Professional Documents
Culture Documents
Using Social Media Mining and PLS-SEM To Examine The Causal Relationship Between Public Environmental Concerns and Adaptation Strategies
Using Social Media Mining and PLS-SEM To Examine The Causal Relationship Between Public Environmental Concerns and Adaptation Strategies
Environmental Research
and Public Health
Article
Using Social Media Mining and PLS-SEM to Examine
the Causal Relationship between Public Environmental
Concerns and Adaptation Strategies
Chia-Lee Yang 1 , Chi-Yo Huang 2, * and Yi-Hao Hsiao 1
1 National Center for High-Performance Computing, Hsinchu City 300, Taiwan; [email protected] (C.-L.Y.);
[email protected] (Y.-H.H.)
2 Program of Technology Management, Department of Industrial Education, National Taiwan Normal University,
Taipei 106, Taiwan
* Correspondence: [email protected]; Tel.: +886-2-77493357
Abstract: With growing scientific evidence showing the harmful impact of air pollution on the
environment and individuals’ health in modern societies, public concern about air pollution has
become a central focus of the development of air pollution prevention policy. Past research has
shown that social media is a useful tool for collecting data about public opinion and conducting
analysis of air pollution. In contrast to statistical sampling based on survey approaches, data
retrieved from social media can provide direct information about behavior and capture long-term
data being generated by the public. However, there is a lack of studies on how to mine social
media to gain valuable insights into the public’s pro-environmental behavior. Therefore, research
is needed to integrate information retrieved from social media sites into an established theoretical
framework on environmental behaviors. Thus, the aim of this paper is to construct a theoretical
Citation: Yang, C.-L.; Huang, C.-Y.;
model by integrating social media mining into a value-belief-norm model of public concerns about
Hsiao, Y.-H. Using Social Media
air pollution. We propose a hybrid method that integrates text mining, topic modeling, hierarchical
Mining and PLS-SEM to Examine the
cluster analysis, and partial least squares structural equation modelling (PLS-SEM). We retrieved
Causal Relationship between Public
data regarding public concerns about air pollution from social media sites. We classified the topics
Environmental Concerns and
Adaptation Strategies. Int. J. Environ.
using hierarchical cluster analysis and interpreted the results in terms of the value-belief-norm
Res. Public Health 2021, 18, 5270. theoretical framework, which encompasses egoistic concerns, altruistic concerns, biospheric concerns,
https://doi.org/10.3390/ijerph18105270 and adaptation strategies regarding air pollution. Then, we used PLS-SEM to confirm the causal
relationships and the effects of mediation. An empirical study based on the concerns of Taiwanese
Academic Editor: Paul B. Tchounwou social media users about air pollution was used to demonstrate the feasibility of the proposed
framework in general and to examine gender differences in particular. Based on the results of the
Received: 12 April 2021 empirical studies, we confirmed the robust effects of egoistic, altruistic, and biospheric concerns of
Accepted: 13 May 2021 public impact on adaptation strategies. Additionally, we found that gender differences can moderate
Published: 15 May 2021
the causal relationship between egoistic concerns, altruistic concerns, and adaptation strategies. These
results demonstrate the effectiveness of enhancing perceptions of air pollution and environmental
Publisher’s Note: MDPI stays neutral
sustainability by the public. The results of the analysis can serve as a basis for environmental policy
with regard to jurisdictional claims in
and environmental education strategies.
published maps and institutional affil-
iations.
Keywords: environmental concerns; air pollution; social media; value-belief-norm model; partial
least squares technique of structural equation modelling (PLS-SEM); gender difference
Int. J. Environ. Res. Public Health 2021, 18, 5270. https://doi.org/10.3390/ijerph18105270 https://www.mdpi.com/journal/ijerph
Int. J. Environ. Res. Public Health 2021, 18, 5270 2 of 23
estimated seven million deaths are associated with air pollution each year according to the
World Health Organization (WHO), much greater than the number who die from acquired
immune deficiency syndrome (AIDS), tuberculosis, and traffic accidents combined [6]. The
economic burden associated with health damage from ambient air pollution internationally
is colossal, estimated to be $5.7 trillion, or 4.8% of global gross domestic product (GDP)
annually, according to a 2020 World Bank report [7].
In response to air pollution, many countries have developed action plans for reducing
air pollution through air quality monitoring and warning systems, the management of
toxic or otherwise potentially harmful chemicals, and energy efficiency and clean energy,
among other measures. These action plans often aim to modify individual behavior that
reduces air pollution, such as encouraging people to consume green and clean products [8].
Previous studies have shown that public environmental concerns are the main function that
is undertaken with the intention to change the environment [9,10]. People who are more
likely to participate in activities to reduce air pollution are those who have been concerned
about and are aware of problems related to air pollution [11]. Thus, public awareness and
support are important for environmental policies designed to resolve air pollution issues.
Public perception of the environment is both a social and an environmental issue.
Recently, numerous scholars have investigated the relationships among public perceptions
and air pollution [12]. The theories of reasoned action [13] and planned behavior [14], the
norm-activation model [15], and value-belief-norm theory [10] were adopted to explain
the relationship between individual beliefs and their pro-environmental behavior, every
probable action intended for protecting the environment. Most research, e.g., [16–19], has
applied these theories to analysis of the relationship between attitudes and behaviors
among survey respondents.
Although statistical sampling methods may have good data quality and provide
powerful support for a theoretical framework, the sampling data sometimes involves
biased selection, and the results may not be generalizable to the entire population. In
addition, stratified sampling captures a wide variety of participant characteristics, but may
not be representative of the underlying population [20]. Mining social media may provide
an opportunity to survey the population directly and retrieve public opinions in real time
and on a massive scale [21].
In comparison with data gathered through approaches based on statistical sampling,
data retrieved from social media can be used to monitor or directly track information
being generated by the public in real time [22]. It also allows for the tracking of individual
interactions with society on environmental issues. Recent studies have used social media to
monitor trends regarding air quality as well as related public attitudes and responses [21,23].
However, data from social media is relatively messy and in an unstructured, raw format.
To the best of our knowledge, no study thus far has mined social media data content on air
pollution and interpreted this data through environmental concern theories. Investigating
the causal relationship between environmental concerns and psychological adaptation
toward air pollution is important for incorporating environmental values retrieved from
social media and behavior in a causal chain.
Thus, a theoretical framework is needed to integrate information retrieved from
social media sites into an established theoretic framework of environmental behaviors.
Therefore, in this paper, we aim to construct a theoretical model by integration of social
media mining into a value-belief-norm theory model of the public’s concerns about air
pollution. Additionally, there is growing epidemiologic evidence of differing associations
between air pollution and respiratory health for men and women [24,25]. Women are also
more aware of environmental hazards that often motivate social action than men because
significant differences exist between the genders in terms of knowledge, awareness, and
attitudes [26,27]. Considerations of gender differences in terms of air-quality issues provide
new insights for researchers and policy makers. Thus, we also integrate gender differences
into the value-belief-norm theory model as a moderating variable.
Int. J. Environ. Res. Public Health 2021, 18, 5270 3 of 23
To retrieve information from social media and confirm the value-belief-norm model,
we propose a hybrid method that integrates text mining, topic modeling, hierarchical
cluster analysis, and partial least squares structural equation modelling (PLS-SEM). We
retrieved data regarding public concerns about air pollution from social media sites. We
classified the topics using hierarchical cluster analysis and interpreted the results in terms
of the value-belief-norm theoretical framework, which encompasses egoistic concerns,
altruistic concerns, biospheric concerns, and adaptation strategies regarding air pollution.
Then, we used PLS-SEM to confirm the causal relationships and the effects of mediation. An
empirical study based on the concerns of Taiwanese social media users about air pollution
was used to demonstrate the feasibility of the proposed framework in general and to
examine gender differences in particular.
2. Literature Review
Environmental concern is the degree to which people are aware of environmental
problems and support efforts to solve these problems and/or indicate a willingness to con-
tribute solutions to such problems [28]. Environmental concern could be the beliefs, affects,
and behavioral intentions a person holds regarding activities or issues related to the envi-
ronment [29]. There is wider recognition that environmental concerns are the main drivers
of the intention to change the environment—that is, to engage in pro-environmental behav-
ior [9]. The value-belief-norm theory was first established by Stern et al. [30] to explain the
influence of human values on behavior in an environmentalist context. Thus, in Section 2.1,
we first review environmental concerns and the value-belief-norm theory. Public concern
and support are one of the most important resources of social movements [30]. There is a
consensus in the literature regarding the strong linkages between public perceptions of air
pollution and emotional and behavioral responses to air pollution [12,31]. Thus, following
the review of the value-belief-norm theory, we review public perceptions of air pollution.
In the past years, social media has emerged as one of the most important sources for
monitoring and detecting environmental problems in general and air pollution problems
in particular. In Section 2.3, we review recent advances in adopting social media for the
detection of air pollution. Then, we review recent research and governmental efforts to con-
trol air pollution in Taiwan, the target of the empirical study. Finally, Section 2.5 will define
the hypotheses based on the path model linking environmental concerns—which include
egoistic, altruistic, and biospheric concerns based on the value-belief-norm theory—to
adaptation for air pollution.
about adverse consequences and posited relationships between values, beliefs, norms, and
behaviors in a causal chain. According to the value-belief-norm theory, people’s attitudes
about environmental issues are determined by the value that they place on themselves
(egoistic value), other people (social-altruistic value), and plants and animals (biospheric
values) [10]. The basic components of the value-belief-norm theory are the three value
orientations regarding environmental concerns, which include egoistic, altruistic, and
biospheric concerns.
Egoistic concerns can be reflected in concerns about environmental problems at the
individual level. These concerns include personal health, financial well-being, quality of
life, availability of resources, and so on. Altruistic concerns, also called social-altruistic
concerns, focus on people other than oneself, including friends, family, community, future
generations, or humanity. People with social-altruistic environmental attitudes care about
environmental problems because the problems affect other people. People with biospheric
concerns focus on all living things, which include plants, animals, ecosystems, and the
biosphere. The value-belief-norm theory has been successfully applied to predicting
people’s pro-environmental behavior in a broad range of circumstances. For example,
De Dominicis et al. [33] examined the different values and found that these values had
different effects on pro-environmental behaviors in the cases of conserving energy, using
public transportation, and participating in a beach clean-up event. Imaningsih et al. [34]
analyzed the influence of egoistic, altruistic and biospheric values on green functional
benefits, green monetary cost, green satisfaction, and green loyalty. They found that egoistic
and supply chain values have a positive effect on green loyalty, while altruistic values do
not affect green loyalty.
disaster risk reduction policy, which may be a critical focus for future collaborative and
applied research [45].
Hypothesis 1 (H1). Altruistic concerns about air pollution are positively associated with egoistic
concerns about air pollution.
Hypothesis 2 (H2). Altruistic concerns about air pollution are positively associated with biosphere
concerns about air pollution.
Hypothesis 3 (H3). Egoistic concerns about air pollution are positively associated with air
pollution adaptation strategies.
Several studies have found that altruistic concerns are stronger among people who
are engaged in pro-environmental activities [60]. Endorsing altruistic values may lead to
perceived environmental stress and coping [51]. As with most prosocial behavior, pro-
environmental behavior has inherent characteristics of altruism [61]. Several studies have
found that altruistic values are stronger among people who engage in pro-environmental
activities [62]. Further research found that altruistic values have a positive impact on
consumers’ attitudes and purchase intentions towards eco-friendly packaged products [63].
Thus, we hypothesize:
Hypothesis 4 (H4). Altruistic concerns about air pollution are positively associated with air
pollution adaptation strategies.
Several studies have found that biospheric concerns are connected with the pro-
environmental behavior intension. Nguyen et al. [64] found biospheric value to encour-
age active engagement in pro-environmental purchase behavior by enhancing consumers’
attitudes towards environmental protection, their subjective norms and environmental
self-identity, and by mitigating their perceived inconvenience associated with eco-friendly
Several studies have found that biospheric concerns are connected with the pro-en-
vironmental behavior intension. Nguyen et al. [64] found biospheric value to encourage
active engagement in pro-environmental purchase behavior by enhancing consumers’ at-
titudes towards environmental protection, their subjective norms and environmental7 of
Int. J. Environ. Res. Public Health 2021, 18, 5270
self-
23
identity, and by mitigating their perceived inconvenience associated with eco-friendly
products. Kiatkawsin et al. [65] found the biospheric values showed a strong influence on
consumers’ likelihood of choosing sustainable products. Imaningsih et al. [34] found the
products.
egoistic andKiatkawsin
biosphericetvalues
al. [65]tofound
have the biospheric
a positive values
effect showed
on green a strong
loyalty. Thus,influence
we hypoth-on
consumers’
esize: likelihood of choosing sustainable products. Imaningsih et al. [34] found the ego-
istic and biospheric values to have a positive effect on green loyalty. Thus, we hypothesize:
Hypothesis 5 (H5). Biosphere concerns about air pollution are positively associated with strate-
Hypothesis to Biosphere
5 (H5).
gies for adaptation concerns about air pollution are positively associated with strategies
air pollution.
for adaptation to air pollution.
Previous studies show that women tended to show higher levels of concern about
Previous studies
environmental problemsshowthanthat
menwomen tended
[60,66]. Mersha toand
showvanhigher levels[67]
Laerhoven of concern
proposed about
that
environmental
gender differences problems
exist than
in themen [60,66].
barriers to Mersha and of
adaptation van Laerhoven
responses to [67] proposed
climate changethatin
gender
Ethiopia. differences exist inetthe
Vicente-Molina al. barriers
[68] foundto adaptation
that womenofare responses to climate change
more altruistically motivated in
Ethiopia.
and behave Vicente-Molina et al. [68]
more responsibly foundthe
towards thatenvironment
women are morethan altruistically
men. Liao etmotivated and
al. [48] found
behave more responsibly towards the environment than men. Liao et al.
that women tend to have a higher knowledge of air pollution than men. Thus, we hypoth- [48] found that
women
esize: tend to have a higher knowledge of air pollution than men. Thus, we hypothesize:
Hypothesis
Hypothesis 6a–e (H6a–e). Gender
6a–e (H6a–e). Gender has
has aa moderated
moderated mediation
mediation effect
effect on
on the
the relationships
relationships among
among
egoistic concerns, altruistic concerns, biosphere concerns, and air pollution adaptation strategies.
egoistic concerns, altruistic concerns, biosphere concerns, and air pollution adaptation strategies.
The
The proposed
proposed research
research model
model is
is presented
presentedin
inFigure
Figure1.1.
Egoistic Ge nde r
Conce rns H3
H6c
H1 H6a H6e
H6d
Altruistic Adaptation
Conce rns H4 Strategies
H2 H6b
H5
Biosphe re
Conce rns
Figure 1.
Figure 1. The
The proposed
proposed research
research model.
model.
3. Research Method
To
To retrieve
retrieve information
information from
from social
social media
media and
and confirm
confirm the
the value-belief-norm
value-belief-norm model,
model,
we
we propose a hybrid method integrating the text mining, topic modeling,
propose a hybrid method integrating the text mining, topic modeling, hierarchical
hierarchical
clustering
clustering and
and the
the partial
partial least
least squares
squares structural
structural equation
equation modelling
modelling (PLS-SEM).
(PLS-SEM). We We first
first
used
used the
the Dcard
Dcard (www.dcard.tw;
(www.dcard.tw; accessed
accessed on
on 20 October
October 20, 2020),
2020), one
one of
of the
the largest
largest social
social
media
media in in Taiwan
Taiwan in in which
which all
all the
the messages
messages are
are tagged
tagged by
by gender,
gender, as
as the
the data
data source
source to
to
collect posts and gender information regarding air pollution. Then, the latent Dirichlet
allocation (LDA) technique was used to derive topic models from Dcard. Based on the
probability scores obtained through topic modeling, hierarchical cluster analysis was
performed to identify the themes of the topic clusters, which includes egoistic concerns,
altruistic concerns, biosphere concerns and adaptation strategies regarding air pollution
by using hierarchical clustering. After all the probability scores were populated for all
the topics, the scores were normalized on a scale of 1 to 5. The resulting matrix is the
dataset for PLS-SEM analyses. Finally, the PLS-SEM was introduced to confirm the causal
relationships among the topics, and the mediation effects additionally. Accordingly, the
theoretical framework consisting of egoistic concerns, altruistic concerns, and adaptation
formed to identify the themes of the topic clusters, which includes egoistic concerns, al-
truistic concerns, biosphere concerns and adaptation strategies regarding air pollution by
using hierarchical clustering. After all the probability scores were populated for all the
topics, the scores were normalized on a scale of 1 to 5. The resulting matrix is the dataset
for PLS-SEM analyses. Finally, the PLS-SEM was introduced to confirm the causal rela-
Int. J. Environ. Res. Public Health 2021, 18, 5270 8 of 23
tionships among the topics, and the mediation effects additionally. Accordingly, the the-
oretical framework consisting of egoistic concerns, altruistic concerns, and adaptation
strategies regarding air pollution is confirmed. Additionally, gender differences were con-
sidered and regarding
strategies adopted asair a moderating
pollution isvariable
confirmed.in theAdditionally,
theoretical model.
gender differences were
In this study,
considered we usedasuser-generated
and adopted textual data
a moderating variable to theoretical
in the generate the dataset for quanti-
model.
tative analysis as well
In this study, we as foruser-generated
used generating thetextual
datasetdata for PLS-SEM.
to generateThe the hybrid
dataset framework
for quantita-
was
tiveinspired
analysisby as previous
well as for research.
generatingSagatheetdataset
al. [69] for
proposed
PLS-SEM.a hybrid framework
The hybrid integrat-
framework was
ing the LDA
inspired and proposed
by previous a process
research. Saga et ofal. analyzing
[69] proposed SEMa hybrid
using LDA. Ray etintegrating
framework al. [70] alsothe
adopted the LDA and the SEM to explore various values affecting
LDA and proposed a process of analyzing SEM using LDA. Ray et al. [70] also adopted thebehavioral intention in
the
LDAcontext of e-learning
and the SEM to exploreservices.
variousBoth studies
values used the
affecting words as
behavioral observation
intention in thevariables
context of
e-learning
and the topicsservices. Both variables
as the latent studies used the words
for SEM. as observation
This paper extends the variables and thetotopics
above method the
as the latent
multi-topic variables
model, usingfor SEM.asThis
topics paper extends
observation the and
variables above method
topic to the
clusters multi-topic
as latent vari-
model,
ables, using topics
avoiding missingas observation variables anddue
too much information topicto clusters as latent
the diverse variables,onavoiding
information social
missing
media. too much information due to the diverse information on social media.
Below,the
Below, themethods
methodswill willbebeintroduced.
introduced.The Thethree
threedata
dataanalytic
analytictechniques,
techniques,namely,
namely,
topicmodelling,
topic modelling, hierarchical
hierarchical cluster
cluster analysis,
analysis, andandthethe
PLSPLS–SEM–SEM methods,
methods, which
which will will
be
be introduced
introduced in theinfollowing
the following subsections
subsections (Sections
(Sections 3.1–3.3),3.1will
andbe3.2), will be
adopted foradopted
retrieving for
retrieving
the problemthe problem
based on thebased onvisualization
idea of the idea of visualization to analyze
to analyze invisible invisible phenomenon
phenomenon as a latent
as a latent
factor in textfactor
data. Thein text data. The
proposed proposed
process consistsprocess
of the consists
following offour
the following
steps (see as four steps
Figure
(see
2): as Figure 2):
Figure
Figure2.2.Research
ResearchFramework.
Framework.
3.1.Text-Mining,
3.1. Text-Mining,Topic
TopicModel
Modeland
andLDA
LDA
Text-mininghas
Text-mining hasbecome
becomea apopular
populartopic
topicacross
acrossa awide
widerange
rangeofoffields
fieldssuch
suchasaspublic
public
health, disaster management, and environmental management [71].
health, disaster management, and environmental management [71]. Text-mining tech- Text-mining tech-
niques are used to extract knowledge to derive meaningful structural
niques are used to extract knowledge to derive meaningful structural information from information from
irregular data patterns or unstructured forms of data to provide meaningful information
irregular data patterns or unstructured forms of data to provide meaningful information
patterns in the shortest time period [72]. Thus, during the past years, text-mining methods
patterns in the shortest time period [72]. Thus, during the past years, text-mining methods
have widely been adopted in numerous research fields such as computational linguistics,
have widely been adopted in numerous research fields such as computational linguistics,
information retrieval, data mining, etc.
information retrieval, data mining, etc.
Topic modelling, a machine learning technique in the field of data mining, is the
Topic modelling, a machine learning technique in the field of data mining, is the most
most popular analytic method for the text-mining of social media. Topic modelling is the
popular analytic method for the text-mining of social media. Topic modelling is the pro-
process of learning, recognizing, and extracting high-level semantic topics across a corpus
cess of learning, recognizing, and extracting high-level semantic topics across a corpus of
of unstructured text. The technique is used to clarify the structure of a group of documents
unstructured text. The technique is used to clarify the structure of a group of documents
by estimating the word-distribution that constitutes a topic based on the premise that each
by estimating the word-distribution that constitutes a topic based on the premise that each
group of documents, which constitutes a corpus belongs to a specific topic [73]. Amongst
group of documents, which constitutes a corpus belongs to a specific topic [73]. Amongst
the techniques, LDA is one of the most popular methods [74].
the techniques, LDA is one
LDA, a generative of the mostmodel
probabilistic popular
of amethods
corpus, is[74].
an unsupervised machine learn-
ing technique that aims to identify the information in is
LDA, a generative probabilistic model of a corpus, an unsupervised
latent machineof
topics in a collection learn-
large
ing technique that aims to identify the information in latent topics in
documents. The basic idea of LDA is that the documents are represented as randoma collection of large
mix-
tures over latent topics, where a topic is characterized by a distribution over words. LDA
treats each document as a vector of words. Each document is represented as a probability
distribution over some topics, while each topic is represented as a probability distribution
over a number of words. Figure 3 shows the graphic model of the LDA where:
α is the per-document topic distributions,
β is the per-topic word distribution,
θd is the topic distribution for document d,
ϕk is the word distribution for topic k,
Zd,n is the topic for the n-th word in document d, and
Wd,n is the specific word.
over a number of words. Figure 3 shows the graphic model of the LDA where:
⍺ is the per-document topic distributions,
β is the per-topic word distribution,
𝜃 is the topic distribution for document d,
𝜑 is the word distribution for topic k,
Int. J. Environ. Res. Public Health 2021, 18, 5270 9 of 23
𝑍 , is the topic for the n-th word in document d, and
𝑊 , is the specific word.
𝛼 𝜃𝑑 𝑧 , 𝑤 , 𝜑 𝛽
N K
D
N: number of words in document d; D: number of documents in corpus; K: number of topics
LDAgraphic
Figure3.3.LDA
Figure graphicmodel.
model.Source:
Source:adapted
adaptedfrom
from[75].
[75].
LDAdefines
LDA definesthe
thefollowing
following generative
generative process
process forfor each
each document
document in the
in the collection:
collection: 1.
1. For each document, pick a topic from its distribution over topics. 2. Sample
For each document, pick a topic from its distribution over topics. 2. Sample a word from a word from
thedistribution
the distributionover
overthe thewords
wordsassociated
associatedwith
withthe
thechosen
chosentopic.
topic.3.3.The
Theprocess
processisisrepeated
repeated
for all the words in the document. The generative process of LDA is as follows:
for all the words in the document. The generative process of LDA is as follows:
For each topic i in (1, K):
Choose per-corpus topic distribution ϕk ∼ Dir ( β)
For each topic i in (1, K):
For each document i in (1, D):
Choose per-corpus topic distribution 𝜑 ~𝐷𝑖𝑟(𝛽)
Choose
For each per-document
document i in (1, D):topic proportion θd ∼ Dir (α)
For each word j in (1, N):
Choose per-document topic proportion 𝜃 ~𝐷𝑖𝑟(⍺)
Choose topic Z ∼ Multinomial (θd ) with Zd,n ∈ [1, K ]
For each word j d,n in (1, N):
Choose word W ∼ Multinomial ( ϕ ) given Z
Choose topic 𝑍 ,d,n~𝑀𝑢𝑙𝑡𝑖𝑛𝑜𝑚𝑖𝑎𝑙(𝜃 ) kwith 𝑍 , d,n ∈ [1, 𝐾]
Choose
3.2. PLS-SEM word 𝑊 , ~𝑀𝑢𝑙𝑡𝑖𝑛𝑜𝑚𝑖𝑎𝑙(𝜑 ) given 𝑍 ,
The SEM methods are widespread in marketing and management research while
3.2. PLS-SEM
analyzing the cause–effect relations between latent constructs. The PLS-SEM is one of the
SEMThe SEM methods
methods, are widespread
which aims to maximize in marketing
the explainedandvariance
management of theresearch while
dependent an-
latent
alyzing
construct [76]. As with other SEM techniques (e.g., the linear structural relations,the
the cause–effect relations between latent constructs. The PLS-SEM is one of the
SEM methods,
LISREL), which
the PLS aims to
approach maximize
allows the explained
researchers variance of
to simultaneously the dependent
assess latentof
the parameters
construct [76]. As with
the measurement modelother
andSEM techniques (e.g.,
the coefficients of thethe linear structural
structural path. Therelations, the LIS-
covariance-based
REL), the PLS approach allows researchers to simultaneously assess
SEM techniques, such as the LISREL and the EQS, use a maximum likelihood function the parameters of the
measurement model and
to obtain estimators in the coefficients
models. of the
Instead, thestructural
PLS-SEMpath. usesThe covariance-based
a least-squares SEM
estimation
techniques,
procedure.such as the avoids
PLS-SEM LISRELmany
and the EQS, useassumptions
restrictive a maximum likelihood
that underlie function to obtain
the covariance-
estimators
based SEMintechniques,
models. Instead,
such asthe PLS-SEM uses
multivariate a least-squares
normality and large estimation
sample size. procedure.
PLS-SEM
PLS-SEM avoids many restrictive assumptions that underlie the covariance-based
analyzes various relationships among several factors, i.e., latent and observed variables. SEM
techniques,
A latent variable is an invisible concept for a target analysis. An observed variable var-
such as multivariate normality and large sample size. PLS-SEM analyzes is an
ious relationships
observable among
item from severalanalysis
a target factors, and
i.e., latent
is usedandto observed
estimate avariables. A latentThese
latent variable. var-
iable is an have
variables invisible concept for
relationships, sucha target
as causalanalysis. An observedrelationships.
and co-occurrence variable is an observable
PLS-SEM was
item
usedfrom a target
as the analysis
research methodandinisthis
used to estimate
work becauseathe latent variable.
sample Thesebevariables
size may have
non-normally
relationships,
distributed due such as causal
to the natureand co-occurrence
of the relationships.
data being retrieved PLS-SEM
from social mediawas used as the
websites.
research method in this work because the sample size may be non-normally distributed
4. Social
due to the Media
nature Mining andbeing
of the data Theoretical
retrievedFramework
from socialConfirmation
media websites. Results
This section presents the process of social media mining and the process of confirming
the theoretical framework being developed in Section 2.5. Based on the proposed hypothe-
ses, we collected data from the largest anonymous social media platform in Taiwan, Dcard,
to test the research model. With over four million registered members, averaging 15 million
unique monthly visitors and 1.5 billion monthly views, Dcard is a wildly popular site
where students and young people go to talk about life, work, politics and social issues. All
the registered members are required to register the gender data and tag on each message.
Thus, the website is very suitable for investigating gender differences in specific domains.
2020, but the messages were written as far back as 2016, when the first messages related
to air pollution were posted on Dcard. The messages are from chat rooms about topics
such as mood, chats, science, news, computers, communication and consumer electronics
products, beauty, life, pets, online shopping, cars, and so on. The text preprocessing of
removing punctuation, common “stop words”, and infrequent words, and performing
Chinese word segmentation were carried out using a program we coded in Python 3.7
(https://www.python.org/downloads/release/python-370/; accessed on 7 June 2020).
The data consists of 3700 messages collected in Dcard. After the data cleaning process
of removing duplicates, errors, and messages unrelated to air pollution, the final sample
consisted of 1043 messages, of which 60% were posted by the male population and 40%
were posted by the female population.
Many researchers have proven that useful information and knowledge can be obtained
from social media data [77]. Although social media mining can extract hidden information
in the text, the social media content is usually affected by its typically noisy nature, such
as abbreviations and/or irregular forms. A rigorous data quality assurance procedure is
needed to reduce the noise of textual data and validity of data quality. To this end, we
applied a four-step procedure in this research. The first step was to eliminate the short
messages, which are unlikely to contain any mentions of air pollution. We calculated the
average messages’ length, which was 482 words in this study, and removed the posts that
were less than 25 words. Those posts being removed were confirmed again by the authors
as useless for further analyses. The second step was to delete meaningless non-words such
as the website URLs, emojis, and special characters [78]. The third step was to remove
unsuitable keywords, such as stop words, by all the authors from the results of topic model
execution, and iterate this step λ times (five times in this study) to assure suitable and
reliable data quality [79]. The fourth step was to assess the validity and reliability of a
measurement model in PLS-SEM. All the selected messages were reviewed by the authors
to filter out irrelevant messages.
Count
Latent Variables Item Code Item Name Top Five Keywords Belonging to Each Topic
(Men/Women)
t3 Fuel 55 (36/19) U.S., Taiwan, natural, fuel, smoking forbidden
t4 Mask 146 (99/47) air, air pollution, air quality, mask, research
Egoistic Concerns
e-cigarette, tobacco, Taiwan, harm
t5 E-cigarette 50 (36/14)
reduction, cigarette
smokes, cigarette smoke, cigarette butts,
t9 Smoking 140 (83/57)
smells, school
Coal-fired power shen’ao power plant, air poolution, govermenal,
t6 61 (42/19)
Altruistic generation EPA( *) , coal burning
Concerns t7 Refuse combustion 62 (39/23) air, garbage, earth, burning, joss paper
tai-power, power plant, power unit, generator
t11 Power generation 68 (50/18)
set, gas
plebiscite, green with nuclear, nuclear,
Biophere t1 Policy ambiguity 30 (26/4)
gavernment, cosignatory
Concerns
climate, energy, global, climate change,
t8 Climate change 83 (63/20)
renewable energy
Wind power Taiwan, wind power, offshore wind power,
t2 64 (39/25)
Adaptation generation policy polar bear,
Strategies t10 Allergy and healthy 165 (56/109) allergy, nose, dortor, feel
t12 Air purifier products 119 (52/67) air purifier, allergy, recommad, air filter
Remark: * EPA is the abbreviation of the Environment Protection Agency, Taiwan.
results in Table A1 also demonstrate good convergent validity and reliability because the
AVE values range from 0.592 to 0.682, which is larger than the threshold value of 0.500
recommended by Bagozzi and Yi [82]. Hence, the convergent validity of the measurement
model is acceptable.
This study uses two common approaches, the Fornell–Larcker criterion and the cross-
loading criterion [83], to examine discriminant validity. The Fornell–Larcker criterion was
examined by comparing the square root of AVE with the correlations between the focal
construct and all the other constructs. In this dataset, all the variables can fulfill the Fornell–
Larcker criterion because the square roots of each AVE are higher than the correlations
between the other latent variables (see Table A2), which demonstrates adequate discrimi-
nant validity for all dimensions. Moreover, discriminant validity can be further verified by
using the cross-loadings assessment proposed by Chin [84]. Table A3 demonstrates that
each of the factor loadings (digits in bold) is greater than all of the cross-loadings.
the factor of adaptation strategies, which is generally considered a moderate effect. The
variances of EC and BC are under 50.400%, which could be considered a weak effect.
For the results of hypothesis testing, H1 was supported: altruistic concerns were
significantly related to egoistic concerns (β = 0.678, p = 0.000). For H2, altruistic concerns
were significantly related to biospheric concerns (β = 0.667, p = 0.000). For H3, egoistic
concerns were significantly related to adaptation strategies (β = 0.351, p = 0.000). H4
predicted an effect of altruistic concerns on adaptation strategies, and this effect was also
significant (β = 0.277, p = 0.000). Finally, for H5, biospheric concerns were significantly
related to adaptation strategies (β = 0.177, p = 0.000). Thus, the PLS-SEM results fully
supported H1, H2, H3, H4, and H5. The hypothesis testing results are presented in Table 3.
Hypotheses Results
H1 (AC → EC) Supported
H2 (AC → BC) Supported
H3 (EC → AS) Supported
H4 (AC → AS) Supported
H5 (BC → AS) Supported
Regarding the implications of the research results, the direct, indirect, and total rela-
tionships with the criteria within each dimension are shown in Figure 2 and Table 4. Here,
the aspect of egoistic concerns (β = 0.351) appears to be the most correlated aspect of AS
while the aspect of altruistic concerns (β = 0.277) is the second. Furthermore, the correlations
between altruistic concerns and adaptation strategies can be derived through three paths:
AC → EC → AS (β = 0.678 × 0.351 = 0.238), AC → BC → AS (β = 0.667 × 0.177 = 0.118),
and AC → AS (β = 0.277). We tested the role of egoistic concerns and biospheric con-
cerns in mediating the relationship between altruistic concerns and adaptation strategies
by using the variance accounted for (VAF). The VAF for egoistic concerns was 0.512%
((0.678 × 0.351)/(0.678 × 0.351 + 0.277) × 100% = 0.512%). The VAF of biospheric concerns
is 0.342% ((0.667 × 0.177)/(0.667 × 0.177 + 0.277) × 100%). Both egoistic concerns and
biospheric concerns are generally considered as partial mediation effects according to
Hair Jr. et al. [89].
(1) biosphere concern to adaptation strategies (BC → AS, β = 0.300), (2) egoistic concern
to adaptation strategies (EC → AS, β = 0.238), and (3) altruistic concern to adaptation
strategies (AC → AS, β = 0.166), respectively.
Path Coefficients-Diff
Relationships p-Value (Men–Women)
(Men–Women)
H6a (AC → EC) 0.004 0.540
H6b (AC → BC) 0.107 0.997
H6c (EC → AS) 0.150 0.044 *
H6d (AC → AS) 0.163 0.036 *
H6e (BC → AS) 0.128 0.919
Notes: *: p < 0.050.
Figure 4. Path analysis results for posts by (a) both genders, (b) men, and (c) women. Notes: *: p < 0.050;
Figure **:
4. Path analysis
p < 0.010; ***: presults for posts by (a) both genders, (b) men, and (c) women. Notes: *: p < 0.050; **: p < 0.010; **
< 0.001.
0.001.
5. Discussion
In this study, we attempted to assess individuals’ concerns about air pollution and
their adaptation behaviors; furthermore, the relationships between individuals’ concerns
5. Discussion
and adaptation strategies were assessed. Based on the analytic results, we would like to dis-
In this study, we attempted to assess individuals’ concerns about air pollu
cuss the topics of most concern among Taiwanese, the relationships between environmental
their adaptationimpacts
concerns and adaption strategies, behaviors; furthermore, the
of environmental relationships
concerns between
on adaptation individuals’
strate-
gies, and moderations of gender in the relationship of altruistic and adaptation strategies we wou
and adaptation strategies were assessed. Based on the analytic results,
discuss the topics
in the following sub-sections: Sectionsof most concern among Taiwanese, the relationships between
5.1–5.4.
mental concerns and adaption strategies, impacts of environmental concerns on
tionConcern
5.1. The Topics of Most strategies,
for and moderations of gender in the relationship of altruistic and ad
Taiwanese
strategies in the following
Based on the results of topic modeling sub-sections: Section
(refer Table 5.1, Section
1), twelve topics5.2, Section
have been5.3 and Se
identified. The inconsistencies with prior research will be discussed in this sub-section.
5.1. The Topics of Most Concern for Taiwanese
Based on the results of topic modeling (refer Table 1), twelve topics have be
tified. The inconsistencies with prior research will be discussed in this sub-sectio
et al. [21] analyzed the contents of a mainland Chinese social media site that wer
Int. J. Environ. Res. Public Health 2021, 18, 5270 15 of 23
Wang et al. [21] analyzed the contents of a mainland Chinese social media site that were
related to air quality and air pollution. Based on that work, people’s major concerns regard-
ing air pollution were prioritized as follows: (1) individual reactive behavior, (2) health
concerns, and (3) requested actions to improve air quality such as reducing carbon emis-
sions [21]. In comparison with the analytic results by Wang et al. [21], our study found that
Taiwanese concerns were more about requesting improved actions such as reducing air
pollutant emissions than individual reactive behavior and health risks.
In addition, according to Wang et al. [21], the most widely adopted individual reactive
behaviors of Chinese people are facemask-wearing, clothes-washing and staying indoors.
However, Taiwanese concerns are primarily about masks, air purifiers and e-cigarettes. The
concerns of Taiwanese people about e-cigarettes (t5 ) could be attributed to the emerging
trend of adopting e-cigarettes by teenagers. After enacting the Tobacco Hazards Prevention
Act [90], the Taiwanese government now prohibits smoking in all workplaces and other
public places, and requires warnings to appear prominently on cigarette packages that
encourage people to stop smoking. However, the diffusion rates of e-cigarettes from middle
school to high school students have increased rapidly. Thus, the problems of e-cigarettes
on polluting indoor air quality, chemical compositions, e-cigarette aerosols, and associated
cardiovascular were broadly discussed on social media sites.
From the perspective of the emission-sources of air pollution, the major sources of
outdoor air pollution include cooking and heating, vehicles, power generation, agricul-
ture/waste incineration, and industry [6]. Lam et al. [91] found that the major stakeholders
(e.g., governments, the environmental groups, and the news media) are mainly concerned
about the emissions of pollutants and the control of end-of-pipe pollution in Hong Kong.
The work by Lam et al. [91] indicated that the public was especially focused on the emis-
sions from power plants and vehicles. According to the results of our analyses (refer to
Table 1), five topics are related to energy: policy ambiguity (t1 ), wind power generation
policy (t2 ), coal-fired power generation (t6 ), refuse combustion (t7 ), and power generation
(t11 ). These results may be attributed to the Taiwan’s energy dilemma between anti-nuclear
energy and anti-air pollution from fossil fuel energy.
According to the Energy Statistics of Taiwan [92], about 93% of Taiwanese energy
consumption was from imported fossil fuels. Taiwan has been facing an energy policy
dilemma due to the desirability of multiple goals: achieving a nuclear-free homeland,
reducing air pollution from fossil fuels, and restricting the receiving capacity of liquefied
natural gas terminals. Some people are against nuclear power because of the public anxiety
from the Japan Fukushima Daiichi nuclear disaster in 2011 and problems with storing
nuclear waste. Anti-nuclear activists advocate turning Taiwan into a nuclear-free country
by 2025. Some people are against the coal-fired power generation plants because of the rise
in air pollution causing 6000 air-pollution related deaths per year in Taiwan [93]. Anti-air
pollution activists promote the view that nuclear energy could potentially provide an
alternative form of power generation to decrease the use of coal-fired thermal power. The
debates regarding energy policy have led to political and legal limbo regarding air pollution
in the 2018 Taiwan referendum. Our research results are in line with the realistic social
situation. The results are also consistent with previous studies showing that air pollution is
closely linked with energy selection. For example, Heinrichs [94] examined the phasing
out of coal-fired power plants in Germany from social, macro-economic, and technical
perspectives. Based on the work on social perspective, the prevailing positive attitude
towards a coal phase-out in German society is also reflected in the willingness to accept the
consequences of the transformation process such as being willing to accept wind turbines
near their homes in order to reduce the emission of air pollution. Severnini [95] exploits the
shutdown of nuclear facilities in the Tennessee Valley after the Three Mile Island accident
in 1979 to estimate its direct impact on coal-fired power generation in the context of particle
pollution. The results show that the total suspended particulate concentration increased
significantly as a result of the substantial increase in coal-fired power generation after the
nuclear shutdown.
Int. J. Environ. Res. Public Health 2021, 18, 5270 16 of 23
Taiwan into a nuclear-free homeland in 2025 [99]. The turnout for the referendum was
50 to 55 percent. In this paper, we present evidence that egoistic, altruism and biosphere
concerns can provide pathways to pro-environmental behavior.
According to the analytic results (refer Figure 4a), the path strength of egoistic con-
cern to adaptation strategies (EC → AS, β = 0.351) is stronger than altruistic concern to
adaptation strategies (AC → AS, β = 0.277) and biosphere concern to adaptation strategies
(BC → AS, β = 0.177). The finding indicates that egoistic concern is strongly related to
adaptation strategies. The result is contrary to those of prior research on climate change.
Steg et al. [57] and Helm et al. [51] suggested that the biosphere value is a stronger predictor
of pro-environmental beliefs, attitudes, norms, and actions than either egoistic or altruistic
environmental concerns for climate change-related issues. In our research, individuals
with strong egoistic concerns are most likely to be concerned about air pollution adaptation
strategies. The perception of risk for different environmental problems may provide the
explanation. According to Bickerstaff et al. [100], public understandings of the risks of air
pollution draw upon evidence from: (1) the physical experience of pollution, (2) fewer
substantial forms of evidence, (3) information relayed through social networks, and (4) ob-
served or reported impacts on the wider environment. In the context of Taiwanese concerns
related to air pollution, public egoistic concerns are based on physical experience of pollu-
tion such as second-hand smoke or e-cigarettes. So, individuals are likely to find solutions
to act on. The major altruistic concerns are related to public administration issues based on
news or social media information such as coal-fired power generation. Individuals have
divergent considerations and hesitate to decide on adoption strategies.
Our findings indicate that egoistic concern regarding air pollution is more related to
adaptation strategies than altruistic or biosphere concerns are to adaptive strategies. This
result is also in line with Bain et al. [101], who argued that appealing to societal and self-
interested benefits could be more effective than focusing on ecological outcomes or altruistic
environmental issues, and motivate behavioral change among climate change deniers.
than women. In the education context, this implies that educators can emphasize egoistic
and altruistic concern topics in promoting an adaption strategy to air pollution to increase
female citizens’ concern.
In addition, our study shows that the priorities of men and women regarding which
environmental concerns were most important are different, as shown in Figure 4b,c. These
findings suggest that the gender differences exist in the relationship between air pollution
concerns and action intentions. The results may be explained by the social construct
expectation [104]. Men are socialized to be protectors of and providers for the family
and may portray environmental pollution as a necessary tradeoff for growth. Men are
more likely to link egoistic and altruistic concerns to actions. In contrast, women are more
concerned about biosphere environmental problems because women are socialized to be
family nurturers and caregivers for children and have greater social responsibility and
empathy towards others. Women are more likely to link biosphere concerns to actions [105].
6. Conclusions
Air pollution is one of the highest environmental risks to health and is regarded as
an urgent issue in the world. There is a growing awareness that environment quality has
been greatly threatened by air pollution. Various studies have explored the relationship
between air pollution, health, and the environment. A key, but generally absent, element
of successful environment policy is a high level of awareness, positive attitudes, and
support from citizens. Such policy requirements are in citizens’ interest. Social media
mining has been shown to be a useful tool for public-opinion extraction and analyses of air
pollution. Compared with traditional questionnaire-based surveys, social media mining
can retrieve knowledge from large unstructured data from booming social media sources.
These social media provide instant public responses and capture long-term data with low
cost. Environmental and social scientists have realized the potential of social media as a
source of data and have conducted numerous studies regarding public opinions. These
studies have examined messages on social media regarding air pollution. However, how to
successfully exploit big data being retrieved from social media and gain valuable insights
into public pro-environmental behaviors still needs further investigation. In pursuit of this
aim, this study has not only analyzed the content of social media but has also transformed
empirical validations of theoretical concepts in environmental research.
The contributions of this research are twofold. First, the combination of theoretical
and practical considerations in this research has important implications for both academics
and policy planners in gaining a better understanding of public concerns related to air
pollution. In the past, many scholars have emphasized the importance of studying the
relationships between human values and environmental behaviors. Various scholars argue
that egoistic, altruistic, and biospheric value orientations are important for understanding
environmental beliefs and behaviors [52]. Values may also affect the extent to which people
are aware of environmental problems being associated with their behaviors. Awareness of
consequences will increase if important environmental values are threatened, and people
may adjust their behavior in order to reduce this threat. This research examines the different
effects of egoistic, altruistic, and biospheric concerns regarding air pollution on adaptation
strategies by using new hybrid social media mining and the PLS-SEM technique. The
results provide support for the generalization of the egoistic, altruistic, and biospheric
interactions and their contributions to adaptation strategies. Second, this paper proposes
a novel hybrid data analysis approach combining two kinds of research models: social
media mining based on big data analysis and the PLS-SEM based on multivariate analysis
techniques. We first used the LDA to analyze un-structured text data retrieved from social
media sites. Then, we clustered the public concerns from LDA results into the theoretical
framework based on the value-belief-norm theory. Finally, we confirmed the causal model
using the PLS-SEM. The novel analytic framework provides researchers and policy-makers
with a new approach to social media data analyses.
Int. J. Environ. Res. Public Health 2021, 18, 5270 19 of 23
This research has several practical applications that are worthy of future study. First,
this study suggests that egoistic concerns are a stronger predictor of adaptation strategies
for air pollution rather than altruistic or biospheric concerns. Additional scientific questions
may arise, leading to a greater understanding of the joint effects of egoistic and altruistic
concerns. Second, this study suggests that the moderating effect of gender difference exists
in the paths from egoistic concerns to adaptation strategies and the path from altruistic
concerns to adaptation strategies. Gender is less evident as a part of current air quality
policy in practice, although women have been identified as vulnerable yet crucial players
in leading their neighborhoods towards a safer environment. Seeing air pollution through
women’s eyes reflects women’s specific needs and interests in air pollution. More research
regarding the potential association between air pollution and gender-related differences
would be worthy of further development and exploration.
This research has two limitations. First, the present study uses the social media site
Dcard as the data source, but Dcard users may not be representative of the population at
large, as social media sites such as Dcard may be more widely adopted by the younger
population. Second, most social media posts may be from subgroups who are likely to
suffer from problems of representativeness [106]. Most of the social media posts examined
in this study were related to personal or social interests. The posts in general and by women
in particular that were related to biosphere concerns were relatively rare. For future studies,
linking social data with other trusted information sources—for example, survey data or air
pollution monitoring data—may be helpful for confirming theoretical frameworks from
other aspects.
Author Contributions: C.-L.Y. wrote the draft. C.-Y.H. revised and finalized the work. C.-L.Y. and
Y.-H.H. analyzed the data from social media site. All authors have read and agreed to the published
version of the manuscript.
Funding: This research was granted by MOST, Taiwan (MOST107-2629-M-492-001-MY2).
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: The data set generated in this study are not publicly available due to
the service-term restrictions of the social media platform.
Conflicts of Interest: The authors declare no conflict of interest.
Appendix A
Latent Variables EC AC BC AS
EC 0.769
AC 0.678 0.790
BC 0.611 0.667 0.826
AS 0.647 0.633 0.576 0.806
Note: The square roots of AVEs are shown diagonally in bold.
Variables
Topics EC AC BC AS
References
1. Hamidi, A.; Ramavandi, B. Evaluation and scientometric analysis of researches on air pollution in developing countries from
1952 to 2018. Air Qual. Atmos. Health 2020, 13, 797–806. [CrossRef]
2. Bailey, M.J.; Naik, N.N.; Wild, L.E.; Patterson, W.B.; Alderete, T.L. Exposure to air pollutants and the gut microbiota: A potential
link between exposure, obesity, and type 2 diabetes. Gut Microbes 2020, 11, 1188–1202. [CrossRef] [PubMed]
3. Jung, C.-R.; Hsieh, H.-Y.; Hwang, B.-F. Air pollution as a potential determinant of rheumatoid arthritis: A population-based
cohort study in Taiwan. Epidemiology 2017, 28, S54–S59. [CrossRef] [PubMed]
4. Lee, P.-C.; Liu, L.-L.; Sun, Y.; Chen, Y.-A.; Liu, C.-C.; Li, C.-Y.; Yu, H.-L.; Ritz, B. Traffic-related air pollution increased the risk of
Parkinson’s disease in Taiwan: A nationwide study. Environ. Int. 2016, 96, 75–81. [CrossRef]
5. Park, J.; Kim, H.-J.; Lee, C.-H.; Lee, C.H.; Lee, H.W. Impact of long-term exposure to ambient air pollution on the incidence of
chronic obstructive pulmonary disease: A systematic review and meta-analysis. Environ. Res. 2021, 194, 110703. [CrossRef]
6. World Health Organization. Air Pollution. Available online: https://www.who.int/health-topics/air-pollution#tab=tab_1
(accessed on 1 April 2021).
7. World Bank. Pollution. Available online: https://www.worldbank.org/en/topic/pollution (accessed on 1 April 2021).
8. Parveen, R.; Ahmad, A. Public behavior in reducing urban air pollution: An application of the theory of planned behavior in
Lahore. Environ. Sci. Pollut. Res. 2020, 27, 17815–17830. [CrossRef]
9. Snelgar, R.S. Egoistic, altruistic, and biospheric environmental concerns: Measurement and structure. J. Environ. Psychol. 2006, 26,
87–99. [CrossRef]
10. Stren, P. Toward a coherent theory of environmentally significant behaviour. J. Soc. Issues 2000, 56, 407–424. [CrossRef]
11. Chin, Y.S.J.; De Pretto, L.; Thuppil, V.; Ashfold, M.J. Public awareness and support for environmental protection—A focus on air
pollution in peninsular Malaysia. PLoS ONE 2019, 14, e0212206. [CrossRef] [PubMed]
12. Sawitri, D.R.; Hadiyanto, H.; Hadi, S.P. Pro-environmental behavior from a socialcognitive theory perspective. Procedia Environ.
Sci. 2015, 23, 27–33. [CrossRef]
13. Ajzen, I.; Fishbein, M. Understanding Attitudes and Predicting Social Behavior; Prentice-Hall: Englewood Cliffs, NJ, USA, 1980.
14. Ajzen, I. The theory of planned behavior. Organ. Behav. Hum. Decis. Process. 1991, 50, 179–211. [CrossRef]
15. Schwartz, S.H. Normative influences on altruism. Adv. Exp. Soc. Psychol. 1977, 10, 221–279.
16. Ramírez, O.; Mura, I.; Franco, J.F. How do people understand urban air pollution? Exploring citizens’ perception on air quality,
its causes and impacts in Colombian cities. Open J. Air Pollut. 2017, 6, 1–17. [CrossRef]
17. Tvinnereim, E.; Liu, X.; Jamelske, E.M. Public perceptions of air pollution and climate change: Different manifestations, similar
causes, and concerns. Clim. Chang. 2017, 140, 399–412. [CrossRef]
18. Oltra, C.; Sala, R. Perception of risk from air pollution and reported behaviors: A cross-sectional survey study in four cities. J.
Risk. Res. 2018, 21, 869–884. [CrossRef]
Int. J. Environ. Res. Public Health 2021, 18, 5270 21 of 23
19. Pu, S.; Shao, Z.; Fang, M.; Yang, L.; Liu, R.; Bi, J.; Ma, Z. Spatial distribution of the public’s risk perception for air pollution: A
nationwide study in China. Sci. Total Environ. 2019, 655, 454–462. [CrossRef]
20. Bornstein, M.H.; Jager, J.; Putnick, D.L. Sampling in developmental science: Situations, shortcomings, solutions, and standards.
Dev. Rev. 2013, 33, 357–370. [CrossRef]
21. Wang, S.; Paul, M.J.; Dredze, M. Social media as a sensor of air quality and public response in China. J. Med. Internet Res. 2015, 17,
e22. [CrossRef] [PubMed]
22. Hswen, Y.; Qin, Q.; Brownstein, J.S.; Hawkins, J.B. Feasibility of using social media to monitor outdoor air pollution in London,
England. Prev. Med. 2019, 121, 86–93. [CrossRef] [PubMed]
23. Guo, B.; Chen, C.; Zhang, D.; Yu, Z.; Chin, A. Mobile crowd sensing and computing: When participatory sensing meets
participatory social media. IEEE Commun. Mag. 2016, 54, 131–137. [CrossRef]
24. Westergaard, N.; Gehring, U.; Slama, R.; Pedersen, M. Ambient air pollution and low birth weight-are some women more
vulnerable than others? Environ. Int. 2017, 104, 146–154. [CrossRef] [PubMed]
25. Clougherty, J.E. A growing role for gender analysis in air pollution epidemiology. Environ. Health Perspect. 2010, 118, 167–176.
[CrossRef] [PubMed]
26. López-Mosquera, N. Gender differences, theory of planned behavior and willingness to pay. J. Environ. Psychol. 2016, 45, 165–175.
[CrossRef]
27. Tenouri, N.F. Understanding the “Who” in Conservation: Why Gender Matters; University of Otago: Dunedin, New Zealand, 2020.
28. Dunlap, R.E.; Jones, R.E. Environmental concern: Conceptual and measurement issues. In Handbook of Environmental Sociology;
Dunlap, R.E., Michelson, W., Eds.; Greenwood: Westport, CT, USA; London, UK, 2002; pp. 482–524.
29. Schultz, P.W.; Gouveia, V.V.; Cameron, L.D.; Tankha, G.; Schmuck, P.; Franěk, M. Values and their relationship to environmental
concern and conservation behavior. J. Cross Cult. Psychol. 2005, 36, 457–475. [CrossRef]
30. Stern, P.C.; Dietz, T.; Abel, T.; Guagnano, G.A.; Kalof, L. A value-belief-norm theory of support for social movements: The case of
environmentalism. Hum. Ecol. Rev. 1999, 6, 81–97.
31. Duan, W.; Sheng, J. How can environmental knowledge transfer into pro-environmental behavior among Chinese individuals?
Environmental pollution perception matters. J. Public Health 2018, 26, 289–300. [CrossRef]
32. Schwartz, S.H.; Howard, J.A. A Normative Decision-Making Model of Altruism; Rushton, J.P., Ed.; Erlbaum: Hillsdale, NJ, USA, 1981;
pp. 189–211.
33. De Dominicis, S.; Schultz, P.; Bonaiuto, M. Protecting the environment for self-interested reasons: Altruism is not the only
pathway to sustainability. Front. Psychol. 2017, 8, 1065. [CrossRef]
34. Imaningsih, E.S.; Tjiptoherijanto, P.; Heruwasto, I.; Aruan, D.T.H. Linking of egoistic, altruistic, and biospheric values to green
loyalty: The role of green functional benefit, green monetary cost and green satisfaction. J. Asian Financ. Econ. Bus. 2019, 6,
277–286. [CrossRef]
35. Manisalidis, I.; Stavropoulou, E.; Stavropoulos, A.; Bezirtzoglou, E. Environmental and health impacts of air pollution: A review.
Front. Public Health 2020, 8, 14. [CrossRef]
36. Represa, N.S.; Fernández-Sarría, A.; Porta, A.; Palomar-Vázquez, J. Data mining paradigm in the study of air quality. Environ.
Process. 2020, 7, 1–21. [CrossRef]
37. Williams, I.D.; Bird, A. Public perceptions of air quality and quality of life in urban and suburban areas of London. J. Envion.
Monitor. 2003, 5, 253–259. [CrossRef] [PubMed]
38. Reames, T.G.; Bravo, M.A. People, place and pollution: Investigating relationships between air quality perceptions, health
concerns, exposure, and individual-and area-level characteristics. Environ. Int. 2019, 122, 244–255. [CrossRef]
39. Howel, D.; Moffatt, S.; Bush, J.; Dunn, C.E.; Prince, H. Public views on the links between air pollution and health in Northeast
England. Environ. Res. 2003, 91, 163–171. [CrossRef]
40. Chakraborty, J.; Collins, T.W.; Grineski, S.E.; Maldonado, A. Racial differences in perceptions of air pollution health risk: Does
environmental exposure matter? Int. J. Environ. Res. Public Health 2017, 14, 116. [CrossRef] [PubMed]
41. Sansom, G.; Berke, P.; McDonald, T.; Shipp, E.; Horney, J. Evaluating the Impact of Race and Gender on Environmental Risk
Perceptions in the Houston Neighborhood of Manchester. Environ. Justice 2019, 12, 92–98. [CrossRef] [PubMed]
42. Paul, B.K. Women’s awareness of and attitudes towards the Flood Action Plan (FAP) of Bangladesh: A comparative study.
Environ. Manag. 1999, 23, 103–114. [CrossRef] [PubMed]
43. Yen, I.H.; Scherzer, T.; Cubbin, C.; Gonzalez, A.; Winkleby, M.A. Women’s perceptions of neighborhood resources and hazards
related to diet, physical activity, and smoking: Focus group results from economically distinct neighborhoods in a mid-sized US
city. Am. J. Health Promot. 2007, 22, 98–106. [CrossRef]
44. Carvajal-Escobar, Y.; Quintero-Angel, M.; Garcia-Vargas, M. Women’s role in adapting to climate change and variability. Adv.
Geosci. 2008, 14, 277–280. [CrossRef]
45. Gaillard, J.-C.; Gorman-Murray, A.; Fordham, M. Sexual and gender minorities in disaster. Gend. Place Cult. 2017, 24, 18–26.
[CrossRef]
46. McCay-Peet, L.; Quan-Haase, A. What is social media and what questions can social media research help us answer. In The SAGE
Handbook of Social Media Research Methods; Sloan, L., Anabel Quan-Haase, A., Eds.; SAGE Publications: New York, NY, USA, 2017;
pp. 13–26.
Int. J. Environ. Res. Public Health 2021, 18, 5270 22 of 23
47. Mansour, E. The adoption and use of social media as a source of information by Egyptian government journalists. Int. J. Lib. Inf.
Sci. 2018, 50, 48–67. [CrossRef]
48. Liao, X.; Tu, H.; Maddock, J.E.; Fan, S.; Lan, G.; Wu, Y.; Yuan, Z.K.; Lu, Y. Residents’ perception of air quality, pollution sources,
and air pollution control in Nanchang, China. Atmos. Pollut. Res. 2015, 6, 835–841. [CrossRef]
49. Tseng, C.-H.; Tsuang, B.-J.; Chiang, C.-J.; Ku, K.-C.; Tseng, J.-S.; Yang, T.-Y.; Hsu, K.-H.; Chen, K.-C.; Yu, S.-L.; Lee, W.-C. The
relationship between air pollution and lung cancer in nonsmokers in Taiwan. J. Thorac. Oncol. 2019, 14, 784–792. [CrossRef]
[PubMed]
50. Environmental Protection Administration. Taiwan Air Quality Monitoring Network. Available online: https://airtw.epa.gov.tw/
ENG/default.aspx (accessed on 1 April 2021).
51. Helm, S.V.; Pollitt, A.; Barnett, M.A.; Curran, M.A.; Craig, Z.R. Differentiating environmental concern in the context of psycholog-
ical adaption to climate change. Glob. Environ. Chang. 2018, 48, 158–167. [CrossRef]
52. De Groot, J.I.; Steg, L. Value orientations and environmental beliefs in five countries: Validity of an instrument to measure egoistic,
altruistic and biospheric value orientations. J. Cross Cult. Psychol. 2007, 38, 318–332. [CrossRef]
53. De Groot, J.I.; Steg, L. Value orientations to explain beliefs related to environmental significant behavior: How to measure egoistic,
altruistic, and biospheric value orientations. Environ. Behav. 2008, 40, 330–354. [CrossRef]
54. De Groot, J.I.; Steg, L.; Keizer, M.; Farsang, A.; Watt, A. Environmental values in post-socialist Hungary: Is it useful to distinguish
egoistic, altruistic and biospheric values? Sociol. Rev. 2012, 48, 421–440. [CrossRef]
55. Kareklas, I.; Carlson, J.R.; Muehling, D.D. “I eat organic for my benefit and yours”: Egoistic and altruistic considerations for
purchasing organic food and their implications for advertising strategists. J. Advert. 2014, 43, 18–32. [CrossRef]
56. IPCC. Climate Change 2014 Synthesis Report; The Intergovernmental Panel on Climate Change (IPCC): Geneva, Switzerland, 2014;
p. 26.
57. Steg, L.; Bolderdijk, J.W.; Keizer, K.; Perlaviciute, G. An integrated framework for encouraging pro-environmental behaviour: The
role of values, situational factors and goals. J. Environ. Psychol. 2014, 38, 104–115. [CrossRef]
58. Sarpong, K.A.; Amankwaa, G.; Frimpong, O.; Xu, W.; Cao, Y.; Ni, X.; Nkrumah, N.K. Consumers’ purchasing intentions for
efficient water-saving products: The mediating effects of altruistic and egoistic values. J. Water Supply Res. T. 2021, 70, 226–238.
[CrossRef]
59. Aprile, M.C.; Fiorillo, D. Water conservation behavior and environmental concerns: Evidence from a representative sample of
Italian individuals. J. Clean. Prod. 2017, 159, 119–129. [CrossRef]
60. Schultz, P.W. The structure of environmental concern: Concern for self, other people, and the biosphere. J. Environ. Psychol. 2001,
21, 327–339. [CrossRef]
61. van den Bergh, B.; Griskevicius, V.; Tybur, J. Consumer choices: Going green to be seen. RSM Discov. Mgmt. Knowl. 2010, 4, 10–11.
62. Schwartz, S.H. Universals in the content and structure of values: Theoretical advances and empirical tests in 20 countries. Adv.
Exp. Soc. Psychol. 1992, 25, 1–65.
63. Prakash, G.; Choudhary, S.; Kumar, A.; Garza-Reyes, J.A.; Khan, S.A.R.; Panda, T.K. Do altruistic and egoistic values influence
consumers’ attitudes and purchase intentions towards eco-friendly packaged products? An empirical investigation. J. Retail.
Consum. Serv. 2019, 50, 163–169. [CrossRef]
64. Nguyen, T.N.; Lobo, A.; Greenland, S. Pro-environmental purchase behaviour: The role of consumers’ biospheric values. J. Retail.
Consum. Serv. 2016, 33, 98–108. [CrossRef]
65. Kiatkawsin, K.; Han, H. Young Travelers’ intention to behave pro-environmentally: Merging the value-belief-norm theory and
the expectancy theory. Tour. Manag. 2017, 59, 76–88. [CrossRef]
66. Swami, V.; Chamorro-Premuzic, T.; Snelgar, R.; Furnham, A. Egoistic, altruistic, and biospheric environmental concerns: A path
analytic investigation of their determinants. Scand. J. Psychol. 2010, 51, 139–145. [CrossRef]
67. Mersha, A.A.; van Laerhoven, F. A gender approach to understanding the differentiated impact of barriers to adaptation:
Responses to climate change in rural Ethiopia. Reg. Environ. Chang. 2016, 16, 1701–1713. [CrossRef]
68. Vicente-Molina, M.; Fernández-Sainz, A.; Izagirre-Olaizola, J. Does gender make a difference in pro-environmental behavior? The
case of the Basque Country University students. J. Clean. Prod. 2018, 176, 89–98. [CrossRef]
69. Saga, R.; Kunimoto, R. LDA-based path model construction process for structure equation modeling. Artif. Life Robot. 2016, 21,
155–159. [CrossRef]
70. Ray, A.; Bala, P.K.; Dwivedi, Y.K. Exploring values affecting e-Learning adoption from the user-generated-content: A consumption-
value-theory perspective. J. Strateg. Mark. 2020, 1–23. [CrossRef]
71. Wang, Z.; Ye, X. Social media analytics for natural disaster management. Int. J. Geogr. Inf. Sci. 2018, 32, 49–72. [CrossRef]
72. Salloum, S.A.; Al-Emran, M.; Monem, A.A.; Shaalan, K. A survey of text mining in social media: Facebook and Twitter
perspectives. Adv. Sci. Technol. Eng. Syst. J 2017, 2, 127–133. [CrossRef]
73. Tong, Z.; Zhang, H. A Text Mining Research Based on LDA Topic Modelling. In Proceedings of the International Conference on
Computer Science, Engineering and Information Technology, Vienna, Austria, 21–22 May 2016.
74. Jelodar, H.; Wang, Y.; Yuan, C.; Feng, X.; Jiang, X.; Li, Y.; Zhao, L. Latent Dirichlet Allocation (LDA) and Topic modeling: Models,
applications, a survey. Multimed. Tools Appl. 2019, 78, 15169–15211. [CrossRef]
75. Blei, D.M.; Ng, A.Y.; Jordan, M.I. Latent Dirichlet Allocation. J. Mach. Learn. Res. 2003, 3, 993–1022.
76. Hair, J.F.; Ringle, C.M.; Sarstedt, M. PLS-SEM: Indeed a silver bullet. J. Mark. Theory Pract. 2011, 19, 139–152. [CrossRef]
Int. J. Environ. Res. Public Health 2021, 18, 5270 23 of 23
77. He, W.; Zha, S.; Li, L. Social media competitive analysis and text mining: A case study in the pizza industry. Int. J. Inf. Manag. Sci.
2013, 33, 464–472. [CrossRef]
78. Krishnan, N.; Thenmozhi, S.; ThamizhArasan, R. Normalization of text in social media: Analyzing the need for pre-processing
techniques and its roles. Int. J. Pure Appl. Math. 2018, 119, 2033–2036.
79. Saif, H.; Fernandez, M.; He, Y.; Alani, H. On stopwords, filtering and data sparsity for sentiment analysis of twitter. In Proceedings
of the ISWC 2014 Posters & Demonstrations Track. 3th International Semantic Web Conference (ISWC 2014), Riva del Garda, Italy,
21 October 2014.
80. Valls Martínez, M.d.C.; Ramírez-Orellana, A. Patient satisfaction in the Spanish national health service: Partial least squares
structural equation modeling. Int. J. Environ. Res. Public Health 2019, 16, 4886. [CrossRef]
81. Hair, J.F.; Sarstedt, M.; Ringle, C.M.; Mena, J.A. An assessment of the use of partial least squares structural equation modeling in
marketing research. J. Acad. Mark. Sci. 2012, 40, 414–433. [CrossRef]
82. Bagozzi, R.P.; Yi, Y. On the evaluation of structural equation models. J. Acad. Mark. Sci. 1988, 16, 74–94. [CrossRef]
83. Gefen, D.; Straub, D.; Boudreau, M.-C. Structural equation modeling and regression: Guidelines for research practice. Commun.
AIS 2000, 4, 7. [CrossRef]
84. Chin, W.W. How to write up and report PLS analyses. In Handbook of Partial Least Squares; Springer: Berlin, Germany, 2010; pp.
655–690.
85. O’brien, R.M. A caution regarding rules of thumb for variance inflation factors. Qual. Quant. 2007, 41, 673–690. [CrossRef]
86. Montgomery, D.C.; Peck, E.A.; Vining, G.G. Introduction to Linear Regression Analysis, 4th ed.; John Wiley & Sons: Hoboken, NJ,
USA, 2012; Volume 821.
87. Shiau, W.-L.; Chau, P.Y. Management, Understanding Behavioral Intention to Use A Cloud Computing Classroom: A Multiple
Model Comparison Approach. Inf. Manag. 2016, 53, 355–365. [CrossRef]
88. Hu, L.-T.; Bentler, P.M. Fit indices in covariance structure modeling: Sensitivity to underparameterized model misspecification.
Psychol. Methods 1998, 3, 424. [CrossRef]
89. Hair, J.F., Jr.; Hult, G.T.M.; Ringle, C.; Sarstedt, M. A Primer on Partial Least Squares Structural Equation Modeling (PLS-SEM), 2nd
ed.; Sage publications: Thousand Oaks, CA, USA, 2017.
90. Chen, M.-L.; Chou, L.-N.; Zheng, Y.-C. Providing a Clean Environment for Adolescents: Evaluation of the Tobacco Hazards
Prevention Act in Taiwan. Int. J. Environ. Res. Public Health 2017, 14, 634. [CrossRef]
91. Lam, J.C.; Cheung, L.Y.; Wang, S.; Li, V.O. Stakeholder concerns of air pollution in Hong Kong and policy implications: A big-data
computational text analysis approach. Environ. Sci. Policy 2019, 101, 374–382. [CrossRef]
92. MOEA. Energy Statistics Handbook 2018. Available online: https://www.moeaboe.gov.tw/ECW/english/content/SubMenu.
aspx?menu_id=1537 (accessed on 1 April 2021).
93. Lo, W.-C.; Shie, R.-H.; Chan, C.-C.; Lin, H.-H. The Attributable Mortality Burden Due to PM2.5 Exposure in Taiwan. Formosan J.
Med. 2016, 20, 396–405.
94. Heinrichs, H.U.; Schumann, D.; Vögele, S.; Biß, K.H.; Shamon, H.; Markewitz, P.; Többen, J.; Gillessen, B.; Gotzens, F.; Ernst, A.
Integrated assessment of a phase-out of coal-fired power plants in Germany. Energy 2017, 126, 285–305. [CrossRef]
95. Severnini, E. Impacts of nuclear plant shutdown on coal-fired power generation and infant health in the Tennessee Valley in the
1980s. Nat. Energy 2017, 2, 1–9. [CrossRef]
96. De Groot, J.I.; Steg, L. Mean or green: Which values can promote stable pro-environmental behavior? Conserv. Lett. 2009, 2, 61–66.
[CrossRef]
97. Crompton, T.; Kasser, T. Meeting Environmental Challenges: The Role of Human Identity; World Wide Fund for Nature Godalming:
Woking, UK, 2009.
98. Song, S.Y.; Kim, Y.-K. Doing good better: Impure altruism in green apparel advertising. Sustainability 2019, 11, 5762. [CrossRef]
99. Wu, K.-Y.; Huang, Y.-H.; Wu, J.-H. Impact of electricity shortages during energy transitions in Taiwan. Energy 2018, 151, 622–632.
[CrossRef]
100. Bickerstaff, K. Risk perception research: Socio-cultural perspectives on the public experience of air pollution. Environ. Int. 2004,
30, 827–840. [CrossRef] [PubMed]
101. Bain, P.G.; Hornsey, M.J.; Bongiorno, R.; Jeffries, C. Promoting pro-environmental action in climate change deniers. Nat. Clim.
Chang. 2012, 2, 600–603. [CrossRef]
102. Bord, R.J.; O’Connor, R.E. The gender gap in environmental attitudes: The case of perceived vulnerability to risk. Soc. Sci. Q.
1997, 78, 830–840.
103. Anderson, M. Men Catch Up with Women on Overall Social Media Use; Pew Research Center: Washington, DC, USA, 2015.
104. Lindsey, L.L. The sociology of gender: Theoretical perspectives and feminist frameworks. In Gender Roles: A Sociological Perspective;
Routledge: New York, NY, USA, 2005; pp. 1–21.
105. Zelezny, L.C.; Chua, P.P.; Aldrich, C. New ways of thinking about environmentalism: Elaborating on gender differences in
environmentalism. J. Soc. Issues 2000, 56, 443–457. [CrossRef]
106. Bright, J.; Margetts, H.; Hale, S.A.; Yasseri, T. The Use of Social Media for Research and Analysis: A Feasibility Study; Department for
Work and Pensions: London, UK, 2014.