Professional Documents
Culture Documents
research-article2018
WMR0010.1177/0734242X18793939Waste Management & ResearchFukuda et al.
Original Article
Abstract
Indonesia is one of the countries that is facing serious solid waste problems. Although the issue has been addressed in the capital city
of Jakarta, many secondary cities have been left behind. This research attempts to analyse the impact of relevant waste management
policy components on respondents’ policy acceptance probabilities in Yogyakarta Province. Using a randomised conjoint field
experiment based on the proposal of modern waste collection services, this article shows that a new waste management policy that
includes waste separation with frequent organic waste collection will gain popular support, especially with the rich. Indeed, if the
government introduces waste collection and processing services that consist of the most preferred levels of these attributes, including
five-times-a-week organic waste collection, a majority of the residents will support this new waste management policy, which will
reduce waste at its source.
Keywords
Waste management policy, solid waste, waste separation, organic waste, randomised conjoint experiment, Indonesia
Received 18th May 2018, accepted 19th July 2018 by Associate Editor Alberto Bezama.
of Yogyakarta Province, 2015). Only 17% of the total waste in people want are frequent waste collection in the evening using
Yogyakarta is managed by the government, and most of the plastic bag containers without instruction.
waste is mismanaged by the community. Inappropriate waste Hazra et al. (2015) included a list of attributes in their random
management causes various environmental problems, such as parameter logit models. They found that education and income
water, soil and air pollution, resulting in serious health prob- are positively related with willingness to pay (WTP) for waste
lems. MSW disposed of in landfills can produce methane gas, collection services, meaning that highly educated and high-
which has been identified as a cause of increasing global warm- income people tend to have higher WTPs than less educated peo-
ing. According to the US Environmental Agency’s Global ple with lower incomes. To understand household preferences for
report, in 2010, Indonesia occupied seventh place on the list of waste management, some researchers, including Sakata (2007),
global methane emissions sources. Nakatani et al. (2008), and Sheau-Ting et al. (2016), used con-
Zhuang et al. (2008) stated that waste sorting at the source is joint analysis in which respondents had the freedom to determine
an effective policy for waste reduction. Germany, the United what type of waste management they prefer. Nakatani et al.
States and Japan have successfully implemented waste separa- (2008) discovered that people would like to support waste sepa-
tion as part of their waste management systems. In Japan, the ration as part of the waste management policy in Kawasaki City.
total MSW was reduced by 69% by implementing a waste sepa- Sheau-Ting et al. (2016) stated that the distance to the nearest
ration system (Ministry of Environment of Japan, 2014). In recycling bin must be considered when encouraging people to
Indonesia, the waste management is based on a simple collect– separate waste.
transport–dispose method, resulting in low participation in reduc- These previous studies utilised methods, such as conventional
ing waste volume at the source. In facing this situation, the conjoint analysis and random utility models, to capture personal
Indonesian government enacted Law No. 18 in 2008 and preferences for waste management. Random utility models for
Government Regulation No. 81 in 2012, which aim ‘3R’ (reduc- example, require parametric assumptions on the preference dis-
tion, reusing and recycling) by promoting waste separation at its tribution, such that the preference for observable characteristics
source, along with an intermediate waste processing and a proper is additive separable and the preference for unobservable charac-
final disposal treatment. However, Statistics of Indonesia 2017 teristics follows the type I extreme value distribution (see, for
indicate that the percentage of households that do not separate example, Small and Rosen, 1981). Rather than using approaches
their waste is still 81%. that rely on more restrictive assumptions, this study uses a ran-
Identifying household preference toward waste management domised conjoint experiment based on Hainmueller et al. (2013).
service components is an important consideration when success- Hainmueller et al. (2013) revised the conventional conjoint anal-
fully introducing waste separation policies (Oosterveer and ysis in the framework of a randomised experimental design to
Spaargaren, 2010). We thus conduct a field experiment to infer estimate the causal effects of policy components on respondent’s
households’ attitude towards modern waste management systems choice probabilities without bias arising from misspecification,
by gathering stated household preferences for hypothetical waste extending the findings from the conventional methodological
collection services that focus on waste separation at the source. literature.
Some research has been conducted to examine personal pref- In this new conjoint analysis, the attribute levels for each
erences for waste management in different areas. Caplan et al. alternative are randomly assigned when constructing the choice
(2002) conducted research on waste management preferences in sets, which allows us to identify the causal effects of each attrib-
Ogden, UT, which faced waste generation problems owing to the ute on the decision non-parametrically. This article thus estimates
closing of its landfill and rapid population growth. In contrast to the causal effects of relevant waste management policy aspects,
the current article, which utilises a randomised conjoint field such as collection frequency and waste processing methods, on
experiment, they used the contingent ranking method to deter- personal acceptance of the new waste separation policy.
mine personal preferences for waste management. Caplan et al. The next section discusses the data, the design of the conjoint
(2002) considered different levels of waste service costs and experiment and the estimation strategy. This is followed by the
waste separations. For waste separation, they provided options of the results and findings, before conclusions.
no separation, green waste and recyclable waste. They also pro-
vided options for different levels of service costs, which included
Survey design, data profiles, and
the current and higher service costs. They found that people in
Ogden supported the new waste management system and stated
estimation strategy
that more than half of the respondents were willing to separate Survey area and data
their waste as garbage and green waste. Yuan and Yabe (2014b)
Survey is conducted in Sleman Regency in the northern part of
conducted another study on the household preferences of kitchen
Yogyakarta Province where population and economy are rapidly
waste source separation in Beijing, China, via a contingent valu-
growing. Population growth has been 1.41% over the past 5 years,
ation method. The research showed that compensation is not a
and in 2015 the economic growth rate was 5.31%, which was the
major factor in residents accepting the proposed separation ser-
highest among all the districts in Yogyakarta Province. The main
vices; rather, the most important service characteristics that
Fukuda et al. 843
survey of this research was conducted from 27 February to 12 stated they do not have time for waste separation without plastic
March 2017 and includes 300 households as respondents. bags for sorting waste being available. The respondents were also
A household questionnaire was used to capture the character- asked about the importance of waste separation. More than half
istics and behaviour concerning the waste management of the of the respondents agree that waste separation will reduce envi-
households. The questionnaire included (1) basic information, ronmental pollution and negative health effects, and nearly 34%
such as age, gender, education, occupation, family size and believe that recyclable waste can be sold to earn money.
monthly income; and (2) household waste management behav-
iour, such as waste separation behaviour, and satisfaction with
Experimental design
the current waste management. The characteristics of the house-
holds are shown in Table 1, while the household behaviour shows As mentioned above, we use a randomised conjoint experiment
that most of the respondents do not separate waste at their home; analysis based on Hainmueller et al. (2013) to assess households’
just 26% separate their waste into organic and inorganic waste. acceptance of new waste management policies. In this experiment,
According to the survey, more than 40% of respondents stated we generate hypothetical waste collection services that aim to
that mixed transport and disposal facilities that do not support improve waste management services specifically in the research
waste separation discourage their willingness to separate waste. area. In each choice set, we provide three alternatives, including
In this area, organic and inorganic waste are still mixed in the two proposed policies and the status quo, and respondents are asked
temporary waste shelter. Furthermore, there is no intermediate to rank each of these three alternatives based on their preference.
processing facility to process both organic and inorganic waste We first show the scenario to each respondent so that they under-
separately. Another potential reason for the respondents to not stand the research aim and design and their role in it. The following
separate waste is a lack of time; a quarter of the respondents scenario is used to explain our experiment to the respondents:
844 Waste Management & Research 36(9)
We will ask your preferences on hypothetical household waste three levels. The fourth attribute is the intermediate processing
collection and disposal service methods. You will be requested to facilities (intermediate), which is given at four levels, namely
rank three alternatives: hypothetical policy A, hypothetical policy none available, composting, recycling and incineration.
B, and neither (or, current status without new policy) as choice C.
Among the three options, the rankings are 1) the most preferable
Intermediate processing facilities are used to process both organic
choice, 2) the second preferable choice and 3) the least preferable and inorganic waste to support the waste separation system.
choice. The common assumptions of household waste collection Composting facilities offer an intermediate processing to process
and disposal methods are 1) each household has the responsibility organic waste into compost. Recycling facilities process recycla-
to perform waste separation at home, 2) each household has to ble inorganic waste. Incineration facilities handle both organic
bring the separated waste to the nearest collection station, 3)
and inorganic waste by burning.
payment per month per household is the only fee that the you will
pay. You will be asked to make rankings four times with different The fifth attribute is the final disposal facilities (final), with
combinations of household waste collection and disposal methods two levels, to accommodate the waste after being processed by
with different conditions. the intermediate processing facilities, if any. The first level is
open dumping, which is a common final disposal method in
Then, we give the respondent the questionnaire including four Indonesia. where there is no treatment to minimise the environ-
choice sets. The proposed policy programme has six attributes mental pollution from the waste. The second level is a sanitary
for which the values (or ‘levels’) are randomly assigned. These landfill, which is a method of waste disposal where the waste is
attributes are organic waste collection frequency, inorganic waste buried either underground or in large piles where waste is con-
collection frequency, walking time to the nearest waste station, trolled and monitored closely.
intermediate processing facility, final processing facility and The last attribute is the monthly payment for waste collection
monthly payment amount. The detailed attributes and their services (payment), which is set at four levels. These levels
respective levels are shown in Table 2. include free, 24,000 rupiahs, 48,000 rupiahs and 96,000 rupiahs
The first and the second attributes are the frequency of the (or approximately, 2, 4, and 8 US dollars, respectively.) Currently,
organic and inorganic waste collection services (labelled as organic payment is zero in the areas that have no waste collection ser-
collection and inorganic collection, respectively), each with three vices, while 24,000 rupiahs is a standard monthly payment for
levels. The government provides the waste collection service on the waste collection services in the research area. Each respond-
the designated days of week, and in principle, residents are allowed ent is required to make four ranking decisions for four different
to bring their organic waste to the temporary waste shelters only on choice sets. Each of the two alternatives (namely, alternatives A
those days when collection service is provided. and B) in each choice set consists a unit of analysis and the data
The third attribute is the walking time to the nearest tempo- set includes 2400 observations with for 300 respondents (i.e.
rary waste shelter in the respondent’s area (walking time), also at 2 × 4 × 300 = 2400).
Fukuda et al. 845
Estimation strategy model whose result is free from the parametric assumptions on
the utility function reduces potential bias owing to misspecifica-
To estimate the causal effects of the policy attributes on the proba- tion. Following Bhattacharya (2015), who provided a point esti-
bility that each proposed policy alternatives is preferable to the sta- mate of WTP, the present article thus estimates the lower bound
tus quo, Hainmueller et al. (2013) proposed the following model: of WTP by using the non-parametric framework via direct appli-
cations of the randomized conjoint experiment methodolgy by
L Dl
Hainmueller et al., 2013 (HHY) conjoint that only requires non-
yitj = β 0 + ∑∑β
l =1 d = 2
ld α itjld + uitj
restrictive behavioural assumptions.
where aitjld is a dummy variable for the lth level of attribute l of Results and findings
policy j in task t of respondent i, L is the number of attributes, Dl
is the number of levels of attribute l, β ld is its coefficient to be
Average component effects of policy
estimated and uitj denotes the error term. (Clustered-robust stand- attributes
ard errors at the respondent level are used in all the regressions.) The results of the choice probability estimation are summarised
Then, yitj is a choice indicator variable for the estimation of the in Figure 1. As can be seen in Figure 1, an organic waste collec-
choice probability; yitj = 1 if the preference rank of policy j is tion frequency of five times a week has a significant impact on
higher than the status quo. the peoples’ acceptance of the proposed policy, unlike three times
a week. This implies that respondents prioritise the frequency of
Welfare analysis by partial identification organic waste collection services over the other attributes, and
that five-times-a-week collection increases the support of the
of WTP
new waste management policy by 5.5%.
As we use independent randomised attribute levels in each choice As for the inorganic waste attribute, in contrast, collection fre-
set, the estimation strategy above provides unbiased and consist- quency does not affect the policy acceptance probability. These
ent average component effect of each of the policy attributes results imply that residents are more concerned with the collec-
indicating its influence on the respondents’ policy acceptance tion frequency of organic waste, not that of inorganic waste.
probability. The above randomised conjoint experiment designed Notably the other attributes, namely, walking time, intermediate,
by Hainmueller et al. (2013) allows us to partially identify the and final, do not have statistically significant effects, meaning
distribution of the WTP and to compute the lower bound of the that they do not have any effect on the probability that the
average WTP for accepting the proposed policy non-parametri- respondents will support the new waste collection services.
cally. (See Appendix for the technical details of the computation In terms of the payment attribute, the lowest payment, that is
of the willingness to pay and its lower bound in our context.) no monthly payment, provides the biggest impact in terms of
WTP is a monetary amount that an individual or a household increasing the probability that a respondent will support a new
agrees to pay in return for the receipt of the benefit that derives from waste management policy relative to other monthly payments.
a proposed policy change or implementation. WTP is a standard No monthly payments increase the respondent choice probability
measurement of social benefit in those fields, including economics by 28%, followed by that of 24,000 rupiahs (approximately 2 US
and policy analysis, allowing to quantitatively assess the welfare dollars) with approximately 18%, and 48,000 rupiahs (approxi-
impact by using monetary measurement. This yields a cost–benefit mately 4 US dollars) with approximately 8%, all of which are
ratio that justifies the policy implementation, if it exceeds unity. relative to the baseline of 96,000 rupiahs (the highest monthly
There are methods allowing the WTP estimation in the stated- payment of approximately 8 US dollars).
preference literature such as contingent valuation method (CVM) The constant term is 17%, meaning that the percentage of peo-
and discrete choice experiments (DCEs). (See, for instance, ple who will support the new waste collection service at the bench-
Alberini and Cooper, 2000; Arrow et, al., 1993; Johnston et al., mark (or the worst set of attributes) is only 17%. However, the
2017.) CVM presents common attribute levels to all respondents implementation of the most preferred waste management policy
in its typical survey design, which makes it difficult to accurately will gain majority support for waste separation. If the government
estimate preferences over multi-attribute policy alternatives. introduces a waste collection service that consists of the most pre-
DCEs has been a popular alternative especially during the last ferred levels of the attributes, including a five-times-a-week
two decades, though it is based on a specific decision model organic waste collection service without monthly payment, then
assuming a parametric random utility, and thus requires untesta- the policy acceptance rate will increase to approximately 59%.
ble assumptions: The additive separability of utility functions
and error terms that follow the type-I extreme-value distribution.
Interaction effects
(See McFadden, 1974, for his original work.)
In the literature, WTP is typically estimated based on the para- Acceptance of the new waste management policy depends on
metric discrete choice model. However, the non-parametric their characteristics. Figure 2 shows that males do not like to
approach of WTP estimation based on a non-parametric choice walk too long to a waste station, whereas female respondents
846 Waste Management & Research 36(9)
Figure 2. Interaction of causal effects on the policy acceptance probability with gender.
Notes: Male is the benchmark and constitutes 52% of the entire sample, while female constitutes the remaining 48%.
care more about the cost of the services. When the waste col- those with master or PhD degrees are also significantly nega-
lection is free, the probability of policy acceptance of the tive, however, this result may be affected by their small sam-
female respondents increases by approximately 15% more ple size. In turn, respondents with higher monthly income
than that of male respondents. Figure 3 shows that retired peo- (above 7.5 million rupiahs, approximately 500 US dollars)
ple and those who rent their houses have a negative tendency have a strong tendency to accept the policy compared with
to accept the new waste separation policy. The results for those in other income classes.
Fukuda et al. 847
Hazra T, Goel S and Maitra B (2015) Willingness-to-pay and preference het- people whose WTP to accept the policy of which all attributes’
erogeneity for service attributes of solid waste management: An experi-
values (except the payment attribute) are given by a vector a is
ence in Kolkata, India. Global NEST Journal 17: 82–92.
Johnston RJ, Boyle KJ, Adamowicz W, et al. (2017) Contemporary guidance equal to X, and let F(X|a) be its cumulative density such that:
for stated preference studies. Journal of the Association of Environmental
and Resource Economists 4: 319–405. X
Local Government of Yogyakarta Province (2015) Report of Accountability
Description of Yogyakarta Province. Yogyakarta: Local Government of
F ( X |a ) =
−∞
∫ f ( x|a ) dx.
Yogyakarta Province.
McFadden D (1974) The measurement of urban travel demand. Journal of
Public Economics 3: 303–328. Note here that if we denote by q(X,a) the choice probability of a
Ministry of Economic Affairs (2015) Study of National Policies and Strategies
policy with payment X and other attributes a against the status
to Accelerate Waste Management. Jakarta: Ministry of Economic Affairs.
Ministry of Environment of Japan (2014) History and Current State of Waste quo (i.e. no policy), then we immediately have:
Management in Japan. Tokyo: Ministry of Environment of Japan.
Nakatani J, Aramaki T and Hanaki K (2008) Evaluating source separation of F ( X |a ) = 1 − q ( X , a )
plastic waste using conjoint analysis. Waste Management 28: 2393–2402.
Oosterveer P and Spaargaren G (2010) Meeting social challenges in develop-
ing sustainable environmental infrastructures in East African cities. Social The marginal distribution of the WTP is then obtained simply
Perspectives on the Sanitation Challenge. Dordrecht: Springer, pp.11–30. by summing the above over the attribute space:
Pokhrel D and Viraraghavan T (2005) Municipal solid waste management in
Nepal: Practices and challenges. Waste Management 25: 555–562.
Rathi S (2007) Optimization model for integrated municipal solid waste man-
F(X ) = ∑F ( X |a ) p ( a ).
a
agement in Mumbai, India. Environment and Development Economics
12: 105–121. where p(a) is the relative frequency of attribute combination a
Sakata Y (2007) A choice experiment of the residential preference of waste given X. Finally, let F-1(F) be the inverse of F(X). Then, the aver-
management services – The example of Kagoshima city, Japan. Waste 1
Management 27: 639–644.
∫F (φ ) dφ . (Applying integra-
−1
age WTP is simply obtained as
Sheau-Ting L, Sin-Yee T and Weng-Wai C (2016) Preferred attributes of
0
waste separation behaviour: An empirical study. Procedia Engineering
145: 738–745. tion by parts and by substitution to the definition of the average
Small KA and Rosen HS (1981) Applied welfare economics with discrete X
choice models. Econometrica 49: 105–130.
Statistics of Indonesia (2017) www.bps.go.id. Available at: www.bps.go.id/
WTP
∫ Xf ( X ) dx
X
by assuming finite minimum and maximum
statictable/2009/02/20/1268/laju-pertumbuhan-penduduk-menurut-
provinsi.html (accesesed 5 August 2017). individual WTP as X and X , respectively, yields the result.)
USEPA. (2012) Global Anthropogenic Non-CO2 Greenhouse Gas Emissions: From the obtained results, we know the estimates of choice
1990–2030. Washington DC: United States Environmental Protection probability for each policy q(X,a) at X = 0, 24,000 rupiahs, 48,000
Agency.
Yuan Y and Yabe M (2014a) Residents’ preferences for household kitchen rupiahs, and 96,000 rupiahs. This identifies the distribution of the
waste source separation services in Beijing: A choice experiment cumulative marginal WTP distribution F(X) piecewise at these
approach. International Journal of Environmental Research and Public values of X as well as X = −∞ and X = +∞, namely, F ( −∞ ) = 0
Health 12: 176–190.
and F ( ∞ ) = 1 . The average WTP can be decomposed at these
Yuan Y and Yabe M (2014b) Residents’ willingness to pay for household
kitchen waste separation services in Haidian and Dongcheng districts, values of X as:
Beijing city. Environments 1: 190–207.
Yuan Y and Yabe M (2015) Residents’ preferences for household kichen 4 F ( X i +1 )
waste source separation services in Beijing: A choice experiment
approach. International Journal of Environmental Research and Public ∑ ∫
i =0 F ( X i )
F −1 (φ ) dφ
Health 12: 176–190.
Zhang DQ, Tan SK and Gersberg RM (2010) Municipal solid waste manage-
ment in China: Status, problems and challenges. Journal of Environmental where X0 = −∞ , X1 = 0, X2 = 24,000 rupiahs, X3 = 48,000 rupiahs,
Management 91: 1623–1633. X4 = 96,000 rupiahs and X = ∞ 5. Because the cumulative distri-
Zhuang Y, Wu SW, Wang YL, et al. (2008) Source separation of household
bution F is monotonically increasing in X, it must be that:
waste: A case study in China. Waste Management 28: 2022–2030.
4 F ( X i +1 ) 4
Appendix ∑ ∫
i =0 F ( X i )
F −1 (φ ) dφ ≥ ∑X F ( X
i =0
i i +1 ) − F ( X i )
Derivation of average WTP
Here we provide the derivation of the average WTP for accepting where the right-hand side provides the lower bound of the aver-
a policy, as well as its lower bound. Let f(X|a) be the density of age WTP for the policy acceptance.