Professional Documents
Culture Documents
Hoffmann 2020 Environ. Res. Lett. 15 014009
Hoffmann 2020 Environ. Res. Lett. 15 014009
To cite this article: Roman Hoffmann and Raya Muttarak 2020 Environ. Res. Lett. 15 014009 - Do protests influence environmental
attitudes? Evidence from Extinction
Rebellion
Yiannis Kountouris and Eleri Williams
LETTER
protective actions (Ortega-Egea et al 2014, Meyer 2015, Imbens 2004). We employ a mediation analysis to esti-
Steg 2016, Chankrajang and Muttarak 2017, Long et al mate whether and to what extent the three considered
2018). However, the exact channels through which edu- mechanisms, knowledge, risk perception and aware-
cation and more broadly learning affect environmental ness of climate change, explain the education effects
behavior require a deeper understanding, especially in on pro-environmental behavior (Breen et al 2013).
resource-scarce and low-income contexts. How and why Mediation analysis is a relevant methodological tool,
education affects environmental behavior is of particular which allows to identify and explore the mechanisms
relevance for policy-makers who can use this knowledge underlying relationships. While the need for a better
to design effective measures and demand-side interven- understanding of the mechanisms is emphasized in
tions aimed at promoting environmental consciousness the literature (e.g. Meyer 2015), these are rarely
and improving environmental protection. empirically explored. Based on a theoretical frame-
To this end, this study investigates the relationship work, our analysis generates insights, which help
between formal education and pro-environmental understanding not only whether, but also why educa-
behavior in the Philippines, a lower-middle income tion may be important for mitigation efforts and how
country, and analyzes the role of climate change it could best be used.
knowledge, risk perceptions, and awareness of the We find that education is positively related to pro-
causes of climate change as possible mechanisms in environmental behavior such as recycling, proper gar-
explaining education effects. To the best of our knowl- bage disposal, and planting trees. Controlling for gen-
edge, there is no other study, which simultaneously eralized propensity scores, an additional year of
considers the link between education and its mediat- schooling is estimated to increase the probability of
ing channels in promoting pro-environmental actions carrying out climate-friendly actions by 3.3%. Like-
in a low-income context. The Philippines represents wise, each of the three tested mediating factors, i.e.
an interesting case for our study. The educational sys- having greater knowledge about climate change,
tem was reformed several times in the past decades higher perception of climate risks, and being aware of
resulting in improved access, high enrolment its causes, has a significant positive impact on green
rates, and a low illiteracy rate of less than 4% behavior. The mediation analysis reveals that educa-
(UNESCO 2018). Moreover, environmental educa- tion effects are mainly driven by differences in the
tion has a long-standing tradition in formal education awareness of climate change being caused by human
curricula and overall awareness of environmental actions. Being aware of the link between anthro-
issues is high in the country. More than 73% of the pogenic activities and climatic change can influence
population can identify with a person who gives the perceived self-efficacy to interfere with and miti-
importance to looking after the environment com- gate climate change, and ultimately the willingness to
pared to 55% and 63% in the neighboring countries take actions against it. Altogether, the three mediating
Thailand and Malaysia, respectively (Inglehart et al factors explain between 30.5% and 32.2% of the total
2014). Nevertheless, the country is faced with severe education effects.
environmental issues, such as pollution, deforestation, The remainder of the paper is structured as fol-
and environmental degradation. The government has lows. Section 2 presents the previous literature and our
responded to these challenges by emphasizing the theoretical framework. Section 3 introduces the
need for action and awareness raising following bot- research setting of our study and the data. Section 4
tom-up community-centered and inclusive approa- presents our identification strategy and discusses the
ches as well as continued environmental education measurement of the key variables. The results are pre-
and learning in the formal schooling system (Republic sented in section 5. Section 6 concludes with a discus-
of the Philippines 2017). sion of the main findings and implications for future
This study uses original survey data collected research. The main text accompanied by supplemen-
among female household heads from low-income tary material available online at stacks.iop.org/ERL/
neighborhoods in Metro Manila and the neighboring 15/014009/mmedia presenting detailed background
province of Rizal who were members of a social devel- information on the study sample (S1), the estimation
opment organization. The surveys, which were con- strategy (S2), the measurement (S3), and further
ducted as part of a project on microfinance and health results (S4).
care, contained a detailed module on environmental
behavior and climate change knowledge and percep-
tions. Although specific in its nature, the sample of 2. Previous literature and conceptual
households represents an interesting case allowing us framework
to study the relationship between pro-environmental
actions and education in a low-income setting. Pro- Previous micro-level studies, mostly from high-
environmental behavior is estimated controlling for a income countries, have found a positive relationship
rich set of personal and contextual background char- between education and a wide range of pro-environ-
acteristics determining educational attainment in mental behaviors, including consumption patterns,
form of generalized propensity scores (Hirano and conservation, and lifestyle choices. For example, there
2
Environ. Res. Lett. 15 (2020) 014009
Figure 1. Conceptual framework explaining the direct and indirect channels through which education influences environmental
behavior. Note: the empirical design controls for the respondents’ pre-education background and indirect channels of influence
allowing us to capture the direct effects of education on pro-environmental behaviors.
is ample empirical evidence showing a correlation (EKC). For example, Inglesi-Lotz and Morales (2017)
between education and environmentally friendly food find that higher education levels are associated with
choices. This includes the purchasing of organic food higher energy consumption in developing countries
products (Blend and van Ravenswaay 1999, Lockie whilst in developed countries, energy consumption
et al 2004, Ngobo 2011), the willingness to pay higher declines as educational levels increase. On the other
prices for eco-labeled seafood (Brécard et al 2009, Xu hand, controlling for per capita income, increasing
et al 2012), the consumption of seasonal fruits and education levels have been shown to compensate for
vegetables (Tobler et al 2011), and a reduced con- the negative effects of economic growth on CO2 emis-
sumption of meat (de Boer et al 2014, De Backer and sion in Australia (Balaguer and Cantavella 2018).
Hudders 2015, Graça et al 2015, Weibel et al 2019). While most of the evidence from low and middle-
At the same time, better educated individuals have income countries is based on macro level data, there
been shown to be more likely to adopt energy con- are only few studies using micro level survey data to
servation behaviors that yield a lower carbon foot- test for the relationship between education and envir-
print, such as purchasing eco-labeled, higher onmental behavior. Moreover, most of the existing
efficiency electrical and heating appliances empirical evidence is based on correlations. Few
(Flamm 2009, Mills and Schleich 2010, Michelsen and recent studies attempt to identify the causal effect of
Madlener 2012, Ma et al 2013, Wijaya and education on green behaviors using changes in com-
Tezuka 2013), adopting renewable energy sources pulsory schooling laws over time (Meyer 2015, Chank-
(Sardianou and Genoudi 2013), and deliberatively sav- rajang and Muttarak 2017). While these studies
ing energy (Poortinga et al 2004, Sardianou 2007, provide evidence for a causal relationship between
Ouyang and Hokao 2009, Mills and Schleich 2012, education and pro-environmental behavior, little is
Lacroix and Gifford 2018, Niamir 2019). A positive known about the actual mechanisms explaining the
relationship has also been reported for recycling and observed effects. Theoretically, education can affect
waste disposal activities (Hage et al 2009, Sidique et al environmental decision-making either directly or
2010) and the adoption of fuel-efficient or alternative indirectly (see figure 1). In this study, we primarily
fuel vehicles (Mannberg et al 2014, Potoglou and focus on direct channels of influence while controlling
Kanaroglou 2007). for potential indirect channels and the pre-education
In low and middle-income countries, on the other background.
hand, increasing education levels often go hand in Formal schooling can directly influence pro-envir-
hand with enhanced economic development and the onmental behavior by improving skills, knowledge
emergence of a wealthier middle class who can afford and awareness that are relevant for climate change
higher consumption levels. In fact, studies that shift mitigation actions. Evidence from FMRI (functional
the focus beyond high-income countries do not neces- magnetic resonance imaging) studies have shown a
sarily find the positive relationship between education positive association between schooling and the devel-
and green environmental behavior (Ameli and opment of neural networks underlying cognitive con-
Brandt 2015, Inglesi-Lotz and Morales 2017). The trol and problem solving (Rosenberg-Lee et al 2011,
educational expansion in many countries has resulted Brod et al 2017). Being in school allows children to
in increased energy and resource consumption, which engage in cognitive activities including learning to
on the macro level led to a nonlinear relationship read, write, understand number, count and solve
between development and sustainable lifestyle, com- numerical problems. Schooling activities in higher
monly referred to as Environmental Kuznets Curve grades involve tasks that require acquisition and
3
Environ. Res. Lett. 15 (2020) 014009
4
Environ. Res. Lett. 15 (2020) 014009
Figure 2. Map of study areas with locations of respondents’ homes. The study areas encompassed both more urban and rural areas
located in Rizal province towards the East of the National Capital Region.
education, which take place outside the traditional and contextual pre-education characteristics. The
educational institutions, also play an important role in generalized propensity score (GPS) derived from this
influencing pro-environmental orientations and exercise is used in the subsequent analysis as an
behaviors (Eshach 2007). Non-formal education and aggregate control variable allowing us to efficiently
learning programs, such as community-based initia- control for potentially confounding influences of
tives, can successfully raise awareness and disseminate respondent’s pre-education characteristics (Hirano
knowledge about the importance of environmental and Imbens 2004, Egger and von Ehrlich 2013). As a
protection. Likewise, informal education, which robustness check we repeated our main analysis
includes learning from daily experience and the edu- including the entire set of pre-treatment controls
cative influences and resources in the immediate instead of the GPS (see supplement table S4). None of
environment, including social interactions, can also our results is sensitive to the changes in the estimation
play a key role (Richardson and Wolfe 2001, procedures. In the estimation of the GPS, we closely
Digby 2013). In this regard, the media and public com- followed the procedures outlined in Hirano and
munication channels are an important source of influ- Imbens (2004) and Kluve et al (2012). Further
ence (Chan 1998, Holbert et al 2003). Although both information on the estimation of the GPS as well as the
non-formal and informal learning are highly relevant tests of the central assumptions can be found in the
and underexplored topics, it was beyond the scope of supplementary material (S2) and in Hoffmann and
this study to empirically explore their influences in the Lutz (2019).
considered setting. In our propensity score estimation, we rely on a
broad set of personal pre-education characteristics:
parental education, parental literacy, cognitive abil-
4. Empirical strategy and measurement ities, age, and early work experience as a proxy for
wealth in childhood. In addition to these personal
4.1. Pre-education background and propensity background variables, we include a set of relevant
score estimation birth province characteristics, which may have influ-
To answer our main research questions on the enced respondents’ likelihood to obtain higher educa-
relationship between education and environmental tion levels. In particular, we are interested in
behavior as well as to explore the mechanisms under- contextual variables, which capture the economic
lying the effect of education (if any), we follow a development level and educational infrastructure in
multiple steps estimation strategy. In a first step, we the province, such as the distance of the birth province
estimate the respondents’ propensity to obtain a to the capital, the literacy rate, population density,
specific level of education based on different personal elementary school completion rate, and the
5
Environ. Res. Lett. 15 (2020) 014009
Figure 3. Percentages distribution of pro-environmental behaviors. Categorization based on open answers by respondents listing all
activities they currently undertake to protect the environment and to reduce human impacts on climate change.
electrification rate in the province. The information environmental mobility seems to not be considered as
on the respondent’s birth province and birth year was a pro-environmental action by the respondents.
merged with these environmental background char- The survey-based subjective measurement of
acteristics of the respective province at the time of the respondent’s environmental behavior may be prone to
respondent’s birth. All contextual data were derived measurement errors and social desirability biases
from the Philippine Census of Population and Hous- (Gatersleben et al 2002). To account for these issues,
ing since 1948. we also collected observational data on respondent’s
littering and garbage disposal behavior, which affects
the immediate environment. When visiting respon-
4.2. Measurement of pro-environmental behavior dents’ homes, interviewers were asked to rate the space
To measure pro-environmental behavior respondents in and around the houses of the respondents in terms
were openly asked to list all activities they currently of littering and garbage disposal on a four-point scale
undertake to protect the environment and to reduce (‘lots of uncollected garbage/very dirty’, ‘some uncol-
impacts on climate change. Interviewers were required lected garbage/dirty’, ‘very little garbage/clean’, ‘no
to probe the respondents several times to obtain garbage visible/very clean’). Based on this informa-
complete answers. The answers were later categorized tion, we generated a dummy variable taking the value
into nine categories, which are displayed in figure 3 zero if there was a lot of or some garbage in the
below. The categorization of the open-ended ques- immediate area around or inside the house. Although
tions was done by two coders in order to ensure this measure serves only as a proxy for environmental
reliability. To test for the robustness of our findings, actions, it allows us to test the robustness of our
we base our analyzes on multiple outcomes: first, we findings.
study whether the respondent undertakes any pro-
environmental actions as a dummy variable; and 4.3. Measurement of mediating factors and
second, we analyze the total sum of all actions under- mechanisms
taken (min 0 to max 4), which can serve as an indicator In the mediation analysis, we consider three factors,
for the intensity of pro-environmental behavior. namely, knowledge about climate change, the percep-
The majority of the respondents (59%) undertake tion of climate change risks, and awareness of the
at least one form of green behavior. On the other hand, causes of climate change, which are expected to
we observe only few respondents who undertake more explain the relationship between education and pro-
than two actions (4.2%). Most of the answers given environmental behavior. These factors were measured
were related to the proper treatment of garbage (e.g. only for those respondents who have at least heard of
not burning, right disposal, segregation, etc). Only few the term ‘climate change’ before (N=860), which
respondents said they changed their consumption served as filter variable in our survey. Having heard of
behavior or tried to save energy or water. Interestingly, climate change naturally is a pre-condition for asses-
very few respondents said that they tried to use envir- sing its causes and associated risks. Therefore, in the
onmentally friendly ways of transportation. Since the mediation analysis in table 2, we control for whether
use of public transportation is common for the inter- the respondents have such a basic awareness of climate
viewed sample, changing to or selecting pro- change, which, in itself, is strongly associated with a
6
Environ. Res. Lett. 15 (2020) 014009
person’s education level (see supplement S4.3). The information about the strength of the mediation by
effects of the main mediators are estimated conditional considering changes in the education effects.
on whether respondents were aware of the term In the analysis we additionally control for a wide
climate change. In the supplementary materials, we set of other potential mediating channels and factors
replicate our mediation analysis only for those respon- that may be directly or indirectly influenced by educa-
dents who have heard of climate change before. The tion and potentially confound our mediation analysis.
results are consistent with the findings for the entire We include measures for economic resources, risk
sample (see supplement tables S8 and S9). preferences, social capital, marital status, number of
The first mediating factor—general knowledge children in the household, religiosity, household size,
about climate change—was assessed with a question and subjective health status. Furthermore, we control
on whether the respondents were able to explain the for the average years of schooling among other
term to the interviewer (minimum acceptable answer respondents from the same neighbourhood and those
was steady or sudden changes in weather due to chan- in the direct peer group as assessed with a social net-
ging climatic conditions). General knowledge was work questionnaire to take potential contextual peer
coded as a dummy variable, which takes the value one effects into account (Durlauf and Ioannides 2010,
if the respondents were able to explain it. In total, Jackson 2011).
70.72% of the reduced sample of 860 respondents Mediation analysis is a valuable tool to identify and
were able to give a correct answer. The second mediat- explore underlying mechanisms explaining the rela-
ing factor is the respondents’ risk perception, which was tionship between educational attainment and pro-
assessed with a question on how serious they perceived environmental behavior. To derive an estimate of the
explanatory power of each of the different mediating
climate change to be on a 10-point scale. The 10 point-
factor, we compare the education coefficients from the
scale was dichotomized at the value 7 resulting in a
baseline model with those in the models controlling
total of 62.2% of the respondents who perceived a high
for the single mediators. If the mediators explain at
risk of a changing climate. Finally, the third mediating
least part of the education effect, we expect the educa-
factor measures the awareness of the causes of climate
tion coefficients to be reduced in the extended models.
change. The respondents were asked for their beliefs
The comparison of effects across nonlinear models is
about the causes of climate change. The responses
not straightforward, as the scale of the underlying
were categorized in a dichotomous variable, which
models may change if an additional factor is included.
takes the value one if the respondents said that climate
To be able to make cross-model comparisons, we
change was (at least in part) caused by human activity
apply the KHB method (Kohler et al 2011, Breen et al
as opposed to being an outcome of natural or super- 2013, Hoffmann and Muttarak 2017), which harmo-
natural processes. Overall, 69% believed climate nizes the scale of the baseline and extended Logit mod-
change to be caused by human activities, which can els. The KHB procedure also provides a test if the
also be interpreted as a measure of the perceived self- difference in the coefficients is significantly different
efficacy to fight climate change by changing own from zero, which would support the mediation
behavior. argument.
7
Environ. Res. Lett. 15 (2020) 014009
Poisson: number of
Logit: any actions actions
Notes: Coefficients are displayed as marginal effects calculated at the mean of all covariates, standard
errors in brackets. Standard errors are clustered at the microfinance group level (m=70). * p0.1, **
p0.05, *** p0.01.
10 years of education (equivalent to the completion of behavior; a result which mirrors findings in the pre-
secondary education compared to a person with no vious literature (Torras and Boyce 1998, Diekmann
education in the Philippines) hence substantially and Franzen 1999, Golley and Meng 2012). Further-
increases the respondents’ likelihood of undertaking more, having a greater number of children is positively
pro-environmental actions by 33%. Comparable related with the outcomes of interest. This may reflect
results are found for the count outcome, indicating an stronger preference of parents to protect the environ-
important role of education net of potentially con- ment for future generations explaining their more for-
founding pre-education factors. While we do not find ward-looking behavior. At the same time, controlling
that the GPS as an aggregate factor itself is significant for the number of children, we observe a decrease in
in the estimated models, some of the variables used for the tendency to carry out environmental actions with
the GPS estimation exert a statistically significant increasing household size.
effect on the considered outcomes (see supplement
table S4, which reports the results of a model using the
full-set of pre-treatment characteristics as controls 5.2. Extended models including mediating factors
instead of the GPS). In the second step of our analysis, we add the
Apart of the education effects, we find that several mediating variables to our baseline models in a
of the additional controls reflecting other potential stepwise fashion to estimate their effect on environ-
mediators have a significant effect on the likelihood mental behavior. Table 2 shows the results of our Logit
and intensity of environmental actions. Higher wealth and Poisson models, which control for area fixed
levels are positively associated with environmental effects and the additional control variables, including
8
Environ. Res. Lett. 15 (2020) 014009
Notes: Coefficients are displayed as marginal effects calculated at the mean of all covariates, standard
errors in brackets. Standard errors are clustered at the microfinance group level (m=70). Additional
controls included in the model, but not displayed: neighborhood dummies, household size, number
of children, marital status, subjective health, social support, subjective religiosity, risk preferences,
education levels in microfinance and direct peer group. * p<0.1, ** p<0.05, *** p<0.01.
wealth. Note that the models additionally control for term ‘climate change’. Especially in the Logit models,
whether the respondents have heard of the term we observe a smaller education effect on the likelihood
‘climate change’, which was the filter requirement for of undertaking environmental actions if our analysis
the successive questions on knowledge, awareness and controls for this additional factor. This suggests that
risk perception of climate change, our key mediating basic awareness of the existence of climate change is
variables. also an important mechanism on its own as it explains
To be able to compare the education effects across part of the education effects reported in table 1. In fur-
the models, we re-calculated the baseline education ther analyzes in the supplementary material (S4.4), we
coefficient (tables 1(b) and (d)) additionally control- further explore this factor and as a sensitivity test re-
ling for whether the respondents have heard of the run our models excluding the respondents who said
9
Environ. Res. Lett. 15 (2020) 014009
Figure 4. Estimated probabilities to undertake environmental actions for different levels of schooling. Notes: the graphs are based on
the Logit models presented in table 2. On average, the respondents with higher climate knowledge, a higher risk perception and an
awareness of the causes of climate change are more likely to report undertaking pro-environmental actions.
they have never heard about climate change (n=174) behavior (though reduced in magnitude) remains sig-
from our analytical sample. All results reported here nificant suggesting that there are other unobserved
remain fully robust to the sample variation. channels that may play a role.
Also after adding further controls, education effects In additional analyzes, we study the influence of edu-
are significant and meaningful. According to the Logit cation on the three mediating factors, climate change
estimates (a), the probability to undertake environ- knowledge, risk perception, and awareness (see supple-
mental actions increases by a substantial 2.1% with each ment table S6). We find a strong positive association
additional year of schooling. Once the mediating vari- between education and the mediators, which is a neces-
ables are introduced to the models (b)–(d), the education sary condition for the mediation argument to hold. An
coefficients are reduced as compared to the baseline spe- increase in education by one year raises the probability to
cification in model (a) suggesting that all mediators can know about climate change by 2.0%. Likewise, the prob-
at least in part explain the observed education effects. ability to perceive climate change as a risk and of being
We find that all three mediating factors exert a sig- aware of the human causes of changing environmental
nificant positive effect on environmental behavior. conditions increase by 2.4% with every additional year of
Having knowledge about climate change, perceiving schooling. An even stronger education effect is found for
climate change as a threat for the livelihood, and being the variable based on which we filter out respondents
aware of the anthropogenic causes of climate change who have not heard of the term ‘climate change’. The
raise the probability to partake pro-environmental probability of having heard of this term increases by a
actions by 8.2%, 9.6% and a substantial 27.2%, respec- substantial 3.9% with every school year revealing a poten-
tively. In the full Logit and Poisson models including tially large role of education in making people aware of
all mediators (e), the risk perception effects are no the very basic existence of global environmental changes.
longer significant in both the Logit and Poisson speci-
fication, suggesting that this mediating factor may be
of less relevance in influencing environmental action. 5.3. Explaining education effects
Likewise, the knowledge effect is slightly reduced and Complementing the previous findings, we derive an
loses its significance in the final Logit specification. estimate for the strength of mediation and test to what
Figure 4 shows the estimated probabilities to extent the considered mediators explain the reported
undertake pro-environmental actions by years of education effects. For this analysis, we compare the
schooling stratified by the levels of the dichotomous size of the education coefficients between the baseline
mediating variables (high versus low). We observe (a) and extended models (b)–(e), which additionally
higher levels of pro-environmental behavior if the control for the mediators. The KHB method employed
respondents have greater knowledge, perceive a higher here makes the coefficients comparable by adjusting
threat of changing climatic conditions, and are aware the underlying scales in the nonlinear estimation. The
of the anthropogenic causes of climate change. The row ‘% change edu. coefficient’ in table 2 shows the
difference is particularly pronounced for the latter, as estimated percentage change in the education effects
indicated by its very large coefficients in table 2. Those due to the inclusion of the mediating factor. Note that
who are aware that human activities contribute to glo- the size of this measure, which can be interpreted as an
bal warming may feel greater self-efficacy that they can indicator for the strength of mediation, is influenced
actually make a difference by engaging in climate- by both the effect of education on the mediator, and its
friendly actions. In the models controlling for all med- influence on the outcome, i.e. pro-environmental
iators the effect of education on pro-environmental behavior.
10
Environ. Res. Lett. 15 (2020) 014009
Table 3. Logit models: effects of education and mediators on alternative outcome measure.
Notes: Coefficients are displayed as marginal effects calculated at the mean of all covariates, standard
errors in brackets. Standard errors are clustered at the microfinance group level (m=70). All models
control for the full set of control variables: neighborhood dummies, household size, number of
children, marital status, subjective health, social support, subjective religiosity, risk preferences,
education levels in microfinance and direct peer group. *p0.1, **p0.05, ***p0.01.
While education effects remain significant even if explaining part of the effect of education on pro-
the additional mediators are included, the magnitude environmental behavior (see supplement table S7).
of the estimates decreases considerably. The difference
between the coefficients of the baseline and extended 5.4. Robustness check using observational data
model is the largest when climate change awareness is The previous models used behavioral information
introduced as a mediator. The former explains about based on the subjective survey responses. In a final
22% of the education effects, revealing its important step, we test for the robustness of our results using
role as potential mechanism explaining the relation- observational data about the environmental behavior
ship between education and pro-environmental beha- of the respondents. For this, we analyze the inter-
vior. While we do find that knowledge and perceptions viewer’s assessments of garbage disposal and littering
of the risks associated with climate change also play a in the space in and around the houses of our
role, the proportion of the education effects explained respondents. Table 3 shows the results of Logit models,
is merely 6.3% and 10.6%, respectively. Hence, they which regress the binary-coded observational variable
on years of education and the considered mediators.
seem to be less relevant factors in explaining the link
All models control for the full set of alternative
between education and environmental actions. Over-
mediators that may confound the relationships.
all, the three main mediators together explain 30.5%
While the effect using the alternative outcome are
and 32.2% in the Logit and Poisson specifications,
smaller than for the subjective measure, the overall
respectively, controlling for the full set of additional
pattern is similar. The probability for showing a pro-
variables and neighborhood fixed effects. environmental behavior increases by 0.9% with each
As mentioned above, education effects are also additional year of education. Adding the mediators to
substantially reduced once the filter variable, measur- the models, we do not see substantial changes when
ing whether respondents have heard of ‘climate knowledge about climate change and respondent’s
change’, is included in the baseline model. As the mea- risk perception are included. However, in line with the
surement of the main three mediators considered in previous models, education effects are significantly
this study depends on the respondents having at least reduced by 10.9% when differences in the awareness
heard of climate change before, we cannot directly of the anthropogenic causes of climatic changes are
compare the importance of this filter variable vis-a-vis controlled for suggesting an important role of this
the other mediators. However, our results suggest that mechanism in explaining education effects on the
already the very basic awareness of the existence of cli- considered outcome. Apart from the role of education,
mate change is an important mechanism on its own we also find a substantial effect of household wealth in
11
Environ. Res. Lett. 15 (2020) 014009
influencing the garbage disposal and littering normative beliefs (Ajzen 1991). The decision to take an
outcome. action includes both the evaluation of a certain out-
come and the estimation of the likelihood to reach this
outcome through one’s actions, i.e. the perceived
6. Discussion and conclusion behavioral control (Ajzen 2002). Being aware that
one’s behavior can attain certain outcomes and the
How does education affect environmental behavior resulting perceived self-efficacy are hence important
and what are the underlying mechanisms? In this prerequisites for individuals to take up a desired beha-
paper, we employ a generalized propensity score vior. This does also apply to pro-environmental beha-
approach and a mediation analysis to answer these vior. Knowing about the anthropogenic causes of
questions. The analysis is performed using original climate change concurrently mean that individuals
survey data that was collected among women in low- also know that they can do something about it. Indeed,
income households in the greater area of Metro Manila the previous environmental psychology literature
and the neighboring province of Rizal. While the finds that poor understanding of the connection
previous literature has a strong focus on advanced between human actions and climate change influences
industrialized countries, our study represents one of the perception of human ability to control and take
the few that explore the determinants of environmen- action against it (Lorenzoni and Pidgeon 2006).
tal behavior in a developing country context. Accordingly, a lack of awareness about the con-
We find education positively influences environ- sequences of human activities on the global climate
mental behaviors. An additional year of schooling rai- system discourages individuals to take any climate-
ses the propensity to undertake environmentally friendly actions. Education thus plays a vital role in
friendly actions by a significant 3.3% (table 1, Logit promoting knowledge about climate change and in
model a) and the number of actions undertaken by raising awareness about the impacts of human
0.073 (table 1, Poisson model c). The education effects activities.
estimated in our baseline models closely resemble the Our study faces some limitations. First, due to the
effects reported in other studies both from high as well cross-sectional and non-experimental nature of our
as low and middle-income countries. Using the Euro- data, we are unable to make causal claims in our analy-
barometer data, Meyer (2015) finds a statistically sig- sis. Although we control for pre-education character-
nificant effect of 0.071 of education on the number of istics using propensity scores and a variety of
environmental actions undertaken, which is almost additional demographic controls, the results could be
identical to our estimate (see Meyer 2015, p. 113, driven by simultaneity issues or omitted variables,
table 4, col 1). Other studies from high-income coun- such as time preferences. To check the validity of our
tries report similar results, such as Mills and Schleich results, we performed various consistency checks,
(2012) Di Maria et al (2010), or Sardianou and Genoudi which did not indicate any problems with the used
(2013). As one of a few studies from a developing identification strategy.
country, Chankrajang and Muttarak (2017) report A second limitation is that our findings are mainly
significant point estimates between 0.035 and 0.059 for based on survey data that are prone to measurement
education effects on different pro-environmental and reporting errors. In particular, we rely on data
actions in Thailand (see Chankrajang and Muttarak about self-reported environmental behavior, which is
2017, page 441, table 5, col 1 to 4). The comparison of potentially subject to social desirability biases. Some
the findings suggests considerable similarities in the respondents may have over-reported their engage-
education effects across settings, which seem to be inde- ment in pro-environmental actions in order to receive
pendent of the respective country wealth level. social approval. For example, when comparing house-
A mediation analysis based on the KHB method hold energy use with self-reported pro-environmental
(Breen et al 2013) is employed to identify the under- behavior among Dutch households, Gatersleben et al
lying mechanisms of the effect of education on pro- (2002) revealed that people who reported to act in a
environmental behavior. Our results show that being more environmentally-friendly way do not necessarily
aware of climate change and its causes explains as use less energy. Other studies, for example by Milfont
much as one-fourth of the education effects suggesting (2009), on the other hand, do not find strong evidence
that education plays an important role in raising of social desirability effects for self-reported pro-
awareness levels, which positively affects environ- environmental behavior. It has also been argued that
mental decision-making. This reveals an important social desirability is likely to represent a minor pro-
role of understanding the causes and consequences of blem when the self-reported measures refer to past/
climate change in triggering pro-environmental present behavior like the ones used in this study rather
actions. than intended/future behavior (Ortega-Egea et al
This result is in line with the theory of planned 2014). To account for potential measurement errors in
behavior which holds that behavioral intentions are a our design, we used different operationalization of our
function of one’s attitude towards performing a part- key outcomes and replicated our main findings using
icular act and subjective norms or the strength of an alternative, observational outcome measure,
12
Environ. Res. Lett. 15 (2020) 014009
garbage disposal and littering, which confirmed our In the context of the 2030 Agenda for Sustainable
main conclusions. Development, Bengtsson et al (2018) have reviewed
Third, our sample is restricted to female house- and presented evidence on the role of education in
hold heads, who were members of a social develop- enabling the progress of the Agenda ranging from
ment organization at the time of the survey. The data improving health, poverty reduction, promoting gen-
hence is not representative for the entire country. We der equality to reducing vulnerability and enhancing
nevertheless believe that the sample and the collected adaptive capacity in the context of climate change
data is highly informative as it allows us to test theor- (Nilsson et al 2016, Weitz et al 2019, Sachs et al 2019).
etical predictions in a setting, which has rarely been Our findings provide further evidence that education
considered in previous research. The aim of our study is an important enabler for reaching sustainable devel-
is to reach a high internal validity and to carefully opment on a global scale. Investments in education
explore some of the mechanisms underlying the iden- can make an important contribution in raising aware-
tified effects. In this regard, we see case studies like ness and ultimately in promoting green behavior con-
ours as a useful tool to understand patterns and tributing to reducing the human impacts on the global
mechanisms on the micro level, which can be com- climate system. In this regard, while it is important to
pared to other studies with a similar empirical design. provide learners with the necessary tools and cap-
Whilst this study focuses mainly on formal school- abilities to undertake pro-environmental actions, it is
ing, knowledge, awareness and risk perception in the also key to raise their perceived self-efficacy. Environ-
context of climate change can be acquired via other mental education curricula should thus not only focus
non-formal and informal channels such as through on the transfer of knowledge and information, but also
the media and social diffusion. These channels are highlight the importance of the individual contrib-
more difficult to assess empirically and require a spe- ution in mitigating the harmful consequences of glo-
cific survey design to be measured. While our study bal environmental change.
was focused on direct mechanisms, the effects of edu-
cation on pro-environmental behavior can also oper- Acknowledgments
ate through other channels, which we do not capture
with our design. Factors like low time discount rates We are very grateful to Wolfgang Lutz, Alejandro
(i.e. focusing on benefits in a longer time horizon) Cunat, Karl Schlag, Bernhard Kittel, Jesus Crespo
have been found to be associated with environmental Cuaresma, Maarten Janssen, Ilona Reindl, and Nadia
behavior and energy consumption (Carson and Roth Steiber for their invaluable feedback and support. We
Tran 2009, Long et al 2018) and are also reported to be thank the Kasagana-Ka Development Center Inc. for
more prominent amongst highly educated people their collaboration and assistance with this study. In
(Chao et al 2009, WhitMarsh 2011). Additional factors particular, we are grateful to the CEO of Kasagana-Ka,
such as biospheric values may also mediate the effect Maria Anna de Rosas Ignacio, and the other members
of education if these values influence pro-environ- of the steering committee. We are furthermore greatly
mental actions and are correlated with education indebted to the staff members in the field without
(Steg 2016). It is thus highly likely that education whom this study would not have been possible. We
affects environmental behavior through other media- thank Aylin Bello, Justine Co, Louise Oblena, and
tors beyond the ones captured in our study. Our study Ricardo Guzman for their outstanding research assis-
thus opens up new enquiries to be pursued in future tance. This study was made possible by funding
research. provided by the University of Vienna and the Vienna
Formal schooling functions not only as an instru- University of Economics and Business. Further fund-
ment to signal the level of education attained but has ing was provided by the Austrian Science Fund (FWF):
an intrinsic value in enhancing the capability to Z171-G11.
achieve the outcomes aspired for (Bengtsson et al
2018). Likewise, investments in education may not Data availability
only have individual-level benefits, but may generate
important spill-overs at the community level such as The data that support the findings of this study are
in the case of reductions in infant mortality (Pamuk available from the corresponding author upon reason-
et al 2011) or disaster preparedness (Witvorapong et al able request. The data are not publicly available for
2015, Hoffmann and Muttarak 2017). Interactions legal and/or ethical reasons.
with well-educated individuals may contribute to the
spread of information and awareness in communities References
and exceed the direct education effects reported in this
study. Education may hence generate further environ- Ajzen I 1991 The theory of planned behavior Organ. Behav. Hum.
Decis. Process. 50 179–211
mental externalities that benefit not only the indivi- Ajzen I 2002 Perceived behavioral control, self-efficacy, locus of
dual and the people in their direct surrounding, but control, and the theory of planned behavior J. Appl. Soc.
also the general public (Lutz et al 2014). Psychol. 32 665–83
13
Environ. Res. Lett. 15 (2020) 014009
Ameli N and Brandt N 2015 Determinants of households’ Eshach H 2007 Bridging in-school and out-of-school learning:
investment in energy efficiency and renewables: evidence formal, non-formal, and informal education J. Sci. Educ.
from the OECD survey on household environmental Technol. 16 171–190
behavior and attitudes Environ. Res. Lett. 10 44015 Flamm B 2009 The impacts of environmental knowledge and
Anderson A 2012 Climate change education for mitigation and attitudes on vehicle ownership and use Transport. Res. D 14
adaptation J. Educ. Sustain. Dev. 6 191–206 272–9
Antonio E, Bass S and Gasgonia D 2012 Philippines Experience, Gatersleben B, Steg L and Vlek C 2002 Measurement and
Lessons and Challenges in Environmental Mainstreaming (UK: determinants of environmentally significant consumer
International Institute for Environment and Development) behavior Environ. Behav. 34 335–62
Balaguer J and Cantavella M 2018 The role of education in the Golley J and Meng X 2012 Income inequality and carbon dioxide
Environmental Kuznets Curve. Evidence from Australian emissions: The case of Chinese urban households Energy
data Energy Econ. 70 289–96 Econ. 34 1864–72
Bengtsson S, Barakat B and Muttarak R 2018 The Role of Education Gough I 2013 Climate change, social policy, and global governance
in Enabling the Sustainable Development Agenda (Abingon- J. Int. Comparative Soc. Policy 29 185–203
on-Thames: Routledge) Graça J, Oliveira A and Calheiros M M 2015 Meat, beyond the plate.
Blend J R and van Ravenswaay E O 1999 Measuring consumer Data-driven hypotheses for understanding consumer
demand for ecolabeled apples Am. J. Agric. Econ. 81 willingness to adopt a more plant-based diet Appetite 90
1072–7 80–90
Brécard D, Hlaimi B, Lucas S, Perraudeau Y and Salladarré F 2009 Hage O, Söderholm P and Berglund C 2009 Norms and economic
Determinants of demand for green products: an application motivation in household recycling: empirical evidence from
to eco-label demand for fish in Europe Ecol. Econ. 69 115–25 Sweden Resour. Conserv. Recycl. 53 155–65
Breen R, Karlson K B and Holm A 2013 Total, direct, and indirect Hirano K and Imbens G W 2004 The propensity score with
effects in logit and probit models Sociol. Methods Res. 42 continuous treatments Applied Bayesian Modeling and Causal
164–91 Inference from Incomplete–Data Perspectives: An Essential
Brod G, Bunge S A and Shing Y L 2017 Does one year of schooling Journey with Donald Rubin’s Statistical Family (Probability
improve children’s cognitive control and alter associated and Statistics Series) ed A Gelman and X-L Meng (Hoboken,
brain activation? Psychol. Sci. 28 967–78 NJ: Wiley) 7 73–84
Carson R T and Roth Tran B 2009 Discounting behavior and Hoffmann R and Muttarak R 2017 Learn from the past, prepare for
environmental decisions J. Neurosci., Psychol., Econ. 2 112–30 the future: impacts of education and experience on disaster
Chan K 1998 Mass communication and pro environmental preparedness in the Philippines and Thailand World Dev. 96
behavior: waste recycling in Hong Kong J. Environ. Manage. 32–51
52 317–25 Hoffmann R and Lutz S U 2019 The health knowledge mechanism:
Chankrajang T and Muttarak R 2017 Green returns to education: evidence on the link between education and health lifestyle in
does schooling contribute to pro-environmental behaviors? the Philippines Eur. J. Health Econ. 20 27–43
Evidence from Thailand Ecol. Econ. 131 434–48 Holbert R L, Kwak N and Shah D V 2003 Environmental concern,
Chao L-W, Szrek H, Pereira N S and Pauly M V 2009 Time patterns of television viewing, and pro-environmental
preference and its relationship with age, health, and survival behaviors: integrating models of media consumption and
probability Judgment Decis. Mak. 4 1–19 effects J. Broadcast. Electron. Media 47 177–96
Cotten S R and Gupta S S 2004 Characteristics of online and offline Huang J, van den Brink H and Groot W 2009 A meta-analysis of the
health information seekers and factors that discriminate effect of education on social capital Econ. Educ. Rev. 8 454–64
between them Soc. Sci. Med. 59 1795–806 Inglehart R et al (ed) 2014 World Values Survey: All Rounds -
Creutzig F et al 2018 Towards demand-side solutions for mitigating Country-Pooled Datafile 1981-2014 (Madrid: JD Systems
climate change Nat. Clim. Change 8 268–71 Institute) (http://worldvaluessurvey.org/
De Backer C J S and Hudders L 2015 Meat morals: relationship WVSDocumentationWVL.jsp)
between meat consumption consumer attitudes towards Inglesi-Lotz R and Morales L D D C 2017 The Effect of Education on a
human and animal welfare and moral behavior Meat Sci. 99 Country’s Energy Consumption: Evidence from Developed and
68–74 Developing Countries Working Papers 201733 Department of
de Boer J, Schösler H and Aiking H 2014 ‘Meatless days’ or ‘less but Economics, University of Pretoria University of Pretoria,
better’? Exploring strategies to adapt Western meat Department of Economics
consumption to health and sustainability challenges Appetite IPCC 2014 Climate Change 2014: Synthesis Report. Contribution of
76 120–8 Working Groups I, II and III to the Fifth Assessment Report of
Del Missier F, Mäntylä T and de Bruin W B 2012 Decision-making the Intergovernmental Panel on Climate Change IPCC,
competence, executive functioning, and general cognitive Geneva, Switzerland p 151
abilities J. Behav. Decis. Mak. 25 331–51 Jackson M O 2011 An overview of social networks and economic
Diekmann A and Franzen A 1999 The wealth of nations and applications Handbook of Social Economics ed J Benhabib,
environmental concern Environ. Behav. 31 540–9 A Bisin and M O Jackson vol 1 (San Diego: North-Holland)
Digby C L B 2013 The influences of socio-demographic factors, and 511–85
non-formal and informal learning participation on adult Kabir M I, Rahman M B, Smith W, Lusha M A F, Azim S and
environmental behaviors Int. Electron. J. Environ. Educ. 3 Milton A H 2016 Knowledge and perception about climate
37–55 change and human health: findings from a baseline survey
Di Maria C, Ferreira S and Lazarova E 2010 Shedding light on the among vulnerable communities in Bangladesh BMC Public
light bulb puzzle: the role of attitudes and perceptions in the Health 16 266–275
adoption of energy efficient light bulbs Scottish J. Political Kluve J, Schneider H, Uhlendorff A and Zhao Z 2012 Evaluating
Econ. 57 48–67 continuous training programmes by using the generalized
Dubois G et al 2019 It starts at home? Climate policies targeting propensity score J. R. Stat. Soc. A 175 587–617
household consumption and behavioral decisions are key to Kohler U, Karlson K B and Holm A 2011 Comparing coefficients of
low-carbon futures Energy Res. Soc. Sci. 52 144–58 nested nonlinear probability models Stata J. 11 420–38.
Durlauf S N and Ioannides Y M 2010 Social Interactions Annu. Rev. Lacroix K and Gifford R 2018 Psychological barriers to energy
Econ. 2 451–78 conservation behavior: the role of worldviews and climate
Egger P H and von Ehrlich M 2013 Generalized propensity scores change risk perception Environ. Behav. 50 749–80
for multiple continuous treatment variables Econ. Lett. 119 Lake R L D and Huckfeldt R 1998 Social capital, social networks, and
32–4 political participation Political Psychol. 19 567–84
14
Environ. Res. Lett. 15 (2020) 014009
Lockie S, Lyons K, Lawrence G and Grice J 2004 Choosing organics: Republic of the Philippines 2017 The National Environmental
a path analysis of factors underlying the selection of organic Education Action Plan 2018–2040 (Republic of the
food among Australian consumers Appetite 43 135–46 Philippines: Environmental Management Bureau)
Long C, Mills B F and Schleich J 2018 Characteristics or culture? Richardson L D and Wolfe M 2001 Principles and Practice of Informal
Determinants of household energy use behavior in Germany Education (London, New York: Routledge)
and the USA Energy Effic. 11 777–98 Rosenberg-Lee M, Barth M and Menon V 2011 What difference
Lorenzoni I and Pidgeon N F 2006 Public views on climate change: does a year of schooling make? Maturation of brain response
european and USA perspectives Clim. Change 77 73–95 and connectivity between 2nd and 3rd grades during
Lutz W, Muttarak R and Striessnig E 2014 Universal education is key arithmetic problem solving NeuroImage 57 796–808
to enhance climate adaption Science 346 1061–3 Rosi A, Bruine de Bruin W, Del Missier F, Cavallini E and Russo R
Lutz W and Striessnig E 2015 Demographic aspects of climate 2019 Decision-making competence in younger and older
change mitigation and adaptation Population Stud. 69 S69–76 adults: which cognitive abilities contribute to the application
Lutz W and Muttarak R 2017 Forecasting societies’ adaptive of decision rules? Aging, Neuropsychol., Cogn. 26
capacities through a demographic metabolism model Nat. 174–89
Clim. Change 7 177–84 Sachs J D, Schmidt-Traub G, Mazzucato M, Messner D,
Ma G, Andrews-Speed P and Zhang J 2013 Chinese consumer Nakicenovic N and Rockström J 2019 Six transformations to
attitudes towards energy saving: the case of household achieve the sustainable development goals Nat. Sustain. 2
electrical appliances in Chongqing Energy Policy 56 591–602 805–14
Mannberg A, Jansson J, Pettersson T, Brännlund R and Lindgren U Sardianou E 2007 Estimating energy conservation patterns of Greek
2014 Do tax incentives affect households’ adoption of ‘green’ households Energy Policy 35 3778–91
cars? A panel study of the Stockholm congestion tax Energy Sardianou E and Genoudi P 2013 Which factors affect the
Policy 74 286–99 willingness of consumers to adopt renewable energies?
McCright A M 2010 The effects of gender on climate change Renew. Energy 57 1–4
knowledge and concern in the American public Population Sidique S F, Lupi F and Joshi S V 2010 The effects of behavior and
Environ. 32 66–87 attitudes on drop-off recycling activities Resour. Conserv.
Meyer A 2015 Does education increase pro-environmental Recycl. 54 163–70
behavior? Evidence from Europe Ecol. Econ. 116 108–21 Steg L and Vlek C 2009 Encouraging pro-environmental behavior :
Michelsen C C and Madlener R 2012 Homeowners’ preferences for An integrative review and research agenda J. Environ. Psychol.
adopting innovative residential heating systems: a discrete 29 309–17
choice analysis for Germany Energy Econ. 34 1271–83 Steg L 2016 Values, norms, and intrinsic motivation to act
Milfont T L 2009 The effects of social desirability on self-reported proenvironmentally Annu. Rev. Environ. Resour. 41 277–92
environmental attitudes and ecological behavior Stevenson R B 2007 Schooling and environmental education:
Environmentalist 29 263–9 contradictions in purpose and practice Environ. Educ. Res. 13
Mills B and Schleich J 2010 What’s driving energy efficient appliance 139–53
label awareness and purchase propensity? Energy Policy 38 814–25 Tobler C, Visschers V H M and Siegrist M 2011 Eating green.
Mills B and Schleich J 2012 Residential energy-efficient technology Consumers’ willingness to adopt ecological food
adoption, energy conservation, knowledge, and attitudes: an consumption behaviors Appetite 57 674–82
analysis of European countries Energy Policy 49 616–28 Torras M and Boyce J K 1998 Income, inequality, and pollution: a
Muttarak R and Chankrajang T 2016 Who is concerned and does reassessment of the environmental Kuznets curve Ecol. Econ.
something about climate change? Gender and education divide 25 147–60
among Thais Vienna Yearbook Population Res. 13 193–220 UNESCO 2018 UNESCO Global Education Monitoring Report:2019:
Neuenschwander L M, Abbott A and Mobley A R 2012 Assessment Migration, Displacement and Education – Building Bridges, not
of low-income adults’ access to technology: implications for Walls Paris, UNESCO
nutrition education J. Nutrition Educ. Behav. 44 60–5 Wals A E J and Benavot A 2017 Can we meet the sustainability
Ngobo P V 2011 What drives household choice of organic products challenges? The role of education and lifelong learning Eur. J.
in grocery stores? J. Retailing 87 90–100 Educ. 52 404–13
Niamir L 2019 Behavioral Climate Change Mitigation: From Weibel C, Ohnmacht T, Schaffner D and Kossmann K 2019
Individual Energy Choices to Demand-Side Potential Reducing individual meat consumption: an integrated phase
(Enschede: Gildeprint) model approach Food Qual. Preference 73 8–18
Nilsson M, Griggs D and Visbeck M 2016 Policy: map the interactions Weitz N, Carlsen H and Trimmer C 2019 SDG synergies: AN
between sustainable development goals Nature 534 320–2 approach for coherent 2030 Agenda implementation Stockholm
Oechssler J, Roider A and Schmitz P W 2009 Cognitive abilities and Environment Institute (https://sei.org/publications/sdg-
behavioral biases J. Econ. Behav. Organ. 72 147–52 synergies-factsheet/)
Ortega-Egea J M, García-de-Frutos N and Antolín-López R 2014 Wen L M, Rissel C, Baur L A, Lee E and Simpson J M 2011 Who is
Why do some people do ‘more’ to mitigate climate change NOT likely to access the Internet for health information?
than others? Exploring heterogeneity in psycho-social Findings from first-time mothers in southwest Sydney,
associations PLoS One 9 e106645 Australia Int. J. Med. Inform. 80 406–11
Ouyang J and Hokao K 2009 Energy-saving potential by improving Whitmarsh L 2011 Scepticism and uncertainty about climate
occupants’ behavior in urban residential sector in Hangzhou change: dimensions, determinants and change over time
City, China Energy Build. 41 711–20 Glob. Environ. Change 21 690–700
Pamuk E R, Fuchs R and Lutz W 2011 Comparing relative effects of Wijaya M E and Tezuka T 2013 Measures for improving the
education and economic resources on infant mortality in adoption of higher efficiency appliances in Indonesian
developing countries Population Dev. Rev. 37 637–64 households: an analysis of lifetime use and decision-making
Poortinga W, Steg L and Vlek C 2004 Values, environmental in the purchase of electrical appliances Appl. Energy 112
concern, and environmental behavior: a study into 981–7
household energy use Environ. Behav. 36 70–93 Witvorapong N, Muttarak R and Pothisiri W 2015 Social
Poortinga W, Whitmarsh L, Steg L, Böhm G and Fisher S 2019 Climate participation and disaster risk reduction behaviors in tsunami
change perceptions and their individual-level determinants: a prone areas PLoS One 10 1–20
cross-European analysis Glob. Environ. Change 55 25–35 Xu P, Zeng Y, Fong Q, Lone T and Liu Y 2012 Chinese consumers’
Potoglou D and Kanaroglou P S 2007 Household demand and willingness to pay for green- and eco-labeled seafood Food
willingness to pay for clean vehicles Transp. Res. D 12 264–74 Control 28 74–82
15