Journal of Cleaner Production: Yongchao Zeng, Peiwu Dong, Yingying Shi, Lingling Wang, Yang Li

Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

Journal of Cleaner Production 256 (2020) 120384

Contents lists available at ScienceDirect

Journal of Cleaner Production


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

Analyzing the co-evolution of green technology diffusion and


consumers’ pro-environmental attitudes: An agent-based model
Yongchao Zeng a, c, *, Peiwu Dong a, **, Yingying Shi a, c, Lingling Wang a, c, Yang Li b, c
a
School of Management and Economics, Beijing Institute of Technology, Haidian, Beijing, 100081, China
b
School of Electrical Engineering, Northeast Electric Power University, Jilin, 132012, China
c
Energy Systems Division, Argonne National Laboratory, Lemont, IL, 60439, USA

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

Article history: Massive diffusion of green technologies is significant for building a cleaner world. However, the process
Received 29 April 2019 of technology diffusion is usually slow and complex. An in-depth understanding of the mechanism
Received in revised form regarding green technology diffusion is an essential precondition for effectively stimulating this process.
10 December 2019
Green technology diffusion heavily involves both social and technological changes. Although existing
Accepted 1 February 2020
Available online 5 February 2020
studies have provided rich knowledge about identifying critical drivers and barriers affecting green
technology diffusion, research that considers consumers’ pro-environmental attitudes and green tech-
Handling editor. M.T. Moreira nology diffusion as an evolving system is still sparse. Aiming at exploring the co-evolution of consumers’
pro-environmental attitudes and green technology diffusion, this paper builds an agent-based model
Keywords: that integrates the relative agreement model with technology diffusion theories to conduct a sequence of
Green technology diffusion controlled numerical experiments, which progressively unveil how attitudinal and technological factors
Relative agreement model impact green technology diffusion. The main findings include that (1) improving consumers’ pro-
Agent-based modeling environmental attitudes is prominently beneficial to green technology diffusion and maturation; (2)
ABM
technology maturity has very limited impact on consumers’ first-time purchases but significantly affects
Agent-based simulation
consumers’ satisfaction, which would further impact consumers’ repeat purchases; (3) consumers that
do not support green technologies frequently emerge during the evolution of attitudes (despite the high
technology maturity), which corresponds to the emergence of the anti-environmental groups observed
in the real world; (4) active interactions between non-adopters enable their attitudes to converge, which
results in the polarization of consumers’ attitudes.
© 2020 Elsevier Ltd. All rights reserved.

1. Introduction stimulate green technology diffusion (Bratanova et al., 2016; Bi


et al., 2017; Zhang et al., 2019). It is believed that the high cost
Green technologies play a critical role in dealing with the will decrease as green technologies progress. Nevertheless, tech-
world’s environmental and resource challenges. However, the nologcial progress alone is insufficient to drive consumers to adopt
diffusion of green technologies is hindered by various difficulties. green technologies (Axsen and Kurani, 2012); social influence is
Technological immaturity accompanied by high cost is a crucial another indispensable driver triggering consumers’ adoption of
barrier to the diffusion of green technologies (Bratanova et al., technologies (Axsen and Kurani, 2012; Arvanitis and Ley, 2013;
2016; Xu et al., 2017). Multiple monetary measures (e.g., taxes Robinson and Rai, 2015). Particularly, green technology diffusion is
and subsidies) have been intensively researched and employed to not purely a commercial process but involves the formation of so-
cial consensus. Many countries have devoted considerable re-
sources to promoting and popularizing environmental awareness
of the public (Gifford and Nilsson, 2014). These efforts broadly in-
* Corresponding author. School of Management and Economics, Beijing Institute
of Technology, 5 South Zhongguancun Street, Haidian District, Beijing, 100081, fluence the aggregate attitude of the public, which can significantly
China. re-shape the diffusion pattern of green technologies. Previous
** Corresponding author. Intelligent Synergistic Management Laboratory, School studies have empirically confirmed that consumers with better
of Management and Economics, Beijing Institute of Technology, 5 South Zhong- pro-environmental attitudes are more willing to adopt green
guancun Street, Haidian District, Beijing, 100081, China.
E-mail addresses: zengyc@bit.edu.cn (Y. Zeng), dongpeiwu@bit.edu.cn (P. Dong).
technologies (Hussain et al., 2014; Norazah Mohd Suki et al., 2016;

https://doi.org/10.1016/j.jclepro.2020.120384
0959-6526/© 2020 Elsevier Ltd. All rights reserved.
2 Y. Zeng et al. / Journal of Cleaner Production 256 (2020) 120384

Degirmenci and Breitner, 2017). As being green is becoming a social empirical observations. For example, despite high technological
norm and consensus, it is necessary to consider frequent in- maturity, attitudes splits are frequently observed during the sim-
teractions and attitude variations among consumers when study- ulations, which corresponds to the emergence of the anti-
ing green technology diffusion. environmental groups during the wise use movement in America
The extreme mutability of populations’ attitudes has been (McCarthy, 2002). Academically, this study contributes to the
noticed and researched with opinion dynamics models (Coates existing literature by clarifying the roles of the attitudinal and
et al., 2018). The relative agreement (RA) model is one of the technological factors in green technology diffusion from a dynamic
most preferred and widely used models that describe the evolution system perspective; also, the novel model could be extended for
of attitudes in large groups. Different from many other opinion further explorations of social-technological systems. Practically,
dynamics models that assume the opinions of agents to be binary this paper could help policymakers distinguish and design effective
(i.e., yes or no), the RA model endows agents with generous het- environmental policies. Moreover, green product producers could
erogeneity: agents’ opinions are represented by continuous extract useful information to adjust relevant marketing strategies.
numbers within a certain interval (Deffuant et al., 2002). The RA This paper is organized as follows. The ensuing section (Mate-
model uses two key variables to describe the opinion of an agent: rials and methods) contains four subsections. Methodology sub-
one reflecting the current opinion of an agent and the other rep- section introduces the features of agent-based modeling. Model
resenting the confidence the agent holds about this opinion. Using framework subsection discusses the structure of this model. Model
continuous variables to describe a consumer’s opinion is more dynamics subsection details how this model operates. In Calibra-
realistic because it captures the ambiguity of opinions (besides tion and validation subsection, the model is calibrated with the
“yes” or “no”, people often say “maybe”), which could effectively Bass diffusion model to replicate the stylized facts of technology
model the diversity of consumers’ attitudes towards green tech- diffusion. The numerical experiments are described in Results
nologies. Another realistic mechanism the RA model mimics is the section. Five controlled numerical experiments are conducted in a
asymmetric influence between two interacting individuals: only sequential order to investigate the roles of technologies and con-
the individual with more confidence can influence the opinion of sumers’ pro-environmental attitudes in green technology diffusion.
an individual with less confidence. Discussion section discusses the implications based on the findings
With the features of heterogeneity and asymmetric influence, of the numerical experiments. The conclusions of this research and
the RA model serves as an effective tool for green technology re- suggestions for future research are given in the final section.
searchers and policymakers to explore the formation and variation
of the pro-environmental attitudes of the public. Robinson and Rai 2. Materials and methods
(2015) incorporated the RA model into a large scale agent-based
model and used rich data to distinguish the necessary de- 2.1. Methodology
terminants affecting the diffusion of photovoltaic systems.
Deffuant et al. (2005) used the RA model to study various sce- Agent-based modeling (ABM) is well known for its advantage of
narios of innovation diffusion and took into account both the simulating complex systems, which consist of numerous interactive
impact of social values and individual benefits. Kaufmann et al. heterogeneous elements. Differing from conventional modeling
(2009) applied the RA model to simulating the diffusion of approaches that model the world from top-down perspectives,
organic farming practices. These studies provide valuable insights agent-based modeling allows researchers to mimic the behaviors
into consumers’ attitude evolution in the context of innovation and attributes of individual elements microscopically and observe
diffusion. Janssen and Jager (2002) considered both technology the phenomena emerging at system level. The ABM approach can
adopters and suppliers. They built a multi-agent model that in- help researchers capture many dynamic features of complex
corporates the interplay between consumers and green product systems.
producers but purely from an economic perspective and left social The RA model is originally an agent-based model that describes
values of consumers underemphasized. the dynamic process of opinion evolution in large groups, in which
The social and technological attributes of green technologies people communicate with and influence each other. Because
resemble the co-existing two sides of one coin. Existing studies technology diffusion is a process involving numerous interactive
have provided considerable knowledge in identifying drivers and and adaptive consumers, ABM is particularly effective and
barriers to green technology diffusion (Trianni et al., 2013; Cagno preferred by many researchers (Kangur et al., 2017; Zeng et al.,
and Trianni, 2014; Minetto et al., 2018). However, the research 2018; Shi et al., 2019). Since this paper applies the RA model to
that integrates consumers’ pro-environmental attitudes and green green technology diffusion, it is natural and proper to select ABM as
technology diffusion as an evolving system is still far from our modeling approach.
adequate. To unveil the roles of these entangled two aspects in
green technology diffusion, an agent-based model is built to 2.2. Model framework
combine the RA model and technology diffusion theories to
conduct an exploratory study, which emphasizes simultaneously Fig. 1 shows the conceptual framework of the model. The
the attitudinal and technological aspects of green technology numbered arrows illustrate the logic flow of the simulated dynamic
diffusion. Agent-based modeling (ABM) is an effective tool to model system. Environmental education and promotion are assumed as
dynamic and heterogeneous complex systems. ABM gives re- exogenous variables that predetermine consumers’ initial pro-
searchers abundant flexibility to conduct controlled numerical environmental attitudes. This assumption is basically consistent
experiments, which is particularly suitable for investigating com- with Deffuant et al. (2005), who initialized the social values of
plex systems. Based on the agent-based model, a sequence of simulated consumers with normal distributions. Consumers are
controlled numerical experiments are conducted to demystify the influenced directly by their neighbors in the network. The in-
roles of technologies and attitudes in green technology diffusion. teractions between consumers may change their opinions about
Besides reproducing the stylized facts about technology diffu- green technologies. The change of each consumer’s opinion in-
sion (e.g., the S-shaped curve as the Bass diffusion model depicts volves two variables e the attitude towards a green technology and
(Bass, 1969)) with different parameter settings, this model can the consumer’s perceived uncertainty about the attitude e which
exhibit behaviors highly consistent with multiple interesting follows the RA model (Deffuant et al., 2002). The attitude and
Y. Zeng et al. / Journal of Cleaner Production 256 (2020) 120384 3

Fig. 1. Conceptual framework of the model.

uncertainty jointly determine the probability that a consumer


adopt the green technology. Once a consumer adopts a green
technology, the consumer has authentic experience of the green
technology and thus changes his/her opinion according to the
satisfaction derived from the green technology. Meanwhile, the
consumer’s perceived uncertainty decreases drastically. As a result
of technology adoption, revenues are generated and received by
green technology suppliers. According to Patterson (1998),
Hartmann (2003), and Hartmann et al. (2006), a corporation nor-
mally uses a proportion of its annual revenues as innovation in-
vestments, which can improve the existing technology. The specific
quantity of innovation investments extracted from revenues de- Fig. 2. A simplified flowchart of the model.
pends on complex factors. For simplicity, a variable is introduced to
directly relate the number of adopters to the improved techno-
logical maturity when the model is implemented (Zeng et al., 2018). with their neighbors. The algorithm is designed as follows. For each
The improved technology with higher maturity then affects the agent, it interacts with its every neighbor with a certain probability
consumers’ satisfactions. Although this conceptual framework in a randomized order. It is assumed that adopters are more active
contains nothing beyond our current knowledge, it incorporates than non-adopters. Hence, if at least one of the two neighboring
the necessary components in the dynamic process of green tech- agents is an adopter, the interaction is more probable to occur (this
nology diffusion. The agent-based model built based on this con- probability is denoted as qlarge ). On the contrary, the interaction
ceptual framework can reproduce many empirical observations as between two neighboring non-adopters is less probable to occur
well as generate interesting but explainable new findings, which (this probability is denoted as qsmall ). During the interaction, agents
could improve the understanding of green technology diffusion. update their opinions according to the RA model. Equations (2) and
(3) describe the opinion update during the interaction between
agent i and j, given that agent j is more confident than agent i.
2.3. Model dynamics !
Oji ðtÞ  
Fig. 2 illustrates the dynamics of the model. When t ¼ 0, the opi ðt þ 1Þ ¼ opi ðtÞ þ step ,  1 , opj ðtÞ  opi ðtÞ (2)
uj ðtÞ
attitude of each consumer is initialized and assigned with a number
between zero and one, which respectively stand for an extremely
where step is a parameter controlling the speed of convergence;
low and high willingness to adopt green technologies. All con-
Oji ðtÞ denotes the opinion overlap of agent i and j. Oji ðtÞ is calculated
sumers originally are potential adopters and reside in a small-world
according to Equation (3).
network. These potential adopters independently adopt green
technologies with a probability of p1i . Once some consumers  
become adopters, their attitudes towards the adopted green tech- Oji ðt þ 1Þ ¼ min opi ðtÞ þ ui ðtÞ; opj ðtÞ þ uj ðtÞ
nologies become unambiguous. The adopters update their opinions  
 max opi ðtÞ  ui ðtÞ; opj ðtÞ  uj ðtÞ (3)
according to Equation (1). Correspondingly, the uncertainties of
these adopters are assigned with a very small value.
It is noteworthy that the influence between each pair of agents
 are not reciprocal e only an agent with less uncertainty (more
1; if satisfied ¼ true
opi ðtÞ ¼ (1) confidence) can influence one with more uncertainty (less
0; if satisfied ¼ false
confidence).
After the independent adoption stage, the population is After interaction, each non-adopter decides whether to adopt
comprised of adopters and non-adopters. They begin to interact the green products with a probability of p2i ðtÞ, which is calculated
4 Y. Zeng et al. / Journal of Cleaner Production 256 (2020) 120384

using Equation (4). model that only considers the unidirectional transmission of dis-
eases: from infected individuals to uninfected individuals (Bass,
p2i ðtÞ ¼ q , ð1  2ui ðtÞÞ,opi ðtÞ (4) 1969).
Based on the RA model and technology experience, the agents’
where q is a parameter waiting for calibration. Equation (4) means behaviors defined in this model correspond in a meaningful way to
the probability of a non-adopter is positively correlated with his/ consumers’ behaviors in the real world, which validates the model
her opinion, but negatively correlated with his/her uncertainty. at the microscopic level. The model can fit well to the S-shaped
Equation (4) reflects the common sense that people tend not to buy diffusion curve admitted as a stylized fact of technology diffusion,
the things they are unfamiliar with because of risk aversion. If non- which validates the model at the macroscopic level. In these senses,
adopters successfully become adopters, they have better knowl- this model has enough validity for the ensuing exploratory tasks.
edge about the technologies to update their opinion according to The results generated by the calibrated model works as the control
Equation (1). The non-adopters in time t enter the next (i.e., t þ 1) group for the following controlled numerical experiments.
iteration. Technology maturity determines the probability that an
adopter feels satisfied with the adopted technology. In this model, 3. Results
consumers are assumed to expect technologies with 100 unit of
maturity to satisfy them. The probability that the technology at 3.1. The overall impact of consumers’ initial pro-environmental
time t satisfies an adopter is updated using equation psi ðtÞ ¼ mðtÞ= attitudes
100, where mðtÞ represents the technology maturity calculated
using Equation (5): The pro-environmental attitudes of populations in different
countries, societies, and communities may differ significantly. To
mðtÞ ¼ minit þ r , ð100  minit Þ,NðtÞ = N (5)
explore the impact of consumers’ pro-environmental attitudes on
green technology diffusion, different initial opinion distributions
where minit is the initial maturity when a technology first enters a
are used to simulate a population’s inclination for green technol-
market; r adjusts the speed of technology maturation; NðtÞ is the
ogies. As shown in Fig. 4, the triangular distributions displayed in
number of adopters at time t; and N is the estimated number of
(a), (b), and (c) respectively refer to the simulated population
total potential adopters. N is not the same as the number of total
having negative, neutral, and positive attitudes towards green
simulated agents. The maximum mðtÞ is controlled to not exceed
products. Different environmental tendencies of the whole popu-
100 in this model, in case of errors. Equation (5) captures the
lation result from different effectiveness of environmental pro-
general regularity of technology development: starting from a low
motion or education. The model is run for 100 times under each of
end, surviving consumers’ choices, and reaching an upper limit
the three conditions to derive statistically meaningful results, and
(Dosi, 1982; Christensen, 2013).
Fig. 5 is obtained. The scenario peakat ¼ 0:65 is also included in
Fig. 5 because it is equivalent to Fig. 3 as a benchmark scenario.
2.4. Calibration and validation Fig. 5 shows two main features: one is that the number of
adopters increases when peakat becomes greater; the other feature
To the best of authors’ knowledge, the empirical data repre- is that the boxes are all short and the outliers are very few, which
sentative enough to depict the diffusion process of green technol- indicates the results are constantly stable. It could be inferred that
ogies is still unavailable. The Bass diffusion curve indicates that the bettering the population’s attitudes towards green products stably
diffusion pattern of technologies is an S-shaped curve, which has produce more adopters, which is consistent with intuition and
been empirically confirmed and widely used by researchers previous research (Norazah Mohd Suki et al., 2016; Hussain et al.,
(Radomes Jr and Arango, 2015; Fan et al., 2017; Reddy, 2018). 2014; Degirmenci and Breitner, 2017). However, it should be
Although the diffusion of green technologies has its uniqueness, it noted that these results are obtained under the condition that the
is relatively appropriate to assume the diffusion curve of green products can well satisfy the consumers (minit ¼ 75 and rate ¼ 2).
technologies is approximately an S-shaped curve in consideration Actually, many green technologies are new to consumers and still at
of the lack of data. Instead of focusing on a specific technology, the the infant stage. It is reasonable to question whether consumers’
model is calibrated with the data derived by Sultan et al. (1990), attitude can save those green technologies with inferior initial
who calculated the average parameter values of the Bass model performance (meaning a small minit ). The next experiment answers
with the data collected from over 200 different technologies. These this question.
parameter values give an S-shaped diffusion curve that hits 95%
market share of the total estimated market scale in the sixteenth 3.2. The seemingly negligible technology maturity
year. Because this research is not concerned with the predictive
accuracy of the model but targets exploring the general dynamics The initial maturity of green technologies is varied to generate
of the interplay between consumers’ pro-environmental attitudes some mutations in the diffusion process. The model is run with
and green technology diffusion, these data could effectively pre- minit ¼ 25, 50, and 75 respectively representing 25%, 50%, and 75%
vent the model from behaving meaninglessly in the huge param- initial technology maturity, and other parameters remain the same
eter space. as those in the first experiment. The results are shown in Fig. 6. A
Netlogo’s Behaviorsearch is employed to find a group of pa- prominent trend is that increasing peakat and minit can gain more
rameters for the agent-based model to fit the Bass diffusion curve adopters but the impact of minit is insignificant in contrast with the
(Wilensky, 1999). The parameter values are listed in Table 1. Given boosting effect brought by peakat. Taking the first row (e.g., sub-
these parameter values, this model shows results approximating to graph (a), (b) (c) in Fig. 6) as an example, increasing green tech-
the Bass diffusion curve (R2 > 0:95, see Fig. 3). It is noteworthy that nologies’ initial maturity while maintaining peakat ¼ 0:25
qsmall is assigned with 0, which seems to imply that a smaller qsmall positively influences the diffusion but with very limited effective-
leads the model to a higher fitness to the Bass diffusion curve. ness. The other two rows exhibit similar features. As to the col-
Actually, the Bass diffusion model only allows adopters to influence umns, each column has three subgraphs showing the results with
non-adopters but leaves the interactions between pure non- the same minit but different peakat. It can be witnessed that
adopters ignored. This mechanism is inherited from the epidemic increasing peakat prominently stimulates the diffusion. This
Y. Zeng et al. / Journal of Cleaner Production 256 (2020) 120384 5

Table 1
Values of parameters through calibration.

Parameter Meaning Value

p the probability that consumers independently adopt green 0.004


technologies
q the parameter adjusting the probability that consumers adopt green 0.28
technologies by their own decisions.
peakat the horizontal coordinate of the peak of a triangular distribution 0.65
minit initial technology maturity 65
r the parameter adjusting the speed of technology maturation 2
qlarge the probability that interaction involving at least one adopter occurs 0.66
qsmall the probability that interaction between two non-adopters occurs 0
uð0Þ the initial uncertainty consumers hold about green technologies 0.46
network density the density of the modeled small-world network 1.25%

3.3. The splitting attitudes

To observe the cause other than technology maturity that


hinders the diffusion, the initial technology maturity is fixed at
100 (indicating the technology completely satisfies consumers
and thus excluding the influence of technology maturity). The
model is run for 300 times in total with peakat ¼ 0:25, 0.50, and
0.75. Fig. 7 displays the attitude distributions at the beginning
(t ¼ 0) and end (t ¼ 30) of a simulation. It can be observed that
there distinctively exist two groups of consumers: one group
converge to the rightmost of the horizontal axis, and the other
group scatter along the left half horizontal axis. The two groups
are distant from each other and have no chance to interact. The
scattering consumers have no experience of green products. They
are just left distantly from the majority during the evolution of
attitudes. This seemingly bizarre behavior of the model actually
captures a phenomenon that has occurred in the real world:
accompanied by the rapid rise of environmental awareness, some
people might be isolated from the majority who have already
achieved consensus on environmental protection. The people
Fig. 3. Fit the agent-based model to the Bass diffusion curve. staying distantly from the environmentalists are not likely to
adopt green technologies anymore. Hence, it is the attitude split
that prevents the green technology from broader diffusion. This
experiment demonstrates that the population’s initial attitudes model behavior corresponds to the emergence of anti-
towards green technologies contribute much more to the diffusion environmental movements, for example, the wise use move-
of green technologies than the initial technology maturity does. ment. In the late 1980s, a trend was stimulated by highly
The results above might not be consistent with the intuition that bureaucratic, and professionalized environmental groups in
immature technologies could affect consumers’ experience and America (McCarthy, 2002). These groups advocate environmental
give rise to negative word-of-mouth, which could be a disaster to protection and resource reservation. However, this advocacy
the diffusion (Goldenberg et al., 2007). On the contrary, technolo- neglected the complaints of rural communities, who once could
gies with high maturity can satisfy consumers better, which could freely use lands and resources in a variety of ways (McCarthy,
produce positive inter-person influence that propels the technol- 2002). These communities might never have authentic experi-
ogy diffusion (Goldenberg et al., 2007). It is natural to speculate that ence of adopting green technologies but were organized to resist
there exists a latent non-technological factor that suppresses the environmental management by their own means.
diffusion, which is confirmed in the next part.

Fig. 4. Different initial opinion distributions of the population.


6 Y. Zeng et al. / Journal of Cleaner Production 256 (2020) 120384

Fig. 5. The impact of initial opinion distributions on green technology diffusion. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web
version of this article.)

3.4. Rethinking the technology maturity assumptions but also an improved understanding of the mecha-
nism underlying the technology diffusion process.
The above results demonstrate that the green market scale is
largely predetermined by the initial attitude distribution of pop- 3.5. Activating the non-adopters’ interactions
ulations. However, technology maturity is not negligible. The
dashed curves shown in Fig. 6 indicate the number of disappointed The activated non-adopters’ interactions might provide another
adopters. As can be seen, a greater minit leads to fewer disappointed necessary piece needed to complete the “story” about the emer-
adopters. It is reasonable to infer that satisfied adopters are willing gence of the anti-environmental groups. As seen in Fig. 7, the
to purchase the technology again. Thus, technology maturity only scattering consumers are not organized or achieve consensus. It is
negligibly impacts consumers’ first-time purchases but might hard to imagine these people could behave uniformly to resist
significantly affect their repeat purchases. It is noteworthy that the environmental protection. However, when the non-adopters (all
initial distributions of attitudes also prominently impact technol- these scattering consumers are non-adopters) are allowed to
ogy maturation. The proportion of disappointed adopters to total interact, their opinions start to converge. Based on the results
adopters is denoted as dis in Fig. 6. Even with the same minit , greater shown in Fig. 7, the model is run for another 30 steps (i.e., t ¼ 60)
peakat can lead to a lower proportion of disappointed adopters, but change qsmall from 0 to 0.33. The scattering consumers begin to
which indicates that a good social environment is beneficial to the cluster (see Fig. 8). This is consistent with the coalition and orga-
maturation of green technologies. nization of the grass-roots participants in the wise use movement
The negligible impact of technology maturity on green tech- (McCarthy, 2002). They were neither organized nor had any clear
nology diffusion might be more comprehensible, if one notes the appeal until they realized the impact from the increasingly pre-
facts about these Bass diffusion model: first, the original Bass vailing trend stimulated by a group of distant and highly profes-
diffusion model only describes the first-time purchases of adopters; sionalized environmentalists. To gather the power to resist the
second, the market scale is estimated as a fixed number; third, environmental trend, the wise use movement participants began to
technology maturation is not considered. In other words, the cluster and formed clear pursuit. These anti-environmentalists are
experiment results that the diffusion pattern has no significant essentially different from the disappointed adopters who should
correlation with technology maturity are consistent with the as- cluster at the left end of the horizontal axis because the anti-
sumptions used in the Bass model. The agent-based model herein environmentalists have no intention and true experience of using
gives not only a possible explanation for the validity of these green technologies. In other words, they just refuse to try green
Y. Zeng et al. / Journal of Cleaner Production 256 (2020) 120384 7

Fig. 6. The combined impact of different initial opinion distributions and technology maturity on green technology diffusion. The dashed curves represent disappointed adopters.
(For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

technologies for some social reasons. However, the disappointed results indicate clearly that increasing the population’s pro-
adopters who have bad experience of using a specific green tech- environmental attitude positively and stably impacts the diffusion
nology are very likely willing to try other green technologies or wait of green technologies, which is consistent with intuitions and the
for the technology maturation because their disappointment is relevant previous research (Norazah Mohd Suki et al., 2016;
caused by technological reasons, not social ones. Hussain et al., 2014; Degirmenci and Breitner, 2017).
The second experiment introduces another critical factor, i.e.,
technological maturity. The results demonstrate that the impact of
4. Discussion
technological maturity seems to be negligible in contrast with the
propelling effect brought by pro-environmental attitudes. As more
As caring about the environment is becoming a norm, the
data are collected, the impact of consumers’ attitudes and tech-
diffusion of green technologies is not only a matter of technology
nology maturity on green technology diffusion becomes more
but also a significant social change. Due to the lack of research
distinct. Fig. 9 shows the mean value of adopters the diffusion can
providing a dynamic perspective on the co-evolution of consumers’
eventually reach within 30 iterations. As can be witnessed, the
pro-environmental attitudes and technology development, this
system is much more sensitive to the variation of the consumers’
paper builds an agent-based model to combine both aspects by
attitudes than to the change of technology maturity. The total
integrating the RA model and relevant technology dynamics the-
number of adopters is largely predetermined by initial attitude
ories. After calibrating the model, a sequence of controlled nu-
distributions. The interactions between consumers give rise to the
merical experiments are conducted to progressively uncover the
split of attitudes: one group hold positive pro-environmental atti-
roles of technology and pro-environmental attitudes in the process
tudes and firmly support green technology, and the other group
of green technology diffusion.
scatter distantly from the supporters. It is necessary to distinguish
Because of the considerable efforts made by governments and
the disappointed adopters from these non-supporters. The disap-
NGOs, the public are broadly educated to be environmentally
pointed adopters are those who have ever tried some green tech-
aware. Therefore, the whole population should manifest certain
nologies but are disappointed for some technological issues (e.g.,
aggregate inclination for green products. The first numerical
immaturity). On the contrary, the non-supporters have no tech-
experiment uses three types of triangular distributions to mimic
nological experience but are gradually isolated during the attitude
the different attitude distributions of the simulated population. The
8 Y. Zeng et al. / Journal of Cleaner Production 256 (2020) 120384

Fig. 7. The divergence of consumers’ attitudes towards green technologies.

evolution. Therefore, it is the split of attitudes that prevents green existing studies, and the information obtained from the official
technologies from diffusing more widely. The emergence of these websites of some leading environmental NGOs (such as Green-
non-supporters mimics the rise of the wise use movement known peace, Conservation International, and the Nature Conservancy),
as an anti-environmental movement launched by rural commu- environmental NGOs use multiple ways to promote the environ-
nities to resist strict environment management policies during the mental awareness of the public, for example, launching educational
late 1980s in America (McCarthy, 2002). projects, collaborating with governments and enterprises, and
Despite the very limited impact technology maturity exerts on advertising, etc. Badruddin, 2015 pointed out that the government
consumers’ first-time purchase, the further experiment shows that alone cannot effectively address environmental issues without the
technologies with higher maturity can prominently improve support from the masses; and NGOs are one of the most effective
adopters’ satisfaction, which would significantly influence con- media to reach the public and play a significant role in environ-
sumers’ repeat purchase decisions. Another important information mental protection. NGOs influence the public through multiple
conveyed by the experiment is that a population with better pro- channels, such as brochures, articles, seminars, lectures and group
environmental attitudes can provide a benign atmosphere for the discussions, for promotion of environmental awareness (Badruddin
maturation of green technologies, which is more evidently dis- (2015)). Yang (2005 investigated the functionality of environmental
played in Fig. 10. Compared with the trivial impact in Fig. 9, NGOs in China and proposed that the environmental NGOs provide
increasing minit notably improves the satisfaction ratio in Fig. 10. In opportunities for self-fulfillment and social experience for the
addition, Fig. 10 explicitly illustrates that a great peakat can lead the public. Particularly, Yang (2005) pointed out that environmental
satisfaction ratio to a high level particularly when minit is small (see NGOs are a site and agent that attract citizens living in a relatively
point A and B in this figure). A noteworthy subtlety is that the constraining political environment to participate in civic actions
improving effect of peakat on the satisfaction ratio is radically and test political limits. Jian-nan (2012) emphasized environmental
constrained by minit ; on the contrary, minit alone can improve the NGOs’ significant contributions to informal environmental educa-
satisfaction ratio to over 0.9. tions that widely influence the public. Therefore, authorities
Improving the pro-environmental attitudes of the public has together with NGOs should continue to promote and educate
multiple merits for green technology diffusion and maturation, people to develop their pro-environmental attitudes. However, for
which implies that the efforts of environmental promotion and the countries where environmental awareness of the public begins
education are crucial for building a cleaner world. According to the to rapidly grow, the authorities should be cautious about the
Y. Zeng et al. / Journal of Cleaner Production 256 (2020) 120384 9

Fig. 8. The non-adopters’ interactions lead the scattering non-adopters to cluster.

Fig. 9. Total adopters the diffusion can reach under the combined influence of minit and peakat
10 Y. Zeng et al. / Journal of Cleaner Production 256 (2020) 120384

Fig. 10. Satisfaction ratio under the combined influence of minit and peakat

divergence of attitudes. Green product manufacturers alone might research contributes to the existing literature by clarifying the roles
have very limited power to change the total scale of green markets played by technologies and attitudes in the context of green tech-
but striving to improve the technologies can gain more repeat nology diffusion from a dynamic system perspective. Combining
purchases from consumers. the RA model and technology dynamics theories, a novel model is
In addition, non-adopters’ interactions might contribute an built, and a sequence of controlled numerical experiments are
important piece needed to complete the formation of anti- conducted.
environmental groups. Combining the above experiments, a The first experiment demonstrates straightforwardly that
possible path of the attitude evolution during the wise use move- improving consumers’ pro-environmental attitudes is beneficial to
ment emerges. At first, some pioneer environmentalists realized green technology diffusion. The second experiment shows that the
the importance of environmental protection and began to influence impact of technology maturity on green technology diffusion is
others, which corresponds to the independent green technology negligible, compared with the significant impact of the pro-
adopters in this model. Nevertheless, the transmission of infor- environmental attitudes. This result might be counter-intuitive.
mation might not be efficient enough, or for some social reasons The ensuing experiment uncovers the latent cause of this
(such as they can only effectively influence the people who have counter-intuitive result and finds that the split of attitudes is a
strong links with them) their influence is limited within local parts dominant factorpreventing green technologies from diffusing more
of the network they reside in (the adopters are only allowed to broadly. The split of attitudes corresponds to the emergence of anti-
communicate with their neighbors when qsmall ¼ 0 in this model). environmentalists. The further experiment demonstrates technol-
After the environmentalists successfully persuade the people ogy maturity significantly impacts the satisfaction of consumers,
accessible in the network, environmental awareness becomes a which is very likely to affect consumers’ repeat purchases. More-
consensus of the majority. The people who are not accessed by the over, the results also indicate that a benign atmosphere of con-
environmental groups are left isolated and gradually realize the sumers’ attitudes is beneficial to the growth of comparatively
prevailing environmental proposition neglects their benefit. They immature technologies. In the last experiment, the interactions
begin to self-organize (through the interactions of non-adopters between non-adopters are activated, and the non-adopters begin to
when qsmall ¼ 0:33 in the model) and put forward clear goals to cluster, which corresponds to the coalition of grass-roots anti-
resist the environmental proposition. As a consequence, the po- environmental groups.
larization of attitudes forms. Although, the wise use movement is Because the essence of models is selective abstractions of re-
history, understanding the mechanism ruling the formation of anti- alities, this model is inevitably simplified from real-world phe-
environmental attitudes is still significant for the countries that nomena. Considering the complexity of decision-making processes
have not yet experienced rapid awakening of environmental and interactions of consumers, future research could extend this
awareness. model by incorporating richer elements (such as research cost and
social capital) to obtain more insights into green technology
5. Conclusions diffusion systems’ evolutionary dynamics. In addition, although the
data used to calibrate this model may be sufficient for an explor-
The severe environmental challenges necessitate the diffusion atory study, researchers still need to consider temporal and
of green technologies. However, technology diffusion is usually a geographical differences when attempting to generalize this cali-
slow process. Understanding the mechanism of green technology brated model to real-world applications.
diffusion can help to propel the progress of sustainability. The
entanglement of technological and social factors veils the regularity Author contributions
underlying green technology diffusion, which entails the explora-
tions into the dynamics of the co-evolution of green technology Yongchao Zeng and Peiwu Dong proposed the idea; Yongchao
diffusion and consumers’ pro-environmental attitudes. This Zeng conceptualized and implemented the agent-based model,
Y. Zeng et al. / Journal of Cleaner Production 256 (2020) 120384 11

analyzed the data and wrote the first draft; Yingying Shi, Lingling Gifford, Robert, Nilsson, Andreas, 2014. Personal and social factors that influence
pro-environmental concern and behaviour: a review. Int. J. Psychol. 49 (3),
Wang and Yang Li verified and validated the model, revised the
141e157.
writing and edited the manuscript; the research was conducted Goldenberg, Jacob, Libai, Barak, Moldovan, Sarit, Muller, Eitan, 2007. The NPV of bad
under the supervision of Peiwu Dong. news. Int. J. Res. Market. 24 (3), 186e200.
Hartmann, George C., 2003. Linking r&d spending to revenue growth. Res. Technol.
Manag. 46 (1), 39e46. https://doi.org/10.1080/08956308.2003.11671543. URL.
Declaration of competing interest Hartmann, George C., Myers, Mark B., Rosenbloom, Richard S., 2006. Planning your
firm’s r&d investment. Res. Technol. Manag. 49 (2), 25e36. https://doi.org/
The authors declare that they have no known competing 10.1080/08956308.2006.11657366. URL.
Hussain, M.A., Khokhar, M.F., Ali, A., 2014. Green awareness effects on consumers
financial interests or personal relationships that could have purchasing decision: a case of Pakistan. Global J. Manag. Bus. 14 (6). https://
appeared to influence the work reported in this paper. journalofbusiness.org/index.php/GJMBR/article/view/1555.
Janssen, Marco A., Jager, Wander, 2002. Stimulating diffusion of green products -
Co-evolution between firms and consumers. J. Evol. Econ. 12 (3), 283e306.
Acknowledgements https://doi.org/10.1007/s00191-002-0120-1. ISSN 09369937.
Jia-nan, C., 2012. Contributions of environmental NGO to environmental education
This work is supported by the National Natural Science Foun- in China. IERI Procedia 2, 901e906.
Kangur, Ayla, Jager, Wander, Verbrugge, Rineke, Bockarjova, Marija, 2017. An agent-
dation of China (No.71873015). In addition, we would like to convey based model for diffusion of electric vehicles. J. Environ. Psychol. 52, 166e182.
special thanks to Wei Meng and Fengming Guo for their irre- Kaufmann, Peter, Stagl, Sigrid, Franks, Daniel W., aug 2009. Simulating the diffusion
placeable assistance with this research. of organic farming practices in two New EU Member States. Ecol. Econ. 68 (10),
2580e2593. https://doi.org/10.1016/j.ecolecon.2009.04.001. ISSN 09218009.
McCarthy, James, 2002. First world political ecology: lessons from the Wise Use
References movement. Environ. Plann. 34 (7), 1281e1302. https://doi.org/10.1068/a3526.
ISSN 0308518X.
Arvanitis, Spyros, Ley, Marius, 2013. Factors determining the adoption of energy- Minetto, Silvia, Marinetti, Sergio, Saglia, Pietro, Masson, Nina, Rossetti, Antonio,
saving technologies in swiss firms: an analysis based on micro data. Environ. 2018. Non-technological barriers to the diffusion of energy-efficient HVAC&R
Resour. Econ. 54 (3), 389e417. solutions in the food retail sector. Int. J. Refrig. 86, 422e434.
Axsen, Jonn, Kurani, Kenneth S., 2012. Social influence, consumer behavior, and Suki, Norazah Mohd, Suki, Norbayah Mohd, Azman, Nur Shahirah, 2016. Impacts of
low-carbon energy transitions. Annu. Rev. Environ. Resour. 37, 311e340. corporate social responsibility on the links between green marketing awareness
Badruddin, 2015. Role of NGOs in the protection of environment. J. Environ. Res. and consumer purchase intentions. Procedia Economics and Finance 37,
Dev. 9(3), 705. 262e268.
Bass, Frank M., 1969. A new product growth for model consumer durables. Manag. Patterson, Marvin L., 1998. From experience: linking product innovation to business
Sci. 15 (5), 215e227. growth. J. Prod. Innovat. Manag.: AN INTERNATIONAL PUBLICATION OF THE
Bi, Gongbing, Jin, Minyue, Ling, Liuyi, Yang, Feng, 2017. Environmental subsidy and PRODUCT DEVELOPMENT & MANAGEMENT ASSOCIATION 15 (5), 390e402.
the choice of green technology in the presence of green consumers. Ann. Oper. Radomes Jr., Amando A., Santiago, Arango, 2015. Renewable energy technology
Res. 255 (1e2), 547e568. diffusion: an analysis of photovoltaic-system support schemes in medellín,
Bratanova, Alexandra, Robinson, Jacqueline, Wagner, Liam, 2016. New technology Colombia. J. Clean. Prod. 92, 152e161.
adoption for Russian energy generation: what does it cost? a case study for Reddy, B Sudhakara, 2018. Economic dynamics and technology diffusion in indian
moscow. Appl. Energy 162, 924e939. power sector. Energy Pol. 120, 425e435.
Cagno, Enrico, Trianni, Andrea, 2014. Evaluating the barriers to specific industrial Robinson, Scott A., Rai, Varun, 2015. Determinants of spatio-temporal patterns of
energy efficiency measures: an exploratory study in small and medium-sized energy technology adoption: an agent-based modeling approach. Appl. Energy
enterprises. J. Clean. Prod. 82, 70e83. 151, 273e284.
Christensen, Clayton, 2013. The Innovator’s Dilemma: when New Technologies Shi, Yingying, 2019. Botang Han, and Yongchao Zeng. Simulating policy in-
Cause Great Firms to Fail. Harvard Business Review Press. terventions in the interfirm diffusion of low-carbon technologies: an agent-
Coates, Adam, 2018. Liangxiu han, and anthony kleerekoper. A unified framework based evolutionary game model. J. Clean. Prod. 119449.
for opinion dynamics. Technical report. www.ifaamas.org. Sultan, Fareena, Farley, John U., Lehmann, Donald R., 1990. A meta-analysis of ap-
Deffuant, Amblard, Weisbuch, et al., 2002. How can extremism prevail? A study plications of diffusion models. J. Market. Res. 70e77.
based on the relative agreement interaction model. J. Artif. Societies Soc. Sim. 5 Trianni, Andrea, Cagno, Enrico, Worrell, Ernst, 2013. Innovation and adoption of
(4). energy efficient technologies: an exploratory analysis of Italian primary metal
Deffuant, Guillaume, Huet, Sylvie, Amblard, Fre deric, 2005. An individual-based manufacturing smes. Energy Pol. 61, 430e440.
model of innovation diffusion mixing social value and individual benefit. Am. Wilensky, U., 1999. Netlogo. http://ccl.northwestern.edu/netlogo/. center for
J. Sociol. 110 (4), 1041e1069. connected learning and computer-based modeling.
Degirmenci, Kenan, Breitner, Michael H., 2017. Consumer purchase intentions for Xu, Xiaoping, He, Ping, Xu, Hao, Zhang, Quanpeng, 2017. Supply chain coordination
electric vehicles: is green more important than price and range? Transport. Res. with green technology under cap-and-trade regulation. Int. J. Prod. Econ. 183,
Transport Environ. 51, 250e260. 433e442.
Dosi, Giovanni, 1982. Technological paradigms and technological trajectories: a Yang, G., 2005. Environmental NGOs and institutional dynamics in China. China
suggested interpretation of the determinants and directions of technical Qtly. 181, 46e66.
change. Res. Pol. 11 (3), 147e162. https://doi.org/10.1016/0048-7333(82)90016- Zeng, Yongchao, Dong, Peiwu, Shi, Yingying, Yang, Li, 2018. On the disruptive
6. URL. http://www.sciencedirect.com/science/article/pii/0048733382900166. innovation strategy of renewable energy technology diffusion: an agent-based
ISSN 0048-7333. model. Energies 11 (11). https://doi.org/10.3390/en11113217. URL. http://www.
Fan, Zhi-Ping, Che, Yu-Jie, Chen, Zhen-Yu, may 2017. Product sales forecasting using mdpi.com/1996-1073/11/11/3217. ISSN 1996-1073.
online reviews and historical sales data: a method combining the Bass model Zhang, Lupeng, Xue, Lan, Zhou, Yuan, 2019. How do low-carbon policies promote
and sentiment analysis. J. Bus. Res. 74, 90e100. https://doi.org/10.1016/ green diffusion among alliance-based firms in China? an evolutionary-game
J.JBUSRES.2017.01.010. URL. https://www.sciencedirect.com/science/article/abs/ model of complex networks. J. Clean. Prod. 210, 518e529.
pii/S0148296317300231. ISSN 0148-2963.

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy