, J.S., Freese, J., 2006. Regression Models for Categorical Dependent Variables
References using Stata, second ed. Stata Press, College Station, TX.
Macauley, M., Palmer, K., Shih, J.-S., 2003. Dealing with electronic waste: modeling
Babu, B.R., Parande, A.K., Basha, C.A., 2007. Electrical and electronic waste: a global the costs and environmental benefits of computer monitor disposal. Journal of
environmental problem. Waste Management & Research 25, 307–318. Environmental Management 68, 13–22.
Cameron, A.C., Trivedi, P.K., 1998. Regression Analysis of Count Data. Cambridge Matthews, H.S., McMichael, F.C., Hendrickson, C.T., Hart, D.J., 1997. Disposition and
University Press, Cambridge, UK. End-of-life Options for Personal Computers. Carnegie-Mellon University, Pitts-
Carnegie-Mellon University, 1991. Design Issues in Waste Avoidance. Carnegie- burgh, PA.
Mellon University, Pittsburgh, PA. National Safety Council, 1999. EPR2 Baseline Report: Recycling of Selected Elec-
Consumer Electronics Association, 2008. Innovation: US Economic Contribution of tronic Products in the United States. National Safety Council, Washington, DC.
Consumer Electronics. A Study of Direct, Indirect, and Induced Effects on Employ- Nixon, H., Saphores, J.-D.M., 2007. Financing electronic waste recycling – Californian
ment and Business Activity. Consumer Electronics Association, Arlington, VA. households’ willingness to pay advanced recycling fees. Journal of Environ-
California Integrated Waste Management Board, 2001. Selected e-Waste Diversion mental Management 84 (4), 547–559.
in California: a Baseline Study (No. 610-01-008). California Integrated Waste Nixon, H., Saphores, J.-D., Ogunseitan, O.A., Shapiro, A.A., 2009. Understanding
Management Board, Sacramento, CA. preferences for recycling electronic waste in California – the influence of
California Integrated Waste Management Board, 2004. Best Management Practices environmental attitudes and beliefs on willingness to pay. Environment and
for Electronic Waste (No. 630-04-005). California Integrated Waste Manage- Behavior 41 (1), 101–124.
ment Board, Sacramento, CA. Osibanjo, O., Nnorom, I.C., 2007. The challenge of electronic waste (e-waste)
Chung, S.-W., Murakami-Suzuki, R., 2008. A comparative study of e-waste recycling management in developing countries. Waste Management & Research 25,
systems in Japan, South Korea and Taiwan from the EPR perspective: implica- 489–501.
tions for developing countries. In: Kojima, M. (Ed.), Promoting 3Rs in Devel- Saphores, J.-D.M., Nixon, H., Ogunseitan, O.A., Shapiro, A.A., 2006. Household
oping Countries: Lessons from the Japanese Experience. IDE-JETRO, Chiba. willingness to recycle electronic waste: an application to California. Environ-
http://www.ide.go.jp/English/Publish/Download/Spot/pdf/30/007.pdf (accessed ment and Behavior 38 (2), 183–208.
03/22/09). Saphores, J.-D., Nixon, H., Ogunseitan, O.A., Shapiro, A., 2007. California households’
Federal Communications Commission, 2009. The Digital TV Transition – What You willingness to pay for ‘‘green’’ electronics. Journal of Environmental Planning
Need to Know About DTV. http://www.dtv.gov/consumercorner.html (accessed and Management 50 (1), 113–133.
03/22/09). Schmidt, C.W., 2006. Unfair trade: e-Waste in Africa. Environmental Health
Florida Department of Environmental Protection, 2003. State of Electronics in Perspectives 114 (4), A232–A235.
Florida Households: Residential Household Electronics Survey. Bureau of Solid Science Daily, 2007. Recycling of Europe’s Electronic Waste Needs Improvement, UN
and Hazardous Waste, Tallahassee, FL. Report Urges. http://www.sciencedaily.com/releases/2007/11/071115113338.
Government Accountability Office, 2007. Digital Television Transition: Preliminary htm (accessed 03/22/09).
Information on Initial Consumer Education Efforts (GAO-07-1248T). Govern- Selin, H., VanDeveer, S.D., 2006. Raising global standards: hazardous substances and
ment Accountability Office, Washington, DC. e-waste management in the European Union. Environment 48 (10), 6–18.
Grossman, E., 2006. High Tech Trash: Digital Devices, Hidden Toxics, and Human Sinha Khetriwal, D., Kraeuchi, P., Widmer, R., 2009. Producer responsibility for e-
Health. Island Press, Washington, DC. waste management: key issues for consideration - learning from the Swiss
Herat, S., 2007. Sustainable management of electronic waste (e-Waste). Clean 35 (4), experience. Journal of Environmental Management 90, 153–165.
305–310. Swann, P., 2007. Digital TV: e-waste tsunami (electronic version). http://www.
Hicks, C., Dietmar, R., Eugster, M., 2005. The recycling and disposal of electrical and tvpredictions.com/dtvtrash060407.htm (accessed 05/16/08).
electronic waste in China – legislative and market responses. Environmental Schwartz, S.H., 1970. Moral decision making and behavior. In: Macauley, J.,
Impact Assessment Review 25 (5), 459–471. Berkowitz, L. (Eds.), Altruism and Helping Behavior. Academic Press, New York,
Horrigan, J.B., 2007. A Typology of Information and Communication Technology pp. 127–141.
Users. Pew Internet & American Life Project, Washington, DC. US Census Bureau, 2007. Current Population Survey, 2006 Annual Social and
International Association of Electronics Recyclers, 2006. IAER Electronics Recycling Economic Supplement. http://www.census.gov/population/www/socdemo/hh-
Industry Report 2006. International Association of Electronics Recyclers, Albany, NY. fam/cps2006.html (accessed 06/05/08).
Jang, Y.-C., Townsend, T.G., 2003. Leaching of lead from computer printed wire US EPA, 2006. Municipal Solid Waste in the United States: 2005 Facts and Figures.
boards and cathode ray tubes by municipal solid waste landfill leachates. Office of Solid Waste. US Environmental Protection Agency, Washington, DC.
Environmental Science & Technology 37, 4778–4784. US EPA, 2007a. Electronics Waste Management in the United States: Approach One
Kang, H.-Y., Schoenung, J.M., 2005. Electronic waste recycling: a review of US (No. 530-R-07-004a). Office of Solid Waste. US Environmental Protection
infrastructure & technology options. Resources Conservation and Recycling 45, Agency, Washington, DC.
368–400. US EPA, 2007b. Management of Electronic Waste in the United States: Approach
Kang, H.-Y., Schoenung, J.M., 2006. Estimation of future outflows and infrastructure Two (No. 530-R-07-004b). US Environmental Protection Agency, Washington,
needed to recycle personal computer systems in California. Journal of DC.
Hazardous Materials 137 (2), 1165–1174. Widmer, R., Oswald-Krapf, H., Sinha Khetriwal, D., Schnellmann, M., Boni, H., 2005.
Kline, P., 1994. An Easy Guide to Factor Analysis. Routledge, New York. Global perspectives on e-waste. Environmental Impact Assessment Review 25
Knowledge Networks (2009a). KNOWLEDGEPANELÒ Overview. http://www. (5), 436–458.
knowledgenetworks.com/knpanel/index.html (accessed 06/06/08). Wong, M.H., Wu, S.C., Deng, W.J., Yu, X.Z., Luo, Q., Leung, A.O.W., Wong, C.S.C.,
Knowledge Networks (2009b). Knowledge Networks bibliography. http://www. Luksemburg, W.J., Wong, A.S., 2007. Export of toxic chemicals: a review of the
knowledgenetworks.com/ganp/docs/KN Bibliography and Journals List - 3-2- case of uncontrolled electronic-waste recycling. Environmental Pollution 149,
09.pdf (accessed 03/22/09). 131–140.