Science of the Total Environment 766 (2021) 142583
Contents lists available at ScienceDirect
Science of the Total Environment
j ourna l homepage: www.e lsev ie r .com/ locate /sc i totenvRenewable energy consumption in Africa: Evidence from a bias corrected
dynamic panelRichmond Silvanus Baye a,⁎, Allesandro Olper a, Albert Ahenkan b, Issa Justice Musah-Surugu b,
Samuel Weniga Anuga b, Samuel Darkwah c
a University of Milan, Italy
b University of Ghana, Ghana
c Mendel University, CzechiaH I G H L I G H T S G R A P H I C A L A B S T R A C T• The study explores the drivers of renew-
able energy consumption in Sub-
Saharan Africa.
• Corrected least squares dummy variable
estimator for 32 African countries
• We abstract two groups of countries-
Carbon Efficient and Least Carbon effi-
cient countries.
• We find that economic, environmental
and socio-economic factors promote re-
newable energy consumption in the
region.⁎ Corresponding author.
E-mail address: richmondbaye@outlook.com (R.S. Bay
https://doi.org/10.1016/j.scitotenv.2020.142583
0048-9697/© 2020 Elsevier B.V. All rights reserved.a b s t r a c ta r t i c l e i n f oArticle history:
Received 11 April 2020
Received in revised form 12 September 2020
Accepted 23 September 2020
Available online 30 September 2020
Editor: Damia Barcelo
Keywords:
Renewable energy consumption
Sub-Saharan Africa
Least squares dummy variable correctedOur study investigates the determinants of renewable energy consumption in Sub-Sahara Africa.We explore the
driving factors of renewable energy consumption in the context of carbon intensity for 32 Sub-Saharan African
countries from1990 to 2015. Using carbon emission intensity to identify group-specific heterogeneity, we recog-
nize carbon-efficient and least carbon-efficient countries in the region. By relying on the corrected least squares
dummy variable estimator (LSDVC), we provide evidence on the driving factors of renewable energy consump-
tion in Sub-Saharan Africa. Consequently, the findings point to varying degrees of impact on renewable energy
consumption in the region. For instance,weobserveadvancement in technology, quality of governance, economic
progress, biomass consumption, and climatic conditions influence renewable energy consumption. With a com-
mon occurrence across all groups, the implications indicate environmental, socio-economic, and climatic factors
playing an important role in renewable energy consumption. The study further shows that urbanization and eco-
nomic globalization depress efforts towards renewable energy consumption. Apart from these common factors,
other controlling variables including; GDP per capita, environmental awareness, and biomass affect each group
differently. We conclude that, policy implications can be drawn from common factors towards harmonization
of clean energy markets and developing a policy mix that combines environmental, economic, and social factors
in attaining the Sustainable Development Goals.
© 2020 Elsevier B.V. All rights reserved.e).1. Introduction
Environmentalists and policymakers are clear on climate change and
global warming, having severe implications on the earth's survivability.
R.S. Baye, A. Olper, A. Ahenkan et al. Science of the Total Environment 766 (2021) 142583Also, they pointed to the trade-off between economic growth and sus-
tainable development as one of the main factors (Marques and
Fuinhas, 2011). As such, global multilateral environmental agreement,
most especially, the United Nations Convention on Climate Change
(1992), the Kyoto protocol (1998), and the Paris agreement (2015), re-
quire countries to enhance the reduction in carbon emitted per unit of a
good produced, while also scaling up clean energy consumption by
2050. In response, there has been call to decouple economic growth
from fossil fuel energy (International Energy Agency, 2016). Conse-
quently, renewable forms of energy such as solar, wind, hydro, geother-
mal, tidal, and biomass have gained attention in policy reforms towards
renewable energy consumption.
Despite the widespread support, the general momentum towards
renewable energy consumption is still at the incipient stage in most
countries. As of 2016, at the global level, the share of renewable energy
consumed from the total primary energy remainswell below18%,while
the share of advanced clean energy stays below 11% (International
Energy Agency, 2016). Given the growth in renewable micro-grids in
many African countries, declining prices of solar-photovoltaic, environ-
mental awareness and policy towards increasing energymix, this trend
is expected to change (Ike et al., 2020; Niyonteze et al., 2020;
Dobrotkova et al., 2018). According to Pillot et al. (2019), the share of re-
newable energy consumption is expected to double over the next
30 years. Nevertheless, this growth in consumption is expected to
vary across regions and countries (Szabó et al., 2016). At the regional
level, country-specific factors such as economic growth, environmental
factors, population growth and urbanization, political factors, techno-
logical progress and trade policies will play a critical role in driving
the change (Balcilar et al., 2018).
Currently, Sub-Sahara Africa lags in energy production and con-
sumption despite the increasing global trend (Hafner et al., 2018).
Thus, a situation many scholars have described as energy poverty. En-
ergy poverty is reflected in consumption trends as the region consumes
3.3% of total world primary energy. In the view of Szabó et al. (2016),
over 600 million people are without access to stable, reliable, and af-
fordable energy supply. Moreover, in the energy economics literature,
many factors cited explain energy poverty in Africa and by extension,
the low levels of clean renewable energy consumption in the region de-
spite the enormous and diverse energy resource in the region.
Our study builds on existing literature and ascribes some of the rea-
sons to underinvestment in clean energy sources (Alola et al., 2019;
Klagge and Nweke-Eze, 2020). Other factors include: political will and
policy decisions towards adoption and consumption (Linnerud et al.,
2014), lack of needed capital investment (Maji et al., 2019), off-grid
solar powered infrastructure (Baurzhan and Jenkins, 2016), and high
population growth and urbanization. Further, under-developed nature
of the regional markets does not promote energy sharing among coun-
tries (Szabo et al., 2011). To gain an understanding into these intricate
phenomena,we explore these driving factors at the regional level before
drilling into group-specific differences and similarities.
Presently, these studies (Oppong et al., 2020; Nathaniel and Iheonu,
2019; Maji et al., 2019; Olanrewaju et al., 2019; Ergun et al., 2019; Da
Silva et al., 2018; Attiaoui et al., 2017; Aïssa et al., 2014) examines the
determinants of renewable energy consumption in Africa by either rely-
ing on a Panel ARDL approach or using a static panel estimator.We pres-
ent our findings using a dynamic panel estimator that is robust to small
sample bias and second order - serial correlation. The dynamic panel es-
timator allows the flexibility of estimating the evolving nature of envi-
ronmental and economic relationships. Besides, our study explores the
drivers through the lens of carbon emissions per unit of a good pro-
duced. Based on this selection,we abstract two groups of renewable en-
ergy consumers in the region. One group is carbon-efficient and the
other group is the least carbon-efficient. We argue, by selecting coun-
tries based on carbon emission intensity, we are able to exploit the
amount of carbon emitted per unit of goods produced and group coun-
tries based on similar carbon intensities. On this account, the study2
contributes to the literature in threefold: (1) add to literature and en-
hance our understanding on renewable energy consumption in SSA,
(2) using a wider time frame and countries, (3) using a bias correcting
dynamic estimation approach.
The study reveals that current levels of renewable energy consump-
tion are contemporaneously correlated with past levels of renewable
energy consumption. In brief, this is true for the two broad categories
and the full sample. Moreover, we observe a positive relationship be-
tween CO2 emissions per capita and renewable energy consumption.
We show that human development improves renewable energy con-
sumption in the full sample. Thus, confirming the importance of income
and general wellbeing for renewable energy consumption. As expected,
urbanization diminished renewable energy consumption. We also ob-
serve a positive relationship between biomass (forest area) and renew-
able energy consumption. Highlighting the importance of climatic
indicators,we show that hydro-electric power and solar energy have in-
duced renewable energy consumption in the region. Considering that
most of the energy consumed in the zone is fromhydro-poweredplants,
our indicators were absolute.
Moving on to the carbon-efficient group, the dynamic assessment of
renewable energy consumption is still relevant to present energy con-
sumption. Interestingly for this group, there is a negative relationship
between income per capita, trade openness, andwood-derived biomass
on the outcome variable. The implications point out that perhaps finan-
cial flows through trade into renewable energy is still lacking in these
countries. Arguably, we agree on the same terms that the share of
wood-derived biomass from renewable sources of power dominates
the share. Hence, as income increases, we expect a lower portion of
the energy consumed from renewable sources. This same connection
highlights the negative relationship between biomass use and
renewable energy use for these countries. In the case of the least
carbon-efficient group, GDP per capita, and biomass (forest area) drive
renewable energy consumption. In summary, these findings affirm his-
toric precedents, present policy relevance, and future implications on
renewable energy consumption in SSA.
2. Literature review
As a result of the constant threat of climate change due to Green
House Gas emissions, there has been a general call towards the produc-
tion of carbon-resilient sources of energy and the reduction in the con-
sumption of fossil-based energies (Marques and Fuinhas, 2011). As
such, scores of empirical studies have emerged to examine the develop-
ment and consumption of renewable energy. For instance, Sadorsky
(2009), points to the investigation of the relationship between renew-
able energy consumption, CO2 emissions, andOil prices for theG7 coun-
tries. The study provides evidence that CO2 per capita and GDP per
capita increased renewable energy consumption. Adding to the evi-
dence, Apergis and Payne (2010) affirmed a long-run relationship be-
tween GDP and renewable energy consumption. On the determinants
of renewable energy consumption with evidence from 64 countries be-
tween the years 1990 and 2011, Omri and Nguyen (2014) showed that
oil price affects renewable energy consumption. According to the afore-
mentioned, they documented that changes in GDP influence renewable
energy use in high and Low-income countries. In the case of the full
sample, trade openness positively drives renewable energy consump-
tion. Other studies that modeled similar relationships include:
(Bhattacharya et al., 2016; Fang, 2011; Omri et al., 2014; Salim and
Rafiq, 2012; Sinha et al., 2018). What is striking in these findings is
the link between renewable energy consumption and economic
growth, trade openness, carbon emissions.
Apergis and Payne (2012) employed a panel time-series model to
establish bidirectional causality among selected variables in their
model. From this exercise, seven Central American countries showed a
positive relationship between CO2 emissions and renewable energy
consumption. Energy prices also positively affected renewable energy
R.S. Baye, A. Olper, A. Ahenkan et al. Science of the Total Environment 766 (2021) 142583consumption. Again, in an extensive study, Nguyen and Kakinaka
(2019) showed a positive association between renewable energy con-
sumption and CO2 emissions for low-income countries. Similarly,
Ohler and Fetters (2014) noted bidirectional causality amid economic
growth and renewable energy consumption. Marques and Fuinhas
(2011) applied a dynamic panel estimation technique for 24 European
countries. Findings from the study showed a contemporaneous effect
of lagged energy use on current energy use. They provide evidence in
support of urbanization, CO2 emissions, and economic growth in moti-
vating renewable energy consumption in the selected countries. In con-
trast to these findings, Ocal and Aslan (2013) show that renewable
energy consumption does not lead to economic growth but indicates a
positive or negative relationship is time or country dependent.
The literature on renewable energy in Africa mainly focused on bio-
mass andfirewood consumption. For instance (Adewuyi andAwodumi,
2017; Sulaiman et al., 2017) have documented that countries in Africa
rely heavily on combustible biomass for heating and cooking. As yet,
empirical studies that examine the share of renewable energy from
total primary energy consumption spans Da Silva et al. (2018). In their
work, they used a Panel ARDL approach to examine the determinants
of renewable energy growth in Sub-Saharan Africa. Evidence selected
in 17 SSA countries from 1990 to 2014 noted the relevance of economic
growth to renewable energy growth in the region. On the relationship
between CO2 and renewable energy consumption, Nathaniel and
Iheonu (2019) showed a negative relationship between these two var-
iables in their study. Olanrewaju et al. (2019) finds a negativeFig. 1. Selected countries
3
relationship between carbon intensity and renewable energy consump-
tion in Africa. Added to this, Oppong et al. (2020) show a long run
granger causality between GDP, CO2 and renewable energy consump-
tion for countries in Africa. This corroborates the findings of Attiaoui
et al. (2017). In similar evidence, Nathaniel and Iheonu (2019) used
an Augmented Mean Group estimation technique to document a uni-
directional causality from renewable energy to CO2 emissions for 19
countries in Africa. Contrary to these findings, Maji et al. (2019) used
a panel dynamic ordinary least squares estimation technique to show
that between 1995 and 2014, renewable energy consumption did not
promote economic growth but rather slowed down economic progress
for 15West African countries. Ergun et al. (2019) used a static panel es-
timation technique for 21 African countries for the period 1990 and
2013 to argue that because most forms of renewable energy use is
from biomass, economic progress or increases in GDP per capita may
translate into lower levels of renewable energy consumption.
3. Material and method
3.1. Source of data and variable description
In investigating the key driving factors of renewable energy con-
sumption in SSA, the study sourced data from the World Bank's World
Development Indicators (WDI), the Quality of Governance database,
and the KOF Swiss economic institute. The data spans from the period
of 1990 to 2015 for selected 32 SSA countries. Fig. 1 shows a map ofused for the study.
R.S. Baye, A. Olper, A. Ahenkan et al. Science of the Total Environment 766 (2021) 142583
Table 1
Summary statistics.
Obs St. dev Min Max Kurtosis Skewness
REN 831 22.042 10.634 98.343 3.662 −1.274
CO2 emissions per capita 800 1.637 0.017 9.979 19.195 3.97
Oil price 832 32.957 12.983 107.272 1.972 0.715
GDP per capita 832 2263.661 164.337 11,937.6 8.32 2.344
HDI 779 0.387 1.03 2.719 2.381 0.51
Urban population growth 832 1.731 −1.477 17.499 13.384 1.315
Trade openness 832 0.401 0 2.078 2.817 0.479
Economic globalization 832 10.117 15.746 85.185 4.361 0.477
Biomass 829 294,000 382 1,600,000 15.782 3.393
Rainfall 832 41.673 9.168 236.563 3.153 0.412
Temperature 832 2.692 17.381 29.541 2.413 −0.085
Governance 650 0.127 0.083 0.898 3.968 0.298
Technology 829 4.829 0 31.067 19.61 3.882the chosen countries- specific carbon intensities. The selection of coun-
tries is limited to the availability of data for the variables selected in the
study. Therefore, Table 1 provides a summary of the reported descrip-
tive statistics of the variables, including the mean, standard deviation,
minimum, maximum, kurtosis and Skewness of the distribution. From
the table, the kurtosis of all the variables remains greater than+1; indi-
cating a very peaked distribution with most of the variables having
heavier tails than a normal distribution.
3.1.1. Environmental variables
As part of the covariates, we source spot oil price measured in $ per
bbl from the yearly statistical reviews of British petroleum. As ameasure
of environmental awareness, we use CO2 emissions per capita. We ex-
pect higher levels of CO2 emissions to give rise to calls for environmental
protection, which in turn drives renewable energy consumption
(Marques and Fuinhas, 2011). We define climatic effects as average an-
nual temperature andmean annual rainfall. We anticipate either a pos-
itive or a negative impact on the outcome variable. From a hydroelectric
generating point of view, there is the likelihood of high average rainfull
needed to fill up water bodies for hydro-electricity generation. On the
temperature side, a high yearly average temperature could signal po-
tential investment into solar infrastructure, but this can potentially con-
strict hydro-electricity generation.
3.1.2. Economic variables
For the economic variables, we use GDP per capita to reflect eco-
nomic progress, with the data taken from theWorld Development Indi-
cators (WDI). Chang et al. (2009) demonstrate that higher-income
countries are likely to consume from renewable sources than those
with low income per capita. To complement the economic indicators,
we use HDI to measure general economic well-being. Additionally, we
also define economic globalization and trade openness as a conduit,
through which investment and technological innovation diffuse to
countries with lower technological capacity. The study expects either
a positive or negative effect on the outcome variable. In the view of
Omri and Nguyen (2014), trade openness facilitates the cross-border
movement and exchange of goods, which is inherently energy reliant.
Added to this, the advent of technology could either push or drag re-
newable energy consumption in the region.
3.1.3. Socio-economic variables
We anticipate a direct effect of urbanization on energy use and by
extension, renewable energy consumption. For instance, Chen (2018)
shows that in urban areas where energy demand is high, urbanization
could increase the demand for alternative sources of energy. Next, as a
proxy for wood-derived biomass, the study anticipates a positive rela-
tionship between forest areas and energy consumption in the region.
The argument remains that, SSA countries use wood derived biomass
for cooking and heating. In another context, Gan and Smith (2011)4
posit that countries endowedwith renewable resources effectively pro-
duce and consume such resources. The reverse also holds but on the
condition of developing a high-level technology to harness the opportu-
nities. We also include the quality of governance indicator to reflect in-
stitutional policies towards renewable energy consumption in the
region. Here, we expect countries with higher quality of governance to
influence renewable energy consumption positively.
3.2. Empirical models
FollowingOmri andNguyen (2014), the study develops an empirical
model that augments standard empirical models by including tempera-
ture, rainfall, urbanization, technology, governance, and biomass as part
of the controls. Expressly, in an econometric form, the determinants of
renewable energy consumption for the full sample and group-specific
cases is stated as follows:
RECit ¼ β0 þ β1CO2per capit þ β2HDIit þ β3Eco globit
þ β4tempit þ β5Rainit þ β6urbanit þ β7Biomassit
þ β8Technologyit þ β9Governanceit þ uit ð1aÞ
RECit ¼ β0 þ β1CO2per capit þ β2GDP per capitait þ β3Eco−globit
þ β4tempit þ β5Rainit þ β6urbanit þ β7Biomassit
þ β8Technologyit þ β9Governanceit þ uit ð1bÞ
where, REC represents renewable energy consumption, which is a func-
tion of; CO2 emissions, HDI, GDP per capita, trade openness, rainfall,
temperature, urbanization, and biomass, technology and governance.
Therefore, applying a panel data dimension to the function in
Eqs. (1a) and (1b) yields the following equation:
RECit ¼ β0 þ βi,t Σ CVi,t þ uit: ð2Þ
where CV captures all the covariates mentioned in Eqs. (1a) and (1b). ui
t is a function of the fixed effects and the unobserved white noise.
3.3. Estimation method
Estimating panel models require certain assumptions: First, the co-
variates should not correlate with each other. Next, the model should
be free from serial correlation. The variables should be exogenous;
whiles the distribution remains normal. In our model, variables like
GDP per capita, CO2 emissions per capita, and urban population growth
change over time. This means, to be able to correctly predict the out-
come of a model, time varying factors should be taken into consider-
ation. Hence using OLS or the fixed effects model may produce biased
coefficients and imprecise standard errors.
One such approach to correcting these measurement errors and un-
observed heterogeneities in static panels is to include dummy variables
or taking first differences. However, a problemwith these approaches is
the persistence of a correlation between the lagged term (RECit-1) and
R.S. Baye, A. Olper, A. Ahenkan et al. Science of the Total Environment 766 (2021) 142583the error term (uit). Blundell and Bond (1998) solved this by proposing
the System GMM estimator. However, recent literature suggests that
the system GMM estimator performs poorly (small sample bias prob-
lem) when the sample is characteristically small.
Our cross-sectional element of the data consists of 16 countries in
each group for a period of 25 years. Given the small size of our panel, es-
timating this with a system GMM regression could lead to a small sam-
ple bias problem. As such, Bruno (2005) proposes a model that corrects
this problem by extending the work of Kiviet et al. (1999) to accommo-
date unbalanced panels. Unlike the conventional system GMM estima-
tor, Bruno (2005) posits that the LSDVC estimator is efficient in bias
correction under unbalanced panels characterized by missing observa-
tions (Flannery andHankins, 2013). Under a strictly exogenous assump-
tion of parameter selection, Bruno (2005) shows that the LSDVC is a
preferred estimator as compared to the system GMM estimator when
the panel is short. Similar to the case of our full sample, Buddelmeyer
et al. (2008) showed that evenwhen T=5 andN=20, the LSDVCyields
consistent and reliable estimates. On this ground, the study implements
the LSDVC estimator with fixed effects that is robust in the presence of
unbalanced panels; second-order serial correlation; and unobserved
heterogeneity in a dynamic panel setting. Our study estimates LSDVC
estimator initialized by a dynamic panel estimate, which relies on a re-
cursive bias correction up to N−1 T−2 with time fixed effects following
100 bootstrapped replications to check the statistical significance of
the coefficients (Bruno, 2005). The empirical strategy for the LSDVC es-
timator is reported in the Eq. (3).
RECi,t ¼ αiRECi,t−1 þ βiCVit þ f i þ ui,t ð3Þ
where RECi,t is the outcomevariable over space and time (REC); RECi,t−1
is the lagged dependent variable with a first-order autoregressive pro-
cess. Also, the inclusion of fixed effects fi is to control for observations
that are not randomly selected. CVit encapsulates the exogenous factors
driving renewable energy consumption in the region. In addition, theUit
is the error term independent of the fi term.
4. Empirical results
4.1. Correlation analysis
Table 2 shows the correlation coefficient of the variables in the
model. Starting with the urbanization coefficient, we observe that the
elasticity of urbanization is positively related to renewable energy
consumption in all groups. Also, we observe a significant negative corre-
lation between technology, trade openness and renewable energy con-
sumption across all groups. The outcome indicates low levels ofTable 2
Correlation table.
REN_full REN_carbon REN_least
sample efficient carbon
efficient
CO2 emissions per cap −0.61*** −0.81*** −0.64***
Oil price −0.12*** −0.16*** −0.12**
GDP per cap −0.59*** −0.73*** −0.54***
HDI −0.49*** −0.18*** −0.45***
urban population 0.38*** 0.24*** 0.44***
growth
Trade openness −0.37*** −0.22*** −0.37***
Economic globalization 0.07** 0.21*** 0.05
Biomass 0.28*** 0.31*** 0.30***
Rainfall 0.11*** −0.28*** 0.35***
Temperature 0.16*** −0.32*** 0.43***
Governance −0.23*** 0.22*** −0.46***
Technology −0.75*** −0.77*** −0.74***
N 832 416 416
***, **, * indicates 1, 5, and 10 percent level of significance.
5
investment in renewable energy in the region. However, economic
globalization and wood derived biomass are positive and significant
for the full sample and the carbon-efficient group. Governance is signif-
icantly negative for the full sample and the least carbon efficient group.
In the case of the full sample and the least carbon-efficient group, we
show a positive correlation between temperature, rainfall and renew-
able energy consumption. Also, in the carbon-efficient group, we ob-
serve a negative correlation.
4.2. Fisher unit root test
Because panel data combines both elements of time series and cross-
sectional data, we need to check the stationarity of the variables to as-
certain that they don't fluctuate the mean. Additionally, the results
from the model may be spurious with nonstationary models and this
could lead to unintended conclusions. Hence, to implement a station-
arity check, we employ the Fisher Unit root test with an Augmented
Dickey-Fuller option under the null hypothesis; all panels have a unit
root, as against the alternative of no unit root. Also, Table 3 presents
the results of the Fisher unit root on the variables used in this study.
For instance, the test result at levels confirms, some variables are sta-
tionary (Urban population growth, biomass, economic globalization,
governance, rainfall, and temperature). Using the first difference, we
transformed non-stationary variables at levels (CO2 emissions per cap,
oil price GDP per cap, and technology) to stationary variables. Meaning
the order of integrations is one.
4.3. Pre-testing of the model
Before estimating the dynamic panel, we first present the drivers of
renewable energy consumption in a static lens thus, estimating a pooled
OLS andfixed effects regression. Our results presented in Table 8 reports
consistent but inefficient outcomes. To show this, we performed several
diagnostic checks. First, to ascertain the suitability of variables in the
model, we performed a test of collinearity of the covariates in the
model. Table 4 reports the output of the VIF test for multicollinearity.
As a rule of thumb, none of the reported values were greater than 10.
Hence, one can be confident of no perfect collinearity in the model.
We further test for cross-sectional dependence in the model using the
Pesaran test for cross-sectional dependence for the full sample (T < N)
and the Breusch Pagan test (T > N) for the groups. To demonstrate,
our results in Table 8, the full sample document no cross sectional de-
pendence among the variables. The group specific cases however report
the presence of cross-sectional dependence. As a final step, we per-
formed the Wooldridge test for serial correlation in the model. At this
point, our result reports the presence of serial correlation in the model
that may cause the standard errors to be bias and the estimates to be
less efficient. As such, using the pooled OLS and the fixed effect resultTable 3
Fisher unit root test.
Levels First difference
Statistic P-value Statistic P-value
REC 55.7816 0.7582 323.9597 0.0000
CO2 emissions per capita 60.206 0.6114 680.3012 0.0000
Oil price 27.3654 1.0000 884.4531 0.0000
GDP per cap 74.4327 0.175 400.7311 0.0000
Urban pop 219.3379 0.0000 403.5189 0.0000
Trade openness 46.1573 0.9548 547.4045 0.0000
Biomass 323.1709 0.0000
Temp 245.5009 0.0000 1267.837 0.0000
Rainfall 680.2298 0.0000 1733.8793 0.0000
Economic glob 114.0764 0.0000 794.9713 0.0000
Governance 84.7850 0.0015 320.1269 0.0000
Technology 53.0149 0.8347 574.9351 0.0000
R.S. Baye, A. Olper, A. Ahenkan et al. Science of the Total Environment 766 (2021) 142583
Table 4
Test for multicollinearity using VIF.
Full sample Carbon efficient Least carbon
efficient
Variable VIF 1/VIF VIF 1/VIF VIF 1/VIF
CO2 emissions per cap 2.18 0.459441 3.89 0.260531 3.76 0.265764
HDI 1.90 0.525603 3.04 0.329209 2.01 0.496525
Log rainfall 1.61 0.622814 4.10 0.243650 2.99 0.334741
Log Temperature 2.05 0.488912 3.34 0.299718 3.73 0.268315
Governance 1.99 0.503704 2.41 0.414785 3.36 0.297423
Economic globalization 1.57 0.637732 3.60 0.277569 2.00 0.500003
Urban Population growth 1.50 0.665495 1.86 0.536377 1.37 0.731201
Log Biomass 1.25 0.801565 2.01 0.498631 1.26 0.794723
Log Technology 3.13 0.319686 2.82 0.335438 5.17 0.193595
Mean VIF 2.05 2.43 2.36may not be ideal for a causal interpretation due to biased standard er-
rors, small sample bias problem and the dynamic nature of relation-
ships. Hence, we use the LSDVC estimator which is robust to second-
order serial correlation, unobserved heterogeneity, corrects the prob-
lem of small sample bias and robust to gaps in unbalanced panels for
our analysis.
4.4. Estimation for the full sample
Firstly, we seek to present the dynamic effect of the determinants of
renewable energy consumption for thewhole sample. Secondly, there is
a discussion on the carbon-efficient group and the least carbon-efficient
group. To sumup, by employing the corrected LSDV estimator, we show
the key factors driving renewable energy consumption in SSA.
The estimator controls for the dynamic effect by including the first
lag of the dependent variable as an independent variable in the model
and the variables assumes a strictly exogenous assumption. Moreover,
concerning the robustness of the result, we bootstrapped the standard
errors to 100 replications to ensure an accurate prediction of the out-
comes (Bruno, 2005). Approximately, the accuracy of the bias is up to
O (1/NT 2) and initialized with the Blundell and Bond (BB) estimator.
Further, using this adjustment ensures the consistency and efficiencyTable 5
Regression results full sample.
LSDVc_1 LSDVc_2 LSDVc_3 LSDVc_4
L.log REN 2.236*** 2.519*** 2.486*** 2.552***
(0.000) (0.000) (0.000) (0.000)
CO2 emissions per cap 0.109*** 0.099*** 0.097*** 0.088***
(0.021) (0.021) (0.021) (0.021)
HDI −0.221*** −0.139* −0.124* −0.119*
(0.070) (0.072) (0.074) (0.072)
Urban pop growth −0.113*** −0.113*** −0.118***
(0.011) (0.011) (0.011)
Trade openness −0.014
(0.017)
Economic globe −0.003***
(0.001)
Log Biomass
Log Rainfall
Log Temp
Log Technology
Governance
N 326 326 326 326
Bias correction initialized by Blundell and Bond estimator. Bias approximation is accurate up
p < 0.01, p < 0.05 and p < 0.1 respctively.
6
of the results. We report our discussion for the full sample in Table 5.
From the results, the first lag of the dependent variable is positive and
significant which implies that current levels of renewable energy con-
sumption is dependent on past patterns of renewable energy consump-
tion. This affirms the appropriateness of using the dynamic panel
estimator. It also validates the idea that changes in consumption pat-
terns take time. All things considered, the study underscores that in-
cluding the lag of the dependent variable in the model improves the
explanatory power of control variables.
Column 9 of Table 5 reports the preferred regression output. Also,
Columns (1–8) report the parsimonious regression with various de-
grees of controls. Moreover, an elaborative way to capture the true ef-
fect on the dependent variable in a dynamic panel setting is to capture
country-specific factors that do not change over time. We achieve this
by controlling for time effects in all the specifications. The first column
which is our baseline regression examines the role of CO2 emissions
per capita on renewable energy consumption and using HDI as control.
Here, our result documents a positive and significant relationship be-
tween CO2 emissions per capita and renewable energy consumption.
However, with the inclusion of several classes of control variables in
the model, the size of the effect diminishes and loses statistical signifi-
cance in column (9). Therefore, we interpret the outcome in column
(7) as (e0.042–1)*100 = 4.29% increase in renewable energy consump-
tion when CO2 emissions per capita increase by a unit. Our result cor-
roborates the findings of (Apergis and Payne, 2014; Sadorsky, 2009,
Salim and Rafiq, 2012). An important note of caution is, as we control
for technological progress and governance, the coefficient - CO2 per
capita losses its significance in explaining renewable energy consump-
tion in the region.
The role of human development is positive in column 9. Our coeffi-
cient of regression shows that a marginal improvement in human de-
velopment drives renewable energy consumption in the region. In
terms of elasticity, we show that the partial elasticity of HDI on renew-
able energy consumption is 15.5%. Our coefficient shows that human
development is prominent for renewable energy consumption and
that continuous improvement in economic development is necessary
for renewable energy consumption. This outcome is consistent with
(Chen, 2018).LSDVc_5 LSDVc_6 LSDVc_7 LSDVc_8 LSDVc_9
2.483*** 2.521*** 2.648*** 2.347*** 2.227***
(0.000) (0.000) (0.000) (0.000) (0.000)
0.080*** 0.084*** 0.042* 0.020 −0.020
(0.021) (0.021) (0.022) (0.023) (0.028)
−0.087 −0.099 −0.019 0.067 0.155*
(0.071) (0.072) (0.074) (0.070) (0.085)
−0.119*** −0.120*** −0.131*** −0.118*** −0.111***
(0.011) (0.010) (0.011) (0.013) (0.017)
−0.003*** −0.002*** −0.004*** −0.004*** −0.004***
(0.001) (0.001) (0.001) (0.001) (0.001)
0.081** 0.092** −0.002 −0.078* 0.161**
(0.039) (0.039) (0.040) (0.044) (0.065)
−0.010 0.036 0.045* 0.067**
(0.023) (0.026) (0.027) (0.033)
2.190*** 1.770*** 1.306**
(0.450) (0.471) (0.604)
0.006 0.020*
(0.007) (0.011)
0.277*
(0.153)
326 326 326 323 243
100 replications. Bootstrapped standard errors reported in paranthesis. ***, **, * indicate
R.S. Baye, A. Olper, A. Ahenkan et al. Science of the Total Environment 766 (2021) 142583In contrast to these findings (Salim and Shafiei, 2014; Ding and Li,
2017), our empirical results show a negative relationship between ur-
banization and renewable energy consumption in the region. Our coef-
ficient of regression shows a 0.11% reduction in renewable energy
consumption in the region. A cautious explanation of this outcome
may be required. A possible reason could be that the growth rate of
urbanization in Nigeria and South Africa may have influenced the
overall outcome. Another reason could be due to rapid urbanization
coupled with rising demand for energy. As such, governments in the
region focus more on fossil-based sources of energy than renewable
forms of energy.
Given that combustible biomass is consumed excessively in the re-
gion, our results are unsurprising as the consumption of biomass (forest
area) increases renewable energy consumption in the region.Wedo not
expect a long term relationship between these two variables due to the
exhaustible nature of forest resources. Based on this result, we interpret
this outcome through the lens of wood-derived biomass consumption.
Further, our coefficient of regression shows a percentage increase in
wood-derived biomass drives renewable energy consumption by
0.16%. Moving on to climatic stress, we present evidence in support of
the growth in solar energy consumption in this region. Our results
show that when daily temperature increases by one (1) degree Celcius,
renewable energy consumption will increase by 1.31%, which is also
true for hydro-forms of energy in region. As hydro-electricity is the
dominant form of energy production in the region, it is not surprising
that the effect on renewable energy consumption improved by 0.067%.
From an economic point of view, our result signals the enormous poten-
tial in solar energy generation in the region. Another possible reason for
the positive impact is the significant drop in solar-PV prices in theworld
market, which has driven off-grid solar investment in the region. Con-
sidered alongside the variables in the model, economic globalization is
significant and has a negative effect on renewable energy consumption.
What this means is that the share of investment into renewable energy,
is dominated by the share of investment into fossil-based fuels.
Moving on, other factors such as technological progress and the
quality of governance could have a direct effect on renewable energyTable 6
Regression results carbon efficient countries.
(1) (2) (3) (4) (
LSDVc_2 LSDVc_2 LSDVc_3 LSDVc_4 L
L.log REN 5.084*** 5.427*** 6.055*** 6.176*** 4
(0.000)* (0.000) (0.000) (0.000) (
Log Oil price −0.103***
(0.020)
Co2 emissions per cap 0.838*** 0.726*** 0.685*** 0
(0.058) (0.062) (0.065) (
Log GDP per capita −0.693*** −0.479*** −0.498*** −
(0.039) (0.051) (0.050) (
Urban population −0.157*** −0.163*** −
(0.016) (0.015) (
Trade openness −0.069**
(0.029)
Economic globalization 0
(
Log Biomass
Log rainfall
Log temp
Log Technology
Governance
Obs. 189 173 173 173 1
Year Dummy Yes Yes Yes Yes Y
Bias correction initialized by Blundell and Bond estimator. Bias approximation is accurate up
p < 0.01, p < 0.05 and p < 0.1 respctively.
7
consumption in the region. As shown in column 9 of Table 5, our result
confirms, the conjecture that technological progress is paramount to re-
newable energy consumption in the region. The coefficient of the re-
gression output point to a 0.02 improvement in renewable energy
consumption when technology improves by 1%. In recent times, due
to advancement in technological innovations over the past decade,
prices of solar-pv has dropped significantly in the world market leading
massive investment in solar energy in the region in recent times. The ef-
fect of governance on renewable energy consumption is undeniable.We
expected a positive association with renewable energy consumption
when the quality of governance improves. This is exactly what we ob-
serve. The coefficient of the regression show an improvement in renew-
able energy by 28% when quality of governance improves by 1%. An
implication from this finding is; when environmental concerns are em-
bedded in long term development plans of countries, a 100% green
economy is achievable in the region.
4.5. Estimation for the carbon efficient group
In this section, we discuss the factors that promote renewable en-
ergy consumption among the carbon-efficient group of countries.
Table 6 reports the estimated effect on the outcome variable with a
focus on the result reported in column 10. The first point of interest is
evidence of theprevious level of renewable energy consumption on cur-
rent levels of renewable energy consumption. Our findings unveil
strong statistical significance, which reveals persistence. Moreover, per-
sistence implies that renewable energy consumption is sustainable over
the years. Next, we show that the partial elasticity of CO2 emissions per
capita improves renewable energy consumption by about 25%. Our co-
efficient suggests that environmental awareness promotes renewable
energy consumption. This finding is consistent with empirical literature
(Sadorsky, 2009), documenting that environmental concerns are im-
portant in driving renewable energy consumption.
Moving on to GDP per capita, the expectation is that as countries
progress, standards of living improves that could translate into better
decisions on environmental quality and protection. Hence, one would5) (6) (7) (8) (9) (10)
SDVc_5 LSDVc_6 LSDVc_7 LSDVc_8 LSDVc_9 LSDVc_10
.400*** 5.292*** 4.877*** 3.711*** 3.823*** 5.233***
0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
.671*** 0.748*** 0.649*** 0.539*** 0.564*** 0.245***
0.066) (0.067) (0.068) (0.067) (0.063) (0.088)
0.349*** −0.560*** −0.488*** −0.396*** −0.379*** −0.524***
0.044) (0.046) (0.047) (0.047) (0.050) (0.130)
0.106*** −0.094*** −0.086*** −0.070*** −0.080*** −0.162***
0.015) (0.015) (0.015) (0.015) (0.015) (0.022)
.004*** 0.006*** 0.005*** 0.002** 0.002*** 0.010***
0.001) (0.001) (0.001) (0.001) (0.001) (0.001)
−0.546*** −0.477*** −0.327*** −0.240*** −0.116
(0.038) (0.039) (0.040) (0.051) (0.108)
−0.039 0.042 0.068** 0.147**
(0.032) (0.033) (0.027) (0.063)
3.492*** 4.035*** 5.346***
(0.424) (0.502) (0.672)
0.021*** 0.029**
(0.008) (0.012)
1.375***
(0.288)
73 173 173 173 170 114
es Yes Yes Yes Yes Yes
100 replications. Bootstrapped standard errors reported in paranthesis. ***, **, * indicate
R.S. Baye, A. Olper, A. Ahenkan et al. Science of the Total Environment 766 (2021) 142583expect a higher share of renewable energy consumption as countries
progress economically. Our result however, portrays a negative rela-
tionship between GDP per capita and renewable energy consumption.
We join Ergun et al. (2019) in arguing that most of the countries in
this category consume a larger fraction of their share of renewable en-
ergy from the “traditional source”. As such, as these countries develop
we expect a negative relationship. Another explanation of this outcome
is because countries are expected to grow in order to improve on the
quality of life of its citizens, fossil fuels which come as a cheaper alterna-
tive to cleaner forms of renewable energy consumption will be inten-
sively used to drive development (Ergun et al., 2019). The estimated
impact reveals that renewable energy consumption declines by 0.52%
when GDP per capita increases by a percentage.
Moving on to economic globalization, our findings reveal that a 1%
improvement in trade improves renewable energy consumption by
1%. This result supports earlier empirical works (Chen, 2018). Indeed,
our finding shows that investment into the renewable sector in these
countries is gaining momentum. From another angle, the impact of
trade on renewable energy consumption could imply that the techno-
logical possibilities for the development of the renewable sector assess-
able for both off-grid and grid installation. We see this impact as the
regression result for technological innovation improves renewable
energy consumption by 0.03%.
In what follows, we discuss the role of socio-economic factors on
renewable energy consumption. Focusing on urbanization, we show
that urban growth depresses renewable energy consumption by
0.16%. This finding highlights the damaging impact of urbanization on
renewable energy consumption in these countries. Also, we argue that
perhaps due to existing energy poverty, national governments are
forced to rely on non-renewable sources, which are relatively cheaper
to produce and distribute. However, a prudent policy action from
national government could promote renewable energy consumption.
The finding in our estimate shows exactly that. That is, quality of gover-
nance improves renewable energy consumption. Our regression model
further focused on wood-derived biomass. Here we point out that
renewable energy consumption declines by about 0.16%. We showTable 7
Regression results for least carbon efficient countries.
(1) (2) (3) (4) (5
LSDVc_2 LSDVc_2 LSDVc_3 LSDVc_4 LS
L.log REN 2.566*** 2.780*** 3.136*** 3.088*** 3
(0.000) (0.000) (0.000) (0.000) (0
Log oil price 0.457***
(0.081)
CO2 emissions per cap 0.077*** 0.051* 0.043 0
(0.026) (0.027) (0.027) (0
Log GDP per cap 0.550*** 0.418*** 0.473*** 0
(0.083) (0.085) (0.092) (0
Urban population growth −0.232*** −0.226*** −
(0.020) (0.020) (0
Trade openness −0.053**
(0.023)
Economic globalization −
(0
Log biomass
Log rainfall
Log temp
Log Technology
Governance
Obs. 193 177 177 177 1
Year Dummy Yes Yes Yes Yes Y
Bias correction initialized by Blundell and Bond estimator. Bias approximation is accurate up
p < 0.01, p < 0.05 and p < 0.1 respctively.
8
that the natural resource base pushes for less consumption of cleaner
forms of renewable energy. Probably, the dependence on biomass
which is cheaper and readily available does not encourage cleaner
forms of renewable energy consumption. All in all, our climatic controls
(temperature and rainfall) show a positive effect on renewable energy
consumption. The strongest of the effect is seen in solar energy (tem-
perature), which highlights the potential for greater renewable energy
consumption in these countries.
4.6. Estimation for the least carbon efficient group
We note that in the preferred regression reported in column 10 of
Table 7, we controlled for time effects in all cases. First of all, as ex-
pected, the lag effect of renewable energy consumption on the outcome
variable is persistent and significant. Our estimates confirm that CO2
emissions depress renewable energy consumption for these countries.
This outcome is not surprising as environmental concerns are perhaps
not paramount in the development plans of these countries. The coeffi-
cient of CO2 emissions per capita shows that renewable energy con-
sumption declines by 7% when CO2 emissions per capita increase by a
metric ton.
With our primary focus on the results reported in column 10 of
Table 7, we show that GDP per capita contributes to renewable energy
consumption among these countries. An obvious implication is financial
development and economic growth is a pre-condition for renewable en-
ergy consumption for this group of countries. Also, the marginal effect
reveals a 0.32% increase in renewable energy consumption when GDP
per capita increases by 1%. Next, we show that urban population growth
deters renewable energy consumption. The partial elasticity coefficient
of urban population growth on renewable energy consumption is ap-
proximately 28%. Considering the magnitude of the decline, it remains
imminent that there is already a depressing energy situation in the
regionwhichhas been exacerbated by population growth and urbaniza-
tion. Trade openness and economic globalization does not contribute to
renewable energy consumption for the select group of countries. Going
further, our results reveal a negative relationship between economic) (6) (7) (8) (9) (10)
DVc_5 LSDVc_6 LSDVc_7 LSDVc_8 LSDVc_9 LSDVc_10
.149*** 3.281*** 3.407*** 3.392*** 3.249*** 5.411***
.000) (0.000) (0.000) (0.000) (0.000) (0.000)
.042 0.034 0.037 0.020 −0.017 −0.077***
.027) (0.027) (0.026) (0.026) (0.038) (0.028)
.412*** 0.437*** 0.458*** 0.465*** 0.371*** 0.320***
.086) (0.086) (0.086) (0.087) (0.083) (0.089)
0.225*** −0.259*** −0.268*** −0.262*** −0.289*** −0.284***
.020) (0.020) (0.020) (0.020) (0.019) (0.021)
0.005*** −0.004** −0.004*** −0.004** −0.003** −0.009***
.001) (0.001) (0.001) (0.001) (0.001) (0.001)
0.535*** 0.560*** 0.552*** 0.400*** 0.932***
(0.073) (0.073) (0.075) (0.077) (0.064)
−0.096*** −0.072** −0.059 0.088**
(0.034) (0.037) (0.047) (0.044)
1.359** 0.564 0.079
(0.633) (0.606) (0.884)
0.061*** 0.077***
(0.018) (0.017)
0.600***
(0.189)
77 177 177 177 174 151
es Yes Yes Yes Yes Yes
100 replications. Bootstrapped standard errors reported in paranthesis. ***, **, * indicate
R.S. Baye, A. Olper, A. Ahenkan et al. Science of the Total Environment 766 (2021) 142583
Table 8
Pre-testing of the panel.
DEP VAR Full model Carbon efficient Least carbon efficient
OLS FE OLS FE OLS FE
CO2 emissions per cap −0.114*** −0.123*** −0.431*** 0.127** −0.0895*** −0.0956***
(0.00662) (0.0127) (0.102) (0.0506) (0.0054) (0.0138)
HDI 0.0437 0.201*** 0.0132 0.743*** 0.316*** 0.0655
(0.0325) (0.0456) (0.0202) (0.0828) (0.0292) (0.0595)
Urban population growth 0.0154* 0.00305 0.0106* 0.00398 0.0440*** 0.0247***
(0.00836) (0.00416) (0.00543) (0.00453) (0.0104) (0.00565)
Economic globalization 0.00209** −0.00011 −0.00464*** −0.00232*** 0.00733*** 0.00240***
(0.00085) (0.00059) (0.00073) (0.00065) (0.00122) (0.00085)
Log Biomass 0.0201*** −0.150*** 0.0396*** 0.525*** 0.0642*** −0.0993**
(0.00577) (0.0336) (0.0045) (0.0844) (0.00678) (0.0388)
Log rainfall 0.191*** −0.0228 −0.107*** 0.0127 0.227*** −0.0423
(0.0225) (0.0243) (0.0161) (0.0335) (0.0265) (0.0298)
Log temp 0.252** −0.246 −0.442*** −0.0568 0.612*** −0.353
(0.125) (0.352) (0.0979) (0.448) (0.121) (0.449)
Log Technology −0.0816*** −0.0349*** −0.0191** 0.0131 −0.133*** −0.0885***
(0.0146) (0.00732) (0.00738) (0.00807) (0.0177) (0.0121)
Governance −0.00084 −0.117*** 0.134** 0.0449 −0.586*** −0.199***
(0.0973) (0.0435) (0.0627) (0.0517) (0.131) (0.0677)
Constant 2.271*** 6.553*** 5.957*** −2.92 −0.0809 6.353***
(0.54) (1.245) (0.369) (2.073) (0.45) (1.543)
Panel serial correlation 246.184*** 62.875*** 235.623***
Observations 569 569 245 245 324 324
R-squared 0.763 0.447 0.759 0.541 0.906 0.603
Mean VIF 2.05 2.43 2.36
Cross Sectional Dependence 0.7065 0.000 0.000
Year Fixed Effects YES YES YES YES YES YES
Robust standard errors in parenthesis. ***, **, * indicates 1, 5, and 10 percent level of significance.globalization and renewable energy consumption. The regression coef-
ficient show a decline in renewable energy consumption by 9% owing to
intensifying economic globalization. However, technological progress
induces renewable energy consumption for these countries. We show
that as technology advances, renewable energy consumption improves
by about 0.08%. Our finding for this group of countries is affirmed as
huge inflows of financial investment has been received in the coal, pe-
troleum and gas sector of these countries over the past decade. Our
work affirms the finding of Amri (2017).
We now turn our attention to the socio-economic implication of re-
newable energy consumption for these countries. The coefficient of re-
gression for urbanization is negative and statistically significant, which
means that ecological civilizationwas likely not included in the national
development plans of these countries. In column 10, renewable energy
consumption reduces by about 28% when the urban population grows
by 1%. Here, we show that people are not aware of the effect of urban
population growth on sustainable energy consumption in the region.
This has further implication on land use change in the era of major
structural change. On climatic shocks, we observe an improvement in
renewable energy consumption by 0.09% when hydro-power is im-
proved in the region. As expected, wood-derived biomass induces re-
newable energy consumption in a positive direction. Our result shows
a 0.9% increase in renewable energy consumption in the region. We
argue again that this outcome is expected given that biomass remains
one of the most consumed form of renewable sources of energy. On
quality of governance, the outcome in this category is no different
from the full sample and the carbon efficient group. This signifies an im-
portant reason to include good institutional reforms towards achieving
the sustainable development goals.
5. Conclusion and policy implication
The study aimed at identifying the key factors that drive the con-
sumption of energy from renewable sources. Using a dynamic estimator
which is robust to second-order serial correlation, heteroscedasticity,
and Nickel bias, the study estimated the determinants of renewable9
energy consumption for 32 Sub Saharan African countries. Further, the
study defined two broad classes of renewable energy consumers in
SSA because of cross country differences in carbon emissions per unit
of a good produced. Emerging questions from the categorization sought
to understandwhat factors contribute to the similarities and differences
in the determinants of renewable energy consumption in SSA. For in-
stance, dwelling on the commonalities, the study observes common fac-
tors such as; quality of governance, technological innovation, and
climatic factors drive renewable energy consumption in both carbon
efficient and least carbon efficient countries. Our study also showurban-
ization depresses renewable energy consumption for both groups and
further points to the relevance of green growth in urban cities in Africa.
Aside from the common factors, our study identified factors such as
environmental awareness (Co2 emissions per cap), GDP per capita, and
economic globalization to having different impacts on renewable en-
ergy consumption in the region. Of first order importance, the study
found that whereas carbon efficient show strong environmental con-
cerns leading to increases in renewable energy consumption, the least
carbon efficient countries show a negative association for environmen-
tal concerns with renewable energy consumption. The stark difference
between these two groups implies that environmental awareness is
country and region dependent. Next, our results show that globaliza-
tion and GDP per capita have different effects on renewable energy
consumption. A focus on the carbon efficient group shows that eco-
nomic globalization is paramount to renewable energy consumption.
Perhaps, under them being carbon efficient signals to development
partners into investing in renewable forms of energy in these countries.
Shifting focus to the least carbon efficient group, we observe that
economic globalization does not promote renewable energy consump-
tion in these countries. Another important difference is the role of
income per capital on renewable energy consumption. Evidently, in
the case of the carbon efficient group, economic progress does not
promote renewable energy consumption; rather discourages renew-
able energy consumption. Moreover, the results indicate income is nec-
essary to decouple carbon emissions from economic growth from the
lenses of the carbon efficient group. In other words, the positive
R.S. Baye, A. Olper, A. Ahenkan et al. Science of the Total Environment 766 (2021) 142583relationship affirms the hypothesis that economic growth is needed for
renewable energy consumption.
From general observations, a critical look at the results showed cli-
matic conditions playing an impactful role on renewable energy con-
sumption. This outcome is consistent across all groups with varying
degrees of impact. More particularly, our results show that the effect
on renewable energy consumption is stronger for the carbon efficient
group than the least carbon-efficient. It is innocuous to argue that cli-
matic conditions positively impact renewable energy consumption
through investment in solar energy and hydro-electricity. The study
also documents the nexus between wood-derived biomass (forest
area) and renewable energy. As an important carbon sink, forested
areas has provided societies with wood for energy consumption. Our
study shares this insight into the least carbon intensive countries. We
show that countries belonging to this category are heavily reliant on tra-
ditional bioenergy for cooking and heating. Our result support the no-
tion that perhaps, this countries overwhelmingly rely on biomass for
their daily livelihood. For the carbon-intensive group, our result did
not find any relationship between traditional biomass and renewable
energy consumption.
In conclusion, we can glean several policy implications from the
findings. First, given that some similarities exist on the drivers of renew-
able energy consumption, national governments should build a com-
mon environmental policy to regulate and foster renewable energy
consumption in the region. For instance, the building of energymarkets
where grid sharing is encouraged could improve renewable energy con-
sumption in other African countries, as well as, ending the energy pov-
erty in the region. Also, on account that GDP per capita improves
renewable energy consumption for thewhole sample implies economic
progress, is a precondition for renewable energy consumption. Knowing
that material use will increase as countries grow, policy implementers
in the region, should look into long term impacts of economic growth
on environmental degradation. Again, as with urban city development,
policy implementers need to encourage environmental protection
through sustainable land-use practices; minimizing biomass consump-
tion at the local level. Likewise, there should be off-grid solar infrastruc-
ture in rural and urban areas and encourage localization of energy
markets in the region. Further, since economic globalization promotes
renewable energy consumption in the region, governments should fos-
ter investments into green technologies, and the management of green
technologies to encourage sustainability in the least carbon efficient
regions.
CRediT authorship contribution statement
Conceptualization: Alessandro Olper; SamuelWeniga Anuga. Acqui-
sition of data: Richmond Baye. Analysis and/or interpretation of data:
Richmond Baye. Drafting the manuscript: Richmond Baye. Revising
the manuscript critically for important intellectual content: Samuel
Darkwah; Issa Justice Musah-Surugu; Albert Ahenkan.
Declaration of competing interest
The authors declare that they have no known competing financial
interests or personal relationships that could have appeared to influ-
ence the work reported in this paper.
References
Adewuyi, A.O., Awodumi, O.B., 2017. Biomass energy consumption, economic growth and
carbon emissions: fresh evidence from West Africa using a simultaneous equation
model. Energy 119, 453–471.
Aïssa, M.S.B., Jebli, M.B., Youssef, S.B., 2014. Output, renewable energy consumption and
trade in Africa. Energy Policy 66, 11–18.
Alola, A.A., Bekun, F.V., Sarkodie, S.A., 2019. Dynamic impact of trade policy, economic
growth, fertility rate, renewable and non-renewable energy consumption on ecolog-
ical footprint in Europe. Sci. Total Environ. 685, 702–709.10Amri, F., 2017. Intercourse across economic growth, trade and renewable energy con-
sumption in developing and developed countries. Renew. Sust. Energ. Rev. 69,
527–534.
Apergis, N., Payne, J.E., 2010. Renewable energy consumption and economic growth: ev-
idence from a panel of OECD countries. Energy Policy 38 (1), 656–660.
Apergis, N., Payne, J.E., 2012. Renewable and non-renewable energy consumption-growth
nexus: evidence from a panel error correction model. Energy Econ. 34 (3), 733–738.
Apergis, N., Payne, J.E., 2014. Renewable energy, output, CO2 emissions, and fossil fuel
prices in Central America: evidence from a nonlinear panel smooth transition vector
error correction model. Energy Econ. 42, 226–232.
Attiaoui, I., Toumi, H., Ammouri, B., Gargouri, I., 2017. Causality links among renewable
energy consumption, CO 2 emissions, and economic growth in Africa: evidence
from a panel ARDL-PMG approach. Environ. Sci. Pollut. Res. 24 (14), 13036–13048.
Balcilar, M., Ozdemir, Z.A., Ozdemir, H., Shahbaz, M., 2018. The renewable energy con-
sumption and growth in the g-7 countries: evidence from historical decomposition
method. Renew. Energy 126, 594–604.
Baurzhan, S., Jenkins, G.P., 2016. Off-grid solar PV: is it an affordable or appropriate solu-
tion for rural electrification in Sub-Saharan African countries? Renew. Sust. Energ.
Rev. 60, 1405–1418.
Bhattacharya, M., Paramati, S.R., Ozturk, I., Bhattacharya, S., 2016. The effect of renewable
energy consumption on economic growth: evidence from top 38 countries. Appl. En-
ergy 162, 733–741.
Blundell, R., Bond, S., 1998. Initial conditions and moment restrictions in dynamic panel
data models. J. Econ. 87 (1), 115–143.
Bruno, G.S., 2005. Estimation and inference in dynamic unbalanced panel-data models
with a small number of individuals. Stata J. 5 (4), 473–500.
Buddelmeyer, H., Jensen, P.H., Oguzoglu, U., Webster, E., 2008. Fixed Effects Bias in Panel
Data Estimators (Available at SSRN 1136288).
Chang, T.H., Huang, C.M., Lee, M.C., 2009. Threshold effect of the economic growth rate on
the renewable energy development from a change in energy price: evidence from
OECD countries. Energy Policy 37 (12), 5796–5802.
Chen, Y., 2018. Factors influencing renewable energy consumption in China: an empirical
analysis based on provincial panel data. J. Clean. Prod. 174, 605–615.
Da Silva, P.P., Cerqueira, P.A., Ogbe, W., 2018. Determinants of renewable energy growth
in sub-Saharan Africa: evidence from panel ardl. Energy 156, 45–54.
Ding, Y., Li, F., 2017. Examining the effects of urbanization and industrialization on carbon
dioxide emission: evidence from China's provincial regions. Energy 125, 533–542.
Dobrotkova, Z., Surana, K., Audinet, P., 2018. The price of solar energy: comparing com-
petitive auctions for utility-scale solar PV in developing countries. Energy Policy
118, 133–148.
Ergun, S.J., Owusu, P.A., Rivas, M.F., 2019. Determinants of renewable energy consumption
in Africa. Environ. Sci. Pollut. Res. 26 (15), 15390–15405.
Fang, Y., 2011. Economic welfare impacts from renewable energy consumption: the China
experience. Renew. Sust. Energ. Rev. 15 (9), 5120–5128.
Flannery, M.J., Hankins, K.W., 2013. Estimating dynamic panel models in corporate fi-
nance. J. Corp. Finan. 19, 1–19.
Gan, J., Smith, C., 2011. Drivers for renewable energy: a comparison among OECD coun-
tries. Biomass Bioenergy 35 (11), 4497–4503.
Hafner, M., Tagliapietra, S., de Strasser, L., 2018. Prospects for renewable energy in Africa.
Energy in Africa. Springer, Cham, pp. 47–75.
Ike, G.N., Usman, O., Alola, A.A., Sarkodie, S.A., 2020. Environmental quality effects of in-
come, energy prices and trade: the role of renewable energy consumption in g-7
countries. Science of The Total Environment 137813.
International Energy Agency, 2016. Decoupling of global emissions and economic growth
confirmed. Retrieved from. https://www.iea.org/news/decoupling-of-global-emis-
sions-and-economic-growth-confirmed.
Kiviet, J.F., et al., 1999. Expectations of Expansions for Estimators in a Dynamic Panel Data
Model; Some Results for Weakly Exogenous Regressors.
Klagge, B., Nweke-Eze, C., 2020. Financing large-scale renewable-energy projects in
Kenya: investor types, international connections, and financialization. Geografiska
Annaler: Ser. B Hum. Geogr. 102 (1), 61–83.
Linnerud, K., Andersson, A.M., Fleten, S.E., 2014. Investment timing under uncertain re-
newable energy policy: an empirical study of small hydropower projects. Energy
78, 154–164.
Maji, I.K., Sulaiman, C., Abdul-Rahim, A.S., 2019. Renewable energy consumption and eco-
nomic growth nexus: a fresh evidence from West Africa. Energy Rep. 5, 384–392.
Marques, A.C., Fuinhas, J.A., 2011. Drivers promoting renewable energy: a dynamic panel
approach. Renew. Sust. Energ. Rev. 15 (3), 1601–1608.
Nathaniel, S.P., Iheonu, C.O., 2019. Carbon dioxide abatement in Africa: the role of renew-
able and non-renewable energy consumption. Sci. Total Environ. 679, 337–345.
Nguyen, K.H., Kakinaka, M., 2019. Renewable energy consumption, carbon emissions, and
development stages: Some evidence from panel cointegration analysis. Renew. En-
ergy 132, 1049–1057.
Niyonteze, J.D.D., Zou, F., Asemota, G.N.O., Bimenyimana, S., Shyirambere, G., 2020. Key
technology development needs and applicability analysis of renewable energy hybrid
technologies in off-grid areas for the Rwanda power sector. Heliyon 6 (1), e03300.
Ocal, O., Aslan, A., 2013. Renewable energy consumption–economic growth nexus in
Turkey. Renew. Sust. Energ. Rev. 28, 494–499.
Ohler, A., Fetters, I., 2014. The causal relationship between renewable electricity genera-
tion and GDP growth: a study of energy sources. Energy Econ. 43, 125–139.
Olanrewaju, B.T., Olubusoye, O.E., Adenikinju, A., Akintande, O.J., 2019. A panel data anal-
ysis of renewable energy consumption in Africa. Renew. Energy 140, 668–679.
Omri, A., Nguyen, D.K., 2014. On the determinants of renewable energy consumption: in-
ternational evidence. Energy 72, 554–560.
R.S. Baye, A. Olper, A. Ahenkan et al. Science of the Total Environment 766 (2021) 142583Omri, A., Nguyen, D.K., Rault, C., 2014. Causal interactions between CO2 emissions, fdi, and
economic growth: evidence from dynamic simultaneous-equation models. Econ.
Model. 42, 382–389.
Oppong, A., Jie, M., Acheampong, K.N., Sakyi, M.A., 2020. Variations in the environment,
energy and macroeconomic interdependencies and related renewable energy transi-
tion policies based on sensitive categorization of countries in Africa. J. Clean. Prod.
255, 119777.
Pillot, B., Muselli, M., Poggi, P., Dias, J.B., 2019. Historical trends in global energy policy and
renewable power system issues in sub-Saharan Africa: the case of solar PV. Energy
Policy 127, 113–124.
Sadorsky, P., 2009. Renewable energy consumption and income in emerging economies.
Energy Policy 37 (10), 4021–4028.
Salim, R.A., Rafiq, S., 2012. Why do some emerging economies proactively accelerate the
adoption of renewable energy? Energy Econ. 34 (4), 1051–1057.11Salim, R.A., Shafiei, S., 2014. Urbanization and renewable and non-renewable energy con-
sumption in OECD countries: an empirical analysis. Ecol. Model. 38, 581–591.
Sinha, A., Shahbaz, M., Sengupta, T., 2018. Renewable energy policies and contradictions
in causality: a case of next 11 countries. J. Clean. Prod. 197, 73–84.
Sulaiman, C., Abdul-Rahim, A.S., Mohd-Shahwahid, H.O., Chin, L., 2017. Wood fuel con-
sumption, institutional quality, and forest degradation in sub-Saharan Africa: evi-
dence from a dynamic panel framework. Ecol. Indic. 74, 414–419.
Szabo, S., Bódis, K., Huld, T., Moner-Girona, M., 2011. Energy solutions in rural Africa:
mapping electrification costs of distributed solar and diesel generation versus grid
extension. Environ. Res. Lett. 6 (3), 034002.
Szabó, S., Moner-Girona, M., Kougias, I., Bailis, R., Bódis, K., 2016. Identification of advan-
tageous electricity generation options in sub-Saharan Africa integrating existing re-
sources. Nat. Energy 1 (10), 1–8.