Вы находитесь на странице: 1из 15

Resources, Conservation and Recycling 60 (2012) 49–63

Contents lists available at SciVerse ScienceDirect

Resources, Conservation and Recycling


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

Willingness to engage in a pro-environmental behavior: An analysis of e-waste


recycling based on a national survey of U.S. households
Jean-Daniel M. Saphores a,b,∗ , Oladele A. Ogunseitan c,1 , Andrew A. Shapiro d,2
a
Civil and Environmental Engineering Department, University of California, Irvine, CA 92697, USA
b
Planning, Policy & Design Department, University of California, Irvine, CA 92697, USA
c
Program in Public Health and School of Social Ecology, University of California, Irvine, CA 92697, USA
d
Electrical Engineering and Computer Science, University of California, Irvine, 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: Using concepts from environmental psychology and economics, we investigate U.S. households’ willing-
Received 12 January 2011 ness to engage in a form of pro-environmental behavior: recycling electronic waste (e-waste) at drop-off
Received in revised form locations. We rely on rich dataset from a 2006 national survey of U.S. households (N = 2136). Our internal
10 December 2011
variables include a modified version of the New Ecological Paradigm scale, a moral norm scale based
Accepted 12 December 2011
on Schwartz’s norm-activation model, and indicators of social pressure for recycling. External variables
consist of detailed socio-demographic characteristics. Our logit model shows that external variables do
Keywords:
not help characterizing people with e-waste recycling experience, except that they tend to have larger
Recycling behavior
Drop-off recycling
families or to be over 60 years old. However, knowing that e-waste contains potentially toxic materials,
Electronic waste recycling conventional materials at work or at school, and especially having strong moral norms helps
Generalized ordered logit explain e-waste recycling behavior. Using a generalized ordered logit model, we then show that the most
Factor analysis important variables for explaining household willingness to recycle e-waste are internal variables, fol-
lowed by recycling convenience, knowledge of the potential toxicity of e-waste, prior e-waste recycling
experience, as well as gender and marital status; education, age, and ethnicity play only a minor role,
while knowledge of e-waste laws, availability of curbside recycling for domestic waste, and income are
not statistically significant. Our results suggest that e-waste recycling can be stimulated by promoting
moral norms, educating the public about the benefits of recycling e-waste, and making e-waste recycling
more convenient but other measures will likely be necessary to tackle the e-waste problem.
© 2011 Elsevier B.V. All rights reserved.

1. Introduction less than 14% of the 2.87 million metric tons of e-waste they gener-
ated, and they are storing larger volumes of e-waste than previously
The need to better understand what motivates people to adopt estimated by the EPA (Saphores et al., 2009).
pro-environmental behaviors (PEB) has taken a new urgency with The exponential growth of e-waste may further be compounded
the increasing emphasis on sustainability. Because of its sheer by a reduction of the useful life of existing devices driven by the
volume, recycling household waste is an important form of pro- ever faster release of products with new features. Moreover, a
environmental behavior and it has been extensively studied. Yet recent review of the recycling literature (Sidique et al., 2010) shows
not all household wastes are alike and electronic waste (e-waste; that drop-off recycling has been largely ignored by academics. This
defined as all broken, obsolete, or out of fashion products contain- is unfortunate because drop-off recycling is easier to implement
ing a circuit board that reach the waste stream) has only received than take-back or other programs involving manufacturers and the
limited attention so far in spite of being the fastest growing segment required facilities are typically less expensive to operate than curb-
of household waste (U.S. EPA, 2009): in 2008, Americans recycled side collection programs. As noted by Bouvier and Wagner (2011),
drop-off recycling is currently by far the most common method for
collecting e-waste from households in the United States.
In this context, our paper makes three contributions based on
∗ Corresponding author at: Civil and Environmental Engineering Department, a national survey of U.S. households: first, we characterize peo-
University of California, Irvine, CA 92697, USA. Tel.: +1 949 824 7334. ple who have recycled e-waste; second, we analyze households’
E-mail addresses: saphores@uci.edu (J.-D.M. Saphores),
willingness to recycle e-waste at drop-off centers; and third, we
Oladele.Ogunseitan@uci.edu (O.A. Ogunseitan), aashapir@uci.edu (A.A. Shapiro).
1
Tel.: +1 949 824 6350. contribute to the empirical literature that brings together eco-
2
Tel.: +1 818 393 7311. nomics and environmental psychology by incorporating in our

0921-3449/$ – see front matter © 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.resconrec.2011.12.003
50 J.-D.M. Saphores et al. / Resources, Conservation and Recycling 60 (2012) 49–63

models concepts from these two disciplines to better understand Various models have been proposed to study pro-
household actual and intended behavior. environmental behaviors (PEB). The two most popular ones,
In addition to saving valuable landfill space and to creating jobs, which serve as a starting point for several other theories, are
properly recycling e-waste is compelling for at least three reasons. Schwartz’s altruism model (Schwartz, 1970, 1973, 1977; Vining
The first one is its impact on resources. The electronics industry uses and Ebreo, 1990; Guagnano et al., 1995), which explains conserva-
a significant percentage of the world’s production of precious and tion efforts by favorable personal norms; and the theory of planned
special metals. In 2007, for example, the manufacturing of mobile behavior (TPB), which relies on the assumption that attitudes have
phones and personal computers consumed 3% of the world’s min- a causal impact on behaviors via behavioral intentions (Ajzen and
ing production of gold and silver. Moreover, the electronics industry Madden, 1986; Ajzen, 1988, 1991; Thøgersen, 1994; Boldero, 1995;
as a whole consumed 13% of the world’s production of palladium, Taylor and Todd, 1995). Other approaches include, for example, the
15% of the cobalt, and 80% of the indium and of the ruthenium value–attitude–behavior model (McCarty and Shrum, 1994), which
(UNEP, 2009). Since many special metals are produced by a hand- shows that attitudes and beliefs mediate between abstract values
ful of countries (USGS, 2010), recycling them from e-waste is an and specific behaviors; operant conditioning theories, which
insurance against the perils of market power (market power could explore how behavior can be altered by providing informational
lead to an increase in the price of a metal above its competitive feedback, rewards or penalties (e.g., see Katzev and Mishima,
value, resulting in a loss of welfare). 1992; Werner et al., 1995); or self-regulation theory (Sansone
The second reason for recycling e-waste is to prevent the release et al., 1992), which proposes that people regulate their behavior
in the environment of toxic materials like cadmium, hexavalent by changing related cognitions, emotions, or perceptions (Werner
chromium, lead, mercury, and brominated flame-retardants. Lead, and Makela, 1998).
for example, can damage children’s nervous system and adults’ In this paper, we rely on the theories of norm-activation and
reproductive system (U.S. EPA, 2010a). Mercury (in the form of value-belief-norms (VBN) because they are well-accepted, empir-
methylmercury) can interfere with the development of babies’ ically supported, and often invoked in rational choice models
brains (U.S. EPA, 2010b). Some flame retardants have been shown (Turaga et al., 2010). Norm-activation theory originated with the
to be toxic and mutagenic, and they have been detected in the envi- theory of moral decision-making (Schwartz, 1970, 1973, 1977). Ini-
ronment far from where they are produced or used (Birnbaum and tially crafted to explain helping behavior, it was extended to explain
Staskal, 2004). Moreover, links between the presence of toxic mate- general altruistic behavior, including PEB. According to Schwartz,
rials in the environment and the improper processing of e-waste altruistic behavior arises from situation-specific personal norms
have recently been established in China and India (Sepulveda et al., that reflect self-expectations derived from shared norms. For altru-
2010). istic behavior to result from personal norms, two criteria must
The third reason is that, when done right, recycling metals in be met: a person must be aware that specific actions (or inac-
e-waste contributes to reducing emissions of greenhouse gases, tions) have consequences for the well-being of others (awareness
which is a pure public good (i.e., a good that is non-excludable of consequences, AC); and he must ascribe responsibility for conse-
and non-rival so everyone consumes the same amount). Indeed, quences of those actions (ascription of responsibility, AR). The joint
recycling these metals has a smaller carbon footprint than pro- presence of AC and AR in a specific situation enables pertinent per-
ducing them from virgin material. For example, recycling 1 kg of sonal norms to motivate behavior, but this is not sufficient because
aluminum avoids generating 2 kg of CO2 , 11 g of SO2 , and 1.3 kg of a person may deny his responsibility to act or the impacts of his
bauxite residue (UNEP, 2009). actions on others.
In spite of these benefits, only a fraction of generated e-waste Schwartz’ norm-activation theory was generalized by Stern
is recycled. In the United States, for example, estimated recycling et al. (1999), who proposed a value-belief-norms (VBN) theory that
rates in 2007 were 18% for televisions and computer products, and incorporates egoistic (i.e., based on self interest) and biospheric (i.e.,
only 10% for cell phones (U.S. EPA, 2008). In the EU, official recy- that value other species) value orientations into personal norms
cling and reuse percentages are much higher, ranging from 60% in addition to altruistic values. According to Stern et al. (1999),
to 90% depending on the product category (Eurostat, 2011) but personal norms are activated when a person believes that envi-
large volume of collected e-waste destined for recycling are illegally ronmental conditions have consequences (AC) and that they can
exported to developing countries (Greenpeace, 2009). Several rea- take action (AR) to reduce those consequences. Moreover, the VBN
sons explain this situation: (1) e-waste laws are not enforced or are theory assumes that AC and AR beliefs are shaped by general beliefs
inadequate (Basel Action Network, 2010); (2) incentives to recycle about human–environment interactions, such as the New Environ-
e-waste are insufficient; (3) recycling programs are not adequate; mental Paradigm (Dunlap and Van Liere, 1978; Dunlap et al., 2000)
and (4) households are not sufficiently aware of the risks linked to and stable basic human values (Schwartz, 1994).
the improper disposal of e-waste or they do not know where to
safely dispose of e-waste. We focus here on e-waste recycling by 2.2. Economics and pro-environmental behavior
households because understanding consumer behavior is central
to improving e-waste collection. By contrast, in economics, pro-environmental behavior has typ-
ically been modeled as the private provision of public goods. In
the standard microeconomics model, individuals are driven by
egoistic motivations and act to maximize their utility from con-
2. Background sumption given budget constraints. Economists mostly focus on
external variables such as prices and convenience so their policy
2.1. Social norms, beliefs, and recycling behavior recommendations typically involve changing monetary incentives
or regulations. In this framework, the incentive to free-ride (since
As highlighted in Clark et al. (2003), a number of researchers everyone has access to the same amount of a public good whether
have drawn on environmental psychology to analyze internal they contribute or not) leads to a limited production of public goods.
(psychological) variables such as beliefs, attitudes, or norms as However, empirical data shows that people make more voluntary
underlying motivations for individual behavior; as a result, their contributions for public goods than implied by the standard theory
policy recommendations typically include providing information, (Piliavin and Charng, 1990). In response, two strands have appeared
trying to persuade, or instilling guilt. in the literature to reconcile theory with empirical evidence.
J.-D.M. Saphores et al. / Resources, Conservation and Recycling 60 (2012) 49–63 51

First, Andreoni (1990) extended the standard economic frame- In an early paper, Tiller et al. (1997) conducted a contingent val-
work by incorporating a “warm glow” benefit from contributing to a uation to quantify household willingness to pay (WTP) to establish
public good in addition to the utility from consuming it. Subsequent drop-off recycling programs in a rural/suburban area of Tennessee.
papers that built on that framework proposed prestige (Harbaugh, After analyzing 481 in-person interviews, they reported that WTP
1998) and social approval (Rege, 2004) to explain the source of depends on age, education, attitudes toward recycling, and house-
private benefits from contributing to public goods. hold income. Two other papers (Speirs and Tucker, 2001; Snyder
In a second strand of the literature that started to bridge the et al., 2004) studied drop-off recycling of conventional household
divide between environmental psychology and economics, Brekke wastes, the first in south-west Scotland and the second in Ohio,
et al. (2003) proposed that people contributing to a public good but they did not estimate models to explain drop-off use. More
are motivated by moral “calculus” rather than pure self-interest. In recently, Sidique et al. (2010) analyzed data from 356 in-person
their model, people derive benefits from being perceived as socially interviews at eight drop-off recycling sites around Lansing, Michi-
responsible and compare the benefits of maintaining a socially gan. Using a Poisson model to explain the number of visits to these
responsible image with the costs of doing so in order to find the sites, they found that location, recycling convenience, familiarity
optimal level of voluntary efforts. Nyborg et al. (2006) expanded with recycling and social pressure are important drivers of recy-
their framework by incorporating insights from Schwartz’ norm- cling behavior. In addition, age, education, income and household
activation model. In their model of an individual’s decision to size were statistically significant.
purchase green products, self-image depends both on the social We found only two papers dealing with e-waste recycling at
external benefits of contributing to a public good and on the drop-off centers. The first one is Saphores et al. (2006), who ana-
perception that buying green products is an individual responsi- lyzed the willingness of California households to recycle e-waste
bility, which is based on internalized moral norms. Their empirical at drop-off centers based on 265 respondents to a mail survey.
application to green electricity programs in Sweden supports the They found that gender, education, convenience (having a recy-
hypothesis that perceived responsibility and perceived external cling center within 5 miles), and environmental beliefs, but neither
benefits are important for explaining households’ decision to par- income nor political affiliation, are useful for explaining household
ticipate. willingness to drop-off e-waste at recycling centers. The second
Our paper builds on both environmental psychology and eco- paper (Bouvier and Wagner, 2011) examined the impact of col-
nomics as we consider both external and internal variables to lection facility attributes (recycling fees, days open, and driving
better explain the determinants of pro-environmental behavior. distance) on the collection rate of household computer monitors
For recent examples of this line of research, see Clark et al. (2003) and televisions in 92 municipal facilities located in the U.S. state of
or Hage et al. (2009) and the references therein. Maine. Using pairwise correlation tests, they found that the num-
ber of items collected decreases with recycling fees and increases
with the number of days open, but driving distance did not seem
2.3. Convenience and knowledge
to matter to households.
Several studies have shown that perceived inconvenience
2.5. Recycling e-waste
reduces participation in recycling activities (DeYoung, 1990;
Boldero, 1995; Ewing, 2001; Domina and Koch, 2002; Do Valle et al.,
In addition, several recent studies have focused on e-waste
2004). Inconvenience typically takes two forms: first, it can be asso-
recycling in different parts of the world. In a 2005 paper, Darby
ciated with lack of storage space, excessive time requirements, or
and Obara assessed consumer attitudes toward the disposal of
perceived risks associated with recycling; second, it can require
small waste electrical and electronic equipment (WEEE) in Cardiff,
bringing recyclables to a drop-off location. Published papers con-
Wales, based on a mail survey and semi-structured interviews. They
sistently indicate that curbside recycling schemes, which are more
found that small WEEE items are not commonly recycled. More-
convenient, have higher participation rates than recycling at drop-
over, because of inconvenient transportation options lower income
off locations (Vining and Ebreo, 1992; Boldero, 1995; Domina and
households are less likely to recycle small WEEE items, and they
Koch, 2002; Woodard et al., 2005).
are more likely to extend the useful life of these items. However,
In addition to convenience, the recycling literature consistently
households who recycle regularly common materials at the curb
reports an increasing relationship between public participation
are more likely to bring small WEEE items to collection sites.
and specific knowledge of a recycling scheme (Vining and Ebreo,
In their study of battery recycling in Switzerland, Hansmann
1990; Gamba and Oskamp, 1994; Scott, 1999; Do Valle et al.,
et al. (2006) proposed a model of recycling behavior based on Sykes
2004). Likewise, specific knowledge of the environmental costs
and Matza’s (1957) Neutralization theory to explain inconsistencies
from not recycling (akin to the awareness of consequences in
between social norms, personal attitudes, and individual behavior.
Schwartz’ framework) has been shown to foster recycling (Hopper
Their analysis of data from 608 households collected via a mail sur-
and Nielsen, 1991; Vining and Ebreo, 1992; Bratt, 1999; Ewing,
vey relied on multivariate regression and analysis of covariance.
2001). Conversely, previous research suggests that there is only a
They found that recycling knowledge and attitudes toward recy-
weak relationship between general concerns for the environment
cling are positively related to battery recycling, while attitudes
and recycling (Vining and Ebreo, 1992; Gamba and Oskamp, 1994;
toward waste disposal, trust in waste management authorities, and
Hornik et al., 1995; Scott, 1999; Domina and Koch, 2002). Further-
level of battery consumption are not statistically significant.
more, given the public good dimension of recycling, Bratt (1999)
Very recently, Wang et al. (2011) investigated the attitudes of
argues that when the environmental impacts of recycling are com-
Beijing residents toward e-waste recycling using a survey admin-
municated to the public, collective benefits should be emphasized
istered by neighborhood organizations. Results of their logistic
rather than the consequences of individual behavior.
regression model of 957 respondents indicate that recycling con-
venience, home ownership, recycling habits, and compensation for
2.4. Recycling at drop-off centers recycling are statistically significant for explaining willingness to
recycle e-waste but family size, education, income, and knowledge
To contextualize our results we reviewed the empirical recycling of environmental laws are not.
literature, starting first with published studies concerned with the In the last couple of years, cell phones recycling has attracted
effectiveness of drop-off recycling. increasing attention from researchers. Using data from 115
52 J.-D.M. Saphores et al. / Resources, Conservation and Recycling 60 (2012) 49–63

survey respondents, Nnorom et al. (2009) investigated the will- study of Preston, Lancashire, Perry and Williams (2007) reported
ingness of Nigerian consumers to participate in a waste phone that British Indians are more likely to participate in recycling than
recycling program and their willingness to pay for more envi- their White British counterparts, with significant differences in atti-
ronmentally friendly cell phones. Approximately 65% of their tudes and participation levels for first and second/third generation
respondents were willing to drop-off obsolete electronics at a ethnic minorities.
nearby recycling facility and half were willing to pay a 20% premium Some research also suggests a positive association between res-
for an environmentally friendlier cell phone. idence type and recycling. According to Granzin and Olsen (1991),
Jang and Kim (2010) focused instead on cell phones use, reuse, recyclers who are not homeowners are more likely to rent a house
and recycling in Korea based on a survey of 1090 consumers, site than an apartment, which may suggest that having more storage
visits to e-waste recycling facilities, as well as interviews with space helps recycling.
representatives from mobile phone companies and environmental Finally, Johnson et al. (2004) reported that liberals are signifi-
regulators. They found that from 2000 to 2007 only approximately cantly more likely than others to recycle in their study of the 2000
5 million cell phones out of the 14.5 million retired annually were National Survey on Recreation and the Environment. Owen and
collected by mobile phone companies and cell phone producers. As Videras (2007) confirmed that result using a dataset that covers
a result, they estimated that 28 million used and broken cell phones 14 OECD regions, and they further showed that religious beliefs
were stored by households by the end of 2006. They recommended play a (small) role in PEB.
more active and more convenient collection programs to collect
used cell phones and for a better recycling infrastructure.
2.7. Model variables

2.6. Socio-economic and demographic variables


A large number of studies found that norms are important for
explaining household recycling behavior (see Turaga et al., 2010
Since relatively few published studies focus either on recy-
for references). However, the evidence is mixed on the impact of
cling e-waste or on recycling at drop-off locations, we reviewed
most other common explanatory variables. In particular, the effect
papers dealing with conventional household recyclables to select
of many socio-economic and demographic variables is ambiguous.
our explanatory variables. We found that the most commonly con-
This apparent dissonance may reflect changes in attitudes over time
sidered variables are gender, age, income, and education. As noted
as more environmentally friendly norms have become internal-
in Do Valle et al. (2004) or Hansmann et al. (2006), findings appear
ized, cohort effects, evolving attitudes toward recycling (Bamberg,
to be inconsistent. Moreover, when socio-demographic variables
2003), local particularities (since many published papers studied
are statistically significant their explanatory power tends to be
local recycling programs), sampling idiosyncrasies, the use of dif-
small (Hornik et al., 1995).
ferent explanatory variables, or methodological limitations (some
Let us start with gender. According to Schultz et al. (1995) and
older papers did not rely on multivariate models). As indicated
Johnson et al. (2004), women are more likely to recycle, but Gamba
below, our study relies on a sample of over 2000 respondents whose
and Oskamp (1994), Werner and Makela (1998), Domina and Koch
socio-economic characteristics were known ahead of time to the
(2002), or Do Valle et al. (2004) detected no relationship between
firm that administered our survey, which helped to better match
gender and recycling.
our respondents with the U.S. population and guarantees better
Several studies suggest that older people are more likely to recy-
socio-economic data.
cle (Vining and Ebreo, 1990; Derksen and Gartrell, 1993; Margai,
We therefore decided to consider general attitudes toward the
1997; Scott, 1999; Ewing, 2001) but others found no statistically
environment, norms, convenience, the availability of storage space,
significant relationship between age and recycling (Oskamp et al.,
as well as a broad range of socio-economic variables in order to
1991; Gamba and Oskamp, 1994; Corral-Verdugo, 1997; Werner
explain experience with e-waste recycling and household willing-
and Makela, 1998). Domina and Koch (2002) argued that this
ness to recycle e-waste at drop-off centers.
apparent discrepancy may result from a generational effect and
Hansmann et al. (2006) further suggest that differences between
countries may play a role. 3. Data
A number of authors (e.g., Vining and Ebreo, 1990; Oskamp et al.,
1991; Gamba and Oskamp, 1994; Domina and Koch, 2002) also Our unique dataset was collected via an Internet-based survey
reported that households with a higher income are more likely to of a random subset of Knowledge Networks’ (KN) online research
recycle, but others disagreed (Derksen and Gartrell, 1993; Scott, panel, which is approximately 43,000 strong. Unlike typical Inter-
1999; Do Valle et al., 2004). net research that rely on self-selected respondents, KN has built
Likewise, the impact of education is ambiguous: some studies a research panel representative of the U.S. population based on a
suggested that more education fosters recycling (e.g., see Vining sampling frame that includes both listed and unlisted numbers, and
and Ebreo, 1990; Folz and Hazlett, 1991; Lansana, 1992; Derksen is not limited to current Web users or computer owners.
and Gartrell, 1993; Owens et al., 2000; Johnson et al., 2004) but not Indeed, prospective panel members are recruited by telephone
others (Gamba and Oskamp, 1994; Werner and Makela, 1998; Do using random digit dialing (RDD) sampling techniques that meet
Valle et al., 2004; Meneses and Palacio, 2005). the Federal Government’s quality standards for RDD surveys. Once
Family size, ethnicity, dwelling type/home ownership, and a selected household accepts to join KN’s panel, its members are
political beliefs have also received some attention. The relationship asked to provide socio-economic and demographic information
between household size and recycling behavior does not appear to such as gender, age, race, income, and education. This core infor-
have been widely studied even though we expect household size mation, which is updated annually, is then available for subsequent
to be correlated with consumption and hence waste generation. surveys.
Scott (1999) did not detect such a relationship, but McQuaid and If needed, household members are provided with a free WebTV
Murdoch (1996), Domina and Koch (2002), and Johnson et al. (2004) and Internet access in exchange for taking part in on-line sur-
found that larger families are more likely to recycle. veys. Households who already have a home computer and Internet
The link between ethnicity and recycling has received less atten- access are asked to answer surveys using their own equipment in
tion. In the U.S., Johnson et al. (2004) concluded that Blacks and exchange for points redeemable for cash. Incentive points may also
foreign-born Latinos are less likely to recycle than Whites. In their be offered to improve the response rates of longer surveys. Panel
J.-D.M. Saphores et al. / Resources, Conservation and Recycling 60 (2012) 49–63 53

members need to complete at least one of every six surveys they are own their home and to have access to a vehicle. Overall, how-
assigned, or they are dropped from KN’s panel, but they are never ever, these differences are relatively minor and our large sample
asked to complete more than four surveys per month. exhibits enough variability to explain e-waste recycling experience
Randomly selected panel members are notified by mail several and willingness to recycle e-waste at drop-off centers in the United
days prior to taking an on-line survey. When the questionnaire States.
is available on-line, they receive an email. This approach per- To capture our respondents’ environmental beliefs, we asked
mits fielding survey quickly and economically; it also allows them to complete a ten-item modified New Ecological Paradigm
respondents to answer survey questions at their convenience. (NEP) scale (Dunlap et al., 2000) that consists of 10 items from
Email reminders are sent to non-respondents, followed by a the original 15-item list. These questions, which use a five-point
reminder phone call after at least three days. For more details, see Likert response scale, are shown in Table 2. Using 5 pairs of
http://www.knowledgenetworks.com/knpanel/index.html. statements, this modified scale investigates five environmental
After getting feedback on a draft survey instrument from col- concerns: the limits of growth (items 1 and 6); the idea that humans
leagues and students and modifying our survey instrument, we are exempt from the constraints of nature (items 2 and 9); anti-
asked Knowledge Networks to administer a pilot survey. This was anthropocentrism (items 3 and 7); the fragility of nature’s balance
done from February 15 to February 20, 2006; 110 of 167 panelists (items 4 and 8); and the possibility of an ecological crisis (items 5
answered our questions. Based on their feedback, a revised survey and 10). We excluded five items from the original scale based on
was fielded from March 28 to April 7, 2006; this time 2,136 of 3,048 low item–total correlations reported in previous studies (Dunlap
panelists answered our survey for a response rate of 70.1%, which et al., 2000; Hawcroft and Milfont, 2010).
is similar to other on-line surveys conducted by KN. We also asked our respondents to complete a seven-item altru-
Our survey had four parts. The first part asked about envi- ism/moral norm scale using the same five-point Likert response
ronmental attitudes and involvement with non-governmental scale as for the NEP (see Table 3). This scale, which was devel-
organizations. The second part gathered information about house- oped specifically for this research, is inspired from Schwartz’
hold trash and recycling activities. In the third part, we asked about norm-activation model (1970, 1977). Our scale incorporates four
willingness to bring e-waste to drop-off recycling centers, the dan- questions designed to capture AC (items 1–4 in Table 3), two to
gers of dumping e-waste, and knowledge of e-waste laws. The last represent AR (items 5 and 6), and one statement that refers to a
part asked additional demographic questions to better understand personal moral norm for recycling (item 7).
the beliefs and behavior of various ethnic groups.
Table 1 shows how our respondents’ key socio-demographic 4. Methods
characteristics compared with U.S. demographic data from the
2006 Community Survey. Our respondents are slightly older (over After analyzing selected survey questions using simple statis-
55 years), more likely to be married and to have children under tical tests, we performed a factor analysis (FA; Thompson, 2004)
18 at home. In addition, we note that African Americans and His- in order to condense the ten survey questions related to the NEP
panics are under-represented. Although our respondents tend to (see Table 2) and the seven questions related to moral norms (see
have more years of formal education, high income earners are Table 3) into a small set of factors. To test the adequacy of our
under-represented. Moreover, our respondents are more likely to results, we performed some standard tests summarized in the notes

Table 1
Respondents selected demographic characteristics versus 2006 census estimates.

Characteristic This survey (2136 respondents) 2006 census estimates

Age 18–24: 8.1%; 45–54: 18.6% ♣ 20–24: 9.7%; 45–54: 20.0%


25–34: 18.1%; 55–64: 17.5% 25–34: 18.4%; 55–64: 14.6%
35–44: 19.5%; ≥65: 18.3.9% 35–44: 20.2%; ≥65: 17.1%

Marital status Married: 58.0%: Divorced: 12.5% Married: 50.4%: Divorced: 10.5%
Widowed: 4.8%; Never married: 22.9% Widowed: 6.4%; Never married: 30.5%
Separated: 1.8% Separated: 2.3%

Household size 1: 20.6%; 2: 38.8%; 3: 17.2% 1: 27.3%; 2: 33.2%; 3: 16.0%


4 or more: 23.3% 4 or more: 23.6%

Number of children under 18 in household 0: 72.3%; 1 or more: 27.7% 0: 65.4%; 1 or more: 34.6%

Ethnicity White: 77.7%; Hispanic: 7.3% White: 66.2%; Hispanic: 14.8%


African-American: 8.5%; African-American: 12.2%;
Other: 6.5% Other: 6.9%

Education Less than high school: 11.4% Less than high school: 15.9%
High school: 33.3% High school: 30.2%
Some college: 29.5% Some college: 26.9%
Bachelor’s or higher: 25.7% Bachelor’s or higher: 27.0%

Household income (in 1000 of $) <25: 26.5%; 75–99.9: 11.6% <25: 17.5%; 75–99.9: 14.3%
25–49.9: 32.8%; 100–149.9: 6.6% 25–49.9: 24.7%; 100–149.9: 13.7%
50–74.9: 19.8%; ≥150: 2.8% 50–74.9: 20.9%; ≥150: 9.0%

Home ownership Own: 69.9%; Rent: 30.1% Own: 67.3%; Rent: 32.7%

Type of dwelling Single-family home: 67.2% Single-family home: 67.1%


Duplex: 5.5%; Mobile home: 7.9% Duplex: 4.0%; Mobile home: 6.9%
Apart./Condo: 17.6%; Other: 1.8% Apart./Condo: 21.8%; Other: 0.1%

Vehicle availability 94.3% of respondents have a vehicle available for their use. 91.2% of households have a vehicle available for their use.

Source of the census data: selected social characteristics in the United States: 2006 – American community survey 1-year estimates (http://factfinder.census.gov/).
♣: The census reports a slightly different age breakdown compared to our survey (15–19 and 20–24 instead of 18–24).
54 J.-D.M. Saphores et al. / Resources, Conservation and Recycling 60 (2012) 49–63

Table 2
Percentage distribution and factor loadings for New Ecological Paradigm (NEPn) items.

Item SA MA U MD SD r Factor loading

1 We are approaching the limit of the number of people the 15.1% 21.4% 35.6% 17.0% 10.9% 0.582 0.5212
earth can support.
2 Human ingenuity will insure that we do not make the 8.6% 24.2% 35.0% 20.7% 11.5% 0.445 0.3936
earth unlivable.
3 Plants and animals have as much right as humans to exist. 48.0% 28.7% 9.6% 9.0% 4.8% 0.605 0.5389
4 The balance of nature is strong enough to cope with the 4.1% 13.8% 26.4% 31.3% 24.4% 0.641 0.5878
impacts of modern industrial nations.
5 The so-called “ecological crisis” facing humankind has 6.6% 18.9% 29.1% 24.3% 21.0% 0.697 0.6331
been greatly exaggerated.
6 The earth is like a spaceship with very limited room and 21.2% 31.9% 20.7% 18.5% 7.7% 0.605 0.5391
resources.
7 Humans were meant to rule over the rest of nature. 16.2% 20.8% 18.3% 22.9% 21.8% 0.579 0.4889
8 The balance of nature is very delicate and easily upset 29.5% 38.1% 18.3% 11.4% 2.8% 0.678 0.6271
9 Humans will eventually learn enough about how nature 4.4% 20.3% 29.2% 24.2% 21.9% 0.238 0.2062
works to be able to control it.
10 If things continue on their present course, we will soon 23.8% 31.9% 27.5% 10.6% 6.2% 0.731 0.6777
experience a major ecological catastrophe.

Notes:
1. SA, MA, U, MD, SD refer respectively to “strongly agree”, “mildly agree”, “unsure”, “mildly disagree”, and “strongly disagree”.
2. “r” designates the inter-item correlation, i.e., the correlation between an item and NPEn. All inter-item correlations are statistically significant (p < 0.01).
3. To select the number of factors to consider (only one here), we adopted the Kaiser criterion (i.e., we retained only factors corresponding to eigenvalues greater than 1).
4. After calculating the NEPn factor using the “loadings” in the last column above, we normalized it so it is between 0 and 1 where 1 indicates strong pro-environmental
attitudes.
5. The NEPn factor passes common specification tests (Thompson, 2004). The Bartlett test for sphericity strongly rejects the null hypothesis that these 10 items are not
correlated (p < 0.001), and the Kaiser–Meyer Olkin statistics (KMO = 0.837 here) indicates that correlations are not excessive (KMO ranges between 0 and 1 and should be at
least 0.6). Cronbach’s alpha, which measures factor reliability, is 0.783, which suggests that NEPn is a reliable factor (a value ≥0.7 is typically required).
6. NEPn explains 28.9% of the variance of constituting items, which is similar to previous studies.

Table 3
Percentage distribution and factor loadings for moral norm (MORn) items.

Item SA MA U MD SD r Factor loading

1 Recycling substantially reduces the use of landfills 42.9% 34.0% 16.6% 4.2% 2.4% 0.674 0.5839
2 Recycling conserves natural resources 47.3% 35.9% 13.3% 1.6% 1.9% 0.720 0.6516
3 Recycling won’t make much difference in the quality of the 2.7% 6.6% 15.4% 33.9% 41.4% 0.735 0.6281
environment
4 Recycling creates jobs 27.8% 43.8% 23.8% 2.9% 1.6% 0.585 0.5202
5 Households like mine should not be blamed for environmental 12.5% 22.7% 25.0% 25.5% 14.3% 0.343 0.2463
problems caused by excessive trash generation
6 My responsibility is to provide only for my family and myself 9.6% 15.5% 13.9% 36.0% 25.0% 0.466 0.3431
7 Given increasing environmental problems with solid waste, I 36.0% 34.1% 17.6% 8.8% 3.5% 0.731 0.5926
feel a moral obligation to recycle

Notes:
1. SA, MA, U, MD, SD refer respectively to “strongly agree”, “mildly agree”, “unsure”, “mildly disagree”, and “strongly disagree”.
2. “r” designates the inter-item correlation, i.e., the correlation between an item and MORn. Inter-item correlations are all statistically significant (p < 0.01).
3. To select the number of factors to consider, we relied on the Kaiser criterion (i.e., we retained only factors corresponding to eigenvalues greater than 1).
4. After calculating the moral norm factor using the coefficients (“loadings”) in the last column above, we normalized it so it is between 0 and 1 where 1 indicates strong
pro-recycling attitudes.
5. The moral norm factor also passes common specification tests (Thompson, 2004). The Bartlett test for sphericity strongly rejects the null hypothesis that these 7 items are
not correlated (p < 0.001), and the Kaiser–Meyer Olkin (KMO) statistics, which equals 0.808 here, indicates that correlations are not excessive (KMO ranges between 0 and 1
and should be at least 0.6. Cronbach’s alpha, which measures factor reliability, is 0.701, which is adequate (a value ≥0.7 is typically required).
6. The MORn factor explains 28.8% of the variance among the underlying items, which is similar to previous studies.

to Tables 2 and 3. These factors were then used in our models that We hypothesized that the propensity to recycle may depend on
characterize e-waste recyclers and our respondents’ willingness to environmental attitudes (summarized by the NEP); altruism/moral
recycle e-waste at drop-off centers. norms; perceived obstacles to recycling (safety, lack of time) and
convenience (including distance to the nearest drop-off recycling
center and access to a vehicle); knowledge of e-waste laws and
4.1. Characterization of e-waste recyclers potential toxicity; income (which impacts the value of time); age;
education; gender; and marital status. For i = 1, . . ., N (where N is
To characterize e-waste recyclers, we included the relevant the number of respondents), this relationship can be written as:
information available about each respondent in a logit model where
the dependent variable yi equals 1 if respondent i had recycled e- 
yi∗ = xi ␤ + i , (2)
waste before and 0 otherwise. Eq. (1) links yi to an unobserved
variable yi∗ that represents the propensity of respondent i to recycle
where xi is a vector of explanatory variables; ␤ is a vector
e-waste:
of unknown parameters common to all respondents; and i is
 an error term with mean zero such that the i s are indepen-
1 if yi∗ > 0,
yi = (1) dently distributed. Descriptive statistics for our models variables
0 if yi∗ ≤ 0.
are presented in Table 4. We estimated both a logit (i has a
J.-D.M. Saphores et al. / Resources, Conservation and Recycling 60 (2012) 49–63 55

Table 4
Descriptive statistics for model variables.

Variable Name Meaning Mean Std. Dev. Min Max

Dependent variables♣
recycledew Has already recycled e-waste at drop-off center? Yes = 1 0.334 0.472 0 1
willrec Willingness to recycle at a drop-off location 1.837 0.934 1 4

Explanatory variables
NEPn Normalized New Ecological Paradigm 0.618 0.181 0.000 0.960
MORn Normalized moral norm 0.734 0.164 0.000 1.000
curbd2p Has had access to curbside recycling for at least 2 years? Yes = 1 0.373 0.484 0 1
nhrec High level of recycling in your neighborhood? Yes = 1 0.205 0.403 0 1
reqrecws Required to recycle at work or at school? Yes = 1 0.146 0.354 0 1
dBRoom Problem finding room to store recyclables (agrees = 1) 0.507 0.500 0 1
dBRctf Recycling drop-off centers are too far away (agrees = 1) 0.325 0.469 0 1
dBSafe Storing recyclable materials at home is safe (agrees = 1) 0.575 0.494 0 1
dBTime Finding time to recycle is a problem (agrees = 1) 0.323 0.468 0 1
dewdrpof12 E-waste drop-off within 5 miles? Yes = 1 0.194 0.396 0 1
dewdrpoff3 E-waste drop-off within 6 to 10 miles? Yes = 1 0.095 0.293 0 1
dewdrpoff5 Does not know how far the nearest e-waste drop-off is: Yes = 1 0.612 0.488 0 1
ewtoxic Knows that e-waste contains toxic materials? Yes = 1 0.606 0.489 0 1
lawew Aware of e-waste laws? No = 1 0.667 0.471 0 1
caravail Has access to a motor vehicle? Yes = 1 0.943 0.231 0 1
age3044 Age 30–44 years old? Yes = 1 0.263 0.440 0 1
age4559 Age 45–59 years old? Yes = 1 0.283 0.451 0 1
age60plus Age 60 years or more? Yes = 1 0.255 0.436 0 1
eduhs High school highest degree in household? Yes = 1 0.332 0.471 0 1
eduscol Some college highest level of ed. in household? Yes = 1 0.296 0.456 0 1
educol A household member has a college education? Yes = 1 0.259 0.438 0 1
eth afa African American? Yes = 1 0.084 0.278 0 1
eth his Hispanic/Latino? Yes = 1 0.071 0.257 0 1
eth oth Other ethnicity? Yes = 1 0.065 0.247 0 1
gender Female? Yes = 1 0.519 0.500 0 1
lhhs Natural logarithm of household size 0.811 0.530 0.000 2.303
lincome Natural logarithm of household income 3.718 0.861 0.916 5.165
incgt175 Annual household income >$175,000? Yes = 1 0.017 0.128 0 1
married Married? Yes = 1 0.578 0.494 0 1
hous sfh Lives in single family house? Yes = 1 0.727 0.445 0 1
denrural Lives in a rural area? Yes = 1 0.313 0.464 0 1

Notes:
These statistics were calculated for the 1995 observations used to estimate our models.
♣: The variable “recycledew” is the dependent variable in our logit model that characterizes recyclers and an explanatory variable in our model that explains willingness to
recycle.

logistic distribution with variance 2 /3) and a probit (i ∼ N(0, 1)) so the probability of outcome j ∈ {1, 2, 3, 4} described in Eq. (4)
model. is given by
⎧ 
⎨ F(1 + zi ), if j = 1,
 
4.2. Willingness to recycle e-waste Pr(yi = j) = F(j + z ) − F(j−1 + z ), if j = 2 or 3, (5)
⎩ 1 − F( i + z ), if j =
i
4,
3 i
We also asked our respondents to indicate their willingness to
recycle e-waste at drop-off centers by choosing one of four cate- where the link function F denotes the cumulative distribution func-
gories: (1) “Very willing;” (2) “Willing;” (3) “Somewhat willing;” tion (cdf) of ε, and the  1 s are unknown “cutpoints” that may be
and (4) “Not willing at all.” To explain our respondents’ willing- estimated jointly with the vector ␥ via maximum likelihood. If F
ness to recycle e-waste, we started from standard ordered response is the cdf of a standard normal distribution, Eqs. (3)–(5) describe
models like the ordered logit or ordered probit (Long and Freese, an ordered probit; if instead F is the logistic cdf with Var(ε) = 2 /3,
2006). To introduce these models, let us consider again an unob- they define an ordered logit model.
served variable y*, which is here a continuous index representing One implication of the ordered response models described by
willingness to recycle e-waste, but to simplify the interpretation of Eqs. (3)–(5) is the “parallel lines” assumption. From Eq. (5), we see
our results we write that for j ∈ {1, 2, 3} cumulative probabilities are

 Pr(yi ≤ j|z) = F(j + zi ), (6)
yi∗ = −zi  + εi . (3)
so probability curves for different values of j are parallel. The gen-
where zi is a vector that includes the same variables as the propen- eralized ordered model relaxes this “parallel lines” assumption by
sity to recycle e-waste plus a binary variable that indicates prior allowing the ␥s to differ for each willingness category. This simply
experience with e-waste recycling. In Eq. (3), y* relates to the entails replacing ␥ with ␥j in Eqs. (5) and (6), so they become:
observed dependent variable y by ⎧ 
⎧ ⎨ F(1 + zi  1 ), if j = 1,
 
⎪ 1 = very willing, if 0 ≡ −∞ < yi∗ < 1 , Pr(yi = j) = F(j + z  ) − F(j−1 + z  ), if j = 2 or 3, (7)
⎨ ∗ ⎩ 1 − F( i +j z  ), if j i= j−1
2 = willing, if 1 ≤ yi < 2 , 3 4,
yi = (4) i 3
⎪ 3 = somewhat willing, if 2 ≤ yi∗ < 3 ,
⎩ 
4 = not willing at all, if 3 ≤ yi∗ < 4 ≡ +∞, Pr(yi ≤ j|z) = F(j + zi  j ). (8)
56 J.-D.M. Saphores et al. / Resources, Conservation and Recycling 60 (2012) 49–63

The generalized ordered model is therefore equivalent to a series Has recycled e-waste before
of binary models where categories of the dependent variable are
A
combined: for j ∈ {1, 2, 3}, the first j categories are lumped together Curbside recycling Has not yet recycled e-waste
available for >=2
and contrasted with categories j + 1, . . ., 4 grouped together. years? All respondents
Allowing parameter values to differ for all categories would 80.0%
be very cumbersome, however. To obtain a more concise model, The problem with 60.0% High level of
recycling is finding recycling in your
Williams (2006) proposed performing a series of Wald tests time to do it 40.0% neighborhood?
(Greene, 2008) to see whether the coefficients of each variable dif-
20.0%
fer across equations. If Wald tests are statistically not significant
0.0%
for one or more variables, the variable with the least significant
Storing materials at Required to recycle at
test values is constrained to have equal coefficients across cate- home is safe school or work?
gories, the model is refitted with constraints, and the process is
repeated until there are no more variables that meet the parallel
Recycling drop-off Finding room to store
lines assumption. A global Wald test is then performed to check centers are too far recyclable materials is
that the final model does not violate the parallel lines assumption. away a problem
As explained in Williams (2006), the choice of a link function is
also important. While logit and probit models are reasonable when
changes in cumulative probabilities are slow, other link functions B Has recycled e-waste before
should be used in the presence of faster changes. The log–log link is
Don’t know distance Has not yet recycled e-waste
better suited when cumulative probabilities increase fairly slowly to nearest e-waste
from 0 and more rapidly when approaching 1. If the opposite holds, drop-off center All respondents
80.0%
the complementary log–log link may be more appropriate. Finally
the cauchit distribution is best when there are extreme values in 60.0%
>10 mi from nearest Aware of e-waste
both tails. We therefore estimated models with different link func- e-waste dropoff center 40.0% laws
tions and selected one on the basis of the Akaike and Bayesian 20.0%
information criteria (AIC and BIC). In addition, we performed a
0.0%
number of specification tests prior to finalizing our results (see
notes in Table 6). Knows that e-waste
6-10 mi from nearest
contains toxic
e-waste dropoff center
materials
5. Results and discussion
<=5 mi from nearest
Our results were estimated using Stata 11.2. They are presented e-waste drop-off
in Tables 2–7 and in Figs. 1–3. To set the stage for our models, let center
us first examine how some likely important variables impacted our
Fig. 1. Impact of key variables on e-waste recycling experience. Panel A: Household
dependent variables.
recycling and e-waste recycling experience. Panel B: E-waste knowledge/recycling
convenience and e-waste recycling experience

5.1. Analysis of answers to some key survey questions

As shown in Table 4, 33.4% of our respondents have recycled


e-waste at a drop-off recycling center; in our discussion below we
call these people e-waste recyclers and we call non recyclers those
who had not brought e-waste to a drop-off recycling center at the
time of this survey.
Fig. 1 shows the impact of some likely important variables on
household e-waste drop-off recycling experience. Results are pre-
sented as a percentage of respondents in each of three categories:
“has recycled e-waste before” at drop-off recycling centers, “has not
yet recycled e-waste”, and “all respondents.” Reported p-values in
the discussion below pertain to 2 (1) tests that compare the pro-
portion of yes/no answers from e-waste recyclers and non recyclers
for each specific question.
From Panel A, we see as expected, that e-waste recyclers are
more likely to have been exposed to curbside recycling for conven-
tional household waste for at least 2 years (p-value ≤ 0.001), to live Fig. 2. E-waste recycling experience and willingness to recycle e-waste at a drop-off
in a neighborhood where recycling is high (p-value ≤ 0.001), and to center.
have required recycling at work or school (p-value ≤ 0.001). Like-
wise, e-waste recyclers are less likely to be concerned with common Panel B displays the impact of e-waste knowledge and recy-
recycling obstacles such as finding room for recyclables (p- cling convenience on e-waste recycling experience. First, we note
value ≤ 0.001), recycling safety (p-value ≤ 0.001), and time needed that many more e-waste non-recyclers than recyclers do not know
to recycle (p-value ≤ 0.001), but not so much with the distance to how far they are from the nearest drop-off recycling center (73.1%
the nearest drop-off recycling center (p-value = 0.072). These differ- versus 37.7%; p-value ≤ 0.001). E-waste recyclers are also more
ences may reflect that people who recycled e-waste before choose aware of e-waste disposal laws than non-recyclers, although by a
to emphasize a sense of satisfaction from contributing to environ- smaller margin (39.9% versus 30.4%; p-value ≤ 0.001). On the other
mental quality in order to overcome the drudgery of that task, as hand, e-waste recyclers are much more likely to know that e-waste
argued by Werner and Makela (1998). contains toxic materials than non-recyclers (76.8% versus 52.7%;
J.-D.M. Saphores et al. / Resources, Conservation and Recycling 60 (2012) 49–63 57

A Very willing categories tend to be larger than for Panel A in Fig. 1. People who
Willing are “very willing” or “willing” are more likely to have had access
Curbside recycling
available for >=2 Somewhat willing to curbside recycling for conventional household recyclables for at
years? Not willing at all least 2 years (p ≤ 0.001), to live in a neighborhood where recycling
80.0% is high (p ≤ 0.001), and to have recycling required at work or school
The problem with 60.0% High level of (p ≤ 0.001). Moreover, people who are more willing to recycle
recycling is finding recycling in your
time to do it 40.0% neighborhood? e-waste are more likely to dismiss obstacles to recycling (find-
ing room (p ≤ 0.001), time (p ≤ 0.001), safety (p ≤ 0.001), and to a
20.0%
smaller extent, the distance to the nearest drop-off recycling center
0.0% (p = 0.021)). This relationship is not monotonic, however, because
Storing materials at Required to recycle of the “not willing at all” category but we note that few respondents
home is safe at school or work?
expressed a complete unwillingness to recycle e-waste (see Fig. 2).
Panel B displays the impact of e-waste knowledge and recycling
Finding room to convenience on willingness to recycle e-waste. We see that the less
Recycling drop-off
store recyclable
centers are too far
materials is a willing categories have the highest fraction of respondents who
away
problem do not know where the nearest e-waste drop-off recycling center
is located (p-value ≤ 0.001). The same applies to knowing e-waste
B disposal laws (p ≤ 0.001) but there is no monotonic relationship
Very willing
with the level of willingness to recycle e-waste because people in
Willing
Don’t know distance the “not willing at all” category seem to know more than those
to nearest e-waste Somewhat willing
in the “somewhat willing” category. We also observe an inverse
drop-off center Not willing at all
80.0% relationship between willingness to recycle and knowledge that e-
60.0%
waste contains toxic materials, with marked differences between
>10 mi from nearest Aware of e-waste the “very willing” (72.2%) and the “not willing at all” categories
e-waste dropoff center 40.0% laws (only 41.0%).
20.0% Simple correlation analyses can be misleading, however, so let
0.0% us go over our models’ results to understand the joint impact of our
explanatory variables after summarizing our factor analysis results.
Knows that e-waste
6-10 mi from nearest
contains toxic
e-waste dropoff center 5.2. Factor analysis
materials

Our analysis of the ten questions related to the New Ecologi-


<=5 mi from nearest
e-waste drop-off cal Paradigm yielded one factor, denoted by NEPn. To facilitate its
center interpretation, it was normalized to vary between 0 and 1 where
1 indicates strong pro-environmental beliefs. More specifically,
Fig. 3. Impact of key variables on willingness to recycle e-waste. Panel A: Household
recycling and willingness to recycle e-waste. Panel B: E-waste knowledge/recycling agreement with items 1, 3, 6, 8 and 10 in Table 2 and disagreement
convenience and willingness to recycle e-waste. with items 2, 4, 5, 7, and 9 indicates pro-environmental beliefs.
Overall, our respondents’ pro-environmental beliefs are not quite
as marked as in previous studies, probably because we surveyed a
p-value ≤ 0.001); this suggests that knowing that e-waste contains wide group of people and used 10 items (see Hawcroft and Milfont,
toxic materials (60.6% of our respondents) is not enough to moti- 2010). Table 2 shows the coefficients used to build the NEPn factor
vate people to recycle e-waste because otherwise the fraction of before normalization. As indicated in Table 2 footnotes, this factor
people with e-waste recycling experience (only 33.4%) would be passes all common specification tests.
much higher. Convenience appears to matter because we observe Results for the moral norm scale are reported in Table 3; fac-
that the fraction of e-waste recyclers decreases with distance to the tor loadings before normalization are listed in the last column. We
nearest drop-off recycling center. again obtained a single normalized factor that spans the interval
Knowing about previous e-waste recycling experience should [0, 1], where values close to 1 suggest a high moral motivation for
help explain people’s willingness to continue recycling e-waste at recycling; it results from agreeing with items 1, 2, 4 and 7 and dis-
drop-off centers and our results confirm that: as shown in Fig. 2, agreeing with items 3, 5, and 6. Overall, our respondents seem to
respondents with e-waste drop-off recycling experience are more be aware of the benefits of recycling but they may not want to carry
willing to recycle e-waste at drop-off centers than those who have the responsibility for excessive trash generation (item 5).
never done it (2 (3) = 120.97; p-value < 0.001). Overall, 75.6% of our
respondents are either very willing or willing to recycle e-waste 5.3. Characterization of e-waste recyclers
at drop-off recycling centers, which is good news although good
intentions do not always lead to action (e.g., see Corral-Verdugo, Results from our logit model that characterizes people who recy-
1997). cled e-waste before are reported in Table 5. We estimated both a
We also assessed the impact of likely important variables on logit and a probit model; they gave very similar results so we kept
people’s willingness to recycle e-waste. Fig. 3 presents results as a the logit.
percentage of respondents in four categories characterized by dif- Let us start with statistically significant variables. Among these,
ferent levels of willingness to recycle e-waste at drop-off centers. the variable measuring moral beliefs stands out with the highest
Reported p-values below pertain to 2 (3) tests that compare the odds ratio at 3.246; its importance is consistent with the findings
proportion of yes/no answers from people in these four different of Hopper and Nielsen (1991), or more recently Fornara et al. (2011).
categories. The variable indicating knowledge that e-waste contains toxic
Panel A illustrates the impact of household recycling experience materials also has a fairly large odds ratio (2.349), which suggests
(for common materials such as glass, newspapers or metal cans) on that specific knowledge about the potential dangers of e-waste
willingness to recycle e-waste. We note that differences between does matter. Interestingly, mandatory recycling at work or school
58 J.-D.M. Saphores et al. / Resources, Conservation and Recycling 60 (2012) 49–63

Table 5
Results of logistic model characterizing e-waste recyclers.

Variable name Coefficient value Robust standard error Odds ratio Odds ratio robust std. error

NEPn −0.177 (0.328) 0.837 (0.274)


MORn 1.177*** (0.396) 3.246*** (1.285)
curbd2p 0.060 (0.132) 1.062 (0.141)
nhrec −0.013 (0.148) 0.987 (0.146)
reqrecws 0.404*** (0.145) 1.497*** (0.218)
dBRoom −0.185 (0.119) 0.831 (0.099)
dBRctf −0.078 (0.124) 0.925 (0.115)
dBSafe −0.037 (0.120) 0.964 (0.115)
dBTime −0.141 (0.133) 0.869 (0.116)
dewdrpof12 0.284 (0.193) 1.329 (0.257)
dewdrpoff3 −0.030 (0.214) 0.971 (0.208)
dewdrpoff5 −1.259*** (0.173) 0.284*** (0.049)
ewtoxic 0.854*** (0.122) 2.349*** (0.287)
lawew −0.164 (0.114) 0.849 (0.097)
caravail 0.371 (0.271) 1.449 (0.392)
age3044 −0.071 (0.169) 0.931 (0.158)
age4559 0.194 (0.168) 1.214 (0.204)
age60pl 0.331* (0.185) 1.392* (0.258)
eduhs −0.315 (0.192) 0.730 (0.140)
eduscol 0.050 (0.197) 1.051 (0.207)
educol 0.065 (0.211) 1.067 (0.225)
eth afa −0.097 (0.223) 0.907 (0.203)
eth his 0.195 (0.227) 1.215 (0.276)
eth oth 0.272 (0.204) 1.312 (0.268)
gender 0.012 (0.109) 1.012 (0.110)
lhhs 0.239** (0.114) 1.270** (0.144)
lincome 0.082 (0.083) 1.085 (0.090)
incgt175 −0.351 (0.440) 0.704 (0.310)
married −0.014 (0.127) 0.986 (0.125)
hous sfh −0.015 (0.130) 0.985 (0.128)
denrural −0.034 (0.125) 0.966 (0.121)

Notes:
Results are based on N = 1995 observations. They were estimated using Stata 11.
The dependent variable is binary: it equals 1 if a respondent has recycled e-waste before and 0 otherwise.
For the meaning of the variables above, see Table 4.
The robust standard errors are the Huber–White–Sandwich estimates; they provide an insurance against mild departures from the homoskedasticity assumption.
A test based on “Collin” in Stata suggests no multicollinearity problem.
An analysis of standardized residuals showed that observations with large residuals are not influential.
Specification tests using Stata’s linktest and likelihood ratio tests showed that models without powers and interactions are adequate here.
McFadden’s R2 = 0.154; the log likelihood of the model without predictors is treated as a total sum of squares, and the log likelihood of the full model is treated as the sum of
squared errors; the ratio of likelihoods suggests the level of improvement over the intercept model offered by the full model.
*
Significance at the 10% level respectively.
**
Significance at the 5% level respectively.
***
Significance at the 1% level respectively.

has a positive spillover on e-waste recycling (odds ratio = 1.497). friendly waste disposal have no impact on batteries recycling in
As expected (e.g., see Vining and Ebreo, 1990; Domina and Koch, Switzerland.
2002; or Sidique et al., 2010), we also find that convenience (here Moreover, experience with recycling of common household
proximity to the nearest recycling drop-off center) is important. materials (such as newspaper, glass or metals) or the neighborhood
More specifically, our respondents do not care about living very level of that activity have no impact on whether or not some-
close from an e-waste drop-off recycling center (the binary variable one recycled e-waste before. This result seems at odds with Darby
indicating that such a center is within 2 miles is not significant) but and Obara (2005) who found that households who regularly recy-
not knowing where the nearest recycling center is located deters cle common household waste at the curb are more likely to bring
e-waste recycling (odds ratio = 0.284). small e-waste items to collection sites, although we found a similar
Of the socio-demographic variables, only two are statistically spillover for people who must recycle at school or work. Knowledge
significant: being over 60 years old and family size. Being older than of e-waste disposal laws, common obstacles to recycling, and even
60 increases the odds that someone recycled e-waste before (odds car availability also do not seem to matter. Several reasons may
ratio = 1.392), which is consistent with previous findings that older explain these results: e-waste disposal laws are rarely enforced at
people are more likely to recycle (Vining and Ebreo, 1990; Domina the household level; recycling e-waste is infrequent, safe and usu-
and Koch, 2002; Sidique et al., 2010). Likewise, family size has a ally does not require a lot of extra space; and 94% of our respondents
positive impact on e-waste recycling (odds ratio = 1.270), which had access to a vehicle.
confirms results by Sidique et al. (2010) for e-waste and by Gamba In addition, education, ethnicity, income, marital status, home
and Oskamp (1994) for general household recycling. ownership, and even gender have no impact on whether or not
Our other explanatory variables are not statistically significant, someone recycled e-waste before.
starting with general environmental beliefs summarized by the
NEPn factor. This supports Ajzen and Fishbein’s (1977) conclusion 5.4. Households’ willingness to recycle e-waste at drop-off centers
that the links between general environmental attitudes/knowledge
and specific behaviors are usually weak. Likewise, Hansmann Let us now examine results from our analysis of households’
et al. (2006) report that general attitudes toward environmentally willingness to recycle e-waste at drop-off recycling centers. A
J.-D.M. Saphores et al. / Resources, Conservation and Recycling 60 (2012) 49–63 59

Table 6
Model estimation results.

Variable Ordered logit (OL) model Generalized ordered logit Concise generalized ordered logit
(GOL) model (cGOL) model

Coefficient Robust std. error Coefficient Robust std. error Coefficient Robust std. error

Coefficients for Eq. (6) Common coefficients for Eq. (8) Common coefficients for Eq. (8)
unless otherwise specified unless otherwise specified
***
NEPn 1.078 (0.290) 1.135*** (0.293) 1.113*** (0.294)
MORn 4.547*** (0.348) 4.571*** (0.353) 4.532*** (0.350)
curbd2p −0.159 (0.114) −0.166 (0.114) –
nhrec 0.368*** (0.142) 0.355** (0.141) 0.323** (0.130)
reqrecws 0.115 (0.140) 0.021 (0.153) −0.006 (0.149)
dBRoom −0.070 (0.105) −0.080 (0.105) –
dBRctf −0.099 (0.102) −0.111 (0.103) –
dBSafe 0.313*** (0.095) 0.316*** (0.096) 0.313*** (0.094)
dBTime −0.098 (0.107) −0.180 (0.122) −0.209* (0.112)
dewdrpof12 0.371* (0.197) 0.367 *
(0.199) 0.416 ***
(0.126)
dewdrpoff3 0.414* (0.233) 0.435* (0.235) 0.475*** (0.178)
dewdrpoff5 −0.047 (0.175) −0.043 (0.177) –
ewtoxic 0.610*** (0.099) 0.604*** (0.099) 0.599*** (0.098)
lawew −0.125 (0.103) −0.130 (0.103) –
recycledew 0.605*** (0.108) 0.606*** (0.109) 0.633*** (0.104)
caravail 0.478** (0.213) 0.442 **
(0.206) 0.450** (0.206)
age3044 0.148 (0.139) 0.109 (0.154) 0.066 (0.119)
age4559 0.057 (0.145) 0.054 (0.145) –
age60pl 0.153 (0.163) 0.152 (0.162) –
eduhs −0.072 (0.175) −0.060 (0.170) –
eduscol −0.028 (0.181) −0.021 (0.176) –
educol 0.157 (0.191) 0.053 (0.196) 0.071 (0.120)
eth afa −0.370** (0.176) −0.365 **
(0.177) −0.380** (0.174)
eth his −0.234 (0.204) 0.020 (0.212) 0.004 (0.199)
eth oth 0.167 (0.205) 0.189 (0.207) –
gender 0.326*** (0.093) 0.341*** (0.093) 0.334*** (0.092)
lhhs 0.111 (0.102) 0.127 (0.102) –
lincome 0.027 (0.066) 0.020 (0.065) –
incgt175 0.146 (0.382) 0.162 (0.401) –
married 0.308*** (0.110) 0.293 ***
(0.110) 0.344*** (0.095)
hous sfh 0.003 (0.112) 0.014 (0.111) –
denrural 0.011 (0.109) 0.010 (0.109) –
Coefficients specific to j = 2 in Eq. (8) Coefficients specific to j = 2 in Eq. (8)
reqrecws 0.115 (0.140) 0.416** (0.205) 0.379* (0.203)
dBTime −0.098 (0.107) −0.098 (0.130) −0.139 (0.121)
age3044 0.148 (0.139) 0.061 (0.162) 0.018 (0.136)
*
educol 0.157 (0.191) 0.261 (0.218) 0.264 (0.151)
eth his −0.234 (0.204) −0.309 (0.224) −0.320 (0.218)
Coefficients specific to j = 3 in Eq. (8) Coefficients specific to j = 3 in Eq. (8)
reqrecws 0.115 (0.140) −0.290 (0.302) −0.325 (0.298)
dBTime −0.098 (0.107) 0.428** (0.208) 0.376* (0.202)
age3044 0.148 (0.139) 0.770*** (0.270) 0.735*** (0.258)
educol 0.157 (0.191) 1.382*** (0.415) 1.372*** (0.377)
eth his −0.234 (0.204) −0.839 ***
(0.301) −0.853 ***
(0.298)
Cutpoints
␶1 −5.941*** (0.492) −5.880*** (0.492) −5.882*** (0.339)
␶2 −4.267*** (0.480) −4.292*** (0.477) −4.292*** (0.324)
␶3 −2.370*** (0.474) −2.763*** (0.474) −2.762*** (0.322)
Log likelihood −2031.17 −2007.42 −2012.35
␹2 (dof)p-value 608.00 (32); p-value < 0.001 637.44 (42); p-value < 0.001 615.76 (27); p-value < 0.001
McFadden’s R2 0.149 0.159 0.157
AIC 4132.34 4104.85 4084.70
BIC 4328.29 4356.78 4252.65

Notes:
Results are based on N = 1995 observations. They were estimated using “gologit2” in Stata (Williams, 2006).
For j = 2 and 3 for the GOL models, only the coefficients shown differ from those in the first equation (j = 1); the coefficient estimates for the ordered logit model are repeated
to better contrast OL with GOL and cGOL results.
For the meaning of the variables above, see Table 4.
For McFadden’s R2 , the log likelihood of the model without predictors is treated as a total sum of squares, and the log likelihood of the full model is treated as the sum of
squared errors; the ratio of likelihoods suggests the level of improvement over the intercept model offered by the full model.
A test based on “Collin” in Stata suggests no multicollinearity problem.
The robust standard errors are the Huber–White–Sandwich estimates; they provide an insurance against mild departures from the homoskedasticity assumption.
An analysis of standardized residuals showed that observations with large residuals are not influential.
Specification tests using Stata’s linktest and likelihood ratio tests showed that models without powers and interactions are adequate here.
*
Significance at the 10% level respectively.
**
Significance at the 5% level respectively.
***
Significance at the 1% level respectively.
60 J.-D.M. Saphores et al. / Resources, Conservation and Recycling 60 (2012) 49–63

number of studies, including Boldero (1995), Cheung et al. (1999), magnitude. This suggests concerns for potential disameties asso-
Mannetti et al. (2004), and Tonglet et al. (2004) report that attitudes ciated with drop-off recycling centers (e.g., pollution, noise, or
toward recycling are significant predictors of recycling intentions, traffic). Knowing that e-waste contains toxic materials but that
although we know that people have a tendency to overstate their recycling is safe also boosts people’s willingness to recycle e-
actions when it comes to recycling (e.g., see Gamba and Oskamp, waste.
1994). Results are shown in Tables 6 and 7. Specification test results Other significant variables include gender (women are more
can be found in the notes below Table 6. willing to recycle as in Saphores et al., 2006; or Sidique et al., 2010);
being married; and prior e-waste recycling experience. In addition,
5.4.1. Contrasting results for ordered models both African Americans and Hispanic appear less willing to recycle
We first estimated simple ordered logit (OL) and ordered pro- e-waste, which confirms the findings of Johnson et al. (2004) for
bit models; they led to similar results so we kept the OL model. recycling.
Using the same explanatory variables, we then estimated gener- On the other hand, education level, age, and mandatory recy-
alized ordered models with 5 different link functions (see Section cling of conventional household wastes (paper, glass, metal) appear
4.2). We found that the models with logit and probit link functions to play a more minor role except for respondents who are less
gave very similar results and performed best from the point of the than very willing to recycle e-waste. We conjecture that educa-
AIC and BIC criteria (results are omitted for brevity), so our pre- tional attainment is an ineffective proxy for specific knowledge
ferred model is a generalized ordered logit (GOL) model. Finally, about the public health/environmental consequences of improp-
we estimated a concise generalized ordered logit model (denoted erly disposing of e-waste. Moreover, retirees may not have an edge
by cGOL) obtained by dropping GOL variables that were not statis- for recycling e-waste at drop-off centers since it is an infrequent
tically significant (p > 0.10) to simplify our discussion of variables activity that requires driving to an unpleasant location.
influencing the willingness to recycle e-waste. Table 6 presents Finally, we see that knowledge of e-waste laws (probably
results for these 3 models. because enforcement is lacking), familiarity with curbside recy-
Let us first focus on the first two models. In general, estimated cling for conventional household waste (possibly because recycling
coefficients for OL and for GOL are close (within one standard error), at drop-off centers is different; see Sidique et al., 2010), household
except for reqrecws, dBTime, age3044, educol, and eth his (see income (as in Saphores et al., 2006 or Scott, 1999), family size (see
Table 4 for the meaning of these variables). For these 5 variables Scott, 1999), and living in a rural area do not appear to matter. The
the parallel lines assumption does not hold and they are not sta- latter is at odds with Saphores et al., 2006, but their finding may be
tistically significant with the OL but they are with the GOL model due to respondents from Mono County, which has a history of envi-
based on Wald tests as explained in Section 4.2. Moreover, a like- ronmental activism. A couple of reasons may explain why wealthier
lihood ratio test comparing the GOL and the OL models rejects the households are not more willing to recycle e-waste: they are likely
null hypothesis that these extra parameters are not statistically sig- to have more space to store e-waste and the value of their time is
nificant (2 (10) = 47.5; p-value < 0.001). This justifies our modeling higher.
choices. Let us now look at average marginal probability effects. They
are obtained by averaging marginal effects over all respondents, so
5.4.2. Generalized ordered logit results they tell us about the impact of a policy that would uniformly affect
Let us now consider results from the concise GOL model (cGOL). all respondents. Let xki denote the kth variable for respondent i. If
Since the probability of a specific outcome depends on the value of xki is continuous, its marginal effect for outcome j ∈ {1, 2, 3, 4} is the
all explanatory variables, we rely on two different approaches to derivative of Pr(yi = j|xi ) with respect to xki and if xki is discrete, its
interpret our results (Long and Freese, 2006): we examine average marginal effect is Pr(yi = j|xki = 1) − Pr(yi = j|xki = 0); other variables
marginal effects and we assess the impact of the sign of the model are left unchanged. Results are presented in Table 7.
coefficients. Let us start with the latter. Right away, we note the large average marginal effect of the
From Eq. (7) for j = 1, simple algebra shows that increasing a normalized moral norm variable: a 0.01 increase in moral norm
positive explanatory variable with a positive coefficient augments raises the probability of being “very willing” to recycle e-waste
the probability of being very willing to recycle e-waste; and for j = 4 by 0.848% and decreases the probability of all other willingness
the reverse holds for the probability of being unwilling to recycle e- categories. This is approximately four times more than for general
waste. For j = 2 or 3, we simply note what variables are statistically environmental beliefs summarized by the New Ecological Paradigm
significant. (NEPn).
Several reasons stand out to explain willingness to recycle e- Variables related to conventional household waste recycling
waste at drop-off centers. The impact of several internal variables have a relatively modest average marginal effect. Increasing social
appears to be particularly strong based on the magnitude of their pressure by changing the perception that neighborhood recycling
coefficients. Personal/moral norms (captured by MORn) come first, is high raises the probability of being “very willing” by 0.061.
followed by environmental beliefs (summarized by NEPn), and However, making recycling compulsory at school or work has no
social expectations (via the variable nhrec; it equals 1 if neighbor- significant impact on the probabilities of being “very willing” or
hood household recycling rates are high). This suggests that people “not willing at all”; it simply shifts people from the “somewhat
are more willing to recycle if they feel a moral responsibility to do willing” to the “willing” categories. Likewise, changing the percep-
so, if they hold pro-environmental views, and if there is social pres- tion that recycling at home is not safe increases the probability of
sure to do so (as in Do Valle et al., 2004 for household recycling or being “very willing” by 0.059, while changing the perception that
in Sidique et al., 2010 for e-waste). recycling is time consuming just adds 0.039.
The statistical significance of external variables highlights the Opening e-waste drop-off recycling centers closer to people’s
importance of recycling convenience (as in Domina and Koch, 2002; homes has a positive (albeit small) impact on their willingness to
Jenkins et al., 2003; Saphores et al., 2006; or Sidique et al., 2010). It recycle e-waste: if it is within 5 miles, it increases the probability
is captured by several variables: having a drop-off recycling cen- of being “very willing” by 0.078. However, our respondents prefer
ter nearby; having access to a car; and finding time to recycle recycling centers that are slightly farther away, possibly because of
e-waste. Both coefficients indicating distance to the nearest drop- associated disamenities such as traffic and noise: having a drop-off
off recycling center are positive and significant (within 5 miles, recycling center between 6 and 10 miles increases the probability
and between 6 and 10 miles), but the latter has a slightly larger of being “very willing” by 0.089.
J.-D.M. Saphores et al. / Resources, Conservation and Recycling 60 (2012) 49–63 61

Table 7
Sample average marginal effects.

Variable Very willing Willing Somewhat willing Not willing at all

Normalized New Ecological Paradigm 0.208*** −0.049*** −0.102*** −0.057***


(0.051) (0.012) (0.025) (0.015)
Normalized moral norm 0.848*** −0.202*** −0.414*** −0.232***
(0.055) (0.019) (0.030) (0.022)
High level of recycling in your neighborhood? Yes = 1 0.061*** −0.017** −0.029*** −0.015***
(0.023) (0.007) (0.011) (0.005)
Required to recycle at work or at school? Yes = 1 −0.001 0.053* −0.070*** 0.018
(0.026) (0.028) (0.021) (0.017)
Storing recyclable materials at home is safe (agrees = 1) 0.059*** −0.016*** −0.028*** −0.014***
(0.017) (0.005) (0.008) (0.004)
Finding time to recycle is a problem (agrees = 1) −0.039** 0.019 0.037** −0.017**
(0.019) (0.019) (0.016) (0.008)
E-waste drop-off within 5 miles? Yes = 1 0.078*** −0.023*** −0.037*** −0.018***
(0.022) (0.008) (0.010) (0.005)
E-waste drop-off within 6–10 miles? Yes = 1 0.089*** −0.027** −0.041*** −0.021***
(0.031) (0.011) (0.014) (0.006)
Knows that e-waste contains toxic materials? Yes = 1 0.112*** −0.035*** −0.052*** −0.025***
(0.017) (0.007) (0.007) (0.003)
Has recycled e-waste at drop-off center? Yes = 1 0.118*** −0.038*** −0.054*** −0.026***
(0.017) (0.007) (0.008) (0.004)

Notes:
Sample average marginal effects, which are based on the concise generalized ordered logit model (last column in Table 6), were generated using the command “margeff”
in Stata (Bartus, 2005). Let xki denote the kth variable for respondent i. If xki is continuous, its marginal effect for outcome j ∈ {1, 2, 3, 4} is the derivative of Pr(yi = j|xi ) with
respect to xki and if xki is discrete, its marginal effect is Pr(yi = j|xki = 1) − Pr(yi = j|xki = 0), with all other variables unchanged.
A number in parenthesis after an average marginal effect is the corresponding standard error.
*
Significance at the 10% level respectively.
**
Significance at the 5% level respectively.
***
Significance at the 1% level respectively.

Two variables make more of a dent: knowing that e-waste con- Although only one-third of our respondents have recycled e-
tains toxic materials and having previously recycled e-waste at a waste before, three quarters of them are either willing or very
drop-off center. The former raises the probability of being “very willing to do so. Women and married people appear more moti-
willing” by 0.112 and the latter by 0.118, while decreasing the prob- vated while African Americans and Hispanic appear less willing
ability of the other levels of willingness. This suggest that educating to recycle e-waste. Knowing that e-waste contains toxic materials
people about the dangers of the improper disposal of e-waste and matters, and so does convenience, although people do not want to
getting them to start recycling e-waste can boost their participation live too close to a drop-off recycling center. Environmental beliefs
in e-waste recycling schemes. and social pressure to recycle in general also have an impact, but the
Overall, these results highlight the importance of internal vari- most important factor underlying willingness to recycle e-waste is
ables that capture moral norms and environmental beliefs for clearly moral norms.
explaining people’s willingness to recycle e-waste, and more gen- These results suggest several courses of action. First, more infor-
erally to engage in pro-environmental behaviors. mation should be provided about the potential toxicity of e-waste
and the potential public health impacts of its improper disposal,
6. Conclusions with a particular attention to African American and Hispanic com-
munities. Drop-off recycling should be expanded, either by creating
The goal of our research was to shed some light on households’ dedicated recycling centers or by concluding agreements with elec-
willingness to recycle e-waste in order to inform policies designed tronics retailers so they take back obsolete or broken electronic
to tackle the growing e-waste problem. Indeed, the prospect of a products. To finance this measure, advanced recycling fees could be
continued increase in the volume of e-waste as well as growing collected, as in California, for example (e.g., see Nixon et al., 2009).
concerns about the loss of rare metals in non-recycled electronic Second, environmental organizations, schools, and universities
products and about the fate of e-waste exported to developing should continue to educate people about the public health and
countries call for better collection and recycling programs. House- environmental impacts of dumping pollutants in the environment.
holds, who consume the bulk of electronic products, have a key role With the help of church groups, they should help reinforce moral
to play. and social norms that condition our environmental behavior. At the
In this context, our generalized ordered logit model illustrates same time, government should improve laws dealing with e-waste
the need to go beyond simple ordered models (such as ordered logit disposal, publicize existing laws and enforce them.
or probit) to better capture the complex impact of various charac- However, the gap between people who state that they are very
teristics on people’s willingness to engage in pro-environmental willing to recycle e-waste and the actual number of recyclers, as
behavior. well as past experience with other pollution problems suggest that
Moreover, our empirical results show that no clear socio- these measures are not going to be sufficient, at least not in the
economic characteristic defines people with e-waste recycling short run.
experience, except that current e-waste recyclers are slightly more In particular, economic instruments, and especially deposit-
likely to have larger families or to be over 60 years old. However, refund systems (where consumers pay a fee when they purchase
they know that e-waste contains potentially toxic materials, they a product and get it back when they return it at the end of its
have been exposed to conventional recycling either at work or at useful life) should be given serious consideration. Indeed, a num-
school, and they are motivated by moral norms. They also need to ber of authors, including Fullerton and Kinnaman (1995) or Palmer
know where the nearest drop-off recycling center is located. et al. (1997) for municipal waste and Sigman (1995) for automobile
62 J.-D.M. Saphores et al. / Resources, Conservation and Recycling 60 (2012) 49–63

battery recycling, have shown the superiority of well designed Darby L, Obara L. Household recycling behaviour and attitudes towards the dis-
deposit refund-system over other economic instruments when the posal of small electrical and electronic equipment. Resources, Conservation and
Recycling 2005;44(1):17–35.
cost of illegal disposal is taken into account. Shinkuma (2003) con- DeYoung R. Recycling as appropriate behavior: a review of survey data from selected
curred when transaction costs are low and Kahhat et al. (2008) recycling education programs in Michigan. Resources, Conservation and Recy-
suggested implementing a deposit-refund system for e-waste after cling 1990;3:1–13.
Derksen L, Gartrell J. The social context of recycling. American Sociological Review
installing radio-frequency identification device (RFID) in electronic 1993;58(3):434–42.
products to lower transaction costs. Domina T, Koch K. Convenience and frequency of recycling: implications for
A second option, which could be used concurrently with the including textiles in curbside recycling programs. Environment and Behavior
2002;34:216–38.
previous one, is to “green” electronics by restricting the use of
Do Valle PO, Reis E, Menezes J, Rebelo E. Behavioral determinants of household
toxic materials in electronics products, with a focus on heavy met- recycling participation. Environment and Behavior 2004;36(4):505–40.
als like lead or mercury. Although this approach does not deal Dunlap RE, Van Liere KD. The new environmental paradigm. Journal of Environmen-
tal Education 1978;9:10–9.
with the recycling of valuable materials in e-waste, it addresses
Dunlap RE, Van Liere KD, Mertig AD, Jones RE. Measuring endorsement of
the public health and environmental risks of improperly process- the New Ecological Paradigm: a revised NEP scale. Journal of Social Issues
ing e-waste. It is attractive because of the very rapid rate of 2000;56(3):425–42.
innovation in electronics, and it has been used by the European Eurostat. Waste Electrical and Electronic Equipment (WEEE),
<http://appsso.eurostat.ec.europa.eu/nui/submitViewTableAction.do>; 2011
Union in its directive on the Restriction of Hazardous Substances [accessed on 12.09.11].
(www.rohs.eu/english/index.html). Results from the removal of Ewing G. Altruistic, egoistic, and normative effects on curbside recycling. Environ-
lead from solder are encouraging although replacements are not ment and Behavior 2001;33(6):733–64.
Folz D, Hazlett J. Public participation and recycling performance: explaining program
environmentally benign (Henshall et al., 2011). success. Public Administration Review 1991;51(6):526–32.
Future research could further explore obstacles to e-waste recy- Fornara F, Carrus G, Passafaro P, Bonnes M. Distinguishing the sources of nor-
cling among minority groups; compare the effectiveness of various mative influence on proenvironmental behaviors: the role of local norms in
household waste recycling. Group Processes & Intergroup Relations 2011;14(5):
e-waste recycling options; and assess the feasibility of economic 623–35.
incentives such as deposit refund programs versus regulations that Fullerton D, Kinnaman T. Garbage, recycling and illegal burning or dumping. Journal
aim at greening electronics. of Environmental Economics and Management 1995;29:78–91.
Gamba RJ, Oskamp S. Factors influencing community residents’ participation
in commingled curbside recycling programs. Environment and Behavior
Acknowledgements 1994;26(5):587–612.
Granzin KL, Olsen JE. Characterizing participants in activities protecting the envi-
ronment: a focus on donating, recycling, and conserving behaviors. Journal of
This research was supported in part by the Program in Indus-
Public Policy & Marketing 1991;10(2):1–27.
trial Ecology at UC-Irvine (UCI). This Program was supported by Greene WH. Econometric analysis. 6th ed. Prentice Hall; 2008.
funds from the National Science Foundation (DMII 0223894) and Greenpeace. Where does e-waste end up? <www.greenpeace.org/international/
en/campaigns/toxics/electronics/the-e-waste-problem/where-does-e-waste-
the University of California Toxic Substances Research and Teach-
end-up/>; 2009 [accessed on 12.09.11].
ing Program (UC 44157); their support is gratefully acknowledged. Guagnano G, Stern P, Dietz T. Influences on the attitude–behavior relationships:
Finally, we are grateful to the Editor and three referees for very a natural experiment with curbside recycling. Environment and Behavior
helpful comments. 1995;27(5):699–718.
Hage O, Söderholm P, Berglund C. Norms and economic motivation in household
recycling: empirical evidence from Sweden. Resources, Conservation and Recy-
References cling 2009;53:155–65.
Hansmann R, Bernasconi P, Smieszek T, Loukopoulos P, Scholz RW. Justifications
Ajzen I. Attitudes, personality and behaviour. Milton Keynes: Open University Press; and self-organization as determinants of recycling behavior: the case of used
1988. batteries. Resources, Conservation and Recycling 2006;47:133–59.
Ajzen I. The theory of planned behaviour. Organizational Behaviour and Human Harbaugh WT. What do donations buy? A model of philanthropy based on prestige
Decision Processes 1991;50:179–211. and warm glow. Journal of Public Economics 1998;67:269–84.
Ajzen I, Fishbein M. Attitude–behavior relations: a theoretical analysis and review Hawcroft LJ, Milfont TL. The use (and abuse) of the new environmental paradigm
of empirical research. Psychological Bulletin 1977;84:888–918. scale over the last 30 years: a meta-analysis. Journal of Environmental Psychol-
Ajzen I, Madden TJ. Prediction of goal-directed behaviour: attitudes, intentions, ogy 2010;30:143–58.
and perceived behavioural control. Journal of Experimental Social Psychology Henshall G, Bath J, Handwerker CA. Lead-free solder process development. Wiley-
1986;22:453–74. IEEE Press; 2011.
Andreoni J. Impure altruism and donations to public goods: a theory of warm-glow Hopper JR, Nielsen JM. Recycling as altruistic behavior normative and behavioral
giving. The Economic Journal 1990;100:464–77. strategies to expand participation in a community recycling program. Environ-
Bamberg S. How does environmental concern influence specific environmentally ment and Behavior 1991;23:195–220.
related behaviors? A new answer to an old question. Journal of Environmental Hornik J, Cherian J, Madansky M, Narayana C. Determinants of recycling behavior: a
Psychology 2003;23(1):21–32. synthesis of research results. Journal of Socio-Economics 1995;24(1):105–27.
Bartus T. Estimation of marginal effects using margeff. The Stata Journal Jang Y-C, Kim M. Management of used & end-of-life mobile phones in Korea: a
2005;5(3):309–29. review. Resources, Conservation and Recycling 2010;55:11–9.
Basel Action Network. The e-waste crisis introduction, <http://e-stewards.org/the- Jenkins RR, Martinez SA, Palmer K, Podolsky MJ. The determinants of house-
e-waste-crisis/>; 2010 [accessed on 12/09.11]. hold recycling: a material-specific analysis of recycling program features and
Birnbaum LS, Staskal DF. Brominated flame retardants: cause for concern? Environ- unit pricing. Journal of Environmental Economics and Management 2003;45:
mental Health Perspectives 2004;112(1):9–17. 294–318.
Boldero J. The prediction of household recycling of newspaper: the role of atti- Johnson CY, Bowker JM, Cordell HK. Ethnic variation in environmental belief and
tudes, intentions and situational factors. Journal of Applied Social Psychology behavior: an examination of the New Ecological Paradigm in a social psycholog-
1995;25:440–62. ical context. Environment and Behavior 2004;36(2):157–86.
Bouvier R, Wagner T. The influence of collection facility attributes on household col- Kahhat R, Kim J, Xu M, Allenby B, Williams E, Zhang P. Exploring e-waste man-
lection rates of electronic waste: the case of televisions and computer monitors. agement systems in the United States. Resource, Conservation and Recycling
Resources, Conservation and Recycling 2011;55:1051–9. 2008;52:955–64.
Bratt C. The impact of norms and assumed consequences on recycling behavior. Katzev RD, Mishima HR. The use of posted feedback to promote recycling. Psycho-
Environment and Behavior 1999;31(5):630–56. logical Reports 1992;71:259–64.
Brekke KA, Kverndokk S, Nyborg K. An economic model of moral motivation. Journal Lansana F. Distinguishing potential recyclers from non-recyclers: a basis for
of Public Economics 2003;87:1967–83. developing recycling strategies. Journal of Environmental Education 1992;23:
Cheung S, Chan D, Wong Z. Reexamining the theory of planned behavior in under- 16–23.
standing wastepaper recycling. Environment and Behavior 1999;31:587–612. Long JS, Freese J. Regression models for categorical dependent variables using Stata.
Clark CF, Kotchen MJ, Moore MR. Internal and external influences on pro- 2nd ed. College Station: Stata Press; 2006.
environmental behavior: participation in a green electricity program. Journal Mannetti L, Pierro A, Livi S. Recycling: planned and self-expressive behavior. Journal
of Environmental Psychology 2003;23:237–46. of Environmental Psychology 2004;24:227–36.
Corral-Verdugo V. Dual realities of conservation behavior: self-reports vs. obser- Margai F. Analyzing changes in waste reduction behavior in a low-income urban
vations of re-use and recycling behavior. Journal of Environmental Psychology community following a public outreach program. Environment and Behavior
1997;17:135–45. 1997;29:769–92.
J.-D.M. Saphores et al. / Resources, Conservation and Recycling 60 (2012) 49–63 63

McCarty JA, Shrum LJ. The recycling of solid wastes: personal values, value orienta- Sigman H. A comparison of public policies for lead recycling. Rand Journal of Eco-
tions, and attitudes about recycling as antecedents of recycling behavior. Journal nomics 1995;26(3):452–78.
of Business Research 1994;30:53–62. Snyder, KC, Kristel, OV, Dhmammarungruang, B, Sang, S,. Report to the Ohio
McQuaid RW, Murdoch A. Recycling policy in areas of low income and Environmental Protection Agency: drop-off recycling—understanding par-
multi-storey housing. Journal of Environmental Planning and Management ticipation and determining an empirically based access credit model,
1996;39(4):547–64. <www.epa.gov/reg5rcra/wptdiv/solidwaste/ohaccesscreditreport.pdf>; 2004
Meneses GD, Palacio AB. Recycling behavior: a multidimensional approach. Envi- [accessed on 15.11.10].
ronment and Behavior 2005;37:837–60. Speirs D, Tucker P. A profile of recyclers making special trips to recycle. Journal of
Nixon H, Saphores J-D, Ogunseitan O, Shapiro AA. Understanding preferences Environmental Management 2001;62:201–20.
for recycling electronic waste in California—the influence of environmen- Stern PC, Dietz T, Abel T, Guagnano GA, Kalof L. A value-belief-norm theory of support
tal attitudes and beliefs on willingness to pay. Environment and Behavior for social movements: the case of environmentalism. Human Ecology Review
2009;41(1):101–24. 1999;6:81–97.
Nnorom IC, Ohakwe J, Osibanjo O. Survey of willingness of residents to participate Sykes GM, Matza D. Techniques of neutralization: a theory of delinquency. American
in electronic waste recycling in Nigeria: a case study of mobile phone recycling. Sociological Review 1957;22(6):664–70.
Journal of Cleaner Production 2009;17(18):1629–37. Taylor S, Todd P. An integrated model of waste management behaviour: a test
Nyborg K, Howarth RB, Brekke KA. Green consumers and public policy: on socially of household recycling and composting intentions. Environment and Behavior
contingent moral motivation. Resource and Energy Economics 2006;28:351–66. 1995;27(5):603–30.
Oskamp S, Harrington MJ, Edwards TC, Sherwood DL, Okuda SM, Swanson DC. Thøgersen J. A model of recycling behaviour: with evidence from Danish
Factors influencing household recycling behavior. Environment and Behavior source separation programmes. International Journal of Research in Marketing
1991;23(4):494–519. 1994;11(1):145–63.
Owen AL, Videras JR. Culture and public goods: the case of religion and the voluntary Thompson B. Exploratory and confirmatory factor analysis. Washington, DC: Amer-
provision of environmental quality. Journal of Environmental Economics and ican Psychological Association; 2004.
Management 2007;54:162–80. Tiller KH, Jakus PM, Park WM. Household willingness to pay for drop-off recycling.
Owens J, Dickerson S, Macintosh D. Demographic covariates of residential recycling Journal of Agricultural and Resource Economics 1997;22:310–20.
efficiency. Environment and Behavior 2000;32(5):637–50. Tonglet M, Phillips PS, Read AD. Using the theory of planned behaviour to investi-
Palmer K, Sigman H, Walls M. The cost of reducing solid waste. Journal of Environ- gate the determinants of recycling behaviour: a case study from Brixworth, UK.
mental Economics and Management 1997;33:128–50. Resources Conservation and Recycling 2004;41:191–214.
Perry GDR, Williams LD. The participation of ethnic minorities in kerbside recycling: Turaga RMR, Howarth RB, Borsuk ME. Pro-environmental behavior—rational choice
a case study. Resources, Conservation and Recycling 2007;49(3):308–23. meets moral motivation. Annals of the New York Academy of Sciences 2010;Eco-
Piliavin JA, Charng H. Altruism: a review of recent theory and research. Annual logical Economics Reviews, ISSN 0077-8923.
Review of Sociology 1990;16:27–65. United Nations Environmental Program (UNEP). Recycling – from
Rege M. Social norms and private provision of public goods. Journal of Public Eco- e-waste to resources. <www.unep.org/PDF/PressReleases/E-
nomic Theory 2004;6:65–77. Waste publication screen FINALVERSION-sml.pdf>; 2009 [accessed on
Saphores J-D, Nixon H, Ogunseitan OA, Shapiro AA. Household willingness to recy- 13.11.10].
cle electronic waste: an application to California. Environment and Behavior U.S. EPA. Electronic waste management in the United States, Approach
2006;38:183–208. 1, Table 3.1. EPA530-R-08-009, <http://www.epa.gov/osw/conserve/
Saphores JM, Nixon H, Ogunseitan OA, Shapiro AA. How much e-waste is there in US materials/ecycling/docs/app-1.pdf>; 2008 [accessed on 09.12.11].
basements and attics? Results from a national survey. Journal of Environmental U.S. EPA. Municipal solid waste generation, recycling, and disposal in the
Management 2009;90:3322–31. United States: facts and figures for 2008. Office of Solid Waste. Report
Sansone C, Weir C, Harpster L, Morgan C. Once a boring task always a boring task? accessed 13.11.10 from <http://www.epa.gov/epawaste/nonhaz/municipal/
Interest as a self-regulatory mechanism. Journal of Personality and Social Psy- pubs/msw2008data.pdf>; 2009.
chology 1992;63:379–90. U.S. EPA. Lead in paint, dust, and soil. <www.epa.gov/lead/pubs/leadinfo.htm>;
Schultz PW, Oskamp S, Mainieri T. Who recycles and when? A review of per- 2010a [accessed on 21.12.10].
sonal and situational factors. Journal of Environmental Psychology 1995;15: U.S. EPA. Mercury. <www.epa.gov/hg/effects.htm>; 2010b [accessed on 21.12.10].
105–21. United States Geological Survey (USGS). Mineral Resources Program website,
Schwartz SH. Moral decision making and behavior. In: Macauley J, Berkowitz L, <http://minerals.usgs.gov/>; 2010 [accessed on 21.12.10].
editors. Altruism and helping behavior. NY: Academic Press; 1970. p. 127–41. Vining J, Ebreo A. What makes a recycler? A comparison of recyclers and nonrecy-
Schwartz SH. Normative explanations of helping behavior: a critique, proposal, and clers. Environment and Behavior 1990;22:55–73.
empirical test. Journal of Experimental Social Psychology 1973;9:349–64. Vining J, Ebreo A. Predicting recycling behavior from global and specific environmen-
Schwartz SH. Normative influences on altruism. Advances in Experimental Social tal attitudes and changes in recycling opportunities. Journal of Applied Social
Psychology 1977;10:221–79. Psychology 1992;22(20):1580–607.
Schwartz SH. Are there universal aspects in the structure and contents of human Wang Z, Zhang B, Yin J, Zhang X. Willingness and behavior towards e-waste
values? Journal of Social Issues 1994;50:19–45. recycling for residents in Beijing city, China. Journal of Cleaner Production
Scott D. Equal opportunity, unequal results: determinants of household recycling 2011;19:977–84.
intensity. Environment and Behavior 1999;31(2):267–90. Werner CM, Turner J, Shipman K, Twitchell FS, Dickson BR, Bruschke GV, et al.
Sepulveda A, Schluep M, Renaud FG, Streicher M, Kuehr R, Hageluken C, et al. A Commitment, behavior, and attitude change: an analysis of voluntary recycling.
review of the environmental fate and effects of hazardous substances released Journal of Environmental Psychology 1995;15:197–208.
from electrical and electronic equipments during recycling: examples from Werner C, Makela E. Motivations and behaviors that support recycling. Journal of
China and India. Environmental Impact Assessment Review 2010;30:28–41. Environmental Psychology 1998;18(4):373–86.
Shinkuma T. On the second-best policy of household’s waste recycling. Environmen- Williams R. Generalized ordered logit /partial proportional odds models for ordinal
tal and Resource Economics 2003;24:77–95. dependent variables. The Stata Journal 2006;6(1):58–82.
Sidique SF, Lupi F, Joshi SV. The effects of behavior and attitudes on drop-off recycling Woodard R, Bench M, Harder MK. The development of a UK kerbside scheme using
activities. Resources, Conservation and Recycling 2010;54:163–70. known practice. Journal of Environmental Management 2005;75:115–27.

Вам также может понравиться