Energy Economics 110 (2022) 106004

Contents lists available at ScienceDirect

Energy Economics
journal homepage: www.elsevier.com/locate/eneeco

Environmental performance, green finance and green innovation: What’s

the long-run relationships among variables?
Quan-Jing Wang a, Hai-Jie Wang a, *, Chun-Ping Chang b, *
a
Business School, Zhengzhou University, 100 Kexue Road, Zhengzhou, Henan 450001, China
b
Shih Chien University, Kaohsiung, Taiwan

A R T I C L E I N F O A B S T R A C T

JEL: What are the relationships among environmental performance, green finance, and green innovation in devel­
C01 oping countries? Existing literatures support the impact of green finance or green innovation on environmental
C32 performance, but rare studies query the cointegration among such three variables. We thus utilize the yearly data
Q40
of 57 developing countries from 2002 to 2016 to empirically examines the relationships among environmental
Keywords: performance, green finance, and green innovation, via panel covariate-augmented Dickey-Fuller unit root test
Green innovation
and the Westerlund and Edgerton (2007) cointegration test. Overall, this study confirms the existence of coin­
Environmental performance
Green finance
tegration relationships among these variables. Moreover, the results of pooled mean group estimation suggest
Cross-section dependence that environmental performance can positively affect the green innovation in the long term in the non-emerging
Cointegration countries and countries with better green innovation or environmental performance. Furthermore, green finance
positively affects green innovation in the emerging countries and countries with lower level of green finance,
while green finance negatively affects green innovation in the countries with better green innovation or envi­
ronmental performance. Our empirical findings offer important policy implications for the sample of developing
countries to promote green innovation and improve environmental performance.

1. Introduction their attitude toward solving environmental degradation and also rep­
resents innovation in green energy production and the improvement of
In response to worldwide climate change and increasing concerns energy utilization efficiency (Wang et al., 2021b; Wang et al., 2022;
about environmental issues from individuals, governments have been Zheng et al., 2021).
carrying out a wide array of policies to promote green innovation such as Green innovation and green finance in recent years have turned into
green brands, environmentally friendly technologies, etc. (Hoffmann, priority measures through the use of resource and carbon intensive
2007; Yung et al., 2011) as well as green finance such as green bonds to growth models, and their positive influence on environmental quality is
deal with environmental degradation (Zhou et al., 2020; Lee et al., widely accepted by scholars (Yin et al., 2019; Kraus et al., 2020; Yang
2021).1,2 Green innovation is essential in public management since the et al., 2021). However, previous studies investigating green innovation
externality of environmental pollution is worsening (Fujii et al., 2013). mainly discussed it from the perspective of financial support on clean
Complementary to the green innovation process for environmental energy R&D (Liu et al., 2019; Yu et al., 2021; Lee et al., 2021), as well as
protection is green finance, which allocates financial capital to the from the perspective of environmental protection expenditure (Yang
research. and development (R&D) of clean energy and environmentally et al., 2021). Scant literature has captured green innovation via green
friendly products and processes (Lee et al., 2020; Wen et al., 2022). trademark or patent applications on environmental management, which
However, while green finance just represents the attitude and the efforts is the true index of green innovation (Zheng et al., 2021; Yu et al., 2021;
of governments to protect the environment, green innovation stands for Tolliver et al., 2021).3 Furthermore, few studies have investigated

* Corresponding authors.
E-mail addresses: [email protected] (H.-J. Wang), [email protected] (C.-P. Chang).
1
Green innovation is defined as: “the development of products (goods and services), processes, marketing methods, organizational structure, and new or improved
institutional arrangements which, intentionally or not, contribute to a reduction of environmental impact in comparison with alternative practices” (OECD, 2009).
2
Green finance is defined as: “all forms of investment or lending that consider environmental effects and enhance environmental sustainability” (Volz et al., 2015).
3
Green innovation is measured by the total amount of patent applications for environment-related technologies, such as air pollution abatement, water pollution
abatement, waste management, soil remediation, and environmental monitoring (OECD, 2019).

https://doi.org/10.1016/j.eneco.2022.106004
Received 18 October 2021; Received in revised form 23 March 2022; Accepted 28 March 2022
Available online 4 April 2022
0140-9883/© 2022 Elsevier B.V. All rights reserved.
Q.-J. Wang et al. Energy Economics 110 (2022) 106004

whether bi-directional cointegration relationships exist among envi­ variables. Once stationarity is determined, we further test the cointe­
ronmental performance, green finance, and green innovation, nor have gration relationship among such variables by utilizing the Westerlund
they empirically tested whether environmental performance and green and Edgerton’s (hereafter WE, 2007) panel cointegration test. Further­
finance can affect green innovation. Identifying the nexus among envi­ more, we study the impacts of environmental performance and green
ronmental performance, green finance, and green innovation can help finance on green innovation based on the Pooled Mean Group (PMG)
governments better understand the relationships among these aspects estimation from the perspective of long-run and short-run. We finally
and spur green products or processes based on the current performance carry out a robustness test by dividing the whole sample into 8 sub-
of environmental quality and development of green finance. Further­ samples.
more, investing these issues can help governments obtain better envi­ The potential contributions of this paper are given as follows. First,
ronmental performance and economic development by conducting unlike previous studies that investigated the unidirectional causal link
corresponding policies to stimulate the green innovation and improve from green innovation to environmental performance or the unidirec­
the allocation of financial resource, which can eventually better promote tional impact of green finance on environmental performance, our study
the whole world to develop in a sustainable, green and environment looks into the cointegration relationships among environmental per­
friendly manner. Hence, this study aims to uncover the nexus among formance, green finance, and green innovation using a yearly balanced
these three variables as well as to identify the short-run and long-run panel dataset for 57 countries during the period from 2002 to 2016 via
influences of environmental performance and green finance on green the panel Westerlund and Edgerton (2007) cointegration test, which can
innovation. incorporate potential CSD (Wang et al., 2021a). Second, we analyze the
There are several reasons to expect a long-run equilibrium among long-run influences of environmental performance and green finance on
environmental performance, green finance, and green innovation in green innovation from the level of the whole panel and two-sub-samples
developing countries. First, green innovation can improve the efficiency of emerging and non-emerging economies. Third, due to the effect of
of energy utilization and bring about more renewable energies, which green finance and environmental performance on green innovation may
benefit for the environmental performance (López et al., 2011; López vary among countries with different levels of environmental perfor­
and Palacios, 2014). Second, more green innovation indicates the better mance, green finance, and environmental management innovation, we
phenomenon of eco-innovation and the ability to bring about more clean further construct 6 sub-samples (High-GI and Low-GI sub-sample, High-
energy, which can attract further international financial support on GF and Low-GF sub-sample, High-EPI and Low-EPI sub-sample), so as to
clean energy R&D and renewable energy production, meaning study whether the relationships among environmental performance,
improving that green finance promotes environmental protection. green finance, and environmental management innovation differ be­
Finally, as suggested by Haller and Murphy (2012), more green inno­ tween them. Finally, previous studies focusing on green innovation often
vation pushes entrepreneurs to invest more into the process of green utilize firms of one country as the sample (Yu et al., 2021), we collect
technologies, thus bringing about more international financial support data of 57 countries, which can offer more common results and policy
and better environmental performance. Based on the above analyses, we implications for developing countries. Our results discover a long-run
prefer that green innovation plays a critical role in the green finance and equilibrium among environmental performance, green finance, and
environmental performance. green innovation, implying that policymakers should consider these
The causal link from environmental performance and green finance factors simultaneously, because a change in one factor may affect the
to green innovation can be understood as follows. First, better envi­ other two.
ronmental performance often promotes the individuals to form habitats The rest of this paper is organized as follows. Section 2 provides a
that need less air pollutants, which may spur policymakers to carry out literature review of related studies. Section 3 offers the measurement of
more green innovation (Yang et al., 2021). Second, Carrión-Flores and variable, data source as well as methodology. Section 4 presents the
Innes (2010) suggest that better environmental performance often main findings and discussions. Section 5 concludes and offers the policy
means more green technologies, the process of which will continue since implication of this study.
the programs of green innovation are usually long-term. Finally, more
green financial support makes governments put up more expenditure on 2. Literature review and hypothetical development
clean energy and renewable energy production R&D, which may bring
about further green innovation. Based on these, we can infer that green 2.1. Environmental performance and green innovation
innovation is affected by environmental performance and green finance.
Based on the above analysis, we thus utilize balanced data of 57 A vast strand in the literatures about green innovation often focused
countries from 2002 to 2016 and employ indicators of environmental on whether the green innovation is beneficial to improve the environ­
performance, green finance, and environmental management innova­ mental performance (Chen and Lee, 2020; Wang and Lee, 2022). For
tion to carry out empirical tests.4 To begin, we investigate whether there instance, Robinson and Stubberud (2015) investigated whether the ef­
is cross-section dependence (CSD) for these variables. Hence, we employ fect of green innovation on environmental performance is constant
the panel covariate-augmented Dickey-Fuller (CADF) unit root test, among small-, medium-, and large-size businesses, concluding that large
which can allow for potential CSD, to test the stationarity of these businesses prefer to protect the environment by promoting the process of
green innovation. By employing a structural equation modeling esti­
mation and data of 669 manufacturing firms in the United Arab Emir­
ates, Singh et al. (2020) pointed out that firm’s environmental
4
The sample countries for our study are limited as data for green finance only performance is affected by its process of green innovation. Kraus et al.
cover 57 developing countries from 2002 to 2016. The other reason why we (2020) utilized data of 244 Malaysia large manufacturing firms and
choose developing countries as the sample countries is that compared to structural equation modeling estimation to study the role of green
developed countries, the former’s producing activities and allocation of finan­ innovation in environmental performance, finding that green innova­
cial resources more likely present some negative externality to the environ­
tion is an important channel through which can both green intellectual
ment. Since developing countries have more pressure and motivation to spur
capital and human resource management promote better environment.
economic growth, under the era of global warming, the manner of economic
development of these countries should be environment friendly. It is thus The viewpoint that green innovation contributes to better environment
necessary to uncover the relationships among environmental performance, is also supported by Chiou et al. (2011) who utilized 124 companies in
green finance, and green innovation for developing countries, which can Taiwan and a structural equation modeling estimation as a test, as well
effectively help the whole world gain sustainable development and a better as Seman et al. (2019) who employed data of 123 Malaysian
environment. manufacturing firms.

2
Q.-J. Wang et al. Energy Economics 110 (2022) 106004

Chen et al. (2015) argued that environment friendly products and investments in the renewable energy sector (Zhou et al., 2020), which
technologies are at the heart of environmental performance, implying can bring about better environmental performance (Romano et al.,
that green innovation benefits environmental performance. Green 2017).
innovation promotes institutions and firms to improve their manner of Aside from these, some scholars have also argued that green finance
production and resource consumption, which can best meet the legal further improves environmental quality by providing support to envi­
criteria of environmental protection (Chan, 2005; Oliva et al., 2019). For ronmentally friendly corporates (Liu et al., 2019; Wen et al., 2021). For
firms, green innovation is often linked with environmental management instance, when banks run green credits, high-polluting corporates are
agendas, indicating that green innovation promotes environmental not permitted in this field (Zhou et al., 2020), and thus financial support
protection (Kammerer, 2009; Adegbile et al., 2017). Zee et al. (2011) on such firms will decrease, which can eventually stimulate the firms to
pointed out that green innovation can reduce a firm’s waste and costs change their manner of production (Chiu and Lee, 2020). Since the
(Day and Schoemaker, 2011; Millard, 2011), which can further help it to profit of green investment is higher than that of traditional projects
reduce the negative environmental impact of its business (Weng et al., (Yuan et al., 2020), stakeholders possess motivation to issue green bonds
2015) as well as improve overall performance (Mahto et al., 2020). or credits, which subsequently promote the firms to produce in an
Other scholars stated that green innovation is an effective way to environmentally friendly manner (Tang and Zhang, 2020).
improve environmental performance, which can bring about a stronger One scholar suggested conversely that green finance negatively af­
competitive advantage (Kratzer et al., 2017; Fousteris et al., 2018). Lin fects environmental performance. He et al. (2019) pointed out that
et al. (2013) proposed that green innovation and environmental stra­ green finance may inhibit a bank’s loan issuances, thus harming the
tegies are critical factors to gain a better environmental performance. efficiency of renewable energy investment and hence decrease envi­
Chan (2005) stated that environmental performance requires the ronmental performance.
participants to meet and exceed societal expectations on natural envi­
ronmental protection (Palazzo and Scherer, 2006). Chen et al. (2015) 2.3. Green finance and green innovation
argued that obeying rules and regulations is not enough to gain better
environmental performance, and that green innovation for products and Owen et al. (2018) investigated the role of green finance in green
processes should be promoted. Similarly, Cai and Zhou (2014) investi­ innovation by studying the longer-term investment requirements of
gated the drivers of green innovation by employing Chinese firms as green innovation activities and pointed out that governments should
samples and concluded that environmental performance can form in­ carry out policies to promote more finance in early-stage green inno­
ternal and external drivers to push firms to conduct more green inno­ vation. Similarly, Liu et al. (2019) pointed out that green finance is the
vation. For instance, environmental performance can bring about key to supporting environmentally friendly enterprises to carry out
stricter environmental regulations, which may promote participants to green projects and processes. Tang and Zhang (2020) also stated that
carry out more green innovation to comply with the regulations and gain green bonds are beneficial for shareholders to promote their firms to
further benefits (Frondel et al., 2007; Darnall et al., 2008; Kammerer, enter into green-related fields, which can then spur green innovation.
2009; Dubey et al., 2015). At the same time, environmental performance Kudratova et al. (2018) argued that green projects bring more profits for
may improve customers’ green demand for better environmental qual­ investors than traditional innovation, and thus green finance may in­
ity, which may spur further environmental innovation (Handfield et al., crease clean energy R&D (Romano et al., 2017). Li et al. (2018) sup­
2002). Moreover, better environmental performance can bring about ported that green bank loans are beneficial to promote green innovation
external competitive pressures for participants to improve environ­ by conducting theoretical analysis according to the green loan theory
mental performance (Li and Ye, 2011), which contributes to better green among government, banks, and firms. Yu et al. (2021) provided that
innovation (Hicks and Dietmar, 2007). green finance can promote firms’ progress of green innovation by
resolving the financing constraints through data of Chinese-listed firms
2.2. Environmental performance and green finance from 2001 to 2017. Zhao and Li (2021) investigated the role of green
finance in green innovation by using a spatial model and proposed that
Some scholars investigated the role of green finance in environ­ green finance benefits green innovation.
mental performance and found that green finance benefits environ­ By browsing the literature above, we realize that that most studies
mental performance. For instance, Zhou et al. (2020) examined the paid attention to the unidirectional causal link from green innovation
influence of green finance on environmental performance by utilizing and green finance to environmental performance and rarely investigates
data of 30 provinces and municipalities in China covering the period the co-movement relationships among these variables, nor have they
2010–2017 and concluded that green finance brings about better envi­ empirically tested the impact of environmental performance and green
ronmental performance. Actually, Nassani et al. (2017) stated that the finance on green innovation. For research about green innovation, most
green financial industry can improve environmental quality by reducing studies utilized data at the firm level for one specific country, such as
emissions of greenhouse gases (GHGs), industrial solid waste, and Malaysia, ignoring any multi-national investigation. Thus, we fill this
wastewater, which is supported by Zhao et al. (2019). Poberezhna gap in the literature by identifying the nexus among environmental
(2018) pointed out that the green economy can better deal with the performance, green finance, and green innovation, by investigating the
threat of environmental degradation such as global water shortages, impacts of environmental performance and green finance on green
because green finance brings environment friendly technologies. innovation from both short-run and long-run perspectives, via multi-
Gianfrate and Peri (2019) argued that the stated that green finance plays national panel data covering 57 countries.
a major role in allocating financial resources to reduce carbon emission,
which is beneficial to achieve the goal of the Paris Agreement. Similarly, 3. Data and methodology
Ziaei (2015) suggested that green financial support helps reduce carbon
dioxide emissions by upgrading the industrial structure. Wang and Zhi 3.1. Variables and data
(2016) proposed that green finance contributes to protect the environ­
ment and sustainably utilize energy and resources. Meanwhile, the 3.1.1. Variable definitions and data source
green finance carried out by governments can spur the private This paper studies the relationships among environmental

3
Q.-J. Wang et al. Energy Economics 110 (2022) 106004

performance, green finance, and innovation via panel country-level data Table 1
covering 57 developing countries during the period 2002–2016 due to Summary of descriptive statistics.
data availability.5 Following Yang et al. (2021), the environmental Variable N Mean SD Min Median Max
performance is captured by the environmental performance index
FULL GI 855 0.127 0.238 0.000 0.013 3.028
(denoted by EPI), data of EPI for some countries is obtained from Yale EPI 855 4.019 0.290 2.843 4.060 4.414
Center for Environmental Law and Policy, while that for other countries GF 855 0.097 0.261 0.000 0.011 3.706
is given by Center for International Earth Science Information Network Emerging GI 240 0.225 0.224 0.000 0.154 1.243
at Columbia University.6 EPI covers two categories such as environ­ EPI 240 4.097 0.270 3.228 4.199 4.396
GF 240 0.026 0.062 0.000 0.005 0.439
mental health and ecosystem vitality, while environmental health Non-emerging GI 615 0.089 0.232 0.000 0.001 3.028
including air quality, sanitation & drinking, water, heavy metals, waste EPI 615 3.989 0.292 2.843 4.039 4.414
management, and so on (Solarin et al., 2017), ecosystem vitality GF 615 0.125 0.300 0.000 0.019 3.706
including biodiversity & habitat, ecosystem services, fisheries, climate
change, pollution emissions, agriculture, and water resources (Wen
0 and 3.028, respectively. For EPI, its mean value, standard error,
et al., 2016). Similar to Buntaine and Pizer (2015) and Lee et al. (2022),
minimum, and maximum are 4.019, 0.290, 2.843, and 4.414, respec­
we capture green finance by the financial support for R&D on clean
tively. For GF, its mean value and standard error are 0.097 and 0.261,
energy and renewable energy production, including hybrid systems
respectively; suggesting that the level of green finance in those countries
(whose unit is million US dollars constant 2016, denoted by GF), which
is too low. For the descriptive statistics of the two sub-samples, we find
is obtained from Our World in Data.7 In line with Zheng et al. (2021),
that the mean value of EPI for the emerging market sub-sample is 4.097,
green innovation is measured by the total amount of patent applications
while that for non-emerging market sub-sample is 3.989. Similarly, the
for environment-related technologies, such as air pollution abatement,
mean of GI for the EM sub-sample is 0.225, while that for the non-EM
water pollution abatement, waste management, soil remediation, and
sub-sample is 0.089, suggesting that the former owns higher green
environmental monitoring technologies, which are obtained from the
innovation and environmental performance than that the latter at the
Organization for Economic Co-operation and Development (OECD)
mean level. For green finance, the mean value of GF for the EM sub-
Statistics (denoted by GI).8 This variable is standardized by the total
sample is 0.026, while that for non-EM sub-sample is 0.125, meaning
population (whose unit is millions). It is worth noting that we take the
that green finance of the former is lower than that of the latter at the
log of data for all these variables.
mean level.
3.1.2. Data description
We first display the scattergram for GI and EPI, as well as GI and GE 3.2. Methodology
in Fig. 1 and Fig. 2, respectively. From Fig. 1, we find that most coun­
tries’ EPI varies from 3.5 to 4.5, while most countries’ GI fluctuates 3.2.1. Cross-sectional dependence
between 0 and 1, and that a small part of observations is larger than 1. Baltagi and Hashem Pesaran (2007) argued that CSD is a common
Furthermore, the fitting lines of GI and EPI suggest that a positive question for panel data. CSD makes the traditional panel estimations
relationship exists between EPI and GI, as better environmental perfor­ invalid (Gengenbach et al., 2009). Based on this reason, our paper first
mance is often linked to a higher level of green innovation. Similarly, we uses the CSD test of Pesaran (2004), as well as that of Pesaran (2015) to
obtain that the distribution of GF is mainly from 0 to 1, and that higher test whether there is CSD.
green finance is often linked to a higher level of green innovation. We define the model as follows:
We next provide the basic statistical distributions of these three
variables in Table 1.9 For the whole sample countries, the mean of GI is yit = αi + βit xit + μit (1)
0.127 and its standard error is 0.238. Its minimum and maximum are Here, i = 1, 2, 3……59, where i stands for individual; t = 1, 2 ……12,
where t represents the time factor; αi denotes the constant parameter
which may vary among individuals; xitis the variables while βit is vector
5
Since the updated data for green innovation end by 2016 and the data for for coefficient of such variables; and μit is the regression residual. The
green finance start at 2002, the sample period is chosen as 2002–2016. Such CSD is calculated as:
countries include Algeria, Argentina, Armenia, Bangladesh, Benin, Bolivia, √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
Brazil, Burkina Faso, Cabo Verde, Cambodia, Cameroon, Chile, China, 2T ∑N− 1 ∑N
Colombia, Costa Rica, Cuba, Dominican Republic, Ecuador, Egypt, El Salvador,
CD = ρ
j=i+1 ij
(2)
N(N − 1) i=1
Ethiopia, Fiji, Georgia, Ghana, Guatemala, Honduras, India, Indonesia, Jordan,
Kazakhstan, Kenya, Madagascar, Malaysia, Mali, Mexico, Mongolia, Morocco, Here, ρij stands for the correlation errors of individuals i and j.
Mozambique, Namibia, Nepal, Nigeria, Pakistan, Peru, Philippines, Senegal, Pesaran (2015) further applied the CSD test which considered
South Africa, Sri Lanka, Tajikistan, Tanzania, Thailand, Tonga, Tunisia, Turkey, heterogenous slope the to a small sample, known by weak CSD test. In
Uganda, Uruguay, Vietnam, and Yemen. this test, the ρij is given by:
6
Website of the Yale Center for Environmental Law & Policy: https://epi.
envirocenter.yale.edu/epi-downloads.Website of the Center for International ∑
T ∧ ∧
uit ujt
Earth Science Information Network: https://sedac.ciesin.columbia.edu/data/
(3)
∧ ∧ t=1
collection/epi. ρ =ρ =(
ij ji )1/2 ( T )1/2
7
∑
T ∧2 ∑ ∧2
https://ourworldindata.org/grapher/international-finance-clean-energy?ta uit ujt
b=table t=1 t=1
8
Website of OECD statistics: https://stats.oecd.org/ Here, u represents the estimation residual.
9
The reason why we divide the full sample into emerging and non-emerging To solve the problem of CSD, the estimations of CADF, Westerlund
economies is given as follows: the emerging markets experience fast economic
and Edgerton (2007)’s cointegration, and PMG are employed to conduct
growth (Feng et al., 2021), which may bring about more environmental
empirical tests, which possess several advantages over other traditional
degradation, the emerging markets possess higher demand in changing the
producing manner via green innovation to balance the economic growth and econometric methods. First, while there exists CSD, traditional estima­
environmental quality, thus the nexus among environmental performance, tions which are constructed under cross-section independence may be
green finance and green innovation is different with that among non-emerging biased. However, the estimations we utilize can incorporate potential
markets. To test the differences between emerging markets and non-emerging CSD that proves the credibility of our empirical results. Second, the
markets, we divide the sample into two sub-samples. estimations of Westerlund and Edgerton (2007)’s cointegration and

4
Q.-J. Wang et al. Energy Economics 110 (2022) 106004

PMG can solve the endogeneity and heterogeneity problems, which are the 1% level, indicating that EPI also exhibits CSD. This conclusion is
common problems that traditional estimations cannot solve. Third, the supported by the result of Pesaran (2015) for EPI. Finally, we further test
PMG estimation can identify the impacts of green finance and environ­ the CSD for GF. From the results of Pesaran (2004) and Pesaran (2015),
mental performance on green innovation from the perspective of long- we find that the statistics are 9.17 and 56.754, respectively, and both are
run and short-run in a single equation. significant at 1%, confirming that GF also exhibits CSD. When we look at
the results of Emerging and Non-emerging sub-samples, the results
3.2.2. Panel unit root test provide similar proof that all three variables exhibit CSD.
If there exists CSD, the unit root test method which can incorporate Since the results of Pesaran (2004) and Pesaran (2015) tests show
CSD should be utilized. This study selects the Pesaran (2007)’s CADF test that there exist CSD problems for these three variables in our panel data,
to examine the stability of all variables, mainly because the CADF test the traditional unit root test and cointegration tests are both valid. To
considers CSD, while the so-called first-generation panel data unit root obtain credible results for the relationships among the three variables,
test does not. The model of CADF unit root test is given as follows: the panel CADF unit root test and Westerlund and Edgerton (2007)
ρ ρ cointegration test (which can incorporate CSD) should be employed.
∑ ∑
△yit = δi + ϑi yit-1 + ϕi yit− 1 + λij △yit− 1 + νij △yt− j + dit + εit (4)
j=1 j=0 4.2. Results of the panel unit root test
∑N
Here, yt = 1
i=1 yit , ϑ is the coefficient of first lag term; ϕi,λij and
N Table 3 provides the results of stationary test for the full sample and
νijrepresents the individual specific effect, linear trend, and common two sub-samples by employing CADF unit root test, which under the null
time effect, respectively; The stationarity test is based on the t-statistic of hypothesis that the variable is not stationary. The statistic of GI for full
ϑ according to the model of CADF unit root test. sample is − 1.265, which does not pass the significance test at the 10%
level, confirming that the level of GI is not stationary. Furthermore, the
3.2.3. Panel cointegration statistic for ΔGI is − 3.086, which passes the significance test at the 1%
We next use the Westerlund and Edgerton (2007) panel cointegration level and suggests that the first difference of GI is stationary. We obtain
test to investigate the long-run relationships among EPI, GF, and GI, that GI follows I (1) progress from an earlier analysis. Similarly, we find
which owns the advantages of a large sample and can solve the problem that the variables of EPI and GF both follow I (1) progress as well. The
of heterogeneity. This model proposes four statistics based on residual, results for the two sub-samples also confirm this conclusion.
such as Gt, Ga, Pt, and Pa. Pτ and Pα are utilized to test whether there is
cointegration in at least one unit, and Gτ and Gα test whether the panel 4.3. Results of the panel cointegration test
as a whole is cointegrated. The model is:

yit = ϑ0i + ϑ1it + ni Dit + ϕi yit− 1 + xit ’ βi + (Dit xit )’ ςi + Ζit (5) Since the panel CADF unit root test shows that these three variables
exhibit I (1) progress, we thus utilize the Westerlund and Edgerton
Here, xit = xit, t− 1 + υit is I (1) progress; Dit represents the potential (2007) panel cointegration which allow for CSD to uncover the nexus
structural break; if t>Tib, then Dit = 1; otherwise Dit = 0; and Tib rep­ among these three variables. The results of four statistics are listed in
resents the location of breakpoint for individual i. Table 4. The dependent variable is GI, while the independent variables
are EPI, GF, as well as EPI and GF jointly. All statistics for full sample are
3.2.4. Panel long-run and short-run estimates significant at 1%, confirming that GI and EPI move together in the long
As Ahmad et al. (2019) declared, the first-generation models would term. Similar findings are obtained from the models of GI versus GF and
be invalid if there exist problems such as non-stationarity and cross- GI versus EPI & GF. All results for the two sub-samples also support this
stationarity. Ma (2015) argued that the estimation of pooled means conclusion except for the model of GI versus EPI & GF in the Emerging
group (PMG) is powerful and efficient to handle the CSD and non- sub-sample.
stationarity (Balcilar et al., 2019). To solve the problem of non- Our results suggest that environmental performance, green finance,
stationarity and heterogeneity, we utilize PMG estimation to uncover and green innovation move together in the long term, which is consis­
the influences of EPI and GF on GI from both the short-run and long-run tent with Zhou et al. (2020) who argued that green innovation and green
perspectives and solve the problem of CSD (Eberhardt and Teal, 2010; finance play a critical role for environmental performance in China, as
Bond and Eberhardt, 2013). The model is calculated as follows: well as the research of Owen et al. (2018) who supported that green
ΔGI it = ξ0 + ξ1 ΔEPI it + ξ1 ΔGFit + d1 (wt ) + μit (6) finance promotes more green innovation. However, unlike those studies
that only targeted the unidirectional causal link from green innovation
Here, wt is subtracted from the dependent variable, it indicates that or green finance to environmental performance, our study offers more
the general-purpose procedure is applied to each set of unit coefficient. evidence on the bi-directional relationships among these variables. This
wt also stands for any potential excluded idiosyncratic process that may finding suggests that policymakers should simultaneously take envi­
evolve over time. ronmental performance, green finance, and innovation into account. If
governments prefer to promote green innovation, then current envi­
4. Empirical results ronmental performance and green finance should be the focus.

4.1. Results of the panel CSD 4.4. The panel PMG estimation analysis

Since CSD is a common question for panel data, which may lead to We next investigate the short-run and long-run impacts of GF and EPI
biased results if we do not incorporate the potential CSD when we utilize on GI via PMG estimator (Pesaran et al., 1999) which can incorporate
traditional estimation. In line with Wang et al. (2021a), we utilize the CSD. Table 5 illustrates the PMG results when the dependent variable is
test of Pesaran (2004) and Pesaran (2015) to carry out CSD test under GI.
the null hypothesis that there is no CSD or weak CSD, respectively. The As can be seen from Table 5, the coefficient of EPI for the full sample
results of the CSD test appear in Table 2. is 0.006, which passes the significance test at the 5% level. This result
The CSD statistic for GI is 7.64 and significant at 1%, confirming that show that environmental performance exerts a long-run positive influ­
GI exhibits CSD. The result of Pesaran (2015) also provides the same ence on green innovation. The coefficient of GF for the full sample is
findings. Hence, we turn to analyze the results of EPI. The CSD statistic − 0.005, which does not pass the significance test at the 10% level,
of Pesaran (204) for EPI is 152.12, which passes the significance test at which offers strong evidence on that green finance exerts no significant

5
Q.-J. Wang et al. Energy Economics 110 (2022) 106004

Table 2
Cross-section dependence tests.
Pesaran (2004) Pesaran (2015)

Variable CSD-test p-value Corr abs(corr) CSD P

Full GI 7.64*** 0 0.049 0.245 62.905*** 0

EPI 152.12*** 0 0.983 0.983 154.669*** 0
GF 9.17*** 0 0.059 0.217 56.754*** 0
Emerging GI 10.62*** 0 0.250 0.414 33.970*** 0
EPI 42.05*** 0 0.991 0.991 42.416*** 0
GF 8.93*** 0 0.210 0.210 19.737*** 0
Non-emerging GI 2.05** 0 0.018 0.214 32.369*** 0
EPI 108.76*** 0 0.981 0.981 110.859*** 0
GF 4.83*** 0 0.044 0.213 39.176*** 0

Note: ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively.

Innes, 2010). For the insignificant influence of green finance on green

Table 3
innovation, the reason may be that such finance may be allocated to­
Results from the CADF panel unit root test.
ward introducing more utilization of environment friendly equipment or
GI ΔGI EPI ΔEPI GF ΔGF existing clean energies to reduce the environmental adverse effect,
Full rather than engaging into the research and development of environ­
− 1.265 − 3.086*** − 1.662 − 2.305*** − 1.795 − 2.809*** mental management technologies.
(3.122) (− 9.376) (0.401) (− 4.013) (− 0.515) (− 7.477)
For the two-sub samples, the results in the Non-emerging group are
in line with that of the full sample, while the results of the Emerging
Emerging group run contrary to that of the full sample. For the Emerging sub-
− 1.472 − 2.921*** − 1.500 − 2.478*** − 1.809 − 2.712***
sample, the coefficient of EPI is 0.002 and does not pass the signifi­
(0.875) (− 4.446) (0.770) (− 2.818) (− 0.364) (− 3.677)
cance test at the 10% level. This result confirms that the green innova­
tion cannot be affected by environmental performance among emerging
Non-emerging
− 1.834 − 3.020*** − 0.992 − 2.220*** − 1.545 − 2.734***
markets. Furthermore, the coefficient of GF is 0.068, which passes the
(− 0.669) (− 7.638) (4.276) (− 2.937) (1.030) (− 5.957) significance test at the 1% level, implying that green finance does exert a
positive impact on green innovation among emerging markets in the
Notes: The statistic of CADF is t-bar, and Z[t-bar] is in parentheses. ***, **, and *
long run. The potential reason for this difference runs as follows. Since
denote statistical significance at the 1%, 5%, and 10% levels, respectively.
emerging markets experience fast economic growth, under the era of
global environmental protection, green finance can help them achieve a
impact on green innovation. The potential reasons for this phenomenon
cleaner environment in order to spur sustainable development, such as
are as follows. First, while the environmental performance is better, then
higher utilization of green technologies or clean energy, as well as
citizens can form life manners that depend on a better environment;
investing in the process of new green technologies, thus the green
conversely, if it turns worse afterwards, then more debates and disputes
finance would benefit for the green innovation in the emerging markets.
about the government will arise. Because citizens and firms in general
Table 5 also provides the short-run PMG estimation results. Taking
care about environment problems, environmental protection pressure
the full sample as one example, the coefficient of ECM is − 0.782, which
on policymakers may increase, which pushes policymakers to target
passes the significance test at the 1% level, meaning that EPI, GF, and GI
more green innovation. Furthermore, better environmental performance
move together. However, the coefficients of ΔEPI and ΔGF are not
is often brought by the expenditures on R&D of environment friendly
passing the significance test at 10%, indicating that green finance and
technologies or projects, since the investment on such programs will
environmental performance exert no significant influence on green
continue even though the environment is better, the process of green
innovation. The results for the two-sub-samples are both similar to that
innovation would be promoted in the long term (Carrión-Flores and
of the full sample. These results support that green innovation involves

Table 4
Panel cointegration test (Westerlund and Edgerton, 2007).
Model Gτ - value Gα- value Pτ- value Pα- value

Full GI V.S. EPI − 3.540*** − 11.431*** − 29.317*** − 13.319***

(− 14.812) (− 5.948) (− 18.494) (− 15.457)
GI V.S. GF − 2.946 *** − 11.040 *** − 24.614 *** − 10.971 ***
(− 9.821) (− 5.406) (− 13.764) (− 11.463)
GI V.S. EPI& GF − 3.927 *** − 9.559 − 24.848 *** − 10.729 ***
(− 15.512) (− 0.522) (− 11.538) (− 6.568)
Emerging GI V.S. EPI − 2.405*** − 8.044 − 7.138** − 5.564
(− 2.794) (− 0.663) (− 1.355) (− 1.202)
GI V.S. GF − 2.183** − 6.747*** − 7.727** − 5.565
(− 1.806) (0.290) (− 1.948) (− 1.203)
GI V.S. EPI& GF − 3.619*** − 0.609 − 5.483 − 0.683
(− 6.880) (5.431) (1.326) (3.707)
Non-emerging GI V.S. EPI − 3.983*** − 12.752*** − 26.181*** − 14.155**
(− 15.718) (− 6.599) (− 17.010) (− 14.314)
GI V.S. GF − 3.244*** − 12.716*** − 21.639*** − 11.527***
(− 10.452) (− 6.556) (− 12.441) (− 10.524)
GI V.S. EPI& GF − 4.574*** − 10.360 − 21.783*** − 11.118***
(− 17.651) (− 1.261) (− 10.472) (− 6.016)

Note: ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively.

6
Q.-J. Wang et al. Energy Economics 110 (2022) 106004

Table 5 Table 6
Results of the PMG panel data estimation method. Westerlund and Edgerton (2007) Panel cointegration test for sub-samples.
Dependent variable: GI Model Gτ - value Gα- value Pτ- value Pα- value

Full Emerging Non-emerging High- GI-EPI − 3.859*** − 10.516*** − 20.675*** − 13.370***

GI (− 12.257) (− 3.279) (− 13.091) (− 10.894)
Long-run
GI - GF − 2.659*** − 8.966** − 17.292*** − 10.986***
EPI 0.006** 0.002 0.006*
(− 5.193) (− 1.773) (− 9.688) (− 8.052)
(2.12) (0.40) (1.86)
GI – EPI& − 4.766*** − 8.840 − 17.479*** − 10.780***
GF − 0.005 0.068*** − 0.006
GF
(− 0.92) (3.27) (− 1.24)
(− 15.694) (0.241) (− 8.149) (− 4.652)
CONS 0.055*** 0.094*** 0.044***
Low-GI GI-EPI − 3.233*** − 12.314 − 17.290*** − 11.777***
(4.67) (4.41) (3.15)
(− 8.722) (− 5.116) (− 9.550) (− 9.154)
GI - GF − 3.223** − 13.043*** − 16.383*** − 10.522***
Short-run (− 8.666) (− 5.837) (− 8.637) (− 7.632)
ECM − 0.782*** − 0.511*** − 0.897*** GI – EPI& − 3.117*** − 10.253 − 15.940*** − 9.368
(− 17.15) (− 6.01) (− 19.66) GF
ΔEPI − 0.045 − 0.059 − 0.039 (− 6.326) (− 0.968) (− 6.503) (− 3.374)
(− 0.47) (− 0.59) (− 0.30) High- GI-EPI − 4.197*** − 12.871*** − 21.679*** − 14.302***
ΔGF 0.606 0.414 0.673 GF (− 14.249) (− 5.568) (− 14.100) (− 12.005)
(1.13) (0.80) (0.93) GI - GF − 3.297*** − 13.769*** − 18.519*** − 12.347***
(− 8.946) (− 6.441) (− 10.922) (− 9.674)
Notes: ***, **, and * denote statistical significance at the 1%, 5%, and 10% GI – EPI& − 5.203*** − 10.390 − 18.102*** − 11.546***
levels, respectively. Z statistics are in parentheses. GF
(− 18.205) (− 1.068) (− 8.752) (− 5.376)
Low- GI-EPI − 2.906*** − 10.040*** − 15.144*** − 8.819***
long-term activities that may take more than 3–5 years for fruition.
GF (− 6.764) (− 2.867) (− 7.391) (− 5.566)
Thus, when governments evaluate the effects of environmental perfor­ GI - GF − 2.608*** − 8.406 − 11.813*** − 5.871**
mance and green finance, they should pay more attention to their long- (− 4.978) (− 1.250) (− 4.040) (− 1.991)
term influence and not any short-run impacts. GI – EPI& − 2.695*** − 8.756 − 14.357*** − 7.232*
GF
(− 3.859) (0.317) (− 4.969) (− 1.317)
High- GI-EPI − 3.894*** − 11.200*** − 21.141*** − 13.809***
4.5. Robustness test EPI (− 12.466) (− 3.944) (− 13.559) (− 11.417)
GI - GF − 2.855*** − 9.788*** − 18.530*** − 12.201***
4.5.1. Sub-samples (− 6.349) (− 2.572) (− 10.933) (− 9.501)
GI – EPI& − 4.852*** − 9.398 − 17.938*** − 11.236***
To prove the credibility of our earlier findings, we carry out empir­
GF
ical test based on several sub-samples based on the levels of GI, EPI, and (− 16.186) (− 0.231) (− 8.593) (− 5.083)
GF. Specifically, if the mean value of GI for one country is larger than the Low- GI-EPI − 3.198*** − 11.653*** − 17.874*** − 10.953***
median value of GI for the whole sample, then we put it into the High-GI EPI (− 8.516) (− 4.462) (− 10.137) (− 8.155)
sub-sample; otherwise, we put it into the Low-GI sub-sample. Similarly, GI - GF − 3.034*** − 12.249*** − 12.135*** − 6.520***
(− 7.531) (− 5.052) (− 4.365) (− 2.777)
we also set other 4 sub-samples according to the median of EPI and GF. GI – EPI& − 3.034*** − 9.714 − 15.063*** − 8.375***
Based on such sub-samples, we re-conduct the empirical investigations. GF
At first, we carry out the panel CADF unit root test for such six sub- (− 5.842) (− 0.505) (− 5.653) (− 2.417)
samples, whose results suggest that the levels of GI, EPI, and GF are Note: ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels,
not stationary, while the first differences of these three variables are respectively.
stationary, offering strong evidence that the three variables follow I (1)
progress.10
performance positively affects green innovation, while green finance
Since these variables exhibit I (1) progress for each sub-sample, we
exerts no influence on green innovation in the long run.
further study the cointegration relationships among the three variables
For the High-GF sub-sample, both environmental performance and
by employing the panel cointegration given by Westerlund and Edgerton
green finance exert no impact on green innovation. For the Low-GF sub-
(2007) for these 6 sub-samples. Table 6 displays the results. We first
sample, the results show that environmental performance exerts no in­
analyze the results of the High-GI sub-sample. For the model of GI versus
fluence on green innovation, while green finance positively influences
EPI, the four statistics are all significant at 1%, confirming the cointe­
green innovation in the long run. For the High-EPI sub-sample, envi­
gration between environmental performance and green innovation. The
ronmental performance exerts no impact on green innovation, while
model of GI versus GF and GI versus EPI & GF offer similar conclusion.
green finance exerts a negative influence on green innovation. For the
The results for the other five sub-samples are also in line with that of the
Low-EPI sub-sample, the results show that environmental performance
model GI versus EPI. These results suggest that environmental perfor­
positively affects green innovation, while green finance exerts no impact
mance, green finance, and green innovation move together, which are
on green innovation.
similar to the baseline results.
The potential reasons for the differences among such sub-samples are
We next examine the short-run and long-run influences of environ­
as follows. Some substitute effects exist between green innovation and
mental performance and green finance on green innovation based on our
green finance when promoting environmental performance (Tolliver
sub-samples by employing the PMG estimation. For the High-GI sub-
et al., 2021). For countries with higher levels of green innovation and
sample, the coefficient of EPI is 0.02, which does not pass the signifi­
better environmental performance, more green finance may be allocated
cance test at the 10% level, implying that the environmental perfor­
to the spread of existing green technologies or equipment in order to
mance can not affect the green innovation in the long term. The
gain sustainable development (Chiu and Lee, 2020), rather than invest
coefficient of GF is − 0.098, which is significant at 1% and suggests that
in R&D activities of new green technologies. This could lead to less
the green finance exerts a negative influence on green innovation. From
attention by governments and society toward green innovation, causing
the result of the Low-GI sub-sample, we find that environmental
a long-term negative impact of green finance on innovation (Wang et al.,
2021a). Similarly, for countries with higher levels of green finance,
10
environmental quality can be improved by its utilization of green
To save space, these results are not reported, but are available upon request.

7
Q.-J. Wang et al. Energy Economics 110 (2022) 106004

finance, which would reduce the importance of green innovation in long run. Furthermore, environmental performance and green finance
environmental protection, thus the environmental performance and exert no impact on green innovation in the short run. Referring to the
green finance cannot promote the green innovation. long-run results of PMG estimation for the 6 sub-samples, environmental
Table 7 also illustrates the short-run PMG estimation results for these performance positively affects green innovation in the High-GI and Low-
sub-samples. The coefficients of ECM are all significant at 1%, offering EPI sub-samples, while environmental performance exerts no impact on
strong evidence that EPI, GF, and GI move together. However, the co­ green innovation in the other sub-samples. Green finance does positively
efficients of ΔEPI and ΔGF are not significant at 10%, except for the affect green innovation in the Low-GF sub-sample and negatively affects
coefficient of ΔEPI in the Low-GI subsample, suggesting that there is no green innovation in the High-GI and High-EPI sub-samples, while it
significant influence from environmental performance and green exerts no impact on green innovation in the other sub-samples.
finance to green innovation in the short run, except for the Low-GI sub- From the authors’ best knowledge, our study is the first one to
sample, in which environmental performance does negatively affect investigate the cointegration among environmental performance, green
green innovation. finance, and green innovation, to examine the impacts of environmental
Since green finance exerts a positive influence on green innovation performance and green finance on green innovation from both short
for countries with higher levels of green finance, governments that have term and long term, and to explore the difference of these relationships
lower green finance should put forth greater efforts to guide its devel­ among various countries.
opment, such as more public transfers or tax incentives. For environ­ As a research study on environmental performance, green finance,
mental performance, which positively affects green innovation among and green innovation, our results have important policy implications to
countries with lower EPI or lower green innovation, governments should policymakers and stakeholders.
target to reduce environmental pollutants and promote the process of
environmental innovation, which can improve environmental perfor­ (1) For policymakers: First, governments should carry out contin­
mance in turn. uous or long-term planning to allocate more financial resources
on green products and processes over a longer horizon. Even
4.5.2. Changing the measurement of the variables though green finance may not bring about green innovation
We finally also conduct a robustness test by changing the measure­ immediately, it can spur the progress of green innovation in the
ment of green innovation as renewable energy innovation, which is future. Furthermore, policymakers in countries with lower levels
similar to Zheng et al. (2021). The PMG results by employing the new of green finance should learn from the experiences of countries at
variable appear in Table 8, from which we find that the main empirical higher levels of green finance and carry out more supportive
results are in line with those in Table 5, again supporting that our policies such as tax incentives that will help guide financial
baseline conclusions are credible. capital to flow into green products or processes. Moreover, gov­
Similarly, we also carry out robustness test by employing the alter­ ernments with lower levels of green finance should ensure the
native measurement of environmental performance by utilizing the stability of their green finance policies in the long term to protect
environmental health index, which covers air quality, sanitation & financial support for green innovation, since green finance spurs
drinking, water, heavy metals, waste management, and so on (Solarin more green innovation in these countries, and green financial
et al., 2017, denoted by EHI), whose data source is similar to that of EPI. investors may have to confront some regulatory risks.
The empirical results by employing alternative measurement of envi­ (2) For investors: As Kudratova et al. (2018) argued, the profit on
ronmental performance are given in Table 9. From Table 9, we can find green projects that benefit sustainable development is higher
that the results are similar to those in Table 5 and Table 8, supporting than that of traditional projects. It is thus reasonable for in­
that our earlier conclusions are reliable.11 stitutions that care about environmental performance to spend
more on R&D covering environmental management innovation,
5. Conclusions as well as allocate greater financial resources to green processes
that help improve environmental quality. Moreover, since sus­
This research investigates whether there exist cointegration rela­ tainable development and environmental protection are essential
tionship among environmental performance, green finance, and green goals of the whole world, investors should thus catch this op­
innovation, as well as presents the short-run and long-run influences of portunity to attach more importance to green finance and green
environmental performance and green finance on green innovation in innovation by investing more in green bonds, green leases, green
developing countries. We employ the data for 57 countries from 2002 to trusts, and green funds that can bring about more profits in the
2016 to conduct empirical investigation, by utilizing estimations such as future. Specifically, as investors prefer to fund a program on
the panel CADF unit root test, Westerlund and Edgerton (2007) coin­ green technologies, they should evaluate the program from a
tegration test, and Pooled Mean Group estimation, which can incorpo­ long-term perspective, because it is an appropriate manner to
rate potential cross-section dependence (CSD). gain profits from green innovation.
The results of panel CSD tests suggest that CSD exists for EPI, GF, and (3) For asset managers: More assets and resources should be allo­
GI. Then the stationarity test based on CADF unit root test confirms that cated to green finance such as green bonds and green funds,
these variables all follow I (1) progress. Hence, the panel cointegration which can be utilized to promote green technologies and envi­
test reveals evidence as to the existence of long-run cointegrating re­ ronmental protection. The weight and share of green finance
lationships among environmental performance, green finance, and should also be recognized by asset managers, and the profit rate
green innovation. Moreover, the long-run PMG estimation suggests that of such green finance should be assessed from the long-term
there is a positive link from environmental performance to green inno­ perspective rather than the short-term one.
vation in the full sample and emerging countries, but green finance (4) For regulators: Because green development is still not mature, the
negatively affects green innovation in non-emerging countries in the regulators should carry out strict control on production activities
that have more environmental adverse effects and promote cor­
porates to boost their green innovation so as to produce in an
11 environmental friendly manner. Since green finance in emerging
We also conduct the empirical test by utilizing the method of mean group
and augmented mean group estimation to investigate the long-run impact of markets negatively affects green innovation, which may be
environmental performance and green finance on green innovation, via the caused by a misallocation or misuse of green finance funding,
measurement of energy innovation and environmental health index, whose regulators in emerging markets should hence pay more effort
results are available upon request. toward improving the allocation and efficiency of green finance

8
Q.-J. Wang et al. Energy Economics 110 (2022) 106004

Table 7
PMG estimating result for sub-samples.
GI GI GI GI GI GI

High-GI Low-GI High-GF Low-GF High-EPI Low-EPI

Long-run
EPI 0.020 0.005* 0.004 0.004 0.007 0.005*
(1.02) (1.83) (0.83) (1.37) (0.48) (1.92)
GF − 0.098*** − 0.001 − 0.006 0.044*** − 0.082*** − 0.001
(− 5.36) (− 0.24) (− 0.99) (2.66) (− 4.77) (− 0.21)
CONS 0.086*** − 0.001 0.059*** 0.061*** 0.099*** 0.015*
(4.56) (− 0.02) (3.07) (4.22) (4.70) (1.83)

Short-run
ECM − 0.611*** − 0.943*** − 0.866*** − 0.716*** − 0.721*** − 0.818***
(− 9.14) (− 20.79) (− 14.42) (− 10.27) (− 11.13) (− 12.65)
ΔEPI − 0.071 − 0.023* − 0.099 0.009 0.033 − 0.123
(− 0.36) (− 1.74) (− 0.51) (0.21) (0.21) (− 1.10)
ΔGF 0.227 1.006 0.078 1.010 1.12 0.131
(0.73) (0.99) (1.63) (1.04) (1.04) (0.97)

Notes: ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively. Z statistics are in parentheses.

by using digital technologies to track the utilization of these re­

sources in order to bring about further environmental manage­
Table 8 ment innovation.
Results of the PMG estimations- changing the measurement of GI.
(5) For corporate sector: As long-term cointegration exists between
Dependent variable: energy innovation environmental performance and green innovation, corporates
Full Emerging Non-emerging should invest more long-term financial capital and human capital
on R&D activities of green technologies, such as constructing a
Long-run
EPI 0.004*** 0.001 0.005** research department on green innovation or setting up special
(3.03) (0.96) (2.36) funding for green technologies, which can enhance their sus­
GF − 0.003 0.073*** − 0.002 tainability and competitiveness in the long run. Similarly, cor­
(− 1.35) (4.62) (− 0.91)
porates can also issue green corporate bonds and introduce long-
CONS 0.062*** 0.103*** 0.051***
(3.24) (4.17) (3.90)
term green investors to the capital market so as to attract more
specific finance to promote the progress of green technologies.
Short-run
ECM − 0.531*** − 0.627*** − 0.902*** Due to our limited time and energy, one may further explore the
(− 7.26) (− 5.81) (− 13.72) interaction between environmental performance and green finance on
ΔEPI − 0.038 − 0.047 − 0.040 green innovation, which can better clarify the nexus among these three
(− 1.26) (− 0.92) (− 0.42) aspects. Furthermore, due to data availability, this study only in­
ΔGF 0.587 0.388 0.604
(0.96) (0.65) (0.73)
vestigates the relationships among environmental performance, green
finance, and green innovation between developing countries. An
Notes: ***, **, and * denote statistical significance at the 1%, 5%, and 10% investigation for developed countries should be further conducted once
levels, respectively. Z statistics are in parentheses.
data become available.

Credit author statement

Table 9
Results of the PMG estimations- changing the measurement of environmental
performance. Quan-Jing Wang: Conceptualization, Data curation, Formal anal­
ysis, Writing - original draft, Methodology. Hai-Jie Wang: Software,
Dependent variable: Energy innovation
Writing - review & editing. Chun-Ping Chang: Writing -review &
Full Emerging Non-emerging editing, Funding acquisition, Investigation, Supervision.
Long-run
EHI 0.009** 0.003 0.006*** Acknowledgements
(2.27) (1.46) (3.08)
GF − 0.002 0.069*** − 0.003
(− 0.96) (3.27) (− 1.20)
The author would like to express their gratitude to reviewers and
CONS 0.048*** 0.099*** 0.042*** editor ‘s comments. Hai-Jie Wang thanks funding support from National
(4.07) (3.43) (6.38) Social Science Fund General Project "Research on the Restructuring
Mechanism and Policy System of China’s Manufacturing Global Value
Short-run Chain under the Background of the COVID-19 Pandemic" (Project No.
ECM − 0.492*** − 0.649*** − 0.816*** 20BJY094) and National Social Science Fund Key Project "Research on
(− 5.10) (− 6.08) (− 9.21) Technology Security Strategy of China’s Manufacturing Industry Chain
ΔEHI − 0.044 − 0.053 − 0.035
(− 1.03) (− 0.78) (− 0.64)
under the New Development Pattern" (Project No. 21AZD109). All errors
ΔGF 0.615 0.413 0.664 are our own.
(1.25) (0.83) (1.19)

Notes: ***, **, and * denote statistical significance at the 1%, 5%, and 10%
levels, respectively. Z statistics are in parentheses.

9
Q.-J. Wang et al. Energy Economics 110 (2022) 106004

Appendix A. Appendix

1.5
Green Innovation (GI)
.5 0 1

3 3.5 4 4.5
Environmental Performance (EPI)

Fig. 1. Scatter plot of green innovation and environmental performance. (For interpretation of the references to colour in this figure legend, the reader is referred to
the web version of this article.)
Note: We drop observations whose GI value is larger than 1.5, since the share of it is too small, which may lead to bias when fitting the line between such variables.
1.5
Green Innovation (GI)
.5 0 1

0 .5 1 1.5
Green Finance (GF)

Fig. 2. Scatter plot of green innovation and green finance. (For interpretation of the references to colour in this figure legend, the reader is referred to the web
version of this article.)
Note: We drop observations whose GF value is larger than 2, due to the same reason as in Fig. 1.

Appendix B. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.eneco.2022.106004.

References Baltagi, B.H., Hashem Pesaran, M., 2007. Heterogeneity and cross section dependence in
panel data models: theory and applications introduction. J. Appl. Econ. 22 (2),
229–232.
Adegbile, A., Sarpong, D., Meissner, D., 2017. Strategic foresight for innovation
Bond, S., Eberhardt, M., 2013. Accounting for unobserved heterogeneity in panel time
management: a review and research agenda. Int. J. Innov. Technol. Manag. 14 (04),
series models. Univ. Oxford 1–11.
1750019.
Buntaine, M.T., Pizer, W.A., 2015. Encouraging clean energy investment in developing
Ahmad, M., Zhao, Z.Y., Li, H., 2019. Revealing stylized empirical interactions among
countries: what role for aid? Clim. Pol. 15 (5), 543–564.
construction sector, urbanization, energy consumption, economic growth and CO2
Cai, W.G., Zhou, X.L., 2014. On the drivers of eco-innovation: empirical evidence from
emissions in China. Sci. Total Environ. 657, 1085–1098.
China. J. Clean. Prod. 79, 239–248.
Balcilar, M., Gungor, H., Olasehinde-Williams, G., 2019. On the impact of globalization
Carrión-Flores, C.E., Innes, R., 2010. Environmental innovation and environmental
on financial development: a multi-country panel study. Eur. J. Sustain. Dev. 8 (1),
performance. J. Environ. Econ. Manag. 59 (1), 27–42.
350.
Chan, R.Y., 2005. Does the natural-resource-based view of the firm apply in an emerging
economy? A survey of foreign invested enterprises in China. J. Manag. Stud. 42 (3),
625–672.

10
Q.-J. Wang et al. Energy Economics 110 (2022) 106004

Chen, Y., Lee, C.C., 2020. Does technological innovation reduce CO2 emissions? Cross- Mahto, R.V., Belousova, O., Ahluwalia, S., 2020. Abundance–a new window on how
country evidence. J. Clean. Prod. 263, 121550. disruptive innovation occurs. Technol. Forecast. Soc. Chang. 155, 119064.
Chen, Y., Tang, G., Jin, J., Li, J., Paillé, P., 2015. Linking market orientation and Millard, D., 2011. Management learning and the greening of SMEs: moving beyond
environmental performance: the influence of environmental strategy, employee’s problem-solving. German J. Human Resour. Manag. 25 (2), 178–195.
environmental involvement, and environmental product quality. J. Bus. Ethics 127 Nassani, A.A., Aldakhil, A.M., Abro, M.M.Q., Zaman, K., 2017. Environmental Kuznets
(2), 479–500. curve among BRICS countries: spot lightening finance, transport, energy and growth
Chiou, T.Y., Chan, H.K., Lettice, F., Chung, S.H., 2011. The influence of greening the factors. J. Clean. Prod. 154, 474–487.
suppliers and green innovation on environmental performance and competitive OECD, 2009. Sustainable manufacturing and eco-innovation: framework, practices, and
advantage in Taiwan. Transp. Res. Part E 47 (6), 822–836. measurement. Synthesis Report-OECD. Organization for Economic Co-Operation and
Chiu, Y.B., Lee, C.C., 2020. Effects of financial development on energy consumption: the Development, Rochester-USA, p. 36.
role of country risks. Energy Econ. 90, 104833. Oliva, F.L., Semensato, B.I., Prioste, D.B., Winandy, E.J.L., Bution, J.L., Couto, M.H.G.,
Darnall, N., Jolley, G.J., Handfield, R., 2008. Environmental management systems and Massaini, S.A., 2019. Innovation in the main Brazilian business sectors:
green supply chain management: complements for sustainability? Bus. Strateg. characteristics, types and comparison of innovation. J. Knowl. Manag. 23 (1),
Environ. 17 (1), 30–45. 135–175.
Day, G.S., Schoemaker, P.J., 2011. Innovating in uncertain markets: 10 lessons for green Owen, R., Brennan, G., Lyon, F., 2018. Enabling investment for the transition to a low
technologies. MIT Sloan Manag. Rev. 52 (4). carbon economy: government policy to finance early stage green innovation. Curr.
Dubey, R., Gunasekaran, A., Ali, S.S., 2015. Exploring the relationship between Opin. Environ. Sustain. 31, 137–145.
leadership, operational practices, institutional pressures and environmental Palazzo, G., Scherer, A.G., 2006. Corporate legitimacy as deliberation: a communicative
performance: a framework for green supply chain. Int. J. Prod. Econ. 160, 120–132. framework. J. Bus. Ethics 66 (1), 71–88.
Eberhardt, M., Teal, F., 2010. Productivity analysis in global manufacturing production. Pesaran, M.H., 2004. General Diagnostic Tests for Cross Section Dependence in Panels.
In: Economics Series Working Papers. University of Oxford. Pesaran, M.H., 2007. A simple panel unit root test in the presence of cross-section
Feng, G.F., Yang, H.C., Gong, Q., Chang, C.P., 2021. What is the exchange rate volatility dependence. J. Appl. Econ. 22 (2), 265–312.
response to COVID-19 and government interventions? Econ. Anal. Policy 69, Pesaran, M.H., 2015. Testing weak cross-sectional dependence in large panels. Econ.
705–719. Rev. 34 (6–10), 1089–1117.
Fousteris, A.E., Didaskalou, E.A., Tsogas, M.M.H., Georgakellos, D.A., 2018. The Pesaran, M.H., Shin, Y., Smith, R.P., 1999. Pooled mean group estimation of dynamic
environmental strategy of businesses as an option under recession in Greece. heterogeneous panels. J. Am. Stat. Assoc. 94 (446), 621–634.
Sustainability 10 (12), 4399. Poberezhna, A., 2018. Addressing water sustainability with blockchain technology and
Frondel, M., Horbach, J., Rennings, K., 2007. End-of-pipe or cleaner production? An green finance. In: Transforming Climate Finance and Green Investment with
empirical comparison of environmental innovation decisions across OECD countries. Blockchains. Academic Press, pp. 189–196.
Bus. Strateg. Environ. 16 (8), 571–584. Robinson, S., Stubberud, H.A., 2015. Green innovation and environmental impact in
Fujii, H., Iwata, K., Kaneko, S., Managi, S., 2013. Corporate environmental and economic Europe. J. Int. Bus. Res. 14 (1), 127.
performance of Japanese manufacturing firms: empirical study for sustainable Romano, A.A., Scandurra, G., Carfora, A., Fodor, M., 2017. Renewable investments: the
development. Bus. Strateg. Environ. 22 (3), 187–201. impact of green policies in developing and developed countries. Renew. Sust. Energ.
Gengenbach, C., Palm, F.C., Urbain, J.P., 2009. Panel unit root tests in the presence of Rev. 68, 738–747.
cross-sectional dependencies: comparison and implications for modelling. Econ. Rev. Seman, N.A.A., Govindan, K., Mardani, A., Zakuan, N., Saman, M.Z.M., Hooker, R.E.,
29 (2), 111–145. Ozkul, S., 2019. The mediating effect of green innovation on the relationship
Gianfrate, G., Peri, M., 2019. The green advantage: exploring the convenience of issuing between green supply chain management and environmental performance. J. Clean.
green bonds. J. Clean. Prod. 219, 127–135. Prod. 229, 115–127.
Haller, S.A., Murphy, L., 2012. Corporate expenditure on environmental protection. Singh, S.K., Del Giudice, M., Chierici, R., Graziano, D., 2020. Green innovation and
Environ. Resour. Econ. 51 (2), 277–296. environmental performance: the role of green transformational leadership and green
Handfield, R., Walton, S.V., Sroufe, R., Melnyk, S.A., 2002. Applying environmental human resource management. Technol. Forecast. Soc. Chang. 150, 119762.
criteria to supplier assessment: a study in the application of the analytical hierarchy Solarin, S.A., Al-Mulali, U., Musah, I., Ozturk, I., 2017. Investigating the pollution haven
process. Eur. J. Oper. Res. 141 (1), 70–87. hypothesis in Ghana: an empirical investigation. Energy 124, 706–719.
He, L., Liu, R., Zhong, Z., Wang, D., Xia, Y., 2019. Can green financial development Tang, D.Y., Zhang, Y., 2020. Do shareholders benefit from green bonds? J. Corp. Finan.
promote renewable energy investment efficiency? A consideration of bank credit. 61, 101427.
Renew. Energy 143, 974–984. Tolliver, C., Fujii, H., Keeley, A.R., Managi, S., 2021. Green innovation and finance in
Hicks, C., Dietmar, R., 2007. Improving cleaner production through the application of Asia. Asian Econ. Policy Rev. 16 (1), 67–87.
environmental management tools in China. J. Clean. Prod. 15 (5), 395–408. Wang, E.Z., Lee, C.C., 2022. The impact of clean energy consumption on economic
Hoffmann, E., 2007. Consumer integration in sustainable product development. Bus. growth in China: is environmental regulation a curse or a blessing? Int. Rev. Econ.
Strateg. Environ. 16 (5), 322–338. Financ. 77, 39–58.
Kammerer, D., 2009. The effects of customer benefit and regulation on environmental Wang, Y., Zhi, Q., 2016. The role of green finance in environmental protection: two
product innovation.: empirical evidence from appliance manufacturers in Germany. aspects of market mechanism and policies. Energy Procedia 104, 311–316.
Ecol. Econ. 68 (8–9), 2285–2295. Volz, U., Böhnke, J., Knierim, L., Richert, K., Roeber, G.M., Eidt, V., 2015. Financing the
Kratzer, J., Meissner, D., Roud, V., 2017. Open innovation and company culture: internal Green Transformation: How to Make Green Finance Work in Indonesia. Springer.
openness makes the difference. Technol. Forecast. Soc. Chang. 119, 128–138. Wang, Q.J., Chen, D., Chang, C.P., 2021a. The impact of COVID-19 on stock prices of
Kraus, S., Rehman, S.U., García, F.J.S., 2020. Corporate social responsibility and solar enterprises: a comprehensive evidence based on the government response and
environmental performance: the mediating role of environmental strategy and green confirmed cases. Int. J. Green Energy 18 (5), 443–456.
innovation. Technol. Forecast. Soc. Chang. 160, 120262. Wang, Q.J., Feng, G.F., Wang, H.J., Chang, C.P., 2021b. The impacts of democracy on
Kudratova, S., Huang, X., Zhou, X., 2018. Sustainable project selection: optimal project innovation: revisited evidence. Technovation 108, 102333.
selection considering sustainability under reinvestment strategy. J. Clean. Prod. 203, Wang, E.Z., Lee, C.C., Li, Y.Y., 2022. Assessing the impact of industrial robots on
469–481. manufacturing energy intensity of 38 countries. Energy Econ. 105, 105748 https://
Lee, C.C., Wang, C.W., Ho, S.J., 2020. Financial inclusion, financial innovation, and doi.org/10.1016/j.eneco.2021.105748.
firms’ sales growth. Int. Rev. Econ. Financ. 66, 189–205. Wen, J., Hao, Y., Feng, G.F., Chang, C.P., 2016. Does government ideology influence
Lee, C.C., Lee, C.C., Li, Y.Y., 2021. Oil price shocks, geopolitical risks, and green bond environmental performance? Evidence based on a new dataset. Econ. Syst. 40 (2),
market dynamics. North Am. J. Econ. Financ. 55, 101309. 232–246.
Lee, C.C., Wang, C.W., Ho, S.J., 2022. Financial aid and financial inclusion: does risk Wen, H., Lee, C.C., Zhou, F., 2021. Green credit policy, credit allocation efficiency and
uncertainty matter? Pac. Basin Financ. J. 101700. upgrade of energy-intensive enterprises. Energy Econ. 94, 105099.
Li, Y.N., Ye, F., 2011. Institutional pressures, environmental innovation practices and Wen, H., Lee, C.C., Zhou, F., 2022. How does fiscal policy uncertainty affect corporate
firm performance-an institutional theory and ecological modernization theory innovation investment? Evidence from China’s new energy industry. Energy Econ.
perspective. Stud. Sci. Sci. 29 (12), 1884–1894. 105, 105767 https://doi.org/10.1016/j.eneco.2021.105767.
Li, Z., Liao, G., Wang, Z., Huang, Z., 2018. Green loan and subsidy for promoting clean Weng, H.H.R., Chen, J.S., Chen, P.C., 2015. Effects of green innovation on environmental
production innovation. J. Clean. Prod. 187, 421–431. and corporate performance: a stakeholder perspective. Sustainability 7 (5),
Lin, R.J., Tan, K.H., Geng, Y., 2013. Market demand, green product innovation, and firm 4997–5026.
performance: evidence from Vietnam motorcycle industry. J. Clean. Prod. 40, Westerlund, J., Edgerton, D.L., 2007. A panel bootstrap cointegration test. Econ. Lett. 97
101–107. (3), 185–190.
Liu, X., Wang, E., Cai, D., 2019. Green credit policy, property rights and debt financing: Yang, Q.C., Feng, G.F., Chang, C.P., Wang, Q.J., 2021. Environmental protection and
quasi-natural experimental evidence from China. Financ. Res. Lett. 29, 129–135. performance: a bi-directional assessment. Sci. Total Environ. 774, 145747.
López, R., Palacios, A., 2014. Why has Europe become environmentally cleaner? Yin, W., Kirkulak-Uludag, B., Zhang, S., 2019. Is financial development in China green?
Decomposing the roles of fiscal, trade and environmental policies. Environ. Resour. Evidence from city level data. J. Clean. Prod. 211, 247–256.
Econ. 58 (1), 91–108. Yu, C.H., Wu, X., Zhang, D., Chen, S., Zhao, J., 2021. Demand for green finance:
López, R., Galinato, G.I., Islam, A., 2011. Fiscal spending and the environment: theory resolving financing constraints on green innovation in China. Energy Policy 153,
and empirics. J. Environ. Econ. Manag. 62 (2), 180–198. 112255.
Ma, B., 2015. Does urbanization affect energy intensities across provinces in China? Yuan, H., Feng, Y., Lee, C.C., Cen, Y., 2020. How does manufacturing agglomeration
Long-run elasticities estimation using dynamic panels with heterogeneous slopes. affect green economic efficiency? Energy Econ. 92, 104944.
Energy Econ. 49, 390–401.

11
Q.-J. Wang et al. Energy Economics 110 (2022) 106004

Yung, W.K., Chan, H.K., So, J.H., Wong, D.W., Choi, A.C., Yue, T.M., 2011. A life-cycle Zhao, J., Zhao, Z., Zhang, H., 2019. The impact of growth, energy and financial
assessment for eco-redesign of a consumer electronic product. J. Eng. Des. 22 (2), development on environmental pollution in China: new evidence from a spatial
69–85. econometric analysis. Energy Econ. 104506.
Zee, S.M., Fok, L.Y., Hartman, S.J., 2011. Exploring the relationships between Zheng, M., Feng, G.F., Jang, C.L., Chang, C.P., 2021. Terrorism and green innovation in
organizational size and market focus and commitment to the green movement and renewable energy. Energy Econ. 104, 105695.
impacts of organizational culture: a comparative study of Jamaica and the United Zhou, X., Tang, X., Zhang, R., 2020. Impact of green finance on economic development
States. Int. J. Bus. Soc. Sci. 2 (22). and environmental quality: a study based on provincial panel data from China.
Zhao, Y., Li, X., 2021. Research on green innovation countermeasures of supporting the Environ. Sci. Pollut. Res. 27 (16), 19915–19932.
circular economy to green finance under big data. J. Enterp. Inf. Manag. https://doi. Ziaei, S.M., 2015. Effects of financial development indicators on energy consumption and
org/10.1108/JEIM-01-2021-0039. CO2 emission of European, east Asian and Oceania countries. Renew. Sust. Energ.
Rev. 42, 752–759.

12