Energy Strategy Reviews 41 (2022) 100847

Contents lists available at ScienceDirect

Energy Strategy Reviews

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

The interaction of finance and innovation for low carbon economy:

Evidence from Saudi Arabia
Nahla Samargandi a, b, d, Kazi Sohag b, c, *
a
Department of Economics, Faculty of Economics and Administration, King Abdulaziz University (KAU), Jeddah, 21589, Saudi Arabia
b
Center of Research Excellence in Renewable Energy and Power Systems, Jeddah, Saudi Arabia
c
Graduate School of Economics and Management, Ural Federal University, Russia
d
K. A. CARE Energy Research and Innovation Center, King Abdulaziz University, Jeddah 21589, Saudi Arabia

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

Keywords: Saudi Arabia sets an ambitious plan to install 58.7 GW renewable capacity along with several mitigation mea­
Low carbon economy sures. For the smooth accomplishment of undertaken projects, the role of the financial market is crucial. This
Financial development paper investigates the role of financial development in explaining the low carbon economy (LCE), incorporating
Quantile ARDL
the role of innovation in the context of Saudi Arabia, where fossil fuel contributes almost 100% share of the total
Saudi Arabia
energy mix. We apply the standard, asymmetric, and quantile Autoregressive Distributed Lag (ARDL) Approaches
to analyse our time-series data from 1981 to 2016 (both yearly and monthly forms). Our empirical analysis
demonstrates that finance asymmetrically fosters the carbon economy, but it helps to achieve a low carbon
economy through the channel of innovation under different economic circumstances. Our empirical evidence
reinforces the role of the financial market in overcoming the financial constraints for launching green projects.

Author credit fuel (e.g. for home heating), and purchase electronic devices (including
air conditioners etc. With on-going high energy requirements), hence
Nahla Samargandi: Conceptualisation, Writing - original draft, Su­ increasing the level of carbon emissions in the atmosphere. Moreover,
pervision. Kazi Sohag: Conceptualisation, Methodology, Formal FD increases financing channels, lowers financing costs and operating
analysis. risk, and optimises assets or liability’s structure to buy new installations,
thus increasing energy consumption and carbon emissions [2].
1. Introduction Conversely, FD can promote a low carbon economy in several ways.
First, financial development provides the market solution to eliminate
The “carbon economy” paradigm refers to the conventional economy the financial constraint scaling up renewable projects and programmes,
based on conventional fossil fuel combustion for transport and elec­ including clean energies and innovation [3]. Many countries integrated
tricity generation, and the search for sustainable development has the national and global goals into local implementations addressing
inspired research interest in developing solutions toward a “low carbon finance and bankability challenges. Small and medium enterprises
economy” archetype, which produces less greenhouse gases. Financial (SMEs) have several potentials in promoting renewable energy due to
market development (FD) has a functional role in the carbon component their knowledge of local markets. Besides, SMEs can be involved in the
of the economy. In general, FD accelerates economic activities by entire value chain of renewable energy technologies, particularly small
mobilising capital, increasing inflows of foreign and local investment, equipment manufacturing activities, installation, civil works, retail and
and consequently increasing energy demand and carbon emissions. FD maintenance. Second, financial market upsurges support research and
increases both local and foreign direct investment to spur economic development (R&D) activities, which augments green or
growth. Hence it indirectly augments carbon emission through the energy-efficient technologies, eventually contributing to a low carbon
channel of higher fossil fuel or carbon content energies consumptions economy [3–5]. Claessens and Feijen [6] document that companies with
[1]. FD augments the credit disbursements to the private sector, sound financial stances often adopt low-carbon development strategies.
including households, which further enables them to buy vehicles, burn Zhang [3] provides a similar proposition, finding that FD fosters

* Corresponding author. Graduate School of Economics and Management, Ural Federal University, Russia.
E-mail addresses: [email protected] (N. Samargandi), [email protected] (K. Sohag).

https://doi.org/10.1016/j.esr.2022.100847
Received 20 October 2020; Received in revised form 28 November 2021; Accepted 10 April 2022
Available online 25 April 2022
2211-467X/© 2022 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
N. Samargandi and K. Sohag Energy Strategy Reviews 41 (2022) 100847

low-carbon technology by energy-efficient technologies, which helps 2. Literature review on finance and low carbon economy
achieve a low-carbon economy in the context of China. Finally, finance
is required to carry out most mitigation and adaptation projects over the Previous literature has examined the nexus between FD and the
globe [7]. The financial market credit supports small and medium en­ environment from different perspectives. Environmental degradation
terprises (SMEs) to run their mitigation and other eco-friendly busi­ has appeared as a central idea of academic studies from the last few
nesses. Moreover, green economic growth, green energies, green decades. Various human activities cause environmental degradation by
technologies, and allied financial markets appear to have the new exploiting natural possessions, weather changes, and damage to the
commercial potential [8,9]. A recent study documents that FD augments ecosystem, hence directly worsening environmental quality [13].
cleaner energy providing financial access to innovative projects [10]. However, relatively few studies examined the ecological footprint (EFP)
These contrasting propositions motivate us to formulate our hy­ perspective. Existing literature on finance and the carbon economy can
pothesis that FD can help achieve a low carbon economy through be divided into three stances. The first stance favours arguments that FD
innovation activities. Our sample country is Saudi Arabia, one of the positively affects environment quality by reducing EFP [14–16]. Yux­
largest producers and exporters of oil and gas in the world. The country iang and Chen [16] argue that FD improves environment quality by
is characterised as a hydrocarbon economy. Saudi Arabia’s oil and gas enhancing income and capitalisation, using new technologies and
sector accounts for 43% of the economy. Consequently, a rapid accel­ implementing environment related regulations. Tamazian et al. [17]
eration in energy demand threatens its energy security due to its indicate that FD reduces environmental degradation by reducing the
depletion attribute. In the long term, due to the industrialisation process, EFP and improving environmental quality.
Saudi Arabia’s economy will incrementally develop, which means that The second strand of literature argues that FD hinders environmental
carbon emissions are also likely to increase irreversibly. Aside from quality by increasing pressure on the EFP [18–21]. These studies find an
economic development and processes of industrialisation, FD also af­ inverse relationship between FD and environmental quality, arguing
fects the amount of carbon emissions. The Saudi national development that FD upsurges carbon emissions by giving credit facilities that induce
plan, Vision 2030, aspires to generate 58.7 GW capacity from RE and customers to buy machinery and vehicles that adversely affect the
foster research and development for economic sustainability. environment. For instance, firms buy new machinery to expand their
Nevertheless, renewable energy (RE) production and associated business and attract more FDI, which increases pollutant emissions and
technologies are itself subject to FD factors [11]. The production process negatively affects the EFP.
of RE requires enabling innovations (storage, EV charging, blockchain, FD has gained significant importance in explaining environmental
Internet of Things and AI and big data), and the entire innovation pro­ quality in recent literature. Nowadays, countries face two global chal­
cess is a long, distinctive, and fickle affair, involving no guarantees of lenges side by side: achieving sustainable development and preventing
success. Besides, the production of RE requires a massive initial invest­ environmental degradation. Energy demand has increased due to rapid
ment, with long payback periods, and private funding agencies are often industrialisation over recent decades, and the current global economy is
incapable of financing commensurately large renewable projects [8]. To predicated on relatively cheap energy sourced from the combustion of
achieve sustainable development, Saudi Arabia aspires to develop 58.7 fossil fuels for production, transport, and consumption of goods and
GW solar and wind projects over the next ten years as part of a budget services. This has caused adverse impacts on environmental quality
allocation of $50 billion to programmes boosting power generation and around the globe, creating a challenge for policymakers to simulta­
cutting oil consumption. The country aspires to develop the entire value neously manage the trade-off between environmental quality and eco­
chain of renewable energy and associated technologies. For instance, nomic development. The fundamental objective of all countries is to
Local experience is accessible in consulting, administration, planning, achieve sustainable development, and the financial sector must perform
and manufacturing solar cells and modules, inverters, tracking systems, an essential role in achieving sustainable development if any develop­
mounting systems, and cabling. For all elements of the PV supply chain, ment at all is to be achieved within the existing economic paradigm
such as public and private sectors, institutions require researchers, [14].
laboratories and workshops. Effective collaboration among the public, The financial sector has adopted new financing techniques, such as
private and financial markets is paramount. green financing and green credit, to promote growth and protect the
In contrast, the production process of RE and green technology (GT) environment [14]. Green credit is one of the new financing techniques
has some constraints. For example, the development of RE and GT are introduced by financial institutions to control pressure on the environ­
the sub-sets of innovation, which is a high-risk activity from a financial ment. The financial institution uses green investments and
perspective, as mentioned previously [12]. This background situates the energy-efficient technologies to reduce environmental pressure (EFP).
interesting research question of how FD and innovation jointly explain Another instrument used by financial institutions to control environ­
the prognosis for the carbon or low carbon economy in Saudi Arabia. mental pollution is to provide loans at low interest rates to industries for
We contribute environmental policy literature in several ways. First, the adoption of clean technologies. Likewise, R&D activities are being
we generate a proxy variable low carbon economy as carbon emission promoted by financial institutions in new energy resources, emphasising
( CO )
from one unit of output GDP 2
without setting any threshold, which im­ environmental protection [13].
plies that the environmental cost of economic growth. Second, we Many researchers have examined the nexus between FD and envi­
develop an empirical model highlighting the interactive role of the ronmental degradation [22–24]. Zhang [3] highlights the role of
financial market and innovation process in facilitating a low carbon well-developed financial institutions in reducing the EFP. He argues that
economy. Our empirical analysis is robust under symmetric, asym­ financial institutions attract and introduce environment-friendly pro­
metric, and quantile distribution. Our empirical analysis provides the jects by investing in research and development and improving envi­
long-run impact of our financial market and the innovation process to ronmental quality. FD promotes research and development activities
explain the carbon or low carbon economy in the context of Saudi Arabia and facilitates clean technologies by increasing FDI inflows, banking
under different economic circumstances and allow asymmetric adjust­ activities, and stock market activities. The investments in RE improve
ments. Our investigation provides several implications for the sustain­ environmental quality and help in reducing the EFP. Likewise, FD im­
able development agenda. proves economic efficiency by reducing financial costs and capital risk.
The next section highlights the review of the literature. Section 3 Contrary to this, mainstream FD is often considered to increase
highlights the model specifications, data and econometric approaches. environmental degradation. Zhang [3] portrays that FD motivates and
Section 4 provides results and discussion. Section 5 concludes with facilitates investors by providing credit facilities to purchase mechanical
policy implications. machinery, automobiles, electronic devices, and houses. Investors
expand their business by setting up machinery and new plants by using

2
N. Samargandi and K. Sohag Energy Strategy Reviews 41 (2022) 100847

these credit facilities. Hence, there is an increase in carbon emissions 3. Model specification, methodology
intensity that worsen environmental quality. Sarkodie and Strezov [24]
argue that FD increases FDI, which further environmental degradation. 3.1. Model specification
This view, which can be said to be the mainstream one among policy­
makers, basically assets underscores that conventional, traditional eco­ We develop the conceptual framework following Zeqiraj, Sohag and
nomic development is predicated on negative environmental impacts; Soytas [35], who document that financial development helps to achieve
consequently, FD within an environmentally destructive economic low carbon through primarily two mitigation measures, including
paradigm is ipso facto detrimental to the environment. However, this cleaner energy and technological innovation (Fig. 1). Given an insig­
over-simplistic view ignores the more complex ways in which FD and nificant share of renewable energy in the total energy mix, this study
economic growth engage with environmental impacts. considers only the innovation channel. Following [35], we generate
( )
Indeed, some eminent economists have argued that public activism CO2
carbon emission from one unit of output Output as proxy of low carbon
by environmental groups exerts an irrational influence on government
economy. The ratio indicates the environmental cost of economic
policies that could ultimately be detrimental to the environment. Based
growth.
on a statistical analysis of the US economy, McCormick [25] concluded
Based on our argument in the introduction section, we develop our
that:
empirical Model whereby a low carbon economy is the function of FD,
“net carbon emissions will decline if economies grow and people get the labour force (LF), trade openness (TO), innovation (GT) and the
richer … higher incomes are associated with a better natural envi­ interaction of FD and innovation (FD*GT). We anticipate that FD aug­
ronment. Fascination with emissions controls and the Kyoto Protocol ments the carbon economy, but along with the GT process, it helps to
is somewhat hard to understand in this context". achieve a low carbon economy. Depending on the magnitude of com­
plementary and substitution effects, LF can positively or negatively
This is corroborated by a recent panel series analysis of 68 countries
impact the carbon economy. TO can be productive or counterproductive
spanning half a century (1961–2020) by Ref. [26], who note the intrinsic
towards the local carbon economy.
paradox between policies that inhibit conventional economic develop­
ment intending to mitigate carbon emissions, while economic develop­
3.2. Data, definition and source
ment per se ultimately lowers carbon emissions over the long term.
Wang et al. [27] examined the relationship between finance, devel­
To measure the interactive impact of FD and GT on the low carbon
opment, energy, and carbon emissions, providing various policy options
economy, we take several variables. Our dependent variable of carbon
to upgrade environmental quality for various regions and countries.
or low carbon economy is generated from carbon emission per unit of
These studies used carbon emissions as a proxy for measuring the en­
GDP or output. To generate this series, we obtain the data on carbon
vironment’s health [28,29]. However, carbon emissions are regarded as
emissions (kt) and GDP constant 2010 in USD from World Development
the outcome of fossil fuel or carbon contented energy consumption and
Indicators (WDI), 2020. As a proxy of FD, we consider credit disburse­
are only one part of environmental degradation [14,30]; thus, there
ment to the private sector share of GDP. The total number of patent
should be a more comprehensive indicator of anthropogenic environ­
applications indicates GT. We also take total LF and TO as control var­
mental pressure [31,32]. EFP combines environmental data into a sin­
iables of this study.
gle, easily comprehended and simple measure [33]. It consists of useful
data concerning several natural resources needed to generate and pro­
3.3. Econometrics approaches
mote the overall economy [23]. It is an efficient measure of depleted
natural resources reproduction and captures the information related to
3.3.1. Standard auto-regressive distributed lags (ARDL) approach
critical natural resources (e.g., water and land utilisation for cultivation
We estimate LCE in KSA using the ARDL approach due to a few
and cropping) [34]. Ecological footprint provides information about
reasons. ARDL approach can provide unbiased coefficients in the case of
resource metabolism and allows governments to compare economic
mixed order of integration. Also, this framework is relatively helpful in
demand and supply resources to handle the distribution of resources
estimating short and finite time series data.
more efficiently.

Fig. 1. Conceptual framework.

∫ Empirical
Model LCE = (FD, LF, TO, INO, FD * INO)

3
N. Samargandi and K. Sohag Energy Strategy Reviews 41 (2022) 100847

LCEt = β0 + β1 LCEt− 1 + β2 FDt− 1 + β3 LFt− 1 + β4 TOt− 1 + β5 INOt− 1 To derive the conditional non-linear ECM, we specify data decom­
∑ p ∑q1 q2
∑ q−
∑1 q−
∑1
+β6 FD * GTt− 1 + γi LCEt− i + δj FDt− i + τl ΔLFt− i (1) posing procedure Δxt = Λj Δxt− j +vt and εt = ω (Δxt −
′
Λj Δxt− j )+ et ,
j=1 j=1
i=1 i=1 i=1
q3
∑ q4
∑ q5
∑ q6
∑ where et is uncorrelated with vt , to reach the following Eq. (5):
+ φl LFt− i + φ2 TOt− i + φ2 INOt− i + φ3 FD * INOt− i + εt
i=1 i=1 i=1 i=1 p− 1
∑ q− 1 (
∑ )
Δyt = ρyt− 1 + θ+ + − −
j xt− j + θj xt− j + yj yt− j + π+j x+t− j − π−j x−t− + εt (5)
Eq. (1) is estimated under ordinary least squared (OLS) framework.
j
j− 1 j=0
At the next step, we determine joint significance of the coefficients of the
lagged variables by Wald test. The null hypothesis (H0): β1 = β2 = β3 = where:
β4 = β5 = β6 = 0 indicates that there exists no co-integrating relation
π+0 = θ+0 + ω, π−0 = θ−0 + ω, π+j
among the regressors and regressand against the alternate hypothesis, ′ ′
(Ha) β1 ∕ = β2 ∕
= β3 ∕= β4 ∕= β5 ∕ = 0. The F-statistics are compared
= β6 ∕ = φ+ −
j − ω Λj and π j = φj − ω Λj for j

against the critical values of the upper and lower bound [36]. Calculated = 1, …‥…q − 1.
F-statistics higher than the upper critical value implies that our null There are two tests including tBDM − statistic and FPSS − statistic. The
hypothesis cannot be rejected (i.e., the existence of co-integrating tBDM − statistic proposed by Banerjee et al. [38] verifies H0 : ρ = 0
relation with LCE and its determinants, and vice-versa). However, if against H1 : ρ < 0 while the FPSS− statistic proposed by Pearsan et al.
the calculated F-statistics are within the lower and upper bounds of [36] tests the joint null H0 : ρ = θ+ = θ− = 0. The bounds testing
critical values, then the test is inconclusive. We then proceed to estimate framework of PSS provides accurate statistical inference for examining
the long-run and short-run coefficients along with the error correction the presence integrated order zero and I(1) variables, which is helpful
term. Finally, we apply a various stability tests including cumulative for partial sum decompositions, and which can help detect complicated
sum of recursive residuals (CUSUM), and the cumulative sum of squares interdependencies. It is also recommended to count the regressors in Xt
of recursive residuals (CUSUMSQ). before decomposition, and to comparing the critical values with those in
PSS [37]. Both the long-run and short-run asymmetries are tested by
3.3.2. Asymmetric non-linear framework applying the following general equation:
The application of non-linear approaches has become increasingly
more popular as the nature of the variables can be non-linear. In this Yt = μ1 + μ2 ftk + βt1 + α1 Xt + α2 Xt ftk + εt (6)
connection, we apply Shin, Yu, and Greenwood Nimmo’s [37] The null hypothesis of π = π has been tested for long-run asym­
+ −
i i
non-linear ARDL framework. This approach has several advantages. It metry using the Wald test of joint significance. The restriction of short-
allows for long-run asymmetric relation and incorporating an error run asymmetry has been tested in either of the two ways π+ = π− for all
correction mechanism. The parameters are estimated through OLS q−
∑1 + q−
∑1 −
approach, employing stationary variables in a consistent pattern. i = 0…….q − 1 , or πj = πj ; the Wald test can evaluate both the
i=0 i=0

y = β+ xt+ + β− ‫ ׳‬xt− + ut, Δxt = vt forms. The symmetry of the impact multipliers is tested through H0 :
π+0 = π−0 , where the asymmetric effect on Yt varies with the unit changes
where yt represents scalar I(1) variable and xt is a vector of regressors in x+
t and xt . Unit changes are measured recursively from the slope
−

with K × 1 dimension, such that xt = x0 + x+

t + xt , where x0 is the initial
−
parameters in level of the NARDL model of the following equation [37]:
∑t ∑t t
∑
value and x+
t = Δx+
j = max(Δxj , 0) and x−t = Δx−j = LCEt = α + β1 FDt + β2 LFt + β3 TOt + β4 INOt + β5 FD*INOt + εt (7)
j=1 j=1 j=1
t
∑
min(Δxj, 0) are partial sum processes of positive and negative change where, LCEt is the low carbon economy, LFt is the labour force, TOt is
j=1 trade openness, INOt in innovation and FD*INOtt is the interaction of FD
in xt . and Innovation.
To have a better econometric interpretation, we assume a single
known threshold value of zero. Asymmetric ARDL (p, q) in the level’s 3.3.3. Quantile ARDL framework
framework equation is written as follows: To analyse the impact of FD on LCE RE across quantiles in a Saudi
p q ( ) Arabian context, we apply the quantile ARDL model proposed by Cho
∑ ∑
yt = θj yt− j + θ+ + − −
j xt− j + θj xt− j + εt (3) et al. [39], which is a combination of the autoregressive distributed lag
j=1 j=0 model of Pesaran and Shin [36], accounting for short- and long-run
relations, and the quantile regression methodology developed by
where θj are autoregressive parameters, θ+
j, θj comprise the asymmetric
−
Koenker and Bassett [40]. The QARDL model offers to estimate the
parameters coming from decomposed terms of the variables and εt is the quantile long-run equilibrium slope coefficients of FD and Control var­
error term follow the normality properties with zero mean value and iables (CV) on LCE, and also helps maintain the consistency of inte­
constant variance, σ2ε . The error correction term is as follows (Eq. (4)): grating coefficients across quantiles. The Wald test examines the
consistency of integrating coefficients across quantiles and time-varying
p− 1
∑ q− 1 (
∑ )
yt = ρyt− 1 + θ+ + − −
θ+ + − −
+ εt (4) integrating relationships. According to Cho et al. [39], the ARDL process
j xt− j + θj xt− j + yj yt− j + j xt− j + θj xt−
can be extended to the quantile regression context. The ARDL process is
j
j=1 j=0
defined as Eq. (8):
where the long-run asymmetric parameters are: p q1
∑ ∑
p
∑ p
∑ q
∑ q
∑ LCEt = α + φi LCEt− i + ωi Xt− i + εt (8)
ρ= ∅j − 1,γj = θi for j=1,‥,p− 1,θ+ = θ+ −
j θ = θ−j ,φ+ + +
0, =θ0 ,φj
i=1 i=0
j=1 i=j+1 j=0 j=0
[ ]
where εt is the error term explained as LCEt − E LCEt/F , with Ft−
q
∑
1
=− θ+ − − −
j for j=1,….,q− 1,φ0 =θ0 ,φj t− 1
i=j+1 being the marginal σ field estimated by { FDt , CVt , LCEt− 1 FDt− 1 , CVt− 1 ,
….}, and p, q1 q2 are lag orders selected by Schwarz information criteria
q
∑ θ+ θ−
=− θ−j for j=1,….,q− 1 and β+ = − and β− = −
i=j+1
ρ ρ (SIC) (CV indicates control variables). The extended form of the Model
in Eq. (8) in the quantile context is the basic form of the QARDL (p, q)

4
N. Samargandi and K. Sohag Energy Strategy Reviews 41 (2022) 100847

model proposed by Cho et al. [39]. ( )(

p
∑ q1
∑ q2
∑ QΔLECt = α(τ) + ρ τ LECt− 1 − βFD (τ)FDt− 1 − βCV (τ)CVt− 1
QLCEt = α(τ) + φi (τ)LCEt− i + ωi (τ)FDt− i + λi (τ)CVt− i + εt (τ) ( )
i=1 i=0 i=0 p− 1
∑ q∑
1− 1 q∑
2− 1

(9) + φi (τ)ΔLECt− 1 + ωi (τ)ΔFDt− 1 λi (τ)ΔCVt− 1 + εt τ

i=1 i=0 i=0

where, εt (τ) = LECt − Q ( ) and QLEC (τF ) is the τth quantile of

t t− 1
(14)
LECt τ/
Ft− 1
The cumulative short-term impact of past FD, other control variables
LECt conditional on the information set Ft− 1 as defined by Kim and White q∑
1− 1 q∑
2− 1
[43] more advanced management levels. We obtain Eq. (10) after and on current LEC, FD, and CV are estimated by w* = wj , λ* = λj
reformulating Eq. (9) for analysing the QARDL model as: j=1 j=1

accordingly. The long-run co-integrating parameters and short-term

q∑
1− 1 q∑
2− 1
cumulative parameters are estimated applying the delta method. The
QLECt = α(τ) + δLCEt (τ)ΔLCEt− 1 + γFD (τ)FDt + δLCEi (τ)ΔLCEt−
i=1 i=1
1
long-run parameters for FD and CV are estimated as βFD * = − φρFD ,
+ γ OP (τ)CVt + εt (τ) βCV* = − φCV
ρ . Here, ECM parameters (ρ) should be significantly negative.
(10) Wald test is used for statistical estimation of the short- and long-term
non-linear, asymmetric impacts of FD on LEC. This Wald test statistic
where asymptotically follows chi-square distribution, which is used to test null
q1
and alternative hypotheses for the short-term and long-term parameters
∑
γFD (τ) = ωi (τ), φ* , β* , ρ* , ω* stated below:
i=0
q1
∑ H0φ : Fφ* (T) = f versus H1φ : Fφ* (T) ∕
=f
δFDt (τ) = − ωi (τ),
j=i+1 H0φ : Sω* (T) = S versus H1φ : Sω* (T) ∕
=S
q1
∑
γCV (τ) = λi (τ),
i=0
H0φ : Sβ* (τ) = S versus H1φ : Sβi* (T) ∕
=S
q1
∑
δCVt (τ) = − λi (τ) H0φ : Sρ* (T) = S versus H1φ : Sρ* (T) ∕
=S
j=i+1
Here F, f, S and s are h* ps, h* 1, h* s, h* 1 are prespecified matrices, with
The parameters in the above-mentioned equation estimate the short- h being the number of restrictions (Cho et al., 2015), and i represents FD
run dynamics. In order to solve for the long-run relationships, we further and CV, respectively. We use Wald test to examine the non-linear impact
reformulate Eq. (10) as follows in Eq. (11): on the speed of adjustment coefficient and the long run integrating slope
QLCE = μ(τ) + X t β(τ) + Mt (τ)
′
(11) coefficients.
The most significant advantage of the quantile regression approach is
X = [GPR, OP] the comprehensive picture it creates of conditional distribution of
dependent variables, highlighting differences in the degree of impact of
[
p
∑
]− 1 independent variables in different market conditions. Quantile ARDL
βFD (τ) = γFD (τ) 1 − φFDi (τ) estimates are robust to outliers, heteroscedasticity, and skewness among
i=1 dependent variables [41].
∞
∑ ∞
∑ p
∑
Mt (τ) = ∂FDj (τ)ΔFDt− 1 + θFDj (τ)Δεt− 1 with μ(τ) = α(τ)[1 − φi 4. Results and discussion
j=0 j=0 i=1

We present our empirical results and discussion in this section.

∞
∑
1
(τ)]− and ∂j (τ) = π1 (τ). βFD (τ) is computed similarly. {θ0 (τ), θ1 (τ),
l=j+1
… and {π 0 (τ), π1 (τ), … are defined such that 4.1. Descriptive statistics
p
( ∑q 1
∑∞ ∑ ∑∞ ωi (τ)Li
θi (τ)Li = (1 − φi (τ)Li )− 1 and. πi (τ)Li = (1 − L)− 1 1− ∑i=0q1 ω (τ)Li −
i=0 i=1 i=0 i=1 i We apply a pairwise correlation test to comprehend the magnitude of
∑q 1 )
ω i ( τ ) the correlation among our variables (Table 1). The correlation coeffi­
∑i=0q1
1− i=1
ω i (τ ) cient of low carbon economy (LCE) and FD appears high at 93%. Like­
Further generalisation of the QARDL model is required to avoid the wise, LF and GT are positively and significantly correlated with LCE.
serial correlation of ε: We examine the order of integration of all variables by applying
p
Phillip-Perron (PP) and Augmented Dickey-Fuller (ADF) approaches by
∑
QΔLECt = α + ρLECt− 1 + φGPR FDt− 1 + φOP CVt− 1 + φi ΔLECt− 1
considering with constant, with constant and trend, and without con­
i=1 stant and trend. Table 2 clearly shows that all our variables are non-
q∑
1− 1 q∑
2− 1 stationary when considering with constant and without constant &
+ ωi ΔGPRt− 1 + λi ΔCVt− 1 + vt (τ) (12) trend under PP and ADF approaches. However, LCE and LF become
i=0 i=0

However, there is a likelihood of contemporaneous correlation be­ Table 1

tween vt and ΔFDt and ΔCVt in Eq. (12). To avoid this correlation, the Correlation matrix.
projection of vt on ΔFDt and ΔCVt can be used in the form: LCE FD LF TO GT
vt = γGPR ΔFDt + γ OP ΔCVt + εt (13) LCE 1.0000
FD 0.9330*** 1.0000
where εt is correlated with ΔFDt and ΔCVt. We obtain quantile ARDL LF 0.8385*** 0.7256*** 1.0000
TO 0.1475 0.1016 − 0.0124 1.0000
-ECM model in Eq. (13) after substituting the previous projection into
GT 0.6733*** 0.6781*** 0.7414*** − 0.2047 1.0000
Eq. (14).
Note: ***, ** & * indicate 1%, 5% & 10% significance level.

5
N. Samargandi and K. Sohag Energy Strategy Reviews 41 (2022) 100847

Table 2
Order of integration.
PP ADF

At Level

LCE FD LF TO GT LCE FD LF TO GT

With constant t-Statistic − 0.4110 0.6680 − 1.6141 − 2.0991 − 1.8520 1.8644 0.1811 − 1.6147 − 2.4030 − 1.8520
Prob. 0.8964 0.9896 0.4649 0.2462 0.3502 0.9996 0.9674 0.4647 0.1484 0.3502
No no no no no no no no no no
With constant & trend t-Statistic − 4.6598 − 2.1848 − 1.7142 − 2.1501 − 3.1769 − 0.5100 − 2.7235 − 96.0913 − 2.4903 − 3.2130
Prob. 0.0035 0.4829 0.7236 0.5012 0.1055 0.9768 0.2341 0.0000 0.3306 0.0984
*** no no no no no no *** no *
Without constant & trend t-Statistic 5.7333 3.5870 1.5176 − 1.0510 1.2679 5.6383 2.2437 17.0175 − 1.1519 0.6428
Prob. 1.0000 0.9998 0.9656 0.2589 0.9450 1.0000 0.9928 1.0000 0.2224 0.8504
No no no no no no no no no no
At First Difference

With constant t-Statistic − 17.8454 − 5.3857 − 5.8708 − 3.6998 − 8.6195 − 7.2446 − 5.3345 − 62.9664 − 3.7610 − 7.9441
Prob. 0.0001 0.0001 0.0000 0.0086 0.0000 0.0000 0.0001 0.0001 0.0074 0.0000
*** *** *** *** *** *** *** *** *** ***
With constant & trend t-Statistic − 17.3517 − 5.7008 − 5.9910 − 3.5842 − 8.5970 − 7.8996 − 5.3965 − 5.9374 − 3.6527 − 7.9061
Prob. 0.0000 0.0002 0.0001 0.0463 0.0000 0.0000 0.0005 0.0001 0.0400 0.0000
*** *** *** ** *** *** *** *** ** ***
Without constant & trend t-Statistic − 7.1525 − 4.7727 − 5.5343 − 3.6390 − 8.1316 − 0.5690 − 4.7708 − 5.5346 − 3.7507 − 7.8549
Prob. 0.0000 0.0000 0.0000 0.0006 0.0000 0.4614 0.0000 0.0000 0.0005 0.0000
*** *** *** *** *** no *** *** *** ***

Note: ***, ** & * indicate 1%, 5% & 10% significance level.

stationary when we take constant and trend in our unit-root equation. LF appears to be positive and significant, indicating it has a comple­
Nevertheless, all the variables become stationary at the first difference, mentary effect in augmenting the carbon economy in the long run,
considering constant and constant and trend. The mix order of inte­ affirming the proposition of [42]. LF insignificantly affects the carbon
gration in our variables confirms the validity of our econometric ap­ economy in the short run. The direct impact of GT on the low carbon
proaches applied in this study. economy is inconclusive in both the short run and the long run. The
interactive impact of the FD and GT appears to be negative but insig­
4.2. Finance and carbon economy under linearity assumption nificant on the low carbon economy in the long run and the short run.
Figs. 1 and 2 plot the cumulative sum of squares of recursive residuals
We estimate our model LCEt = f(FDt ,LFt ,TOt ,INOt ,FD *GTt ). Where (see Fig. 3).
measure the direct impact of FD and mediating impact through GT.
Table 3 reports the results obtained from standard ARDL approach. The 4.3. Finance and carbon economy under asymmetric impact assumption
coefficient of error correction is negative and significant, which con­
firms the short-run adjustments towards the long-run equilibrium of our At this stage, we re-estimated our Model by applying asymmetric
Model. Precisely, the Model adjusts 67% per year after any shocks to­ ARDL approach. Table 4 reports the result. A positive shock in FD has a
wards the long-run equilibrium. Perhaps higher factor mobility and less positive but insignificant influence on the carbon economy. In contrast,
friction in the financial market contribute to a higher adjustment rate. the coefficient of negative shock in FD is negative and significant,
Table 3 exhibits that FD significantly spurs carbon economy by indicating that it helps achieving LCE. The lower part of Table 4 con­
fostering carbon intensity, energy intensity and energy uses per output firms that the impact of FD on the carbon economy follows an asym­
unit in the long run. Our long-run result concurs with the views of metric nature. Likewise, LF has an asymmetric impact on the carbon
Frankel and Romer [1] and Zhang [3]. Javed and Sharif [2] documented economy. A positive shock in labour supply augments the carbon
that FD augments carbon emission through higher fossil fuel or carbon economy, whereas a negative shock insignificantly influences it. This
content energies consumption. However, our empirical findings provide empirical finding implies that the LF has a complementary rather than
some additional insight that FD insignificantly influences the carbon substitute effect on the carbon economy. A positive shock in GT insig­
economy in the short run, which differs from Zhang [3]. The coefficient nificantly promotes low carbon economy whereas a negative shock in
GT positively and significantly promotes carbon economy. The lower
Table 3 part of Table 4 confirms that GT has an asymmetric impact on the carbon
Finance and Carbon Economy under Standard ARDL approach. economy. The coefficient of the interaction between FD and GT is
Regressor Coefficient Standard Error T-Ratio [Prob]
negative but insignificant under the shock regime, indicating that they
jointly insignificantly help to achieve a low carbon economy. Interest­
Error Correction − 0.67980 0.14033 − 4.8443[.000]
ingly, an interaction between FD and GT augments the carbon economy
FD 1.3908 0.5659 2.4574[.021] under the negative shock regime. Findings under asymmetric ARDL are
LF 2.6041 0.9989 2.6068[.015] partially in line with the findings under the standard ARDL approach.
TO 0.0483 0.0545 0.8865[.384]
GT 1.0959 2.9471 0.37184[.713]
FD*GT − 0.1089 0.0767 − 1.4195[.168] 4.4. Finance and carbon economy under quantile distribution assumption
C − 24.475 12.8616 − 1.9030[.069]

Short Run Quantile ARDL approach is used to finally re-estimate our Model
ΔFD 0.59354 0.4019 1.4765[.151] considering three quantiles: 25%, 50%, and 75%. Coefficients of error
ΔLF − 0.48542 1.0055 − 0.4827[.633] correction under each quantile are negative and significant, confirming
ΔTO − 0.15074 0.08787 − 1.7153[.097] the long-run equilibrium through adjusting in the short-run (Table 5).
ΔGT 0.74496 2.0189 0.3689[.715] The coefficient of FD is positive and significant under each quantile,
− 0.07403 0.05293 − 1.3985[.173]
indicating that it promotes a carbon economy. The magnitude of the
ΔFD*GT

6
N. Samargandi and K. Sohag Energy Strategy Reviews 41 (2022) 100847

Fig. 2. Plot of cumulative sum of recursive residuals.

Fig. 3. Plot of cumulative sum of squares of recursive residuals.

Table 4 Table 5
Finance and carbon economy under asymmetric ARDL approach. Finance and carbon economy under quantile ARDL approach.
Variable Long Run Effect [+] Long Run Effect [− ] VARIABLES q25 q50 q75

Coefficient F-Stat P-value Coefficient F-Stat P-value ECM(ρ*) − 0.028*** − 0.0127*** − 0.0108***
(0.005) (0.003) (0.003)
FD 1.247 1.768 0.184 − 0.895 5.899 0.016
Long Run
LF 0.207 3.843 0.051 0.249 2.045 0.154
TO 0.097 0.946 0.331 − 0.099 0.617 0.433 βFD (τ) 0.588*** 1.223*** 2.599***
GT − 0.488 0.005 0.941 0.075 4.909 0.027 (0.157) (0.373) (0.755)
FD*GT − 1.078 0.514 0.474 0.629 5.381 0.021 βLF (τ) 1.890*** 3.300*** 4.566***
(0.246) (0.793) (1.052)
Long-Run Asymmetry Long-Run Asymmetry
βTRD (τ) 0.060*** 0.061* 0.085
F-Stat P-value F-Stat P-value (0.024) (0.035) (0.069)
βGT (τ) − 1.040 0.482 2.283
FD 5.209 0.023 2.399 0.122
(1.143) (2.126) (3.525)
LF 3.731 0.054 3.271 0.071
βFD*GT (τ) − 0.498 − 2.420*** − 5.196**
TO 0.001 0.991 0.337 0.562
(0.385) (0.768) (1.726)
GT 4.001 0.046 2.221 0.137
Short Run
FD*GT 5.715 0.017 2.739 0.099
ΔFD − 0.060 0.011 0.086
(0.177) (0.161) (0.260)
coefficient is higher in the higher quantile. Our findings using quantile ΔLF − 0.767 − 0.874* − 1.080**
(0.526) (0.506) (0.522)
ARDL are consistent with the findings under the standard ARDL and
ΔTRD − 0.085*** − 0.101*** − 0.095**
asymmetric ARDL in terms of the sign and significance of coefficients. (0.028) (0.026) (0.031)
Likewise, LF augments the carbon economy under each quantile where ΔGT − 0.610 − 0.315 0.178
its impact is higher at the higher quantile, implying that LF is comple­ (0.463) (0.314) (0.531)
mentary with carbon content energy in the Saudi economy. TO fosters ΔFD*GT 0.040 − 0.044 − 0.246
(0.180) (0.152) (0.291)
carbon economy at the first and second quantiles, while its impact dis­
appears at the higher quantile. Trade is found to be negative and sig­ Constant − 0.040 − 0.067*** − 0.092***
(0.026) (0.021) (0.023)
nificant toward the carbon economy. GT insignificantly influences the
Observations 430 430 430
carbon economy under each quantile. The striking finding is that the
interaction between FD and GT helps realise the low carbon economy, F-Test 7.71*** 3.24* 3.83**

which is consistent with our views and exiting literature [35].

7
N. Samargandi and K. Sohag Energy Strategy Reviews 41 (2022) 100847

5. Conclusion [6] S. Claessens, E. Feijen, Financial Sector Development and the Millennium
Development Goals, The World Bank, 2007.
[7] J. Linnerooth-Bayer, S. Hochrainer-Stigler, Financial instruments for disaster risk
Given the critical role of financial market for energy transition in management and climate change adaptation, Climatic Change 133 (1) (2015)
achieving low carbon economy, we conducted this study to investigate 85–100.
its impact on carbon intensity in Saudi Arabia. This study generates a [8] M. Al Mamun, K. Sohag, M. Shahbaz, S. Hammoudeh, Financial markets,
innovations and cleaner energy production in OECD countries, Energy Econ.
series to make a proxy of carbon economy by considering the carbon (2018), https://doi.org/10.1016/j.eneco.2018.04.011.
emission from per unit of GDP. The higher value of this series indicates [9] K. Sohag, M. Al Mamun, G.S. Uddin, A.M. Ahmed, Sectoral output, energy use, and
the carbon economy, while the lower value indicates the low carbon CO2emission in middle-income countries, Environ. Sci. Pollut. Control Ser. 24 (10)
(2017), https://doi.org/10.1007/s11356-017-8599-z.
economy. We empirically assessed our proposition that FD has a certain [10] B. Holmstrom, Agency costs and innovation, J. Econ. Behav. Organ. 12 (3) (1989)
role in influencing carbon intensity along with the GT process in the 305–327.
context of Saudi Arabia, where cleaner energy plays no role. We consider [11] Muhammad Shahbaz, Ilham Haouas, Kazi Sohag, Ilhan Ozturk, The financial
development-environmental degradation nexus in the United Arab Emirates: the
the role of the LF and TO as control variables in our empirical Model. We importance of growth, globalisation and structural breaks, Environ. Sci. Pollut.
applied the ARDL approach under the assumptions of symmetric, Control Ser. (2020) 1–15.
asymmetric, and quantile distributions, leading to several interesting [12] P.-H. Hsu, X. Tian, Y. Xu, Financial development and innovation: cross-country
evidence, J. Financ. Econ. 112 (1) (2014) 116–135.
findings. [13] M.T. Majeed, M. Mazhar, Financial development and ecological footprint: a global
Firstly, FD arguments carbon economy in the long run in an asym­ panel data analysis, Pakistan J. Commer. Soc. Sci. (PJCSS) 13 (2) (2019) 487–514.
metric manner. Moreover, the impact of FD on the carbon economy has a [14] M.A. Boutabba, The impact of financial development, income, energy and trade on
carbon emissions: evidence from the Indian economy, Econ. Modell. 40 (2014)
higher magnitude at the higher quantile. Secondly, GT is yet to be
33–41.
augmented to facilitate achieving a low carbon economy. Thirdly, our [15] Ugur Korkut Pata, Veli Yilanci, Financial development, globalisation and
empirical findings are in line with our proposition that FD, along with ecological footprint in G7: further evidence from threshold cointegration and
GT, helps achieve a low carbon economy in the long run. Interestingly, fractional frequency causality tests, Environ. Ecol. Stat. 27 (4) (2020) 803–825.
[16] K. Yuxiang, Z. Chen, Resource abundance and financial development: evidence
the interactive effect of FD and GT follows the asymmetric distribution. from China, Resour. Pol. 36 (1) (2011) 72–79.
Moreover, the interactive impact is higher in the higher quantile. [17] A. Tamazian, J.P. Chousa, K.C. Vadlamannati, Does higher economic and financial
Fourth, LF augments the carbon economy, indicating it has a comple­ development lead to environmental degradation: evidence from BRIC countries,
Energy Pol. 37 (1) (2009) 246–253.
mentary effect rather than a substitute of capital or the carbon content of [18] M.A. Baloch, J. Zhang, K. Iqbal, Z. Iqbal, The effect of financial development on
energy towards the carbon economy. TO somewhat spurs the carbon ecological footprint in BRI countries: evidence from panel data estimation,
economy. Environ. Sci. Pollut. Control Ser. 26 (6) (2019) 6199–6208.
[19] F. Ganda, The environmental impacts of financial development in OECD countries:
Our empirical findings provide several policy implications. First, FD a panel GMM approach, Environ. Sci. Pollut. Control Ser. 26 (7) (2019)
should focus on innovative and RE projects to achieve a low carbon 6758–6772.
economy. Therefore, the financial market can play an important role to [20] A. Jalil, M. Feridun, The impact of growth, energy and financial development on
the environment in China: a cointegration analysis, Energy Econ. 33 (2) (2011)
overcome the financial constraints for launching mitigation projects. 284–291.
Also, the financial market can support the value chains of renewable [21] M. Zakaria, S. Bibi, Financial development and environment in South Asia: the role
energy production capacity to achieve a low carbon economy. of institutional quality, Environ. Sci. Pollut. Control Ser. 26 (8) (2019) 7926–7937.
[22] K. Sohag, F.D. Taşkın, M.N. Malik, Green economic growth, cleaner energy and
militarisation: evidence from Turkey, Resour. Pol. 63 (Oct) (2019), https://doi.
Declaration of competing interest org/10.1016/j.resourpol.2019.101407.
[23] S. Katircioglu, K.K. Gokmenoglu, B.M. Eren, Testing the role of tourism
development in ecological footprint quality: evidence from top 10 tourist
The authors declare that they have no known competing financial destinations, Environ. Sci. Pollut. Control Ser. 25 (33) (2018) 33611–33619.
interests or personal relationships that could have appeared to influence [24] S.A. Sarkodie, V. Strezov, Effect of foreign direct investments, economic
the work reported in this paper. development and energy consumption on greenhouse gas emissions in developing
countries, Sci. Total Environ. 646 (2019) 862–871.
[25] R.E. McCormick, The Relation between Net Carbon Emissions and Income, You
Acknowledgements Have to Admit Its Getting Better, Hoover Institution Press, Stanford, CA, 2004,
pp. 173–202.
[26] J. Rojas-Vallejos, A. Lastuka, The income inequality and carbon emissions trade-off
This project was funded by King Abdulaziz University, Jeddah, Saudi
revisited, Energy Pol. 139 (2020) 111302.
Arabia, and King Abdullah City for Atomic and Renewable Energy, [27] L. Fangmin, W. Jun, Financial system and renewable energy development: analysis
Riyadh, Saudi Arabia under grant no. (KCR-KFL-06-20). Therefore, the based on different types of renewable energy situation, Energy Proc. 5 (2011)
authors gratefully acknowledge their technical and financial support. 829–833.
[28] H.A. Bekhet, A. Matar, T. Yasmin, CO2 emissions, energy consumption, economic
growth, and financial development in GCC countries: dynamic simultaneous
References equation models, Renew. Sustain. Energy Rev. 70 (2017) 117–132.
[29] I.K. Maji, M.S. Habibullah, M.Y. Saari, Financial development and sectoral CO 2
[1] J.A. Frankel, D.H. Romer, Does trade cause growth? Am. Econ. Rev. 89 (3) (1999) emissions in Malaysia, Environ. Sci. Pollut. Control Ser. 24 (8) (2017) 7160–7176.
379–399. [30] U. Al-Mulali, B. Saboori, I. Ozturk, Investigating the environmental Kuznets curve
[2] S. Dasgupta, B. Laplante, N. Mamingi, Pollution and capital markets in developing hypothesis in Vietnam, Energy Pol. 76 (2015) 123–131.
countries, J. Environ. Econ. Manag. 42 (3) (2001) 310–335. [31] A. Erosa, T. Koreshkova, D. Restuccia, How important is human capital? A
[3] Y.-J. Zhang, The impact of financial development on carbon emissions: an quantitative theory assessment of world income inequality, Rev. Econ. Stud. 77 (4)
empirical analysis in China, Energy Pol. 39 (4) (2011) 2197–2203. (2010) 1421–1449.
[4] A. Tamazian, J.P. Chousa, K.C. Vadlamannati, Does higher economic and financial [32] M. Bagliani, G. Bravo, S. Dalmazzone, A consumption-based approach to
development lead to environmental degradation: evidence from BRIC countries, environmental Kuznets curves using the ecological footprint indicator, Ecol. Econ.
Energy Pol. 37 (1) (2009) 246–253. 65 (3) (2008) 650–661.
[5] M. Salahuddin, K. Alam, I. Ozturk, K. Sohag, The effects of electricity consumption, [33] M. Senbel, T. McDaniels, H. Dowlatabadi, The ecological footprint: a non-monetary
economic growth, financial development and foreign direct investment on CO2 metric of human consumption applied to North America, Global Environ. Change
emissions in Kuwait, Renew. Sustain. Energy Rev. 81 (2018) 2002–2010. 13 (2) (2003) 83–100.

8
N. Samargandi and K. Sohag Energy Strategy Reviews 41 (2022) 100847

[34] I. Ari, M.A. Koksal, Carbon dioxide emission from the Turkish electricity sector and [39] J.S. Cho, T. Kim, Y. Shin, Quantile cointegration in the autoregressive distributed-
its mitigation options, Energy Pol. 39 (10) (2011) 6120–6135. lag modeling framework, J. Econom. 188 (1) (2015) 281–300.
[35] V. Zeqiraj, K. Sohag, U. Soytas, Stock market development and low-carbon [40] R. Koenker, G. Bassett Jr., Regression quantiles, Econometrica: J. Econom. Soc.
economy: the role of innovation and renewable energy, Energy Econ. 91 (Sep) (1978) 33–50.
(2020), https://doi.org/10.1016/j.eneco.2020.104908. [41] R. Koenker, K.F. Hallock, Quantile regression, J. Econ. Perspect. 15 (4) (2001)
[36] M.H. Pesaran, Y. Shin, R.J. Smith, Bounds testing approaches to the analysis of 143–156.
level relationships, J. Appl. Econom. 16 (3) (2001) 289–326. [42] K. Sohag, O. Kalugina, N. Samargandi, Re-visiting environmental Kuznets curve:
[37] Y. Shin, B. Yu, M. Greenwood-Nimmo, Modelling asymmetric cointegration and role of scale, composite, and technology factors in OECD countries, Environ. Sci.
dynamic multipliers in a nonlinear ARDL framework, in: Festschrift in Honor of Pollut. Control Ser. 26 (27) (2019) 27726–27737.
Peter Schmidt, Springer, 2014, pp. 281–314. [43] Tae-Hwan Kim, Halbert White, Estimation, Inference, and Specification Testing for
[38] A. Banerjee, J. Dolado, R. Mestre, Error-correction mechanism tests for Possibly Misspecified Quantile Regression, Emerald Group Publishing Limited,
cointegration in a single-equation framework, J. Time Anal. 19 (3) (1998) 2003.
267–283.