Content uploaded by Milja Heikkinen
Author content
All content in this area was uploaded by Milja Heikkinen on Feb 18, 2020
Content may be subject to copyright.
Transnational municipal networks and climate change adaptation: A
study of 377 cities
Milja Heikkinen
a
,
*
, Aasa Karimo
b
, Johannes Klein
c
, Sirkku Juhola
a
, Tuomas Yl€
a-Anttila
d
a
Helsinki Institute of Sustainability Science, Ecology and Environment Research Programme, Faculty of Biological and Environmental Sciences, PL 65
(Viikinkaari 2a), 00014, University of Helsinki, Finland
b
Faculty of Social Sciences, PL 54 (Unioninkatu 37), 00014, University of Helsinki, Finland
c
Geological Survey of Finland, Espoo, Finland, PL 96 (Vuorimiehentie 5), 02151, Espoo, Finland
d
Helsinki Institute of Sustainability Science, Faculty of Social Sciences, PL 54 (Unioninkatu 37), 00014, University of Helsinki, Finland
article info
Article history:
Received 26 June 2019
Received in revised form
3 December 2019
Accepted 6 February 2020
Available online 7 February 2020
Handling editor: Yutao Wang
Keywords:
Climate change adaptation
City networks
Adaptation planning
C40 network
Global covenant of mayors
ICLEI
abstract
Cities have increasingly recognised the risks posed by climate change and the need to adapt. To support
climate action, cities have formed cooperative networks such as the C40 Cities Climate Leadership Group,
the Global Covenant of Mayors and the International Council for Local Environmental Initiatives. How-
ever, a lack of scientific evidence exists when it comes to the actual impact of network participation,
especially in the context of adaptation. This study is the first to test statistically the association between
network membership and progress in adaptation planning in 377 cities globally. The results show that
network members are more likely to have started the adaptation process than other cities, and that being
a member of multiple networks is associated with higher levels of adaptation planning. Moreover, cities
in wealthier countries are more likely to be more advanced in adaptation planning than others. We
consider the possible explanations for these results based on the previous literature and information
gathered from the networks. The main implications of our study are that network organisations should
consider how to encourage the adaptation process among their members and the increased involvement
of cities from lower-income countries.
©2020 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/by/4.0/).
1. Introduction
Cities are important actors in climate change adaptation (Revi
et al., 2014;van der Heijden, 2018;van der Heijden et al., 2018).
High hopes have been expressed about the benefits of networking
between cities for climate action. With the support from inter-city
networks such as the C40 Cities Climate Leadership Group (C40),
the International Council for Local Environmental Initiatives (ICLEI)
and the Global Covenant of Mayors (GCoM), cities are pictured as
leading the way to a climate-safe future, combining ambitious
mitigation and adaptation efforts (C40, Solecki et al., 2018;van der
Heijden et al., 2019). Cities are also seen as the drivers of globally
sustainable development (UNEP, 2011;Barber, 2017;ICLEI, 2018;
Solecki et al., 2018;van der Heijden et al., 2019).
Meanwhile, the Intergovernmental Panel on Climate Change
(IPCC) Fifth Assessment Report (FAR) states that there is medium
confidence, based on medium evidence and with medium agree-
ment, that horizontal learning through networks of cities benefits
urban adaptation (Revi et al., 2014: 539). Despite the enthusiasm
surrounding these networks, there is little systematic evidence
concerning the effects of network participation (Wolfram et al.,
2019). There has been no research to show whether network
participation is associated with progress in planning of climate
change policies at the city level, especially when it comes to
adaptation (Fünfgeld, 2015;Woodruff, 2018).
We define transnational municipal networks (TMNs) related to
climate change as organisations that aim to support cooperation
between cities to improve their climate change mitigation and
adaptation work. TMNs can require cities to adopt certain quanti-
tative or qualitative climate goals. They organise events, produce
information (e.g. reports on their members’climate actions), offer
tools and/or resources and represent cities internationally. TMNs
originally concentrated on mitigation, but adaptation has increas-
ingly been on their agenda. Although some scholars have begun to
*Corresponding author.
E-mail addresses: milja.e.heikkinen@helsinki.fi(M. Heikkinen), aasa.karimo@
helsinki.fi(A. Karimo), johannes.klein@gtk.fi(J. Klein), sirkku.juhola@helsinki.fi
(S. Juhola), tuomas.yla-anttila@helsinki.fi(T. Yl€
a-Anttila).
Contents lists available at ScienceDirect
Journal of Cleaner Production
journal homepage: www.elsevier.com/locate/jclepro
https://doi.org/10.1016/j.jclepro.2020.120474
0959-6526/©2020 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
Journal of Cleaner Production 257 (2020) 120474
study their role in adaptation, it remains understudied (Juhola and
Westerhoff, 2011;Busch, 2015;Woodruff, 2018). The big question
regarding adaptation is whether these networks live up to the high
expectations to increase local adaptation efforts (Woodruff, 2018).
To our knowledge, this study is the first effort to answer this
question through a statistical analysis of a global sample of cities.
We analysed the connection between TMN membership and
progress in climate adaptation planning worldwide using the
Adaptation Process Index (API), which is the first global measure of
city-level climate change adaptation planning efforts, developed by
Araos et al. (2016b). We have employed their data set in this study,
analysing the connection between memberships in three key global
TMNsdC40, ICLEI and GCoMdand the progress of cities in their
adaptation planning processes.
The study covers 377 large cities (at least one million in-
habitants) for which the API is currently available (Araos et al.,
2016b). Large cities are of interest in climate adaptation research
due to their economic importance and the risks related to popu-
lation concentration (Hunt and Watkiss, 2011). Moreover, earlier
studies have found that large cities seem to be drivers in TMNs
(Woodruff, 2018).
Previous empirical studies have predominantly focused on a
handful of cities in wealthy western countries (van der Heijden,
2018), in the form of case studies of individual cities or limited
geographical areas (Kern and Bulkeley, 2009). The samples of the
few earlier large-N studies have either been from Europe (Reckien
et al., 2018) or from North America (Woodruff, 2018). The existing
studies also reflect differences in focus. Reckien et al. (2018) re-
ported on the status and drivers of European city-level climate
change mitigation and adaptation planning in relationship to na-
tional- and international-level planning and legislation. Woodruff
(2018) analysed how factors like planning capacity affect the
probability of a city to join TMNs.
This study goes beyond the existing qualitative literature by
statistically testing the association between TMN participation and
climate adaptation enot just showing how some cities may benefit
from these networks but testing whether this is the case at the
aggregate level and by seeing what kind of cities are more likely to
benefit than others. We also go beyond the limited existing quan-
titative work by analysing a global sample which includes a sig-
nificant number of non-western cities, especially from China and
India. This approach provides new insights into the global role of
TMNs by significantly widening the geographical scope of the
analysis compared to earlier studies.
We selected three TMNs
1
to analyse: C40, ICLEI, and GCoM
(formerly the Covenant of Mayors and the Compact of Mayors). C40
is a network of global mega-cities concentrating especially on
climate action. GCoM also concentrates on climate issues. ICLEI has
a broader aim to support the sustainability of cities, but it also
strongly promotes climate action, and adaptation was included in
its strategic plan for 2006.
2
These networks have been identified as
important global climate change-related TMNs in previous studies
(Kern and Bulkeley, 2009;Heidrich et al., 2016;Busch, 2015;Busch
et al., 2018;Woodruff, 2018;Reckien et al., 2018;van der Heijden
et al., 2019).
The C40 was established in London in 2005. Its target is to
develop and implement policies and programmes that generate
measurable reductions in both greenhouse gas emissions and
climate risks. The network promotes cities as leaders of change, and
cities need to pass an application process to become members
(www.c40.org). According to Davidson and Gleeson (2015), C40
represents a new strategic urbanism phase of transnational urban
governance, because it ties together the most influential and
economically-powerful mayors of global mega-cities to adopt a
more visible political stance. C40 also has several private sector
partners, like Bloomberg Philanthropies (https://www.c40.org/
partners).
ICLEI is the oldest of the three networks, founded in 1990. When
compared with the C40, the climate agenda of ICLEI is more diverse,
targeting also smaller urban areas and connecting the work in
general themes of sustainability. It has an annual membership fee
(depending on the region, population and per capita gross national
income), but membership is open to all cities and regions (www.
iclei.org).
GCoM differs from the other two. It is a combination of initia-
tives, including Covenant of Mayors (launched 2008), Compact of
Mayors (launched 2014) and Mayors Adapt (launched 2014). These
were combined in 2015. GCoM aims to give political support to the
climate work of the cities by supporting the engagementof mayors.
According to them, they are world’s largest global alliance for city
climate leadership with over 9000 members (https://www.
globalcovenantofmayors.org/about/).
Although these networks operate independently, they have in-
terconnections. C40 and ICLEI had a role in creating the Compact of
Mayors, and they also work together on a range of projects, like
mitigation related Carbon Disclosure Project.
Next, we review the existing literature related to the impact of
TMNs, and draw our hypotheses, before describing our dataset and
the methods. In the results section, we have presented the average
and median APIs for cities with different combinations of network
memberships, as well as the results of statistical analysis. Our
findings indicate that there is a statistically significant connection
between network participation and starting the adaptation process,
and that being a member of multiple networks is associated with
higher levels of adaptation planning. Finally, we discuss this in light
of the existing literature, and draw a conclusion.
2. Literature review and hypotheses
The role of TMNs has been studied mostly from the point of view
of mitigation (Fünfgeld, 2015). We provide a short overview of
recongnized benefits in Table 1. The benefits are often connected to
information sharing, learning, shaping mitigation initiatives, and
increased resources. It has been found that networks help munic-
ipalities to act when state-level action is lacking, and that they act
as city advocates shaping the political environment and legal frame.
Networks may cause similar effects when it comes to adaptation
(Cast
an Broto and Bulkeley, 2013).
Overall, we identify two specific topics in this discussion on
which we wish to shed light. First, claims have been made that
networking cities are pioneers of mitigation (Kern and Bulkeley,
2009), and the same may apply when it comes to adaptation.
Relatively strong claims state that networks do support local
climate adaptation (Revi et al., 2014;Woodruff, 2018), even though
little empirical evidence exists showing that networks have an ef-
fect on actual city-level climate action. Proving this claim is diffi-
cult, and scholars should be critical of highly normative claims on
the impact of networks (van der Heijden, 2018). If city networks do
support local climate change adaptation, it would be reasonable to
expect members would have a higher API. Hence, we hypothesise
that
H1. Network members are more advanced in their climate change
adaptation planning processes than non-members.
1
We decided to leave 100 Resilient Cities out since it was founded 2013 and it
concentrates on resilience, a different concept than adaptation. Also, their future
plans are unclear: http://www.100resilientcities.org/closing-note/.
2
http://archive.iclei.org/index.php?id¼10832.
M. Heikkinen et al. / Journal of Cleaner Production 257 (2020) 1204742
Many cities are members of more than one network. It is
possible that different networks offer different kinds of support,
making it reasonable to use resources in multiple memberships. If
networks do support climate change adaptation, it would be logical
to assume that the more networks a city participates in, the more
support it gets. Hence, we hypothesise that
H2. The more networks a city is a member of, the more advanced it is
in its adaptation planning process.
Second, TMNs seem to be biased towards the wealthy western
countries. When studying mitigation networks, Bansard et al.
(2017) found that cities in Europe and North America are over-
represented, and the cities which participate in several networks
connecting them together come from this region. Cities in wealthier
countries seem to end up at the core of networks defining best
practices, which may exclude the cities in less wealthy countries
and increase differences between cities (Kern and Bulkeley, 2009;
Shi et al., 2016). This is because cities in wealthy countries tend to
have higher overall administrative capacities to design policies, as
well as higher capacities to implement adaptation and mitigation
measures. For example, C40 may seem horizontal, but it is still
largely dominated by cities like New York and London (Acuto, 2013,
see also Bouteligier, 2013;Lee, 2018).
Cities with lower capacities may adopt passive roles, with their
membership becoming mostly symbolic (Kern and Bulkeley, 2009).
Not all members have access to the benefits offered by the net-
works (Lee, 2015;van der Heijden, 2018), and although networks
have been active for some decades now, a great number of cities
have not joined (van der Heijden, 2018). On the other hand, it has
been argued that cities lagging in climate action reap the greatest
benefits from the networks (Busch et al., 2018;Reckien et al., 2018).
Therefore, we test the hypothesis
H3. The wealthier the country in which a city is located, the more
advanced the city is in its adaptation planning process.
Our dataset is the first global, large-N sample of city-level
climate adaptation measures. With these data, we can not only
test H3, but also control for wealth differences when testing H1 and
H2. Further, we can assess whether our two key independent var-
iables, network participation and wealth, are correlated.
3. Material and methods
We used data on 997 adaptation initiatives in 402 urban areas
around the world (Araos et al. (2016a,b). The dataset includes in-
formation about public adaptation planning in urban areas
3
with
more than one million people. The researchers considered material
in the following 13 languages, with a minimum of four cities per
language: English, Spanish, French, Chinese, Arabic, Russian,
German, Portuguese, Farsi, Korean, Japanese, Turkish, and Indone-
sian (Araos et al., 2016b).
Araos et al. (2016b) collected the data from climate change
planning documents in a web-based search using the Google search
engine. The search terms were “climate change”and the city’s
name. Thus, the search focussed on “highly intentional”adaptation
policies as defined by Dupuis and Biesbroek (2013: p. 1480). The
data was collected between January 2 and March 29, 2014 (Araos
et al., 2016b). This method is consistent with other studies col-
lecting information about adaptation planning (Reckien et al., 2014,
2018;Lesnikowski et al., 2016).
3.1. Variables
Our dependent variable is the adaptation process index (API,
Araos et al., 2016b), drawn from the data described above. The in-
dex includes the following criteria: presence of climateprojections,
presence of vulnerability assessments, consideration of multiple
sectors, reassessment of development priorities in the face of
climate change, availability of climate change planning documents,
consultations and stakeholder engagement, management of bar-
riers and uncertainty, and monitoring and evaluation of adaptation
activities. The more criteria a city fulfilled, the higher its API. The
values range from 0 to 8. For further theoretical justifications on
why these particular criteria are included in the API, see (Araos
et al., 2016b).
The main independent variable of interest is network partici-
pation. For each city in the dataset, we coded whether it is (or is
not) a member of a network on spring 2019, and if it is a member,
whether the city joined the network before 2014. We received the
information about memberships and when the cities joined
through personal communication with the networks, and through
webpage searches when necessary. Due to some missing data, we
had to drop some cities out of the analyses leaving us with 377
cities. We use network participation to explain the variation in API
in two ways: first, by using each network membership as an in-
dependent variable separately, and then by combining network
memberships into one variable that measures the number of net-
works of which the city is a member.
As national-level control variables, we used location at the
continent level (Africa being the reference category), level of na-
tional adaptation legislation at the time based on Climate Change
Laws of the World database, 2017 (0 ¼no legislation,
1¼executive 2 ¼legislative) and gross domestic product. As a city
level control variable, we used the size of the city. We controlled for
location and GDP because earlier research found TNMs to be biased
towards wealthy countries in Europe and North America and cities
in wealthy countries have better resources for climate action, as
explained in section 2. We controlled for the existence of national
legislation because cities in countries in which legislation requires
cities to plan for adaptation are likely to be more advanced in their
adaptation planning, and for city size, because larger cities may
have better capacities for adaptation than smaller ones (cf. Reckien
Table 1
Overview of benefits of networking. MLG ¼multi-level governance, SNA ¼Social Network Analysis.
Author(s): Year: Method(s): Scope: Main results concerning (potential) network benefits:
Bulkeley et al. 2003 theoretical TMNs as part of MLG in Europe preliminarily: lobbying, information sharing, learning, policy initiative creation
Kern &Bulkeley 2009 case study TMNs as part of MLG in Europe pioneers benefit: information sharing, learning, access to funding, legitimacy
Andonova et al. 2009 theoretical networks as transnational governance information sharing/diffusion, learning, possibly increased resources
Lee &van de Meene 2012 SNA learning in C40 learning, information sharing
Busch 2015 case study Influence of TMNs in German cities information sharing, learning, goal setting, city advocate
Lee &Koski 2015 statistical methods mitigation in C40 member cities motivation of local policy &action, spills over to non-members
Busch et al. 2018 survey, interviews Influence of TMNs in German cities information sharing, learning, possibly increased resources
3
The terms “city”and “urban area”do not necessarily refer to areas defined by
administrative boundaries, but the data collection is based on the United Nations’
definition of “urban agglomeration”.
M. Heikkinen et al. / Journal of Cleaner Production 257 (2020) 120474 3
et al., 2018).
Table 2 presents descriptive statistics of all variables used in the
analyses. Of the sample of 377 large cities, 202 were members of at
least one network, which is more than half of the sample. Out of
these 202 cities, 116 had joined at least one network before 2014
(30.77% of the whole sample and 57.43% of the network members).
Overall, the GCoM has the most members among the cities in our
sample with 150 members, 110 cities in our sample are members of
ICLEI and 74 are members of C40. The mean API score of the whole
sample is very low, only 1.58 on the scale from 0 to 8.
3.2. Models
To test binary relationships between network participation and
country-level GDP and API, we first used one-way analysis of
variance (Raykov and Markoulides, 2013). We then analysed the
relationship further by using zero-inflated negative binomial
regression models. We used this model type to take into account
the type and distribution of the dependent variable, the city’s API.
4
The API can only take non-negative integer values that arise
from counting rather than ranking (Araos et al., 2016b). This kind of
count data is typically analysed using Poisson regression models
(Greene, 1994). However, the distribution of API across the cities
analysed is overdispersed and contains excessive zeros, which
creates a violation of the assumptions of a conventional Poisson
model. The zero-inflated negative binomial model takes the large
number of zeros into account and allows for the variance of the
dependent variable to be greater than the mean, which is not the
case in traditional Poisson regression models for count data
(Greene, 1994).
Zero-inflated count regression models are a mixture of a
generalized linear model for the dichotomous outcome, and a
conventional event-count generalized linear model, such as nega-
tive binomial regression (Desmarais and Harden, 2013). This is why
a zero-inflated negative binomial model divides the analysis into
two distinct parts, a zero-inflation model and a count model
(Greene, 1994). This means that two different processes are ana-
lysed at once, one analysing the probability of the dependent var-
iable getting a value zero (a reversed binomial model), and the
other analysing the distribution of the count variable including zero
(Greene, 1994). In case of the API, the two processes are whether a
city has started the adaptation process in the first place, and if so,
how far along in the process it is. The expected value of the API is
thus expressed as a combination of both processes:
m
i
¼
p
i
*0þð1
p
i
Þ*expx
T
i
b
;
where
p
i
is the probability that a city has not started the adaptation
process and x
i
is the count component of the model (Zeileis et al.,
2008). We conducted the analysis using the pscl package in an R
environment for statistical computing.
The API includes two parts describing the awareness of
vulnerability and exposure: vulnerability and climate trend (Araos
et al., 2016b). It is not probable that vulnerability and/or exposure
that a city is unaware of will affect the adaptation planning process.
We did not include vulnerability as an explicit control variable,
because there is no reliable global index which would describe
vulnerability at the city level before 2014 (Araos et al., 2016b) and
vulnerability is partly covered by controlling for economic capacity,
since existing patterns of development have profound effects on
vulnerability (Shi et al., 2016).
4. Results
Our first and second hypotheses are that network members
have higher APIs than other cities and that the more memberships,
the higher the API. Results of the analysis of variance showed a
Table 2
Variables used in the analyses.
Continuous variable information Minimum Maximum Mean Std. Deviation
Adaptation process index 0 8 1.58 2.47
GDP 2016 (in thousands) 0.80 87.86 22.54 18.19
Population (100 000) 10.02 369.33 31.14 37.91
Categorical variable information N Percent
Memberships:
C40 member (total) 74 19.89
C40 member 2014 or later 32 8.49
C40 member before 2014 42 11.14
ICLEI member (total) 110 29.18
ICLEI member 2014 or later 25 6.63
ICLEI member before 2014 85 22.55
GCoM member (total) 150 39.79
GCoM member 2014 or later 127 33.69
GCoM member before 2014 23 6.10
Member of 1 network before 2014 85 22.55
Member of 2 networks before 2014 28 7.43
Member of 3 networks before 2014 3 0.80
Level of adaptation legislation:
Executive 212 56.23
Legislative 47 12.47
Continent:
Africa 46 12.20
Asia 185 49.07
Europe 34 9.02
Latin America 56 14.85
North America 50 13.26
Oceania 6 1.59
4
We also conducted the analyses using multilevel negative binomial regression
but ended up with the single level model because the model estimate for country
level variance was 0. This might be due to the fact that more than half of the cities
in our data are the only observation from their country.
M. Heikkinen et al. / Journal of Cleaner Production 257 (2020) 1204744
connection between network membership and adaptation plan-
ning progress, giving preliminary support to both hypotheses 1 and
2.
Table 3 presents the means and analysis of variance in API for
the cities with different network membership combinations.
5
The
numbers show that there is a connection between network mem-
berships and higher APIs, especially when the city is member of
two or three networks, or when it is member in C40 or GCoM. Our
third hypothesis is that wealthier cities have higher APIs regardless
of network participation, so we also analysed the correlation be-
tween API and GDP. Our results indicate a significant association
between these two variables with a Pearson correlation of 0.339
(p <0.000) between API and GDP.
Table 3 also presents the results from analysis of variance in GDP
for cities with different network membership combinations. Cities
that are not members of any of these networks have a significantly
lower GDP compared to other cities. Specifically, members of GCoM
and C40 have a higher GDP than non-members, but also members
of ICLEI that joined before 2014 have a statistically significantly
higher mean GDP compared to cities that are not members.
These analyses offer preliminary support to all three hypothe-
ses. Wealth is connected to both network membership and the
adaptation planning progress. However, differences in GDP be-
tween network members and non-members raise a question of
whether these connections are independent or not. The results
from zero-inflated negative binominal models give us a more
detailed picture (Table 4).
Table 4 presents the results of the zero-inflated negative bino-
mial models (models 0 to 6). They also support our hypotheses.
Network members have been more likely to start their climate
change adaptation planning process than other cities (H1), and
being a member of multiple networks is connected to having a
higher API (H2). These results hold when wealth, geographical
location and legislation are controlled for. Also, cities in wealthy
countries have higher API scores when network memberships are
controlled for (H3). We also conducted the analyses using standard
negative binomial regression as a robustness check. However, ac-
cording to the Vuong test (Vuong, 1989) results, the zero-inflated
model is a significant improvement over the standard negative
binomial model (see supplement 2 and 3 for details).
The upper part of Table 4 shows the count part of the zero-
inflated regression models, which are to be read like regular
count regression models, i.e. a one unit increase in an independent
variable results in exp(B) increase in the dependent variable. The
lower part of the table shows the zero-inflation model, which in-
dicates the probability of the dependent variable (API) being a
certain zero. This indicates that the coefficient 1.87 for C40
members in Model 3 means that being a member of C40 decreases
the odds of having an API value of zero by exp (1.87) compared to
other cities.
We began regression modelling by estimating a null model that
includes only control variables, namely population, legislation, and
continent. Model 0 shows that indeed, there are differences in the
API of cities by continent. Asian and Latin American cities have
significantly lower total APIs compared to the African cities in our
sample (Model 0, count model), and all but Asian cities have a
Table 3
Comparing means of API and GDP by network participation.
API GDP (in thousands)
Mean 95% confidence interval Mean 95% confidence interval
Lower Upper Lower Upper
No network memberships before 2014 0.89 0.68 1.10 18.80 16.84 20.77
Member of 1 network before 2014 2.21 1.62 2.79 29.24 24.83 33.64
Member of 2 networks before 2014 5.21 4.12 6.31 34.86 27.86 41.85
Member of 3 networks before 2014 4.33 5.71 14.37 43.02 27.44 58.60
ANOVA F Sig. F Sig.
Between groups 41.051 0.000 14.461 0.000
95% confidence interval 95% confidence interval
Mean Lower Upper Mean Lower Upper
Not a member of C40 1.14 0.91 1.37 21.56 19.58 23.54
Member of C40 before 2014 4.09 3.17 5.02 33.70 27.24 40.16
Member of C40 2014 or after 1.69 0.81 2.57 17.15 11.49 22.82
ANOVA F Sig. F Sig.
Between groups 3.,527 0.000 10.223 0.000
95% confidence interval 95% confidence interval
Mean Lower Upper Mean Lower Upper
Not a member of ICLEI 1.23 0.98 1.48 21.12 18.99 23.25
Member of ICLEI before 2014 2.64 1.99 3.29 28.35 24.21 32.48
Member of ICLEI 2014 or after 0.76 0.05 1.47 17.97 11.08 24.86
ANOVA F Sig. F Sig.
Between groups 13.169 0.000 6.094 0.002
95% confidence interval 95% confidence interval
Mean Lower Upper Mean Lower Upper
Not a member of CoM 1.01 0.77 1.25 16.77 14.98 18.56
Member of GCoM before 2014 5.22 3,98 6.45 43.45 39.04 47.86
Member of GCoM 2014 or after 1.78 1.33 2.24 29.07 25.35 32.79
ANOVA F Sig. F Sig.
Between groups 39.082 0.000 42.496 0.000
5
We analysed the means of API separately for cities with API >0. The results
point to similar direction. For details, please see the supplement.
M. Heikkinen et al. / Journal of Cleaner Production 257 (2020) 120474 5
significantly lower probability of having an API value zero
compared to African cities (Model 0, zero-inflation model).
The existence of national legislation on climate adaptation is not
significant in any model. This is not surprising, since only five out of
80 countries had legislation in place before 2014. The results con-
cerning the association between the existence of a national non-
binding strategic or guiding document (laws 2 in Table 4)are
mixed. The existence of a guiding document is associated with an
increased likelihood of having an API higher than zero, but nega-
tively associated with API in the count model when GCoM mem-
bership is included as a covariate.
Next, we test H3 on the association between the wealth of the
country that a city is located in and the city’s API. The reason for
testing H3 before H1 and H2 is that adding GDP into the models at
this stage ensures that the wealth of the country is controlled for in
all subsequent models that test H1 and H2. Since H1 and H2 are our
primary hypotheses of interest, we wanted to present them first in
the hypotheses section above.
The connection between GDP and higher API is statistically
significant also when the city’s location is controlled for (Model 1).
GDP remained significant in all count and zero-inflation models
(Models 2e6), indicating there is an association between GDP and
higher API regardless of network participation, and the probability
of having a value zero is lower in wealthier cities. All in all, these
results support H3.
H1 was tested in zero-inflated Models 2 to 5 and count Models 2
to 5. The zero-inflated Models 2 and 3 show that membership of
C40 and ICLEI decrease the probability of API being zero when a city
has joined the network before 2014. Model 4 shows that mem-
bership of GCoM is not significant. The results are similar when all
memberships are in the same model (Model 5).
Count Models 2 and 3 show that membership of ICLEI or C40 is
not connected to higher API. However, cities which joined the
GCoM network before 2014 have significantly higher API scores
Table 4
Parameter estimates and model fit for zero-inflated negative binomial models explaining adaptation process index.
Parameter estimates: count model Model 0 Model 1 Model 2 Model 3 Model 4 Model 5 Model 6
Intercept 1.82 (0.23) *** 1.78 (0.23) *** 1.82 (0.24) *** 1.78 (0.23) *** 1.92 (0.25) *** 1.95 (0.25) *** 1.83 (0.24) ***
H1 C40 member before 2014 0.23 (0.14) 0.20 (0.14)
C40 member 2014 or after 0.03 (0.17) 0.03 (0.18)
ICLEI member before 2014 0.03 (0.10) 0.03 (0.10)
ICLEI member 2014 or after 0.07 (0.28) 0.05 (0.30)
GCOM member before 2014 0.62 (0.23) ** 0.60 (0.23) *
GCoM member 2014 or after 0.09 (0.13) 0.09 (0.14)
H2 Member in 1 network 0.09 (0.12)
Member in 2 networks 0.33 (0.15) *
Member in 3 networks 0.30 (0.31)
H3 GDP 2016 (in thousands) 0.01 (0.00) *** 0.01 (0.00) ** 0.01 (0.00) ** 0.01 (0.00) * 0.01 (0.00) * 0.01 (0.00) *
Population (100 000) 0.00 (0.00) 0.00 (0.00) 0.00 (0.00) 0.00 (0.00) 0.00 (0.00) 0.00 (0.00) 0.00 (0.00)
Level of adaptation legislation: executive 0.19 (0.17) 0.30 (0.16) 0.33 (0.17) 0.30 (0.16) 0.43 (0.17) ** 0.45 (0.17) ** 0.37 (0.17) *
Level of adaptation legislation: legislative 0.28 (0.16) 0.24 (0.16) 0.26 (0.16) 0.23 (0.16) 0.11 (0.17) 0.13 (0.17) 0.18 (0.16)
Asia 0.44 (0.17) * 0.53 (0.17) ** 0.49 (0.18) ** 0.51 (0.18) ** 0.48 (0.18) ** 0.45 (0.19) * 0.46 (0.18) *
Europe 0.30 (0.22) 0.71 (0.24) ** 0.72 (0.24) ** 0.71 (0.24) ** 1.12 (0.30) *** 1.12 (0.30) *** 0.79 (0.25) **
Latin America 0.47 (0.23) * 0.53 (0.23) * 0.56 (0.23) * 0.54 (0.23) * 0.58 (0.23) * 0.60 (0.23) ** 0.56 (0.24) *
North America 0.36 (0.25) 0.98 (0.30) ** 0.99 (0.30) *** 0.98 (0.30) ** 1.04 (0.30) *** 1.06 (0.30) *** 1.01 (0.31) **
Oceania 0.24 (0.30) 0.73 (0.33) * 0.77 (0.32) * 0.72 (0.34) * 0.76 (0.33) * 0.79 (0.35) * 0.82 (0.34) *
Log(theta) 10.17 (62.98) 10.68 (44.10) 10.80 (39.43) 10.68 (44.32) 10.87 (38.33) 10.94 (35.26) 10.87 (38.38)
Parameter estimates: zero-inflation model Model 0 Model 1 Model 2 Model 3 Model 4 Model 5 Model 6
Intercept 3.96 (0.74) *** 5.22 (0.90) *** 5.39 (0.93) *** 5.29 (0.92) *** 5.11 (0.90) *** 5.19 (0.97) *** 5.28 (0.93) ***
H1 C40 member before 2014 1.87 (0.65) ** 1.75 (0.67) **
C40 member 2014 or after 0.62 (0.45) 0.71 (0.48)
ICLEI member before 2014 0.80 (0.33) * 0.79 (0.35) *
ICLEI member 2014 or after 0.44 (0.68) 0.36 (0.71)
GCOM member before 2014 1.05 (1.14) 0.59 (1.29)
GCoM member 2014 or after 0.15 (0.37) 0.35 (0.41)
H2 Member in 1 network 0.68 (0.33) *
Member in 2 networks 2.51 (0.87) **
Member in 3 networks 18.28 (6456.69)
H3 GDP 2016 (in thousands) 0.06 (0.02) *** 0.06 (0.02) *** 0.06 (0.02) *** 0.06 (0.02) ** 0.06 (0.02) ** 0.06 (0.02) **
Population (100 000) 0.02 (0.00) *** 0.02 (0.00) *** 0.01 (0.01) 0.02 (0.00) *** 0.02 (0.00) *** 0.01 (0.01) 0.01 (0.00) *
Laws 2 2.31 (0.62) *** 3.12 (0.70) *** 3.21 (0.72) *** 3.07 (0.70) *** 2.99 (0.70) *** 3.00 (0.73) *** 3.11 (0.71) ***
Laws 3 0.05 (0.63) 0.10 (0.70) 0.23 (0.72) 0.08 (0.73) 0.16 (0.76) 0.24 (0.82) 0.34 (0.79)
Asia 0.58 (0.45) 0.10 (0.48) 0.35 (0.50) 0.18 (0.48) 0.17 (0.49) 0.34 (0.52) 0.30 (0.50)
Europe 4.70 (0.83) *** 4.09 (0.88) *** 4.32 (0.91) *** 4.26 (0.89) *** 3.93 (0.95) *** 4.23 (0.99) *** 4.15 (0.90) ***
Latin America 1.35 (0.66) * 1.34 (0.71) 1.33 (0.73) 1.35 (0.73) 1.34 (0.72) 1.55 (0.77) * 1.35 (0.75)
North America 3.47 (0.77) *** 1.32 (0.92) 1.27 (0.96) 1.34 (0.92) 1.31 (0.92) 1.37 (0.95) 1.25 (0.95)
Oceania 5.21 (1.32) *** 3.69 (1.37) ** 3.50 (1.41) * 3.70 (1.46) * 3.70 (1.37) ** 3.67 (1.51) * 3.47 (1.44) *
AIC 999.68 978.64 973.37 979.28 977.67 975.77 970.94
Log Likelihood 480.84 468.32 461.68 464.64 463.83 454.88 458.47
N 377 377 377 377 377 377 377
***p <0.001, **p <0.01, *p <0.05.
M. Heikkinen et al. / Journal of Cleaner Production 257 (2020) 1204746
than others, controlling for geographic location and wealth (Model
4). These results are reproduced when all memberships are in the
same model (Model 5). This is because almost all of these cities are
members of the former Covenant of Mayors: out of 23 cities, 21 are
in western Europe. The remaining two cities are in Canada and
Mexico. In the dataset, 13 European cities are not members of
GCoM, 12 of them in Russia. Their average API is clearly lower than
among the European members of GCoM, which is probably why
controlling by continent has not made the result insignificant. The
effect of GCoM on API is not supported by standard negative
binomial models (Supplement 2), so this result needs to be inter-
preted with caution.
H2 was tested in Model 6 by studying whether membership of
multiple networks is associated with a higher API than member-
ship of just one network. Being a member of two networks before
2014 was significantly connected to both higher API and a smaller
probability of having an API value equal to zero (Model 6). Being a
member of one network decreases the probability of a zero value
(Model 6, zero-inflation part), but is not significant in increasing API
(the count-part). Being a member of all three networks is not sig-
nificant, but this is probably because only three cities were mem-
bers of all three networks before 2014. These results are also
supported by standard negative binomial models (Supplement 2).
Network membership was not significant in any of the models, if
the city had joined 2014 or later.
Adding GDP, memberships of C40 and GCoM, and the number of
network memberships improves the model fit according to both
measures used (AIC and Log Likelihood). Adding ICLEI memberhip
improves only the Log Likelihood of the model. Analysis of Pearson
residuals indicate a reasonably good fit to the data for all models
(Supplement 4).
5. Discussion
This article began with the observation that there are high ex-
pectations that TMNs will propel cities into ambitious climate
adaptation. Few studies, however, have looked into the validity of
this claim, and the ones that have are limited to a few cases or
specific geographic areas (Woodruff, 2018). We statistically tested
whether network members are more advanced in climate change
adaptation planning than other large cities, using a dataset that
includes cities from all over the world and estimating a set of zero-
inflated negative binomial regression models.
Our results support the claim that TMN members are more
likely to have started climate change adaptation planning process
than other cities. Members of C40 and ICLEI are less likely to have
an API of zero than other cities when wealth, geographical location
and legislation are controlled for. Members of the GCoM network
have overall higher APIs than other cities. Being a member of two
networks is associated with higher API scores and being a member
of one of them increases the likelihood of a city’s API being higher
than zero.
We found that cities in wealthier countries are members of
these networks more often than those in less wealthy countries.
Even though wealth partly explains higher API scores through its
effects on network membership, wealth also has a direct effect on
API (i.e. wealthy cities have a higher API scores when network
membership is controlled for).
Does all this mean that network membership supports cities in
their climate change adaptation planning? With the kind of cross-
national data used here, it is not possible to establish firmly the
direction of causality. However, we did test the hypotheses sepa-
rately for cities that joined before our data collection (2014) and
after, and these tests lent further support to the interpretation that
network membership does support the adaptation process. Those
cities that had joined networks before 2014 had been more likely to
have started the adaptation process than non-members, but this
did not apply to those cities which joined later. Also, being a
member of two networks before 2014 was connected to higher API,
while being a member of two after that date was not.
If it were the case that cities already active in adaptation would
be more likely to join the network, it would also be logical to find an
association between higher APIs for cities that joined the networks
after Araos et al. (2016b) measured their adaptation progress in
2014. We found no such association.
Kern and Bulkeley (2009) found that the most active core
members of TMNs are often the founding members or those who
join early, while the cities joining only after their neighbours or
collaborators may adopt more passive roles. To establish whether
this is the case, or whether networks actually push cities to action,
would necessitate a full second round of data collection on adap-
tation initiatives in all cities included in our sample and analysis of
their API scores both before and after joining the network. This
should be done in future research.
In addition to the shortcomings imposed by our cross-sectional
dataset, this study has other limitations. First, we concentrated on
large cities. Results could be different for small or medium-sized
cities. Second, several large urban areas with adaptation activities
were not included in the dataset, since they did not offer infor-
mation in the thirteen languages used in the data collection. Third,
there is always the possibility of human error. Some cities may have
had adaptation documents that were not found, leading to under-
estimated API. Fourth, country-level factors like GDP and location
may not give a perfect picture about the capacities of the cities.
However, they are the best proxies for which reliable data were
available. Fifth, API is based on planning documents. Therefore, this
study does not reveal if cities implement these plans or not. Global
level analysis of implementation is another important topic for
future research. Also, independent adaptation by citizens, the pri-
vate sector and NGOs were beyond the scope of this study.
At the beginning, we noted that research on TMNs has often
focused on certain, often high-capacity, cities and regions (Bansard
et al., 2017) and may create an illusion about the cities on the front
line of adaptation, while in reality only a handful of cities partici-
pate (van der Heijden, 2018). The cities in less wealthy countries
may lack the necessary resources to join networks. Even among the
ones that do participate, not all are active (Kern and Bulkeley,
2009), nor gain access to the benefits (Lee, 2015).
Our results show that many large cities remain outside the
global networks, especially in less wealthy countries. Also, they
show that cities located in poorer countries, especially outside the
networks, have advanced less in the adaptation process. This is
problematic, since urbanisation and population growth are rapid in
less wealthy countries (UN-DESA, 2015), where climate change
vulnerability is often high as well (Shi et al., 2016;van der Heijden,
2018). Considering the strong role cities from wealthy countries
have in the networks, it is probable that simply joining the net-
works is not a solution.
Overall, our results show that much improvement is needed in
adaptation planning in cities. This is true also for cities that belong
to TMNs and cities in wealthy countries. Each of the networks we
studied had many members with API scores of zero, and none of
them had an average API close to 8, the highest figure on our scale.
In our results, network membership was significant even when
GDP was controlled for. This lends some support to Lee (2015)
argument that the attributes of the cities are more important
drivers of their climate policy practices than the attributes of the
host country. However, we did find that wealth, as a country
attribute, does influence adaptation planning progress indepen-
dent of the effect occurring through network membership. Wealth
M. Heikkinen et al. / Journal of Cleaner Production 257 (2020) 120474 7
has also been found to be a driver of national level adaptation
(Berrang-Ford et al., 2014). It has also been found that participation
in networks supports cities especially when they lack state-level
support (Bulkeley et al., 2003;Heidrich et al., 2016;Busch et al.,
2018;Reckien et al., 2018). In our data, all cities from less wealthy
countries with an API score of a perfect 8 participated at least in one
network.
6
It has been noted that the climate plans of C40 members from
different contexts are remarkably similar, even though one could
assume that different practices are needed (Heikkinen et al., 2018).
While this similarity does not necessarily follow from network
membership (Heikkinen et al., 2018), previous studies have also
criticised networks for being too conservative and reinforcing the
status quo (Acuto, 2013;Bouteligier, 2013;van der Heijden, 2017;
Heikkinen et al., 2018). Also, a good solution may be less efficient or
even harmful when exported to another context (Gupta et al., 2015;
van der Heijden, 2017). It should be critically considered whether
traditional large-scale master planning is the best way to tackle
climate change-related risks in cities located in less wealthy
countries (Shi et al., 2016).
Therefore, networks should consider developing new ways to
act that would support context-based actions and more funda-
mental changes. C40 has taken steps in this direction by intro-
ducing personalised climate advisors. However, this is a costly
method and therefore not offered to all members (C40, 2017 per-
sonal communication
7
). This again raises the question of accessi-
bility to the benefits (Lee, 2015), which may lead to a situation in
which the cities most needing support do not get it. Also, to have
greater impact, networks might want to consider how non-
members could benefit from their work.
6. Conclusion
Our analysis shows that the networks do have potential to
support urban adaptation, but there is also room for improvement.
Even among network members, the average and median APIs are
not close to the maximum. Since API only measures the planning
process, there is a risk that performance is even worse when it
comes to actual action (see also Revi et al., 2014;Woodruff and
Stults, 2016). The main implication of these results is that TMNs
should consider how to further encourage adaptation among their
member cities. Our comparison of countries with different levels of
wealth and across geographic locations suggest that the networks
should also consider giving special attention to cities in the less
wealthy countries. Current and future mega-cities, facing a
considerable need for adaptation, are located there (Bulkeley et al.,
2011;Shi et al., 2016). It also seems that combining ambitious
emission reductions and sustainable well-being in these countries
is challenging (Sugiawan et al., 2019), which presents a further
challenge to TMNs operating there. Based on this study, the level of
adaptation process or networking in these countries is not high.
Thus, the networks should be careful in planning how to support
these regions so that cities are empowered to find the solutions that
work best in their contexts.
Declaration of competing interest
The authors declare that they have no known competing
financial interests or personal relationships that could have
appeared to influence the work reported in this paper.
CRediT authorship contribution statement
Milja Heikkinen: Conceptualization, Methodology, Investiga-
tion, Project administration, Writing - original draft. Aasa Karimo:
Methodology, Formal analysis, Validation, Visualization, Writing -
original draft. Johannes Klein: Conceptualization, Writing - orig-
inal draft, Writing - review &editing. Sirkku Juhola: Conceptual-
ization, Writing - review &editing, Supervision. Tuomas Yl€
a-
Anttila: Conceptualization, Methodology, Writing - review &
editing, Supervision.
Acknowledgements
First, we thank Malcolm Araos and colleagues, who gave us the
access to their data. Data on climate change adaptation policies are
from the Tracking Research on Adaptation to Climate Change
Consortium (TRAC3). The analysis and findings represent the work
and view of the authors. We thank the participants of ESG 2018
conference session “Agency 18 eCities and City Networks as Agents
in Environmental Governance”, Onerva Korhonen, Malcolm Araos,
the anonymous reviewers and the professional proofreaders for
their useful feedback. We take responsibility for remaining errors.
Appendix A. Supplementary data
Supplementary data to this article can be found online at
https://doi.org/10.1016/j.jclepro.2020.120474.
Funding
This work was supported by the Tiina and Antti Herlin Foun-
dation, Finland [grant no. 20170005]; Kone Foundation, Finland
[grants no. 085319 and no. 090022]; and University of Helsinki
Research Funds, Finland.
References
Acuto, M., 2013. The new climate leaders? Rev. Int. Stud. 39, 835e857. https://
doi.org/10.1017/S0260210512000502.
Andonova, L.B., Betsill, M.M., Bulkeley, H., 2009. Transnational climate governance.
Global Environ. Polit. 9, 52e73. https://doi.org/10.1162/glep.2009.9.2.52.
Araos, M., Austin, S.E., Berrang-Ford, L., Ford, J.D., 2016a. Public health adaptation to
climate change in large cities: a global baseline. Int. J. Health Serv. 46 (1),
53e78. https://doi.org/10.1177/0020731415621458.
Araos, M., Berrang-Ford, L., Ford, J.D., Austin, S.E., Biesbroek, R., Lesnikowski, A.,
2016b. Climate change adaptation planning in large cities: a systematic global
assessment. Environ. Sci. Pol. 66, 375e382. https://doi.org/10.1016/
j.envsci.2016.06.009.
Bansard, J., Pattberg, P., Widerberg, O., 2017. Cities to the rescue? Assessing the
performance of transnational municipal networks in global climate governance.
Int Environ Agreements 17, 229e246. https://doi.org/10.1007/s10784-016-
9318-9.
Barber, B., 2017. Cool Cities. Urban Sovereignty and the Fix for Global Warming. Yale
University Press, ISBN 9780300228113.
Berrang-Ford, L., Ford, J.D., Lesnikowski, A., Poutiainen, C., Barrera, M.,
Heymann, S.J., 2014. What drives national adaptation? A global assessment.
Climatic Change 124 (1e2), 441e450.
Bouteligier, S., 2013. Inequality in new global governance arrangements: the
NortheSouth divide in transnational municipal networks. Innovat. Eur. J. Soc.
Sci. Res. 26 (3), 251e267. https://doi.org/10.1080/13511610.2013.771890.
Bulkeley, H., Davies, A., Evans, B., Gibbs, D., Kern, K., Theobalde, K., 2003. Envi-
ronmental governance and transnational municipal networks in Europe.
J. Environ. Pol. Plann. 5 (3), 235e254. https://doi.org/10.1080/
1523908032000154179.
Bulkeley, H., Schroeder, H., Janda, K., Zhao, J., Armstrong, A., Chu, S., Ghosh, Shibani,
2011. The role of institutions, governance, and urban planning for mitigation
and adaptation. In: Hoorneweg, D., Freire, L., Lee, M., Bhada-Tata, P., Yuen, B.
(Eds.), Cities and Climate Change: Responding to an Urgent Agenda. World
Bank, Washington DC, pp. 125e159. https://doi.org/10.1596/978-0-8213-8493-
0.
Busch, H., 2015. Linked for action? An analysis of transnational municipal climate
networks in Germany. Int. J. Urban Sustain. Dev. 7 (2), 213e231. https://doi.org/
10.1080/19463138.2015.1057144.
6
The two cities having API scores of 8 without participating in any of the net-
works analysed are: Marseille, France, and Sacramento, USA.
7
See also: C40 City Advisors: https://www.c40.org/programmes/city_advisers.
M. Heikkinen et al. / Journal of Cleaner Production 257 (2020) 1204748
Busch, H., Bendlin, L., Fenton, P., 2018. Shaping local response ethe influence of
transnational municipal climate networks on urban climate governance. Urban
Climate 24, 221e230. https://doi.org/10.1016/j.uclim.2018.03.004.
C40, 2017a. Personal Communication. A Recorded Conversation with an Employee
on 22.5.2017. For Further Information, Please Contact the Corresponding
Author.
C40. Cities leading the way: seven climate action plans to deliver on the Paris
Agreement. report. https://resourcecentre.c40.org/resources#cities-leading-
the-way (accessed 21.11.2018).
Cast
an Broto, V., Bulkeley, H., 2013. A survey of urban climate change experiments
in 100 cities. Global Environ. Change 23, 92e102.
Climate Change Laws of the World database. Grantham Research Institute on
Climate Change Environment and Sabin Center for Climate Change Law.
Available at: climate-laws.org.
Davidson, K., Gleeson, B., 2015. Interrogating urban climate leadership: toward a
political ecology of the C40 network. Global Environ. Polit. 15 (4), 21e38.
https://doi.org/10.1162/GLEP_a_00321.
Desmarais, B.A., Harden, J.J., 2013. Testing for zero inflation in count models: bias
correlation for the Vuong test. STATA J. 13 (4), 810e835.
Dupuis, J., Biesbroek, R., 2013. Comparing apples and oranges: the dependent var-
iable problem in comparing and evaluating climate change adaptation policies.
Global Environ. Change 23 (6), 1476e1487. https://doi.org/10.1016/
j.gloenvcha.2013.07.022.
Fünfgeld, H., 2015. Facilitating local climate change adaptation through trans-
national municipal networks. Curr. Opin. Environ. Sustain. 12, 67e73.
Greene, W.H., 1994. Accounting for Excess Zeros and Sample Selection in Poisson
and Negative Binomial Regression Models, Working Papers94-10. New York
University, Leonard N. Stern School of Business, Department of Economics.
Gupta, J., Pfeffer, K., Verrest, H., Ros-Tonen, M., 2015. Geographies of Urban
Governance. Advanced Theories, Methods and Practices. Springer. https://
doi.org/10.1007/978-3-319-21272-2.
Heidrich, O., Reckien, D., Olazabal, M., Foley, A., Salvia, M., de Gregorio Hurtado, S.,
Orru, H., Flacke, J., Geneletti, D., Pietrapertosa, F., Hamann, J.J.-P., Tiwary, A.,
Feliu, E., Dawson, R., 2016. National climate policies across Europe and their
impacts on cities strategies. J. Environ. Manag. 168, 36e45. https://doi.org/
10.1016/j.jenvman.2015.11.043.
Heikkinen, M., Yl€
a-Anttila, T., Juhola, S., 2018. Incremental, reformistic or trans-
formational: what kind of change do C40 cities advocate to deal with climate
change? J. Environ. Pol. Plann. https://doi.org/10.1080/1523908X.2018.1473151.
Hunt, A., Watkiss, P., 2011. Climate change impacts and adaptation in cities: a re-
view of the literature. Climatic Change 104 (2011), 13e49. https://doi.org/
10.1007/s10584-010-9975-6.
ICLEI - Local Governments for Sustainability, 2018. The ICLEI Montr
eal Commitment
and Strategic Vision 2018-2024 (Bonn, Germany).
Juhola, S., Westerhoff, L., 2011. Challenges of adaptation to climate change across
multiple scales: a case study of network governance in two European countries.
Environ. Sci. Pol. 14 (3), 239-24.
Kern, K., Bulkeley, H., 2009. Cities, europeanization and multi-level governance:
governing climate change through transnational municipal networks.
J. Commun. Media Stud. 47 (2), 309e332.
Lee, T., Koski, C., 2015. Multilevel governance and urban climate change mitigation.
Environ. Plann. C Govern. Pol. 33, 1501e1517. https://doi.org/10.1177/
0263774X15614700.
Lee, T., 2015. Global Cities and Climate Change. The Translocal Relations of Envi-
ronmental Governance. Routledge, New York.
Lee, T., 2018. A network comparison of socialization, learning and collaboration in
the C40 cities climate leadership group. J. Environ. Pol. Plann. 21, 104e115.
https://doi.org/10.1080/1523908X.2018.1433998.
Lee, T., van de Meene, S., 2012. Who teaches and who learns? Policy learning
through the C40 cities climate network. Pol. Sci. 45, 199e220. https://doi.org/
10.1007/s11077-012-9159-5.
Lesnikowski, A., Ford, J., Biesbroek, R., Berrang-Ford, L., Heymann, S.J., 2016. Na-
tional-level progress on adaptation. Nat. Clim. Change 6 (3), 261e264. https://
doi.org/10.1038/nclimate2863.
Raykov, T., Markoulides, G.A., 2013. Basic Statistics: an Introduction with R. Rowman
&Littlefield publishers.
Reckien, D., Flacke, J., Dawson, R.J., Heidrich, O., Olazabal, M., Foley, A., Hamann, J.,
Orru, H., Salvia, S., De Gregorio Hurtado, S., Geneletti, D., Pietrapertosa, F., 2014.
Climate change response in Europe: what’s the reality? Analysis of adaptation
and mitigation plans from 200 urban areas in 11 countries. Climatic Change 122
(1e2), 331e340. https://doi.org/10.1007/s10584-013-0989-8.
Reckien, D., Salvia, M., Heidrich, O., Church, J.M., Pietrapertosa, F., De Gregorio-
Hurtado, S., D’Alonzo, V., Foley, A., Simoes, S.G., Eli
sKrko
s, Lorencov
a, Orru, H.,
Orru, K., Wejs, A., Flacke, J., Olazabal, M., Geneletti, D., Feliu, Efr
e, Vasilie, S.,
Nador, C., Krook-Riekkola, A., Matosovi
c, M., Fokaides, P.A., Ioannou, B.I.,
Flamos, A., Spyridaki, N.-A., Balzan, M.V., Fül€
op, O., Paspaldzhiev, I., Grafakos, S.,
Dawson, R., 2018. How are cities planning to respond to climate change?
Assessment of local climate plans from 885 cities in the EU-28. J. Clean. Prod.
https://doi.org/10.1016/j.jclepro.2018.03.220.
Revi, A., Satterthwaite, D.E., Arag
on-Durand, F., Corfee-Morlot, J., Kiunsi, R.B.R.,
Pelling, M., Roberts, D.C., Solecki, W., 2014. Urban areas. In: Field, C.B.,
Barros, V.R., Dokken, D.J., Mach, K.J., Mastrandrea, M.D., Bilir, T.E., Chatterjee, M.,
Ebi, K.L., Estrada, Y.O., Genova, R.C., Girma, B., Kissel, E.S., Levy, A.N.,
MacCracken, S., Mastrandrea, P.R., White, L.L. (Eds.), Climate Change 2014:
Impacts, Adaptation, and Vulnerability. Part A: Global and Sectoral Aspects.
Contribution of Working Group II to the Fifth Assessment Report of the Inter-
governmental Panel on Climate Change [Field. Cambridge University Press,
Cambridge, United Kingdom and New York, NY, USA, pp. 535e612.
Shi, L., Chu, E., Anguelovski, I., Aylett, A., Debats, J., Goh, K., Schenk, T., Seto, K.,
Dodman, D., Roberts, D., Timmons Roberts, J., VanDeveer, S.D., 2016. Roadmap
towards justice in urban climate adaptation research. Nat. Clim. Change 6 (2),
131e137. https://doi.org/10.1038/nclimate2841.
Solecki, W., Rosenzweig, C., Dhakal, S., Roberts, D., Salisu Barau, A., Schultz, S., Ürge-
Vorsatz, D., 2018. City transformations in a 1.5
C warmer world. Nat. Clim.
Change 8, 174e185.
Sugiawan, Y., Kurniawan, R., Managi, S., 2019. Are carbon dioxide emission re-
ductions compatible with sustainable wellbeing? Applied Energy 242, 1e11.
https://doi.org/10.1016/j.apenergy.2019.03.113.
UN-DESA, 2015. World Urbanization Prospects: the 2014 Revision. ST/ESA/SER.A/
366, United Nations, New York.
UNEP, 2011. In: Fischer-Kowalski, M., Swilling, M., von Weizs€
acker, E.U., Ren, Y.,
Moriguchi, Y., Crane, W., Krausmann, F., Eisenmenger, N., Giljum, S.,
Hennicke, P., Romero Lankao, P., Siriban Manalang, A., Sewerin, S. (Eds.),
Decoupling Natural Resource Use and Environmental Impacts from Economic
Growth, A Report of the Working Group on Decoupling to the International
Resource Panel. United Nations Environment Programme, ISBN 978-92-807-
3167-5.
van der Heijden, J., 2017. Innovations in Urban Climate Governance. Cambridge
University Press, Cambridge.
van der Heijden, J., 2018. City and subnational governance. High ambitions, inno-
vative instruments and polycentric collaborations? In: Jordan, A. (Ed.), Gov-
erning Climate Change. Polycentricity in Action. Cambridge University Press
(edit.).
van der Heijden, J., Bulkeley, H., Certom
a, C., 2019. Promises and concerns of the
urban century. Increasing agency and contested empowerment. In: van der
Heijden, J., Bulkeley, H., Certom
a, C. (Eds.), Urban Climate Politics: Agency and
Empowerment. Cambridge University Press. https://doi.org/10.1017/
9781108632157.
van der Heijden, J., Patterson, J., Juhola, S., Wolfram, M., 2018. Advancing the role of
cities in climate go- vernanceepromise, limits, politics. J. Environ. Plann.
Manag. 62, 365e373. https://doi.org/10.1080/09640568.2018.1513832.
Vuong, Q., 1989. Likelihood ratio tests for model selection and non-nested hy-
potheses. Econometrica 57 (2), 307e333.
Wolfram, M., van der Heijden, J., Juhola, S., Patterson, J., 2019. Learning in urban
climate governance: concepts, key issues and challenges. J. Environ. Pol. Plann.
21 (1), 1e15.
Woodruff, S., Stults, M., 2016. Numerous strategies but limited implementation
guidance in US local adaptation plans. Nat. Clim. Change 6, 796e802. https://
doi.org/10.1038/NCLIMATE3012.
Woodruff, S., 2018. City membership in climate change adaptation networks. En-
viron. Sci. Pol. 84, 60e68.
Zeileis, A., Kleiber, C., Jackman, S., 2008. Regression models for count data in R.
J. Stat. Software 27 (8).
M. Heikkinen et al. / Journal of Cleaner Production 257 (2020) 120474 9
