Energy Consumption and Environmental Quality in Africa: Does Energy Efficiency Make Any Difference? ^†

Abstract

In line with the quest by policymakers to reduce greenhouse gas emissions towards Agenda 2050 and environmental sustainability, this study examines whether in the remit of Sustainable Development Goal 7, energy efficiency plays a significant role in mitigating environmental concerns associated with energy consumption. We do this by drawing macro-data on 20 sub−Saharan African countries for the period 2000–2020. Evidence based on the dynamic Generalize Method of Moments estimator shows that although overall, energy consumption triggers remarkable environmental setbacks, renewable energy consumption shows a favourable environmental effect. The results further show that energy efficiency is both directly and indirectly effective for reducing environmental pollution. Notably, the study finds that energy efficiency interacts with energy consumption to yield marked greenhouse gas emission reductions measured against carbon and nitrous emissions. In particular, we find that while renewable energy is significant for propelling Africa towards environmental sustainability, non-renewable energy shows a harmful effect. We provide policy recommendations based on the finding that investments in energy efficiency and renewable energy provide solutions to maintaining environmental sustainability. African countries should strive to include renewable energy in their energy mix and improve investments in line with SDG7 and Aspiration 1.7 of Africa’s Agenda 2063.

1. Introduction

Since the ground-breaking report of the Brundtland committee in the current Sustainable Development Goals (SGDs) era, policymakers worldwide have stepped up efforts aimed at achieving environmental sustainability (Sarkodie [1]; Sachs et al. [2]; Brundtland [3]). Broad-based commitment to this effect is seen in the institution of Agenda 2030, dubbed “The Future We Want”, which, in a nutshell, seeks to foster shared prosperity, ease the stress on natural capital and protect ecosystems for the sustainability for the planet (United Nations [4]). Beside this commitment, Africa leaders are discussing a growth course dubbed “The Africa We Want”, with the long-term objective of streamlining the continent’s glaring lags in economic opportunities, human capital development, institutional effectiveness and environmental progress by 2063 (African Union [5]).

For instance, Goal 1.4 of “The Africa We Want” (hereafter: Agenda 2063) seeks to ensure that African countries create and share wealth by leveraging the abundant natural resources of the continent through industrialisation, trade and entrepreneurship. In line with this goal is the implementation of the African Continental Free Trade Area (AfCFTA), which is a total unification of Africa’s merchandise and non-merchandise market of about 1.5 billion people for freer trade, forward and backward linkages and improvement in the continent’s global value chain participation. Indeed, the United Nations Conference on Trade and Development (UNCTAD) [6] report projects that there would be a rebounding of Foreign Direct Investment (FDI) to Africa from 2022 following the implementation of the AfCFTA and the successful finalisation of its attendant investment protocol.

This brings to the fore the issue of energy consumption for growth and environmental sustainability. While on the one hand, energy can sustain economic activity (e.g., trade and FDI) to spur growth and equitable distribution of incomes as in Wang et al. [7], Igawa and Managi [8], Nchofoung and Asongu [9], Mazzucato and Semieniuk [10], on the other hand, high energy intensity, particularly with the non-renewable fuel characteristics of trade and FDI, can also be a drawback for environmental sustainability. This is underlined by the fact that projections by the International Energy Agency (IEA) report [11,12] and International Renewable Energy Agency (IRENA) report [13] show that energy consumption will rise in Africa in line with the growing urbanisation and incomes. However, although in general, the environmental concerns regarding renewable energy consumption are negligible, as shown by Bouyghrissi et al. [14], Shahnaz [15], and Acheampong [16], concerns about Africa’s high non-renewable energy consumption and poor energy systems have been raised (Mensah et al. [17]; Hanif and Raza [18]). For instance, information gleaned from Nathaniel and Iheonu [19] and Gu et al. [20] indicates that Africa’s share of global greenhouse gases and carbon dioxide (CO₂) emissions in particular has been rising since 2015, raising concerns about the achievement of SDGS 13, 14 and 15 in the medium term and the net zero (0) carbon emission target by 2050 in the longer term.

Indeed, a survey of the extant scholarship on the energy−environment relationship reveals that the former affects the latter directly or indirectly. The direct effect stems from the increase in energy intensity arising as result of economic activities, especially industrialisation, commerce and resource exploitation. For instance, Arouri et al. [21], Gao et al. [22], Esso and Keho [23] argue that the quest to spur development contributes to environmental footprint as high energy consumption becomes a key factor to run machines, sustain production and distribution of goods and services. However, in a continent where the non-renewable energy consumption (i.e., fossil fuel) is high and there is growing competition among African governments to attract FDI, environmental progress should not be sacrificed for economic growth. Additionally, though renewable energy has been identified as a major step towards the flattening of the Environmental Kuznets Curve (EKC), concerns have been raised that there are some industries (e.g., the petrochemical, cement, print media) whose activities run largely on non-renewable energy and could trigger massive environmental sustainability setbacks as African leaders seek to develop the industrial base of the continent (Wang et al. [7]; Adom et al. [24]; Agradi et al. [25]). The indirect effect also lends itself to the materialisation effect of rising incomes (Adom et al. [24]). This occurs when economies create wealth, and thus raise the standard of living of the masses, which in effect empowers them to acquire more energy-intensive materials (Knight et al. [26]; Magazzino [27]). Especially in Africa, where precarity is widespread, rising income levels can heighten involvement in activities that are energy-intensive.

This brings to the fore the role of energy efficiency (hereafter: EE) in mediating the energy-consumption-environmental quality nexus. We pay attention to EE considering the admonition by the United Nations that identifying synergies or trade-offs among the SDGs will be crucial for the attainment of multidimensional sustainability by 2030. One such possible key element that can engender relevant synergies with energy consumption is EE. EE means using less energy for production, as discussed by Adom et al. [28] and Liddle and Sadorsky [29]. In particular, the SDG 7, which highlights the most affordable, reliable and modern forms of energy for sustainable development, positions EE as one of the core modalities for realising environmental progress. Indeed, a survey of the literature shows that EE can foster environmental progress from two perspectives. First, EE fosters environmental sustainability through reducing energy intensity, the protection of human life and natural capital (Arouri et al. [21], aside from its much-emphasized role in the mitigation of climate change through reductions in carbon footprint and the achievement of net-zero emissions in the broader perspective by 2050. As reckoned by the International Energy Agency (IEA) [12], the Organisation for Economic Co−operation and Development (OECD) [30,31], and the Green Growth Knowledge Platform [32], EE can also support eco-friendly practices and innovations that have been found to be effective for enhancing environmental quality of life. Particularly in Africa, where resources for building resilience and/or mitigating climate change concerns are lacking (OECD [33,34]), EE presents a medium through which policymakers can flatten the EKC.

This is anchored in the argument by Razzaq et al. [35], Adom et al. [24] and Ohene-Asare et al. [36] that EE promotes environmental sustainability by spurring green growth as it decreases carbon intensity and greenhouse gas emissions. In other words, EE can reduce the level of energy intensity in the production process, which could go a long way to promote green growth and durable employment creation. EE is crucial for addressing income inequality through the reduction in energy and energy-related cost savings for low-income households, thereby freeing up resources for other productive activities in this regard. EE offers a key channel to reduce the cost of production, improve economic growth and improve living conditions, which requires environmental quality of life (Njiru and Letema [37]).

Despite these plausible synergistic relationships between energy consumption and EE for environmental progress, rigorous empirical content exploring this in the case of Africa is difficult to find, indicating a gap in the literature. Moreover, the question of whether EE interacts with both renewable and non-renewable energy to reduce greenhouse gas emissions remains unanswered in the literature. Further, considering the relatively low level of EE in Africa, as shown by Agradi et al. [25], Adom et al. [24] and Ohene-Asare et al. [36], prior contributions have not been explored in the short-term to long-term environmental sustainability gains from improving EE in Africa. This study seeks to contribute to filling these gaps. The first contribution the study adds to the literature is to examine the unconditional effects of energy consumption on environmental quality. Africa has vast renewable energy potentials; however, the continent tends to depend more on non-renewable energy (Wang et al. [7]; Adom et al. [24]; Agradi et al. [25]). Hence, it is worthwhile to disaggregate energy into renewable and non-renewable energy to see which component hampers environmental quality. The current study adds to the literature by examining the differential effect of renewable and non-renewable energy consumption on environmental quality in Africa. The paper also determines whether EE can contribute to environmental quality. The present study is novel in the sense that its focus is not only on the energy-consumption−environmental quality in Africa, but it also explores the moderating role of EE on this relationship. Extant literature uses a component of greenhouse gas emissions as a measure of environmental quality. However, the current study used the total greenhouse emissions and later employed two components, carbon and nitrous emissions, as robustness checks.

The rest of the paper is organised as follows: the next section is the literature review, which provides a theoretical link between FDI, energy consumption and inclusive growth, while Section 3 outlines our methodology. We present our results and discussion in Section 4, and that of our conclusion and policy recommendations in Section 5.

2. Literature Review

This section reviews literature on theoretical reviews on energy efficiency and environmental sustainability on one hand and an empirical review on energy consumption and environmental sustainability.

2.1. Theoretical Review on Energy Efficiency and Environmental Sustainability

The theoretical relationship between energy efficiency and the environment emanates from ecological modernization theory, which states that environmental concerns due to economic activities can be mitigated by improving energy efficiency measures through technological innovation and green practices that harmonise environmental and economic performance (Razzaq et al. [35]; Gouldson and Murphy [38]). Empirically, the literature, which includes Marques et al. [39], Akram et al. [40], Razzaq et al. [35], Ohene-Asare et al. [36], Adom et al. [24], Agradi et al. [25], Opoku and Boachie [41], Go et al. [42] and Dauda et al. [43] highlights the significant role energy efficiency plays in spurring sustainable development and environmental quality in both developing and developed countries. A few studies, such as Pan et al. [44], Mahmood and Kanwal [45], and Sinha [46], however, find a negative impact of EE on sustainable development.

In recent times, Dercon [47] argued that environmentally friendly growth strategies such as EE are more of a menace than a potential for advancement and development. Additionally, the link between environmental and the energy reducing effect of EE is opposed by Khazzoom [48], who claims that EE is not cost-effective in reducing pollution because improvements in EE, which lower the implicit cost of energy, also induce higher energy demand and thus higher pollution. This is known as the rebound effect (Herring and Sorrell [49]). Thus, the rebound effect of improvements in EE can nullify potential environmental progress impacts (Gillingham et al. [50]). However, these findings are often susceptible to the sample, the period and the method of estimation employed by different researchers. For instance, by using a fixed-effect panel quantile regression technique for a sample of 66 developing countries from 1990 to 2014, Akram [40] found that EE significantly reduced CO2 emissions. Notably, the study provided robust evidence that an improvement in EE had a greater emission-reducing effect compared to renewable energy and nuclear energy in all model specifications.

Likewise, Marques et al. [39] and Rajbhandar and Zhang [51] argue that beyond the socio-economic benefits, economies may also gain an environmental dividend through energy efficiency improvement. In another study, Özbuğday and Erbas [52] examined the dynamic effects of economic activity, energy efficiency, economic structure, and renewable energy consumption on CO₂ emissions in 36 countries for the period 1971–2009. Employing the common correlated effects estimator, the study found that EE reduced CO₂ emissions in the long term. Additionally, Tajudeen et al. [33] found a remarkable reduction in CO₂ emissions due to an improvement in energy efficiency in 30 OECD economies from 1971 to 2015. Furthermore, Ponce and Khan [34] revealed a negative relationship between EE and CO₂ emissions in the nine developed countries analysed from 1995 to 2019. This result corroborates those of Lei et al. [53], who employed the non-linear ARDL to estimate the effects of renewable energy consumption and energy efficiency on CO₂ emission in China for the period covering 1991 to 2019. Javid and Khan [54] investigated the effect of energy efficiency on environmental quality in the top five greenhouse gas-emitting countries (China, USA, India, Germany, and Japan). Using an annual dataset spanning from 1971 to 2016 and employing a structural time series model technique, the authors provide evidence that efficient energy-saving appliances are crucial measures for improving environmental progress in the face of economic growth. This study therefore extends the theoretical literature on energy consumption, environment and energy efficiency by considering the African perspective of energy consumption, economic growth and environmental progress.

2.2. Empirical Review on Energy Consumption and Environmental Sustainability

The neoclassical growth theory implicitly identifies energy as input in the economic growth process, as explained by Agradi [37], Howarth [55] and Saunders [56]. Since economic growth is driven by an accumulation of production factors, of which energy consumption is key, it can have profound implications for environmental quality (Lopez [57]). In other words, the relationship between energy consumption and environmental sustainability can be examined within the trilemma of energy consumption, economic growth and greenhouse gas emissions. This is consistent with the EKC hypothesis, which suggests that there is an inverted U−shaped relationship between environmental pollution and economic growth (see Grossman [58], Panayotou [59], Shafiei and Salim [60]).

Moreover, in the conventional growth framework (neoclassical and endogenous), because technological progress can lead to changes in energy consumption, there is the notion that a directed technological change in the form of changes in energy consumption or intensity can indirectly alter environmental quality. For example, Eriksson [61] asserts, the direction of technological change can be pursued with a focus on energy conservation or reduction in the volume of greenhouse gas emissions, for instance, through carbon capture or abatement. This is an indication of the scale effect, which states that in the early stages of development, increasing economic activity causes more environmental damage, because more resources, such as energy, are required to increase growth (Brock and Taylor [62]). Despite not pointing to a direct relationship between energy consumption and environmental quality, the EKC also offers a simple and easy-to-understand framework based on the relationship between economic growth and environmental quality (Swain et al. [63]).

A survey of the literature on the energy–environmental quality relationship shows that prior contributions have focused on two key areas, with the first paying attention to overall energy consumption, and that of the second on the effects of renewable and non-renewable energy consumption (Sarkodie and Adams [64]). Many authors, including Dercon [47], Khazzoom [48], Herring and Sorrell [49], Gillingham et al. [50], Rajbhandari and Zhang [51], Özbuğday and Erbas [52], Tajudeen et al. [33], Ponce and Khan [34], Lei et al. [53], Javid and Khan [54] and Dogan and Seker [63], among others, argue that renewable energy is beneficial to environmental quality; however, excessive use of non−renewable energy can degrade the environment. For instance, Dogan and Seker [65] examine the impacts of energy, real income and trade openness on CO₂ emissions in the European Union (EU) for the period 1980–2012. The estimated results of the dynamic ordinary least square method show that renewable energy consumption represses CO₂ emissions, while non-renewable energy proved otherwise. The study also confirmed the EKC hypothesis in the sampled countries. In a similar study, Alola et al. [66] employed the Panel Pool Mean Group Autoregressive Distributed Lag (PMG−ARDL) approach on a balanced panel of 16−EU countries from 1997 to 2014. The study also reveals the suppressing-effect of renewable energy consumption on environmental degradation, while non-renewable energy triggers environmental setbacks. The study, however, used an ecological footprint as a measure of environmental depletion contrary to the conventional use of CO₂ emission. Another study by Sarkodie et al. [67] paid attention to the emission of particulate matter (P.M2.5) in the 54 countries. Investigating over the period 2000–2016, the authors found evidence from a generalized least square random effects model estimation with first regressive [AR (1)] disturbance to show that an increase in energy consumption increased PM_2.5 emissions per capita. Analysing the energy−environment relations in the OECD countries from 1980 to 2011, Shafiei and Salim [60] used the STIRPAT model as the empirical basis. While the authors established that non-renewable energy consumption increased CO₂ emissions, renewable energy consumption dampened CO₂ emissions. This is a result that has been validated in a recent contribution by Zafar et al. [68] showing that investments in renewable energy by OECD countries have played a salient role in improving environmental sustainability through a reduction in CO₂ emissions. Sinha and Sengupta [69] also examined the impact of renewable energy and fossil fuel energy consumptions on nitrous emissions using Asia-Pacific Economic Cooperation (APEC) countries as a case study. The study, which covered the period 1990–2015, revealed that renewable energy measures had a positive impact on environmental quality (i.e., reducing nitrous emissions).

It has been advanced by Khan et al. [70] that the energy−environment literature examining the long-run relationship between energy consumption and ghgs for different income groups comprising lower middle income, upper middle income and heavily indebted countries and regions (i.e., East Europe, and Central Asia, Latin America and Caribbean, Middle East and North Africa, South Asia and sub−Saharan Africa). Using the Johnson cointegration, a modified version of Granger causality, and variance decomposition analysis for the period 1975–2011, the authors show that energy consumption Granger causes ghgs emission.

In Africa, Zoundi [71] employed Dynamic Ordinary Least Squares (DOLS) with a set of robustness tests to determine the short-run and long-run impacts of renewable energy on CO₂ emissions in 25 African countries over the period 1980–2012. Although the study found that renewable energy had a negative effect on CO₂ emissions, the authors pointed out that the impact was undermined by the greater effects of primary energy consumption on CO₂ emissions. Additionally, contrary to the study by Sarkodie and Ozturk [72] that found the existence of the EKC, Zoundi [71] found no evidence of the EKC hypothesis.

Furthermore, Sharif et al. [73] relied on data for 74 countries and the fully modified ordinary least squares (FMOLS) and panel causality techniques to analyse the effects of green energy and fossil fuel consumption on the environment from 1990 to 2015. The evidence from the study suggested that fossil fuel consumption had an increasing effect on environmental degradation, while green/renewable energy consumption had a significant negative effect on environmental degradation. The study also validated the EKC hypothesis in the sampled countries. In a similar study where a balanced panel dataset for 120 countries was used, Dong et al. [74] examined the role of renewable energy consumption on CO₂ emissions across the globe. The study, which spanned 1995–2015, revealed that renewable energy consumption had a negative effect on CO₂ emissions in sampled countries. The result remained consistent, even when the authors disaggregated the sample into four sub-panels (high income, upper middle income, lower middle income and lower income). Further, the estimated results showed that CO₂ emissions in the lower middle income and lower income subpanels were substantially lower than those in the high income and upper middle subpanels. Drawing on a sample of seven selected regions, Al-Mulali et al. [75] applied the DOLS and vector error correction model Granger causality to show that renewable energy consumption had a significant negative influence on pollution in Central and Eastern Europe, Western Europe, East Asia and Pacific, South Asia and the Americas. Conversely, the authors found no evidence of the significant impact on environmental progress in the case of the Middle East and North Africa and sub-Saharan Africa.

The potential impact of renewable energy consumption on environmental quality (CO₂ emission) within the EKC model was also examined by Bilgil et al. [76], employing a panel dataset of 17 OECD countries over the period 1977–2010. The results from the FMOLS and panel DOLS techniques showed that renewable energy consumption yielded a negative effect on CO₂ emissions. Consequently, the study encouraged policies to increase renewable energy consumption via improved renewable energy technologies. In a recent study Mujtaba et al. [77], autoregressive distributed lags (ARDL) and nonlinear ARDL techniques were used to estimate the symmetric and asymmetric effects of economic growth, capital formation and renewable and non−renewable energy consumption on CO₂ emissions and the ecological footprints of 17 OECD countries spanning from 1970 to 2016. The results for the ARDL estimation indicated that a percentage increase in renewable energy consumption corresponded to an expected 0.2% reduction in CO₂ emissions, while a 1% increase in non-renewable energy increased CO₂ emissions by 1.08%. In particular, the conventional energy from fossil fuel was observed to worsen environmental conditions, but renewable energy improved environmental quality for both CO₂ emissions and ecological footprint. A similar outcome was found by Balsalobre-Lorente et al. [78], who reported the environmental quality-improving impact of renewable electricity consumption in five EU countries (Germany, France, Italy, Spain and the UK) for the period from 1985 to 2016. In addition, contrary to Sarkodie et al. [67], some studies, such as Dogan and Seker [65], Danish et al. [79] and Sarkodie et al. [67], found an N-shaped EKC.

Despite the aforementioned findings, some authors found that renewable energy consumption increased carbon emissions. For instance, by applying the FMOLS and DOLS on data from 1980 to 2009, Farhani and Shahbaz [80] found that both renewable and non-renewable energy accelerated the level of emissions in Middle Eastern and North African (MENA) countries. Similar findings were reported by Mert and Bölük [81] in the case of the EU. Evidence from the study showed that both renewable and non−renewable energy consumption contributed to environmental degradation in the EU, although renewable energy contributed around 50% less per unit of energy in terms of greenhouse gas emissions in 16 EU countries.

In sum, the extent of literature on the energy-consumption−environment nexus yields a myriad of mixed results. The review of the literature so far shows that the relationship between energy consumption, energy efficiency and environmental quality has not received attention, leaving a gap in the literature. The current study seeks to bridge the gap by first examining the relationship that exists between energy consumption and environmental quality. Accordingly, the current study also makes advances in energy-consumption−environmental quality scholarship by investigating the potential role that energy efficiency can play in the energy-consumption−environment quality nexus. To understand which form of energy consumption influences environmental quality in Africa, we disaggregated energy consumption into renewable and non-renewable energy consumption. Unlike other studies, such as Akram et al. [40], Tajudeen et al. [5] and Shafiei and Salim [60], who focused on only a portion of greenhouse gas emissions, we used the whole range of greenhouses gas emissions, and later used two components (carbon and nitrous emission) of it as robustness checks.

3. Materials and Methods

3.1. Data

This study was entirely macro-data-based, and spanned 2000–2020 for 23 African countries due to the availability of data. The main outcome variable in this study was greenhouse gas emissions (ghgs). That is, unlike prior contributions, such as Khan et al. [70], Akram et al. [40] and Kisswani and Zaitoun [82], which rely on CO₂ emissions as the main environmental quality indicator, this study considered other ghgs, including other pollutants, such as methane, biomass, nitrous oxide and fluorinated greenhouse gases, such as hydrofluorocarbons, perfluorocarbons, sulphur hexafluoride and nitrogen trifluoride. The main independent variable of interest in this study was energy consumption, which is measured as the total primary energy consumption of a country.

The essence of primary energy demand for environmental progress is clearly articulated in the Section 1 and Section 2. It is imperative to point out that we disaggregated energy consumption into renewable and non-renewable energy consumption to contribute to the policy discourse on the extent to which non-renewable energy (e.g., crude oil, coal, etc) and sustainable energy sources, such as solar and hydropower, impact environmental progress.

For our moderator, we considered energy efficiency, which according to the remit of SDG 7 and empirical evidence in Adom et al. [24] and Filippini and Zhang [83], is a key determinant of the extent to which energy consumption impacts the environment. Unlike other variables, which are readily available in databases, that of energy efficiency is computed based on sound econometric techniques. In this study, we follow the approach of Kumbhakar et al. [84] in computing our energy efficiency indices for the sampled countries.

Moreover, we controlled for variables such as development assistance, economic growth and FDI to capture the essence of donor support, economic expansion and capital flows on environmental progress. To begin with, we paid attention to economic growth, as per the claim espoused in the environmental Kuznets hypothesis that in study areas such as Africa, where economic development is in the early stages, sustained income growth is achieved at the expense of environmental progress (Khan et al. [70]; Muhammad et al. [85]; Alvarado et al. [86], Grossman and Krueger [58]). In other words, in the early stages of economic development, growth is achieved with little attention to environmental sustainability. However, in the latter stages of economic development, where incomes are high, governments, firms and households spend more on addressing environmental pollution Lorente and Álvarez-Herranz [87]; Song et al. [88].

Additionally, we addressed the implication of FDI, considering the theoretical arguments conveyed in the pollution halo and pollution haven hypotheses. While according to the former hypothesis, FDI can trigger green innovations to spur environmental progress, as revealed by Muhammad et al. [85], and Kisswani and Zaitouni [82], the latter suggests that FDI can heighten energy intensity and environmental degradations in developing countries, as foreign firms avoid stringent environmental standards in developed countries by operating in developing countries where environmental laws are laxer (Sabir et al. [89]; Shahbaz et al. [90]).

Finally, we considered development assistance, taking into account the essence of donor support in the form of monetary aid and environmentally progressive technologies aimed at assisting developing countries in (1) mitigating environmental degradation and (2) building resilience to withstand environmental shocks (Mahalik et al. [91]; Opršal and Harmáček [92]; Lim et al. [93]). Indeed, some previous work, such as Boamah et al. [94] points out that African countries receive high inflows of development assistance. A description of the variables used for the empirical analysis is presented in

Table 1. Description of variables and data sources.

Variables	Symbol	Descriptions	Sources
Outcome variables
Greenhouse gas emissions	ghgs	Total greenhouse gas emissions in kiloton of CO2 equivalent. This comprises carbon emissions and other pollutants, such as methane, biomass, nitrous oxide and fluorinated greenhouse gases, such as hydrofluorocarbons, perfluorocarbons, sulphur hexafluoride and nitrogen trifluoride	WDI
CO2 emissions	CO2	Carbon emissions per capita produced during consumption of solid, liquid and gas fuels and gas flaring.	WDI
Nitrous emissions	N2O	Nitrous oxide emissions in thousands of CO2 emissions from agricultural biomass burning, industrial activities and livestock management	WDI
Variables of interest
Energy consumption	ener	Denotes primary energy consumption from combustible renewables and waste—solid biomass and animal products, gas and liquid from biomass and industrial and municipal waste	WDI
Renewable energy	rener	Renewable energy consumption (% of total final energy consumption)	WDI
Non-renewable energy	nonre	Fossil fuel consumption comprising coal, oil, petroleum and natural gas products.	WDI
Moderating variable
Energy efficiency	EE	Energy efficiency index calculated following the Kumbhakar et al. (2014) approach	Authors
Control variables
FDI	Fdi	Net inflow of foreign direct investment as a percentage of gross domestic product	WDI
Foreign aid	oda	The inflow of official development assistance as a share of gross national income.	WDI
Economic growth	gpc	GDP per capita growth	WDI
Economic growth squared	${\mathrm{gpc}}^2$	GDP per capita growth squared	Authors

Note: WDI—World Development Indicators.

.

3.2. Computation of Energy Efficiency Scores

Although several techniques have been put forward for the computation of energy efficiency, it is clear from the existing scholarship that two techniques—the stochastic frontier approach (SFA) and the data envelopment analysis (DEA) have been widely used. In this study, however, we relied on the former, based on the objectives of this contribution and some econometric concerns associated with the latter. To begin with, a major advantage of the parametric SFA vis-à-vis the non−parametric DEA is that it is designed to split EE scores into persistent and transient EE scores (Kumbhakar et al. [84]). This decomposition is particularly relevant in this study, as it enables us to inform policy as to whether short-term or long-term policies are worthwhile.

Additionally, unlike the DEA, the SFA is designed to handle omitted variables better (Hu et al. [95]). Moreover, as Agradi et al. [25] and Mutz et al. [96] point out, the SFA accommodates data outliers, measurement errors and data uncertainty better than the DEA, making it more robust and reliable.

We took cues from Filippini and Zhang [83] by specifying a conditional energy demand function, as apparent in Equation (1), where the energy used is internally determined by factors such as economic growth, industrialisation and trade openness. That said, we followed Adom et al. [24] by adopting a Cobb−Douglas energy demand function, as specified in Equation (2) to capture the elasticity of energy consumption to changes in our energy demand covariates. To allow linear computation, we proceeded by linearizing Equation (2) through logarithm transformation, as seen in Equation (3).

$$e_{it}^{ED}=f\left(p_{it},q_{it},W_{it},\omega \right)e^{v_{it}-u_{it}}$$

(1)

where: ${\omega }_p<0$ and ${\omega }_q>0$

$$e_{it}^{ED}=f\left(p_{it},q_{it},W_{it},\omega \right)=A{p}_{it}^{{\omega }_p}{q}_{it}^{{\omega }_q}{W}_{it}^{{\omega }_w}$$

(2)

$$ln{e}_{it}^{ED}={\delta }_0+{\omega }_{p}ln{p}_{it}+{\omega }_{q}ln{q}_{it}+{\omega }_{w}ln{W}_{it}+{\epsilon }_{it}$$

(3)

where $e_{it}^{ED}$ is the energy demand, which is driven primarily by income ($q_{it}$) in country $i$ at time $t$, price of energy ($p_{it})$ and some energy demand controls denoted by $W_{it}$.

By implication, f ($p_{it}$, $q_{it}$,$W_{it}$, $\omega$)$e^{v_{it}-u_{it}}$ is our benchmark energy demand function, $\omega$ represents the energy−demand elasticities and $e$ is the Euler’s mathematical constant.

Additionally, A is a constant and ${\epsilon }_{it}$ is the error term, which on the one hand, measures the level of energy inefficiency ($u_{it}$) (assumed by half−normally distributed) and on the other hand, an indicator for the idiosyncratic noise term symbolised by $v_{it}$ with a normal distribution. With all that said, we followed the logic of Greene [97], computing our technical EE scores denoted by $e{s}_{it}$ by augmenting Equation (3) to allow for our energy inefficiency term ($-{\mathrm{u}}_{\mathrm{it}}$), as seen in Equation (4)

$$ln{e}_{it}^{ED}={\delta }_0+{\omega }_{p}ln{p}_{it}+{\omega }_{q}ln{q}_{it}+{\omega }_{w}ln{W}_{it}+{\epsilon }_{it}-u_{it}$$

(4)

It is imperative to point out that $q_{it}$ and p_it are defined as the level of national income (US$) and crude oil price (US$), respectively, in each country $i$ at time $t$ and p_it. Based on the empirical evidence on the drivers of energy demand (see, Adom et al. [24]; Adom et al. [28]; Zhang and Adom [98], the study paid attention to the following energy demand controls: (i) industrialisation (lnind), (ii) trade openness (lntrade), (iii) urbanisation (lub) and (iv) human capital (hc). Next, we obtained the energy efficiency scores by taking the exponential of ($-u_{it}$):

$$e{s}_{it}=\exp \left(-u_{it}\right)$$

(5)

where $0\le e{s}_{it}\le 1$.

Finally, to decompose the inefficiency term $-u_{it}$ into permanent/time invariant (${\alpha }_i$) and transient/time-variant (${\mathsf{\pi }}_{\mathrm{it}})$ while accounting for unobserved country-specific heterogeneities [$u_i$] in Equations (6) and (7), we followed Kumbhakar et al. [84] by specifying Equation (7). This computation involved 4 key steps, which we took from Kumbhakar et al. [84]. First, the full energy demand function is estimated by applying the random-effect or fixed-effect estimator, the better of which is decided based on the Hausman test (see the attendant results in

Table A1. Hausman test on Equation (6).

	Coefficients
Variables	Fixed Effect (b)	Random Effect (B)	Difference	Standard Error
Trade openness	−0.0209	−0.0343	0.0133	0.0000
Urbanisation	−0.312 ***	−0.439 ***	0.1273	0.0646
Economic growth	0.0867	0.0641	0.0226	0.0371
Crude oil price	0.0042	−0.0273 **	0.0315	0.0126
Industrialisation	0.0494	0.0713 **	−0.0218	0.0065
Human capital	0.680 ***	0.654 ***	0.0256	0.0476
t	−2.312 ***	−0.0051 **	−2.3064	0.8885
t2	0.0005 ***	0.0001	0.0004	0.0001

Note: t is time; t² is time squared. b = consistent under Ho and Ha; obtained from xtreg. B = inconsistent under Ha, efficient under Ho; obtained from xtreg. Test: Ho: difference in coefficients not systematic. Chi Statistic: 9.02; Chi(p−value): 0.2512. where: *** p < 0.01, ** p < 0.05,

), as shown in the Appendices section. The second and third steps involve the estimation of the transient ($\exp \left(-{\pi }_{it}\right)$) and persistent ($\exp \left(-{\alpha }_i\right)$) components of EE, using the stochastic frontier residuals. The final step has to do with the computation of the overall EE scores $(\exp \left({\pi }_{it}-u_i\right)$, which is obtained by taking a product of the transient and persistent components of EE.

$$u_{it}=u_i+{\alpha }_i+{\pi }_{it}$$

(6)

$$ln{e}_{it}^{ED}={\delta }_0+{\varnothing }_{p}ln{p}_{it}+{\varnothing }_{h}ln{h}_{it}+{\varnothing }_{k}ln{K}_{it}+{\epsilon }_{it}-u_i-{\alpha }_i-{\pi }_{it}$$

(7)

In the estimation of EE scores via the stochastic frontier approach, a decision was required as whether to adopt a cost-type or production-type function. Accordingly, Schmidt and Sickles [99] suggest a skewness test to inform a choice. On the basis of this suggestion, studies of this kind should adopt a production-type stochastic frontier if there are negatively skewed least squares residuals, whereas a cost-type stochastic frontier should be adopted for positively skewed residuals (Adom et al. [24]). The attendant result, which we disclose in

Table 2. Skewness test of the energy demand function (Equation (7)).

Skewness	Kurtosis	Pr(Skewness)	Pr(Kurtosis)	Joint Chi−Square Test
−0.677	2.449	0.000	0.001	33.2 ***

where; *** p < 0.01.

, indicates that the former is preferred.

Drawing on the ecological modernization theory and the pollution halo hypothesis, we modelled environmental sustainability as primarily driven by energy consumption, energy efficiency and a set of control variables (FDI, foreign aid and economic growth). Accordingly, we followed the functional form approach of Akram et al. [40], and Opoku and Boachie [41] by specifying environmental quality model as:

$$envs=f\left(ener,EE, fdi,oda, gpc,gp{c}^2, Z\right)$$

(8)

where $envs$ is an environmental quality variable, $ener$ is primary energy consumption, which is further disaggregated into renewable ($renr$) and non-renewable energy ($nonre$), $fdi$ is foreign direct investment, introduced to capture the existence or otherwise of the PHH or PH in Africa, gpc is per capita income and $gp{c}^2$ is per capita income squared capturing the EKC. We proceeded, therefore, by transforming Equation (8) in a standard panel model, where the lag of environmental sustainability was introduced in the model to take into account the implication of the previous year’s environmental progress efforts on current environmental quality.

$$ghg{s}_{it}={\lambda }_0+{\delta }_{1}ghg{s}_{it-1}+{\beta }_{1}ene{r}_{it}+{\beta }_{2}E{E}_{it}+{\beta }_{3}fd{i}_{it}+{\beta }_{4}od{a}_{it}+{\beta }_{5}gp{c}_{it}+{\beta }_{6}gp{c}_{it}^2+{\mathcal{I}}_i+{\mu }_t+{\epsilon }_{it}$$

(9)

We shifted attention to the specification of a full panel model in the remit of Hypothesis 2, where we introduced an interaction term for energy consumption and energy efficiency. To this end, we modified Equation (9) to obtain Equation (10):

$$ghg{s}_{it}={\lambda }_0+{\delta }_{1}ghg{s}_{it-1}+{\beta }_{1}ene{r}_{it}+{\beta }_{2}E{E}_{it}+{\beta }_{3}fd{i}_{it}+{\beta }_{4}od{a}_{it}+{\beta }_{5}gp{c}_{it}+{\beta }_{6}gp{c}_{it}^2+{\beta }_7\left(ene{r}_{it}\times E{E}_{ir}\right)+{\mathcal{I}}_i+{\mu }_t+{\epsilon }_{it}$$

(10)

Though on the advice of Stock and Watson [100], the pooled least square estimator can be employed to estimate Equation (10), we opt for the instrumental variable approach of Blundell and Bond [101].

The choice of approach is explained as follows. First, the number of countries considered in this study (i.e., 23 countries) exceeds the period of time under consideration (i.e., 21 years) (Asongu and Odhiambo [102]; Ofori and Grechyna [103]). Second, there are concerns of endogeneity in Equation (10), which if unresolved, can bias the attendant estimates and render confidence intervals invalid. The inclusion of the lag of the dependent variable (i.e., ghgs emission) in Equation (10) causes endogeneity as $ghg{s}_{it-1}$ depends on ${ϵ}_{it-1}$, which also depends on the country-specific impact ${\mu }_i$ (Arellano and Bover [104]). This kind of endogeneity arises in the first difference estimation as the system generalised method of moment (GMM) estimator sweeps away the country-specific effects leading to a correlation between the error terms and the lag of ${ϵ}_{it-1}$ and $ghg{s}_{it-1}$ (Kripfganz [105]; Roodman [106]).

Further, in the remit of the Green Solow model, another endogeneity concern arising as a result of a bi-causal relationship between environmental progress and economic growth is apparent.

Additionally, Arellano and Bond [107] argue that by instrumenting the endogenous variables—the lag of ghgs and economic growth with their past values—the endogeneity problems can be addressed. Nonetheless, doing so plunges the estimation procedure into the realm of the first-difference GMM estimator, which is also not without limitation. The first-difference GMM estimator does not account for the possible information contained in the level relationship and in the relationships between the level and the first differences, rendering the estimator inconsistent and unreliable in the presence of strong endogeneity, such as the one identified in this study (Ahn and Schmidt [108]).

In this regard, Blundell and Bond [102] suggested that the system GMM estimator, which estimates the level and first-difference regressions as a system, be preferred to the first-difference estimator. This is re-echoed by Bond et al. [109] and Windmeijer [110], who argued that the system-GMM estimator yielded asymptotically consistent and reliable estimates (i.e., lower bias and standard errors) compared to the first-difference GMM. Accordingly, this study instruments the level equation with the lagged first-differenced covariates and that of the first-differenced estimation with the lagged level variables. Moreover, we followed Roodman [106] and used a two-step system GMM by collapsing our instruments to take care of possible instrument proliferation and overfitting.

We proceeded therefore by modifying Equation (10) in consonance with the specifications encapsulating the level and first difference (two step) system GMM, as seen in Equations (11) and (12), respectively.

$$ghg{s}_{it}={\alpha }_0+{\delta }_{1}ghg{s}_{it-1}+{\beta }_{1}ene{r}_{it}+{\beta }_{1}E{E}_{it}+{\beta }_3\left(ene{r}_{it}\times E{E}_{it}\right)+{{\displaystyle \sum }}_1^4{\varphi }_k(V_{kit-\tau }+V_{kit-2\tau })+{\mathcal{I}}_i+{\mu }_t+{\epsilon }_{it}$$

(11)

$$\begin{gathered} ghg{s}_{it}-ghg{s}_{it-\tau }={\delta }_1\left(gh{g}_{it}-gh{g}_{it-2\tau }\right)+{\beta }_1\left(ene{r}_{it}-ene{r}_{it-\tau }\right)+{\beta }_2\left(E{E}_{it}-E{E}_{it-\tau }\right)++{\beta }_3(ene{r}_{it}\times E{E}_{it}- \\ ene{r}_{it}\times E{E}_{it-\tau })+{{\displaystyle \sum }}_1^4{\theta }_k(V_{kit-\tau }+V_{kit-2\tau })+\left({\mu }_t-{\mu }_{it-\tau }\right)+({\epsilon }_{it}-{\epsilon }_{it-\tau }) \end{gathered}$$

(12)

We conclude the model specifications by presenting an expression for computing the net effect from the interactions between energy consumption ($ener$) and energy efficiency (EE) as:

$$\frac{\partial \left(ghg{s}_{it}-ghg{s}_{it-\tau }\right) }{\partial \left(ene{r}_{it}\right)}={\beta }_1+{\beta }_3\overline{\left(E{E}_{it}\right)}$$

(13)

where $\overline{E{E}_{it}}$ is the average energy efficiency score in country $i$ at time $t$. Additionally, $ene{r}_{it}$ is energy consumption, EE is energy efficiency and ($ene{r}_{it}\times E{E}_{it})$is the interaction term for energy consumption and energy efficiency, ${\mathcal{I}}_i$represents the country−specific effects, ${\epsilon }_{it}$ is the idiosyncratic error term and $V_k$ is a vector of the control variables (GDP per capita, GDP per capita square, FDI, and foreign aid).

It is imperative to point out that the effectiveness of the system GMM estimator in reliable estimates, as Ofori et al. [111] state, depends on several post-estimation tests, which we paid attention to. Therefore, we evaluated the validity of the instrument using the Hansen test of over-identification. The Hansen test of instrument validity is premised on the null hypothesis that the set of identified instruments and the residuals are uncorrelated. Therefore, the suitability of the instruments and hence the robustness of our estimates depends on the failure to reject the null hypothesis. On the other hand, if the null hypothesis is rejected, then the instruments are not robust because the restrictions imposed by relying on the instruments are invalid. Additional caveats for evaluating the robustness of the estimates are the post-estimation tests of (i) whether there is evidence of second-order serial correlation in the residuals or not; (ii) the significance of the interaction terms; and (iii) the Wald test for the overall model significance. Finally, we checked the robustness of our estimates by proxying environmental quality by (1) CO₂ emissions and (2) nitrous emissions, as specified as:

$$\begin{array}{c}C{O}_{2it}-C{O}_{2it-\tau }={\delta }_1\left(C{O}_{2it}-C{O}_{2it-2\tau }\right)+{\beta }_1\left(ene{r}_{it}-ene{r}_{it-\tau }\right)+{\beta }_2\left(E{E}_{it}-E{E}_{it-\tau }\right)++{\beta }_3(ene{r}_{it}\times E{E}_{it}-\\ ene{r}_{it}\times E{E}_{it-\tau })+{\displaystyle \sum }_1^4{\theta }_k(V_{kit-\tau }+V_{kit-2\tau })+\left({\mu }_t-{\mu }_{it-\tau }\right)+({\epsilon }_{it}-{\epsilon }_{it-\tau })\end{array}$$

(14)

$$\begin{array}{c}N2O-N2O{t}_{it-\tau }={\delta }_1\left(N2O_{it}-N2O{t}_{it-2\tau }\right)+{\beta }_1\left(ene{r}_{it}-ene{r}_{it-\tau }\right)+{\beta }_2\left(E{E}_{it}-E{E}_{it-\tau }\right)++{\beta }_3(ene{r}_{it}\times E{E}_{it}-\\ ene{r}_{it}\times E{E}_{it-\tau })+{{\displaystyle \sum }}_1^4{\theta }_k(V_{kit-\tau }+V_{kit-2\tau })+\left({\mu }_t-{\mu }_{it-\tau }\right)+({\epsilon }_{it}-{\epsilon }_{it-\tau })\end{array}$$

(15)

4. Results and Discussion

The study results are presented on

Table 3. Descriptive Statistics.

Variables	N	Mean	Std. Dev.	Minimum	Maximum
Outcome variables
Greenhouse gas emissions	281	118.853	184.514	−85.278	828.871
CO2 emissions	437	34,766	83,058.39	660	447,980
Nitrous emissions	437	12,755.652	14,798.338	220	62,990
Key independent variables
Energy consumption (primary)	462	20.836	34.425	0.710	157.511
Renewable energy consumption	437	56.944	30.394	0.059	98.343
Non-renewable energy consumption	345	40.944	30.138	1.640	99.978
Moderating variable
Energy efficiency	483	0.550	0.213	0.124	0.984
Control variables
Foreign aid	460	4.666	5.750	−0.251	62.187
Foreign direct investment	460	3.725	5.487	−6.370	39.760
Economic Growth	483	1.725	3.622	−14.87	12.457
Economic Growth squared	483	16.065	25.048	0.000	221.111

Note: Std. Dev. is standard deviation.

with the descriptive statistics of the variables employed for the analysis. The correlations between these variables are also reported in

Table A2. Pairwise correlations.

Variables	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)
(1) Greenhouse gas emission	1
(2) Nitrous emissions	−0.135 *	1
(3) CO2 emission per capita	−0.0760	0.121 *	1
(4) FDI	−0.117 *	0.793 ***	0.337 ***	1
(5) Foreign aid	−0.00664	−0.428 ***	0.349 ***	−0.0633	1
6) GDP per capita	−0.00961	0.448 ***	−0.296 ***	0.102	−0.989 ***	1
(7) GDP per capita square	−0.0315	−0.137 *	0.529 ***	0.311 ***	0.755 ***	−0.750 ***	1
(8) Primary energy consumptions	0.274 ***	−0.276 ***	0.0791	−0.173 **	0.483 ***	−0.504 ***	0.389 ***	1
(9) Renewable energy	0.260 ***	−0.131 *	−0.126 *	−0.141 *	0.0733	−0.0871	0.0543	0.168 **	1
(10) Non-renewable energy consumption	0.203 ***	0.0989	0.184 **	0.191 **	−0.0460	0.0459	0.146 *	0.141 *	0.0912	1
(11) Energy efficiency	0.0754	−0.0383	0.0785	0.0643	0.109	−0.116	0.210 ***	0.0578	0.0887	0.480 ***	1

Where: * p < 0.05, ** p < 0.01, *** p < 0.001.

in the Appendices section. The average ghgs equivalence is 118.85 Kilotons, while that of nitrous emissions and carbon emissions averaged 34,766 and 12,755.6 Kilotons, respectively.

The data show that the average primary energy consumption over the study period is 20.8. At the disaggregated level, we observed an average renewable energy consumption of 56.9 compared to 40.9 for that of non-renewable energy consumption. This is an indication of the high renewable energy potentials across the continent. For our moderator, energy efficiency, we found a mean value of 0.550.

Scrutinizing the data further, we present Figure 1 to show some development regarding renewable and non-renewable energy consumption in Africa. Figure 1 shows that while Mauritius (0.24) is the country with the least energy consumption (primary), South Africa emerged as the country with the highest energy consumption (4.84). For non-renewable energy consumption, it is clear in Figure 1 that Senegal (51.2), Mauritius (79.2), Botswana (66.1) and South Africa (86.7) are African countries with high energy intensity. With regard to renewable energy consumption, the majority of African countries, especially Congo Democratic Republic (96.7), Ethiopia (93.2), Gabon (82.9), Cameroon (81.3), Kenya (77.7), Mozambique (85.4), Niger (82.6), Nigeria (84.2), and Tanzania (85.2) record high renewable energy consumption in their respective energy portfolios. In Figure 2, there is an indication that the high energy consumption is increasing the emission of greenhouse gases of most African countries.

While it is widely documented that energy is an indispensable catalyst for sustainable development, as evidenced in the improving livelihoods and well-being in developing economies, rising levels of greenhouse gas emissions through anthropogenic activities can pose a significant threat to development (Alola et al. [66]). In a setting such as Africa, where the population keeps rising and urbanization is projected to triple by 2050 (Cartwright [112]), and the adoption of green technologies and clean fuels is in its early stages, as apparent in Figure 2, it is imperative to examine whether energy efficiency can play any significant role in the energy-environment nexus.

4.1. Energy Efficiency: Estimation of Persistent, Transient and Total Scores

In this section, we address the results on the determinants of energy demand, the EE scores and the correspondingly persistent and transient efficiency scores. As reported in

Table 4. Determinants of energy demand frontier.

Variable	Coefficient	Standard Error	t−Value
Trade openness	−0.0343	0.0244	−1.40
Urbanization	−0.439	0.0973	−4.52
Economic growth	0.0641	0.0411	1.56
Crude oil price	−0.0273	0.0112	−2.45
Industrialisation	0.0713	0.0328	2.17
Human capital	0.654	0.159	4.11
T	−0.0051	0.0020	−2.51
t2	0.0001	0.0001	0.16
Constant	13.116	3.287	−0.15
Observations	451	–	–
Countries	23	–	–
F−stats [p−value]	205.9 [0.000]	–	–

Note: (Dependent variable: energy consumptions (all sectors), OECD data.

, the study found that while variables such as human capital, trade openness, urbanization and industrialisation were significant determinants of energy demand in Africa, the crude oil price showed otherwise. Clearly, our findings align with prior contributions, such as Agradi et al. [25], Adom et al. [24] and Filippini and Zhang [83].

That said, we turn our attention to the presentation of the results on the EE scores (overall) and that of the persistent and transient components (see

Table 5. Energy efficiency estimates.

Variable	Obs	Mean	Std. Dev.	Min	Max
$\mathrm{Transient}\mathrm{EE}(-{\tau }_{it}$)	483	0.550	0.213	0.124	0.984
$\mathrm{Persistent}\mathrm{EE}(-{\mathcal{U}}_i$)	483	0.570	0.215	0.125	1.000
$\mathrm{Overall}\mathrm{EE}(-{\tau }_{it}-{\mathcal{U}}_i$)	483	0.963	0.040	0.797	0.992

Note: All EE estimates are generated following the SFA by Kumbhakar et al. (2014).

). The results as disclosed in Table 5 show that the mean persistent EE is 0.570 relative to 0.550 for that of transient EE, culminating in an overall EE score of 0.963. Conspicuously, our result is higher than that of [24] but close to that of [36], which was obtained using the DEA approach.

We contend that the sharp difference in our EE scores relative to those of Adom et al. [24] is plausibly due to our study period, which coincided with efforts by African countries to improve efficiency. More crucially, the results suggest that EE in Africa is more transient than persistent (i.e., high persistent inefficiency). This also indicates that policymakers should channel resources towards the streamlining structural inefficiencies.

4.2. System GMM Results: Effect of Energy Consumption and Energy Efficiency on Environmental Sustainability in Africa

Table 6. GMM results for the effects of energy consumption and energy efficiency on environmental sustainability in Africa (Dependent variable: Greenhouse gas emissions).

Variables	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
Greenhouse gas emissions (−1)	0.9456 ***	0.8255 ***	0.9521 ***	0.9395 ***	0.9185 ***	0.6499 ***	0.5799 ***	0.7156 ***
	(0.0066)	(0.0666)	(0.0144)	(0.0196)	(0.0287)	(0.0640)	(0.0820)	(0.0777)
FDI	−0.0015 ***	−0.0015	−0.0023 ***	−0.0006	−0.0038 ***	−0.0058 **	−0.0103 ***	−0.0053
	(0.0005)	(0.0019)	(0.0006)	(0.0019)	(0.0011)	(0.0026)	(0.0020)	(0.0031)
Foreign aid	0.0009	0.0061	−0.0057 **	−0.0052 *	−0.0058 *	0.0003	0.0053	0.0069
	(0.0007)	(0.0041)	(0.0026)	(0.0027)	(0.0030)	(0.0028)	(0.0038)	(0.0041)
Economic growth	0.0078 ***	0.0055 *	0.0090 ***	0.0103 **	0.0073 ***	0.0018	0.0059 **	−0.0024
	(0.0012)	(0.0028)	(0.0020)	(0.0043)	(0.0014)	(0.0029)	(0.0026)	(0.0047)
Economic growth squared	−0.0002	0.0009	0.0000	−0.0009	0.0003	0.0006	0.0004	0.0014
	(0.0004)	(0.0007)	(0.0006)	(0.0007)	(0.0006)	(0.0007)	(0.0007)	(0.0011)
Primary energy consumption		0.0041				0.0043
		(0.0054)				(0.0123)
Renewable energy consumption			−0.0025 **				−0.0138 **
			(0.0011)				(0.0066)
Non-renewable energy consumption				0.0022 *				0.0109
				(0.0012)				(0.0088)
Energy efficiency (EE)					−0.2647	−0.7716	−5.3022 ***	−2.0507 ***
					(0.2848)	(0.6327)	(0.8499)	(0.6854)
Primary energy consumption * EE						−0.0007
						(0.0220)
Renewable energy consumption EE							−0.0211 **
							(0.0084)
Non-renewable energy consumption EE								0.0136
								(0.0214)
Constant	0.5845 ***	1.6567 **	0.4045 **	0.5449 **	0.7830 ***	3.1134 ***	3.0915 ***	1.1815 *
	(0.0684)	(0.6156)	(0.1921)	(0.2477)	(0.2114)	(0.6109)	(0.7845)	(0.6658)
Observations	414	414	414	322	414	414	414	322
Countries	23	23	23	23	23	23	23	23
Instruments	22	22	22	18	22	22	22	18
Wald chi statistics	6.506 × 106 ***	157281 ***	4.511 × 106 ***	1.140 × 106 ***	2.517 × 106 ***	49610 ***	54166 ***	28,130 ***
Wald p-value	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000
Net effect	Na	na	na	Na	na	–	−0.0984	–
Joint significance test	Na	na	na	Na	na	–	13.16 **	–
p-value	Na	na	na	Na	na	–	0.0018	–
Hansen p−value	0.195	0.573	0.181	0.603	0.564	0.399	0.826	0.568
AR(1)	0.049	0.035	0.026	0.013	0.034	0.058	0.061	0.073
AR(2)	0.325	0.372	0.321	0.231	0.335	0.339	0.373	0.626

Standard errors in parentheses where; *** p < 0.01, ** p < 0.05, * p < 0.1.

shows the results of our main regression—the effect of energy consumption and energy efficiency on greenhouse gas emissions. The results, as apparent in Column 2, show that although statistically significant, primary energy consumption is positively related with ghgs. In other words, energy consumption is positively related to environmental progress. This result corroborates with Khan et al. [70], and Sarkodie et al. [67]. According to these studies, the high rate of emissions was due to an increase in the share of non-renewable energy consumption for economic activities. This is plausible as per the information in Figure 1, which indicates that most countries in Africa, for example South Africa, depend heavily on non-renewable energy consumption (especially coal) to drive their economic activities.

Further, the study found a negative relationship between renewable energy consumption and ghgs emissions, thus providing support for the statement of SDG 7 that renewable energy consumption could promote environmental sustainability (see Column 3). Specifically, we found that an increase in renewable energy consumption by 1 per cent would help to decrease ghgs emissions by 0.003 per cent. The result obtained in this study is in line with the findings of Sharif et al. [73], Wang and Dong [113] and Danish and Wang [114], and with the argument that increased consumption of renewable energy leads to a reduction in environmental deterioration. In a setting endowed with numerous low carbon energy sources, such as hydro, solar, biofuel, and wind systems, as Wang and Dong [113] and Suberu et al. [115] point out, the results suggest that Africa can achieve remarkable environmental sustainability by increasing the share of renewable energy consumption in its overall energy mix.

By contrast, the results show that non-renewable energy has a positive effect on ghgs emissions. In terms of magnitude, we found that for every 1 per cent increase in non-renewable energy (Column 4), there was an increase in ghgs emissions of 0.002 per cent. This result is consistent with Bekun et al. [116], who investigated the role of fossil fuel consumption on environmental degradation, pointing out that a high proportion of fossil fuels in the energy mix increased emissions and accelerated climate change challenges.

With respect to the effect of EE on ghgs, we found a negative but statistically insignificant effect (Column 5). Although statistically insignificant, the relationship suggests that EE measures have the potential to foster environmental progress. In this regard, our evidence lends itself to suggestions by Sarkodie [1] that EE is a potential channel for reducing ghgs emissions and improvements in quality of life.

In the light of our second hypothesis, we found that EE interacted with renewable energy consumption to yield remarkable environmental sustainability effects (see Column 7). We report a net effect of −0.0984, which is computed as:

$$\frac{\partial (ghg{s}_{it}) }{\partial \left(rene{r}_{it}\right)}=-0.0138+\left(-0.0211*4.009\right)=-0.0984,$$

where −0.0138 is the direct effect of renewable energy consumption on ghgs emissions, the coefficient of the interaction between renewable energy consumption and energy efficiency is −0.021 and −0.581 is the mean value of energy efficiency. This result implies that although unconditional, renewable energy consumption promotes environmental sustainability, and the effect is pronounced in the presence of EE.

For the control variables, we found that FDI improved the environment as it reduced ghgs (Column 1). The results indicate that ghgs reduce by −0.0015 per cent with every additional percentage increase in direct inflows to Africa. In this sense, our result shows evidence of the pollution halo hypothesis, which maintains that FDI inflows in the form of modern, clean technologies improve environmental quality (Zhu et al. [117] and Zhang and Zhou [118]). Our result confirms the findings of some previous empirical studies, such as Jiang et al. [119] and Hakimi and Hamdi [120], but contradicts that of Liu et al. [121] and Solarin et al. [122].

Additionally, the findings support the argument that FDI can help to clean up the environment, particularly if it is accompanied by green technologies that produce spillover for domestic industries (Demena and Afesorgbo [123]). Furthermore, we found that the coefficient of official development assistance on ghgs was negative and statistically significant in Column 5. Though environmentally progressive, the effect of foreign aid is weak, and suggests that greater environmental sustainability gains can be achieved if development partners support the region by way of green technology and climate change mitigation. Additionally, the results on economic growth and its squared term (although being statistically insignificant), give an indication of the EKC for African countries. Our results suggest that, in the early stages, economic growth is achieved at the expense of environment, but as incomes rise and environmental awareness grows, commitment to environmental progress is demanded by the public. This is accelerated by the increased usage of cleaner energy solutions to replace the prevalent fossil fuel energy (i.e., non-renewable energy consumption) solutions in both households and some manufacturing processes. This result is consistent with the conclusions of Sarkodie and Adams [64] and Zafar et al. [68].

The validity and reliability are confirmed subject to the post-estimation test results. First, the coefficients of the Hansen p−value are not significant at any of the conventional levels, and thus we do not reject the null hypothesis of no correlation between the set of identified instruments and the residuals (Ofori et al. [111]). Second, AR (2) statistics indicate the absence of second-order serial correlation in the residuals.

4.3. Robustness Checks

We check the robustness of the results using the two key components of ghgs—carbon dioxide (CO₂) and nitrous oxide (N₂O) emissions—as dependent variables. We first pay attention to the results on the former.

4.3.1. Effects of Energy Consumption and Energy Efficiency on the Carbon Dioxide Emissions

This section presents the results for the effects of energy consumption and energy efficiency on CO₂ emissions as indicated on

Table 7. GMM results for the effects of energy consumption and energy efficiency on environmental sustainability in Africa.

Variables	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
CO2 emissions (−1)	0.9738 ***	0.9890 ***	0.9686 ***	1.0196 ***	0.9765 ***	0.9919 ***	0.9386 ***	0.9463 ***
	(0.0102)	(0.0213)	(0.0177)	(0.0318)	(0.0190)	(0.0297)	(0.0471)	(0.0454)
FDI	−0.0002	−0.0004	0.0004	0.0015	0.0001	0.0006	−0.0026	0.0004
	(0.0007)	(0.0008)	(0.0005)	(0.0015)	(0.0011)	(0.0021)	(0.0017)	(0.0017)
Foreign aid	0.0003	0.0011	0.0049	0.0001	0.0013	0.0011	−0.0010	−0.0023
	(0.0011)	(0.0015)	(0.0032)	(0.0072)	(0.0049)	(0.0056)	(0.0049)	(0.0081)
Economic growth	0.0098 ***	0.0098 ***	0.0108 ***	0.0074	0.0104 ***	0.0148 ***	0.0157 ***	0.0052
	(0.0011)	(0.0012)	(0.0014)	(0.0051)	(0.0015)	(0.0022)	(0.0024)	(0.0066)
Economic growth squared	−0.0010 **	−0.0011 **	−0.0012 **	−0.0005	−0.0011 **	−0.0007	−0.0002	0.0004
	(0.0005)	(0.0005)	(0.0005)	(0.0009)	(0.0005)	(0.0006)	(0.0006)	(0.0011)
Primary energy consumption		0.0008				0.0077
		(0.0013)				(0.0104)
Renewable energy consumption			−0.0020				0.0074
			(0.0019)				(0.0055)
Non-renewable energy consumption				0.0017				−0.0056
				(0.0025)				(0.0071)
Energy efficiency (EE)					−0.0310	−0.3269	−1.7672 ***	0.3959
					(0.2361)	(0.4028)	(0.5723)	(0.5071)
Primary energy consumption * EE						−0.0146
						(0.0129)
Renewable energy consumption * EE							−0.0210 ***
							(0.0065)
Non-renewable energy consumption * EE								0.0240
								(0.0166)
Constant	0.2787 **	0.1551	0.4122	−0.0694	0.2643 **	−0.0873	−0.0294	0.1126
	(0.1001)	(0.1787)	(0.2447)	(0.3338)	(0.1112)	(0.3256)	(0.4908)	(0.4349)
Observations	414	414	414	322	414	414	414	322
Countries	23	23	23	23	23	23	23	23
Wald chi statistics	2.630 × 107 ***	2.750 × 107 ***	1.030 × 107 ***	162117 ***	1.870 × 107 ***	1.746 × 106 ***	808,827 ***	154,430 ***
Wald p−value	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000
Net effect	na	na	na	na	na	–	–	–
Joint significance test	na	na	na	na	na	–	–	–
p−value	na	na	na	na	na	–	–	–
Hansen p−value	0.714	0.670	0.663	0.896	0.708	0.581	0.691	0.949
AR(1)	0.001	0.001	0.001	0.001	0.002	0.001	0.000	0.001
AR(2)	0.561	0.548	0.529	0.891	0.572	0.577	0.511	0.652

(Dependent variable: Carbon Dioxide (CO₂) emissions). Standard errors in parentheses where; *** p < 0.01, ** p < 0.05, * p < 0.1.

. The results indicate that primary energy consumption is positive, albeit statistically insignificant. The sign of the coefficient suggests that more primary energy consumption is inimical to environmental quality (Javid and Khan [54]). The result suggests that given the level of development of most African countries (early stages of development), and the fact that a greater share of the energy consumption needed to meet development goals comes from non-renewable energy consumption, an increase in CO₂ emissions is imminent.

The results further show that renewable energy consumption has a negative influence on CO₂ emissions, although we did not find empirical backing. The results are in line with Al-Mulali et al. [75], who revealed that renewable energy consumption has no significant effect on pollution in sub-Saharan Africa. This means that African countries can possibly lower CO₂ emissions by raising the share of renewable power in their total energy demand. On the other hand, we report a positive coefficient of non-renewable energy consumption on CO₂ emissions, implying that the consumption of non-renewable energy triggers considerable environmental setbacks in the sampled countries.

We find a negative and statistically significant relationship between EE and CO₂ emissions. Specifically, a 1 per cent improvement in EE yields a 1.77 per cent reduction in CO₂ emissions ceteris paribus (Column 7 of Table 7). The implication of this result is that pursuing EE measures could possibly be one of the most effective ways of mitigating environmental pollution and its attendant threat to human life. Especially in a setting such as Africa where the informal sector is likely to operate under a number of non-state norms and conventions (Swain et al. [63]), an increase in EE in the form of a reduction in the use of unclean cooking fuels (e.g., use of fuelwood, charcoal, dung in the open) and energy-intensive productive industries, such as paper manufacturing, printing and basic organic and inorganic chemicals (Adom et al. [24]), could trigger remarkable environmental sustainability setbacks.

We now pay attention to our second objective, where we examine the net effect of energy consumption on CO₂ emissions (see Columns 3–5 of Table 7). We found negative and statistically significant coefficients of interaction between renewable energy consumption and energy efficiency, indicating that an improvement in energy efficiency is a beneficial factor that propels renewable energy consumption to foster environmental sustainability. The finding of the notable role of EE in the renewable energy-consumption–environmental quality relationship is consistent with the claims of Javid and Khan [54] and IEA [11] that in highly informal settings such as Africa, there is an extensive margin for EE to support the quest to reduce indoor pollution and stress on the physical environment and environmentally related mortalities.

The coefficients of economic growth and economic growth squared (i.e., proxied with GDP per capita growth and GDP per capita growth squared) on CO₂ emissions were positive and negative, respectively. The findings provide strong support for the presence of the EKC and corroborate the findings of Sarkodie and Adams [64] and Aboagye [124]. FDI also has a negative relationship with CO₂ emissions, albeit not statistically significant at any conventional level. Additionally, consistently with our main results, foreign aid had a positive but insignificant relationship with CO₂ emissions. A possible explanation for this result is the fact that the recipient government may use the aid as leverage against pro-environmental pressures emanating from donor organisations to drive economic growth, which may take time before they improve environmental quality [94].

4.3.2. Effects of Energy Consumption, Energy Efficiency on Nitrogen Gas Emissions

Table 8. GMM results for the effects of energy consumption and energy efficiency on environmental sustainability in Africa.

Variables	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
Nitrous oxide (−1)	0.8893 ***	0.7067 ***	0.9704 ***	0.9165 ***	0.8098 ***	0.1576	0.4199 ***	0.7355 ***
	(0.0293)	(0.0734)	(0.0462)	(0.0800)	(0.0430)	(0.1535)	(0.1433)	(0.1052)
FDI	−0.0002	0.0008 **	0.0006 ***	0.0005	−0.0006 ***	0.0002	−0.0014 **	−0.0003
	(0.0002)	(0.0004)	(0.0002)	(0.0004)	(0.0002)	(0.0007)	(0.0006)	(0.0005)
Foreign aid	0.0007 ***	0.0035 **	0.0045 ***	0.0045 ***	0.0001	0.0038 ***	0.0029 **	0.0032 ***
	(0.0002)	(0.0014)	(0.0008)	(0.0011)	(0.0001)	(0.0010)	(0.0011)	(0.0011)
Economic growth	0.0017 ***	0.0003	−0.0000	0.0015	0.0021 ***	0.0008	0.0009	0.0011
	(0.0004)	(0.0006)	(0.0003)	(0.0010)	(0.0004)	(0.0008)	(0.0006)	(0.0008)
Economic growth squared	−0.0002 **	0.0002	−0.0000	−0.0001	−0.0002 ***	0.0002	−0.0002	−0.0002
	(0.0001)	(0.0001)	(0.0001)	(0.0002)	(0.0001)	(0.0002)	(0.0002)	(0.0001)
Primary energy consumption		0.0020 ***				0.0018
		(0.0004)				(0.0058)
Renewable energy consumption			−0.0014 ***				−0.0040 **
			(0.0003)				(0.0016)
Non-renewable energy consumption				0.0016 ***				0.0027 *
				(0.0005)				(0.0014)
Energy efficiency (EE)					0.0685 **	0.4238	0.5626 *	0.2500
					(0.0255)	(0.2738)	(0.3135)	(0.1554)
Primary energy consumption * EE						0.0013
						(0.0086)
Renewable energy consumption * EE							−0.0150 ***
							(0.0031)
Non-renewable energy consumption * EE								−0.0018
								(0.0028)
Constant	0.2369 ***	0.5666 ***	0.1199	0.0893	0.3778 ***	1.5033 ***	1.1094 ***	0.3443 *
	(0.0633)	(0.1524)	(0.0985)	(0.1808)	(0.0801)	(0.2466)	(0.2425)	(0.1856)
Observations	414	414	414	322	414	414	414	322
Countries	23	23	23	23	23	23	23	23
Instruments	22	22	22	18	22	22	22	18
Wald chi statistics	4.170 × 107 ***	639,448 ***	339,980 ***	140,314 ***	4.718 × 106 ***	14619 ***	2.495 ×106 ***	104,812 ***
Wald p−value	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000
Net effect	na	na	na	na	na	–	−0.0127	–
Joint significance test	na	na	na	na	na	–	4.55 **	–
p−value	na	Nna	na	na	na	–	0.0462	–
Hansen p−value	0.304	0.777	0.857	0.557	0.167	0.560	0.559	0.221
AR(1)	0.097	0.047	0.046	0.028	0.078	0.018	0.033	0.034
AR(2)	0.200	0.362	0.459	0.805	0.167	0.618	0.307	0.426

(Dependent variable: Nitrous Oxide (N₂O) emissions). Standard errors in parentheses where; *** p < 0.01, ** p < 0.05, * p < 0.1.

reports the results where nitrous oxide is used as the dependent variable. We first examine the effect of the control variables on nitrous emissions. The evidence suggests that FDI has a negative effect on nitrogen gas emissions, although the effect is statistically insignificant. It can be deduced from the relationship that FDI can contribute to improved environmental sustainability in Africa plausibly through the transfer of clean technologies in host countries. Conversely, we find that foreign aid has a positive and statistically significant effect on nitrous oxide emissions. In specifics, foreign aid adversely affects Africa’s environment although this is low in magnitude. A possible explanation for this outcome is because foreign aid can negatively hamper the host country’s environmental quality by reducing institutional efficiency (Asongu and Nnanna [125]. That is, through foreign aid, donor countries undermine governance in recipient countries by exerting undue influence on environmental regulation, often to allow for the penetration of polluting firms. In such cases, government control over local industrial units regarding emissions is restrained, resulting in unrestricted harmful emissions into the atmosphere (Farooq [126]).

Additionally, in respective terms, we found that the coefficients of GDP per capita and GDP per capita squared were positive and negative, validating the EKC hypothesis in Africa. This is in line with Sinha and Sengupta [69] in the case of APEC countries.

With reference to the first hypothesis, we examined the effects of energy consumption and energy efficiency on nitrous oxide emission. The results for the effect of primary energy consumption suggest that nitrous oxide emission rises by 0.002 per cent for every 1 per cent increase in primary energy consumption. Our finding conforms to that of Yu and Liu [127], who found a positive relationship between energy consumption and nitrous oxide emissions in China.

On the other hand, we found a significant negative relationship between renewable energy consumption and nitrous emissions, suggesting the environmental sustainability-enhancing effect. In particular, an increase in renewable energy consumption was associated with a 0.0014 per cent improvement in the environment. This result corroborates the conclusion of Sinha and Sengupta [69] that renewable energy solutions have the potential to improve environmental quality, since they help reduce the level of anthropogenic emissions. Consistently with our results on ghgs, we found that non-renewable energy consumption had a positive and significant relationship with nitrous emissions. With a coefficient of 0.0016, the results suggest that for every 1 percent increase in the consumption of non-renewable energy (e.g., crude oil, coal, etc.) there is an upsurge in nitrous emissions by 0.002 per cent.

The relationship between energy efficiency and nitrous emissions is positive and statistically significant. The magnitude of the coefficient shows that every 1 per cent improvement in energy efficiency leads to an increase in nitrous emissions by 0.07 per cent. This result provides evidence in support of a potential rebound effect on the environment. However, regarding whether energy efficiency forms a significant synergy with energy consumption, the results from our estimates show a positive relationship. This result can be attributed to the fact that the impact of renewable energy is undermined by the higher effect of primary energy consumption on CO₂ emissions.

5. Conclusions and Recommendations

In line with efforts geared towards the realisation of Agenda 2050 and the Aspiration 1.7 of Africa’s Agenda 2063, this study, in the remit of SDG 7, examines whether energy efficiency has a direct or indirect mitigating effect on the environmental impacts of energy consumption in Sub-Saharan Africa. To this end, we employed macro-data from the period 2000–2020 for 23 African countries for the analysis. Robust evidence based on the dynamic GMM estimator shows that overall, energy consumption is positively associated with environmental degradation in Africa. At the disaggregated level, however, the effects differ substantially. In particular, we found that while renewable energy is significant for propelling Africa towards environmental sustainability, non-renewable energy shows a harmful effect. Additionally, the evidence suggests vis-à-vis energy consumption, energy efficiency is remarkable for promoting environmental sustainability. More crucially, the study found that energy efficiency interacted with renewable energy consumption to yield notable environmental sustainability effects. Furthermore, evidence from the transient and persistent energy efficiency scores indicates that Africa’s energy problem pertains more to structural (long-term) inefficiencies than short-term issues. This indicates that policymakers in African countries should channel resources towards streamlining structural inefficiencies in energy consumption and supply.

The findings emanating from the study reveal several implications for policymakers, which we outline as follows. First, EE can be seen as one solution for mitigating the environmental sustainability concerns of energy consumption. Since EE reduces CO₂ emissions, we therefore recommend that African countries should be interested in integrating EE measures into their energy policies, since EE yields greater environmental sustainability benefits. Moreover, the findings support the argument that FDI can help to clean up the environment, particularly if it is accompanied by green technologies that produce spillover for domestic industries. Therefore, considering the implantation of the African Continental Free Trade Area (AfCFTA), which the UNCTAD (2021) reports will trigger greater FDI inflow and high energy demand, appropriate policy options aimed at ensuring energy efficiency will prove key for promoting (mitigating) renewable energy (non-renewable energy) to foster environmental progress. Second, the study recommends that African governments adopt policies that encourage investment in energy efficiency, and such policies should not just be restricted to foreign companies, but should also target local firms as well as households. Third, our evidence on the crucial role of renewable energy consumption for environmental progress suggests that African leaders should strive to improve the share of renewable energy in Africa’s overall energy mix.

In this regard, the study recommends that African leaders should prioritise environmentally sustainable investments in the region’s energy production and systems. The study result further implies that renewable energy consumption promotes environmental sustainability, and the effect is pronounced in the presence of EE. This calls for integration of EE and renewable energy policies by African countries in efforts geared towards promoting environmental sustainability. Fourthly, the result suggests that, given the level of development of most African countries (early stages of development), and the fact that a greater share of the energy consumption needed to meet development goals comes from non-renewable energy consumption, an increase in CO₂ emissions is imminent. However, this study recommends a steady approach to the consumption of non-renewable energies, since these are unsustainable and environmentally degrading. The study recommends a careful energy transitional pathway for African countries to consider in their energy consumption mix. This means that African countries can possibly lower CO₂ emissions by raising the share of renewable power in their total energy demand. For instance, in addition to an existing energy policy, Ghana has recently adapted an energy transition pathway policy in this new era of energy transition, and this can be replicated in other countries to provide a guide towards future energy consumption. Finally, we recommend that organisations such as the African Development Bank and the World Bank support African leaders in building capacity to cushion poor and vulnerable households to transition to renewable energy options for production and consumption.

This study was limited by the short time span available for the researchers to conduct the study and the absence of country-specific data for smaller countries without their energy consumption records. The study relied more on BP Statistics, International Energy Agency, International Renewable Energy Data sources. Even though these data sources are reliable, internal country specific energy consumption data would have been more reliable and up to date. For further studies, we suggest more specific research on energy consumption and environmental quality and the effects of energy efficiency for country-specific analysis focusing on countries such as South Africa and possibly Ghana and Nigeria. There could also be a comparative analysis between Ghana, Nigeria and South Africa to understand the experiences of different countries and the lessons to be learned from each.