Energy Efficiency, Consumption, and Economic Growth: A Causal Analysis in the South African Economy

Abstract

Energy efficiency potentially reduces global carbon emissions, whereas the need of emerging countries to maintain economic growth and development entails a sharp increase in energy consumption. However, to meet this, current energy systems need to be transformed. Several studies find different conclusions on the short-run and long-run relationship and the direction of causality, and none of the studies have considered energy efficiency in their model. This study investigates the direction of causality between energy efficiency, energy consumption, and economic growth in South Africa. To determine if a long-run relationship between the variables exists, the Johanson cointegration test is used, and the results indicate that there is a long-run relationship between economic growth, energy depletion, energy efficiency, non-renewable energy consumption, renewable energy consumption, and energy security, with trace statistics suggesting that the null hypothesis of no cointegration should be rejected at a 5% level of significance. The Toda and Yamamoto procedure of the Granger causality approach was then applied. This study finds a unidirectional causality between energy efficiency, non-renewable energy consumption, and economic growth and no causality between renewable energy consumption, energy depletion, energy security, and economic growth. The growth hypothesis is supported, while the neutrality hypothesis is only confirmed regarding renewable energy consumption and economic growth. The results further suggest that a unidirectional Granger causality exists between non-renewable consumption and energy efficiency, and economic growth in South Africa. In South Africa, energy efficiency is a significant tool to enhance sustainable growth and attain climate objectives. Also, energy efficiency helps to lower the costs of mitigating carbon emissions and further advance both social and economic development.

1. Introduction

Recently, countries worldwide have committed to new targets to reach net zero by 2050 or sooner. A total of 75% of global emissions come from the energy sector; therefore, transforming it will be crucial to addressing the climate crisis. Countries agree that a swift shift from the business-as-usual culture will do more to attain change at the pace that is needed. Energy efficiency can potentially reduce global carbon emissions by more than 40% by 2040, according to the International Energy Agency (2023). The need for emerging countries to maintain economic growth and development entails a sharp increase in energy consumption. However, to meet this, current energy systems need to be transformed. Energy efficiency is the initial transformation lubricant, and it enhances the attainment of net-zero energy objectives at close to zero costs and provides a wide range of advantages for society. The International Energy Agency (2023) further mentions that the key objective of doubling the progress on energy intensity is just a strong wish, and for the past ten years, no country achieved a sustained yearly average improvement greater than 4%. Worldwide, an improvement in energy efficiency is generally recommended to increase production and enhance environmental quality.

The transformation of existing energy systems is important because of the increase in the global demand for primary energy. Raimi et al. (2023) noted that, compared to 2021 levels, the global electricity demand is expected to increase by between 62% and 185% by 2050. This will put an extra strain on the energy systems and exacerbate weaknesses in the country’s energy infrastructure, including pipelines and other ways of transporting oil and power. The International Energy Agency (2022) reports that economic growth prospects, energy efficiency, energy security, and environmental sustainability are key factors which influence the future of the global energy demand and the transformation of energy systems and emphasises the critical role of energy efficiency. In order for countries to mitigate the climate change crisis and attain a net zero target by 2050, the causal relationship between economic growth, energy efficiency, energy security, and energy consumption should be extensively investigated and understood.

The objective of the investigation of the present study is to examine the causal relationship between energy efficiency, economic growth, and energy consumption in South Africa. Since the dawn of democracy in 1994, South Africa’s initial Energy Efficiency Strategy (ESS) was compiled in 2005, and the main objective was geared towards sustained economic growth and development and the promotion of energy efficiency practices (Department of Minerals and Energy, 2005). The ESS borrowed its guidelines from the publication of the White Paper on Energy Policy in 1998. The White Paper on Energy Policy in 1998 linked the developments in the energy sector with South Africa’s socio-economic development strategies. As suggested by the Department of Minerals and Energy (2005), ESS’s strategic objectives included the three foundations of sustainable development: environmental, social, and economic sustainability. The ESS established a nationwide objective of a 12% energy efficiency improvement by 2015. While energy efficiency declined from 2010 to 2014, resulting in 2% greater energy usage in 2014, this trend reversed from 2014 to 2018. Combined with a shift to less energy-intensive sectors in the South African economy, this resulted in a decrease in energy consumption in 2018, despite a growth in economic activity (International Energy Agency, 2023). South Africa’s post-2015 strategy emphasised the capture of the multiple benefits of energy efficiency by improving energy security, safeguarding environmental sustainability, reducing energy depletion, and stimulating economic growth, but the empirical literature outlining this relationship is sparse.

In South Africa, several studies focused on the causality between energy consumption and economic growth (see Menyah & Wolde-Rufael, 2010; Lin & Wesseh, 2014) and suggested a unidirectional causality between energy consumption and economic growth, while H. B. Khobai and Le Roux (2017) found a two-way causality between energy consumption and economic growth. Also, another pool of empirical studies investigated the link between energy consumption and environmental quality (Akadiri et al., 2019; Bekun et al., 2019; Mapapu & Phiri, 2018; H. B. Khobai & Le Roux, 2017; H. Khobai & Sithole, 2022; Ganda, 2018; Ndlovu & Inglesi-Lotz, 2020); in addition to the fact that the studies find different conclusions on the short-run and long-run relationship and the direction of causality, none of the studies have considered energy efficiency in their models. Furthermore, this study includes energy security, environmental sustainability, energy depletion, and economic growth causal relationships in a single model in South Africa.

The principal findings of this study are as follows: (1) The Johanson cointegration test indicates that there is a long-run relationship between economic growth, energy depletion, energy efficiency, non-renewable energy consumption, renewable energy consumption, and energy security, with trace statistics suggesting that the null hypothesis of no cointegration should be rejected at a 5% level of significance. (2) The Toda and Yamamoto (1995) procedure of the Granger causality approach indicates a unidirectional causality between energy efficiency, non-renewable energy consumption, and economic growth and no causality between renewable energy consumption, energy depletion, energy security, and economic growth. The growth hypothesis is supported, while the neutrality hypothesis is only confirmed regarding renewable energy consumption and economic growth. The results suggest that a unidirectional Granger causality between non-renewable consumption and economic growth, and energy efficiency and economic growth in South Africa exists. The empirical results appear relatively robust with this estimation technique.

Energy plays a crucial role in South Africa’s economic development, with the relationship between energy efficiency, energy consumption, and economic growth being of particular interest to policymakers and researchers. As South Africa transitions towards a more sustainable and energy-secure future, understanding the causal interactions between these variables is essential for informed decision-making. This study examines the direction of causality between energy efficiency, energy consumption, and economic growth to provide insights into policy implications for sustainable economic development. The closest study to our study is by Chen et al. (2023). However, the current study differs from this in various ways. Unlike Chen et al. (2023), who used data from the ten most energy-efficient countries, time-series data (1985 to 2021) from South Africa are used in the current study to examine this linkage. In addition, since the Sustainable Development Goal 7.3 target calls for global progress on energy efficiency by doubling the rate of improvement in energy efficiency globally by 2030, the current study uses primary energy consumption/use per GDP as a proxy for energy efficiency. Also, studies in South Africa failed to recognise the role of energy efficiency, energy security, and energy depletion in the energy consumption–economic growth nexus and the environmental implications. There are few studies, such as Fu et al. (2021) and Chen et al. (2023), that concentrated on energy efficiency dynamics, focusing on environmental sustainability, energy security, energy efficiency, and energy depletion, on one side and economic growth on the other; nonetheless, the studies are not country specific. Therefore, this study attempts to address this gap.

The rest of this study is organised as follows. Section 2 reviews the related literature on the theoretical and empirical relationship between energy efficiency, energy consumption, and economic growth. Section 3 presents the baseline model, data, variables, and methodologies. Section 4 reports the empirical results and analysis. Section 5 concludes the study with policy implications.

2. Theoretical Framework and Literature Review

The relationship between energy and economic growth has been widely studied using different economic theories. The growth hypothesis suggests that energy consumption drives economic growth, implying that energy constraints could hinder development (Stern, 2011). On the other hand, the conservation hypothesis implies that economic growth leads to lower energy consumption as economies shift toward more energy-efficient technologies (Apergis & Payne, 2010); whereas the feedback hypothesis proposes a bidirectional relationship where energy consumption and economic growth influence each other (Omri, 2013).

Empirical studies on South Africa have produced mixed findings. Inglesi-Lotz and Pouris (2016) found that energy efficiency improvements contribute to economic growth, while Odhiambo (2023) showed that energy consumption remains a significant driver of industrial productivity. The theoretical literature offers a variety of justifications for the significance of energy efficiency on the energy–economic growth relationship. Adom et al. (2021) pointed out that the relationship between economic growth and energy efficiency is explained in the traditional classical, neoclassical, endogenous growth, new endogenous growth, and catch-up theories of economic growth.

However, the theories downplay the importance of energy efficiency. The four hypotheses (growth hypothesis, conservation hypothesis, feedback hypothesis, and neutrality hypothesis) that explain the economic growth and energy consumption relationship failed to explain the pertinency of energy efficiency clearly. Menegaki and Tsani (2018) demonstrated the importance of including energy efficiency in the growth and energy consumption relationship using the four hypotheses as follows.

Firstly, the growth hypothesis postulates that energy consumption is a crucial complement to economic growth. It assumes a one-way causal relationship between energy consumption and economic growth (energy consumption causes economic growth) (Rajaguru & Khan, 2021). Furthermore, energy conservation reduces economic growth. With the inclusion of energy efficiency, the higher the efficiency, the more economic growth decreases. Secondly, Alper and Oguz (2016) noted that the conversation hypothesis assumes a one-way causality from economic growth to energy consumption and, therefore, no feedback effects. Energy consumption will not cause economic growth. Including energy efficiency further implies that a higher energy efficiency increases energy conservation (Menegaki & Tsani, 2018). Thirdly, the feedback hypothesis, which includes energy efficiency, suggests that energy conservation reduces economic growth; in turn, reduced economic growth will lead to poor energy consumption too. Therefore, a higher energy efficiency lowers economic growth, and the consequent hindered, or lower, economic growth can hardly produce more energy efficiency; this is an inevitable outcome when an investment in technology, which ameliorates energy efficiency, does not take place. Finally, if there is no causal relationship between energy consumption and economic growth, the neutrality hypothesis is upheld. According to the neutrality hypothesis, the decrease in energy consumption does not hurt economic growth. Consequently, energy efficiency has no impact on the causality (Menegaki & Tsani, 2018).

Economic growth also influences environmental quality. Among other theoretical frameworks, which explain various factors that influence the environment, studies by Wang et al. (2017); Song et al. (2011); and Danish et al. (2021), just to mention a few, note the importance of the IPAT equation by Holdren and Ehrlich (1974) as a crucial framework through which the various factors influence the environment. The equation is important because it shows how complicated the human–environment interface is by quantifying the contribution of various important factors (such as economic growth, energy security, energy depletion, etc.) to the overall environmental impact. This crude equation implies that the influence (I) of society on the environment is related to the population size (P) and its function (f), which is influenced by factors like affluence (A) and the availability of technology (T) (Chertow, 2000).

Lastly, other important hypotheses which also describe the link between economic growth and energy consumption are the decoupling hypothesis, energy innovation hypothesis, and the Jevons Paradox: rebound effect (Liu et al., 2020). Jin et al. (2018) noted that technological progress plays a major role in enhancing energy efficiency as well as saving energy, and further contributes to economic growth, which in turn stimulates the energy demand and finally encourages long-term energy consumption, a phenomenon termed the rebound effect. This claim was presented by Khazzoom (1987) and Brookes and Grubb (1992). Also, with greater technological improvement, energy efficiency will improve and The decoupling hypothesis posits that when there is higher economic activity in an economy, it is likely to promote energy consumption in both the long run and the short run. energy demand will increase, and energy is viewed as a factor of production, which is substitutable for productivity and technology as key drivers of growth and living standards, a suggestion which is in line with neoclassical economics. Decoupling vitally provides a prospective resolution to the puzzle of striking a balance between economic growth and the preservation of the ecological and planetary system. In a case where decoupling outpaces economic growth, it is possible and feasible to attain an absolute reduction in energy use and environmental emissions. According to Tenaw and Hawitibo (2021), this implies the ability to reduce environmental effects without affecting economic growth. Therefore, decoupling entails a scenario where the growth rate of emissions is stable and/or is lower than the growth rate of the economy.

The empirical debate regarding the link between economic growth and environmental quality continues, but given the aim of this study, this section reviews empirical studies augmented with the inclusion of the energy efficiency variable in the economic growth–energy consumption relationship, which has environmental consequences; the energy depletion variable in the economic growth–energy consumption relationship and its environmental implications; and lastly the energy security variable into economic growth–energy consumption and its environmental implications. Therefore, in this study, the literature is characterised by three classifications.

The first classification focuses on empirical studies that explored the energy efficiency–economic growth–energy consumption nexus and its environmental implications; for example, J. Zhang et al. (2020), F. Zhang et al. (2022), S. Khan et al. (2022), and L. Xu et al. (2022) suggest that economic growth harms environmental quality, but energy efficiency reduces carbon emissions. Contrarily, Ibrahim and Alola (2020) and Alam et al. (2022) suggest that energy efficiency worsens environmental quality.

Applying the cointegration test developed by Westerlund (2008) and the causality test by Dumitrescu and Hurlin (2012), Bayar and Gavriletea (2019) examined the effects of energy efficiency and renewable energy on economic growth between 1992 and 2014 in emerging economies. They found a positive long-term relationship between energy efficiency and economic growth and an insignificant long-term relationship between renewable energy and economic growth. The Dumitrescu and Hurlin (2012) causality test found a short-run unidirectional causality from both energy efficiency and renewable energy use to economic growth.

J. Zhang et al. (2020) employed the Panel Autoregressive Distributed Lag (ARDL) and the Data Envelopment Analysis (DEA) model to investigate the Environmental Kuznets Curve (EKC) hypothesis in 15 developing countries for 1990–2013. They found that economic growth (GDP) positively impacts carbon emissions, but the GDP squared is inversely related to carbon emissions. Additionally, the DEA-based results revealed that these emerging economies had comparatively poor environmental conditions because of their high energy intensity and low energy efficiency.

In BRICS countries, Akram et al. (2021) studied how the heterogeneous effects of renewable energy consumption and energy efficiency affect economic growth in 1990–2014, using fixed-effect panel quantile regression and the Dumitrescu–Hurlin (D-H) heterogeneous panel causality test. The results show that energy efficiency has a significant impact across all quantiles, although the impact is strongest at the 50th quantile and 60th quantile of economic growth. Renewable energy consumption has a significant negative impact on economic growth; however, this effect is robust at upper economic growth quantiles (0.60–0.90). The panel causality test supports the feedback hypothesis between economic growth and energy efficiency. The results showed that renewable energy consumption and economic growth have a bidirectional causality. Additionally, the results demonstrated a causality from energy efficiency to renewable energy consumption.

Using the data from 2004 to 2019, F. Zhang et al. (2022) investigated how energy efficiency, financial inclusion, economic growth, environmental-related technical innovation, and the human capital index affect carbon emissions in the five RCEP member countries, utilising the panel quantile regression estimator. The results demonstrate that financial inclusion and economic growth significantly aggravate environmental degradation by enhancing the carbon emission level, among which the robust carbon emission growth is found in the second quantile. Also, the results demonstrate that carbon emissions are significantly lowered by environmentally related technical innovation, energy efficiency, and the human capital index.

Also, S. Khan et al. (2022) investigated whether boosting renewable energy production, increasing energy efficiency rates, and promoting financial inclusion can lower carbon dioxide emissions in the Next Eleven countries using panel regression analysis (the Augmented Mean Group (AMG) technique). The results demonstrate that energy efficiency and increases in renewable electricity shares in total electricity outputs mitigate carbon dioxide emissions in the long run. On the other hand, it has been found that economic growth, financial inclusion, and trade increase carbon dioxide emissions. Gang et al. (2023) also found that green investment, energy efficiency, and urbanisation are negatively related to environmental sustainability.

L. Xu et al. (2022) employed the Quantile-on-Quantile (Q.Q.) regression method to examine the global factors that influenced environmental quality between 1990Q4 and 2020Q4. They found that economic growth degrades the environment by increasing carbon emissions. Energy efficiency, on the other hand, is a key component of environmental sustainability because it lowers the amount of carbon emissions in the atmosphere. Using the pairwise Granger causality test, the study demonstrated a two-way causal connection between economic growth–energy efficiency and energy efficiency–carbon emissions, as well as a unidirectional causality from economic growth to carbon emissions.

Contrary to studies by J. Zhang et al. (2020); F. Zhang et al. (2022); L. Xu et al. (2022); and S. Khan et al. (2022), which suggested that energy efficiency is a key component of environmental sustainability because it lowers the amount of carbon emissions, Ibrahim and Alola (2020) analysed the nexus between energy, economic growth, and environmental sustainability in MENA countries using the ARDL Pooled Mean Group and Data Envelopment Analysis models. They found that energy efficiency worsens environmental quality, and economic growth has a significant negative impact on the environment.

Also, Alam et al. (2022) examined how Oman’s possibilities of achieving an environmentally sustainable growth between 1972 and 2019 were influenced by energy use, energy efficiency, and financial development. They found a long-run relationship between the investigated variables. Additionally, it has been established that an increasing energy efficiency increases carbon outputs in Oman, whereas an increasing energy use and financial development decrease carbon outputs. In South Africa, Aye et al. (2015) investigated the drivers of energy efficiency and found that the manufacturing employment index and the move from the overall productive structure towards low-value-added services, intensive energy consumption, and CO₂ emissions affect energy efficiency in the economy.

The second classification focuses on empirical studies, which are the energy security–economic growth–energy consumption nexus and its implications on environmental quality. Ali et al. (2023) found that the use of renewable energy increases energy security, reduces carbon emissions, and enhances environmental protection without sacrificing economic growth. Nepal et al. (2021) and Shittu et al. (2021) found a negative relationship between energy security and the ecological footprint, suggesting that prioritising energy security benefits environmental quality. In India, Nepal et al. (2021) used the ARDL model to examine the dynamic relationships between energy security, captured by national-level energy use; foreign direct investment; economic output; carbon emissions; and trade during the period of 1978–2016. They found a short-run and long-run nexus among the variables. A 1% increase in FDI results in a 0.013% reduction in energy use. Using the VECM Granger causality tests, the results demonstrated that in the long run energy use is Granger caused by all the variables. In BRICS countries, from 1976 to 2016, Hsu et al. (2021) used the OLS method to quantify the link between energy efficiency, energy security, and economic development. Macroeconomic factors, like energy consumption, economic growth, and environmental degradation, were also included. Energy efficiency was assessed for the period spanning from 2010 to 2018. They found that Brazil and Russia are the member countries with less energy use in successive years. In terms of energy and economic development, Brazil, Russia, and South Africa came out on top, while China and India were among the nations with the worst scores. Results for energy efficiency showed that China had the highest score of 1, while India and South Africa had scores of roughly 0.623 and 0.64, respectively.

Also, Shittu et al. (2021) investigated the factors that contribute to environmental degradation (proxied by the ecological footprint) in 45 resource-rich Asian countries. Over the years from 1990 to 2018, the study used a variety of panel methodologies, including the instrumental variable two-stage least square technique. The results demonstrated an inverse relationship between energy security and the ecological footprint, indicating that countries need to be secured in terms of energy to reduce the degradation of the environment. They also found a non-linear relationship between the ecological footprint and economic growth, even though it is opposed to the EKC claim. Uche et al. (2024) examined the effects of energy security on environmental quality in the Emerging Seven (E7) countries, using panel quantile regression. They found that energy security accelerates pollution at the 75th and 95th quantiles, while improving environmental equity only at the 25th and 50th quantiles. Ibrahiem and Hanafy (2021) found a two-way causality between energy security and economic growth. The study further suggests that energy security, foreign direct investments, and deterioration in the environment cause renewable energy.

The last classification is based on empirical studies that investigated energy depletion, economic growth, and implications on environmental quality. For a heterogeneous panel of nations, using the data from 1995 to 2016, Bhuiyan et al. (2018) investigated the dynamic relationships between alternative energy use and nuclear energy consumption, tourism revenues, bank-specific characteristics, and environmental and resource issues. A two-step differenced Generalised Method of Moments (GMM) estimator was used. They found that bank-specific factors significantly lessen resource/energy depletion, while worldwide travel increases carbon emissions globally and depletes energy resources. Nuclear energy consumption slows down the depletion of resources, whereas industrial value-added accelerates both carbon emissions and energy resource depletion. The panel causality results support various causality patterns among the variables. In the MENA, an EKC hypothesis was tested using a GMM econometrics technique by Amirnejad et al. (2021). They found a U- and inverted-U-shaped relationship between economic growth and energy-forest depletion. The health and energy-forest depletion indexes were adversely impacted by carbon emissions. Hussain et al. (2020) analysed the effects of natural resource depletion on energy use and carbon dioxide emissions for a panel of 56 “Belt & Road Initiative” (BRI) countries from 1990 to 2014.The Augmented Mean Group (AMG) panel estimator and Common Correlated Effects Mean Group (CCEMG) results suggest a positive relationship between energy use, carbon emissions, and natural resource depletion. The findings of the VECM Granger causality test highlighted a feedback hypothesis between carbon emissions, energy use, economic growth, resource depletion, urbanisation, and trade openness. In China, Y. Xu and Zhao (2023) used the Robust GEE population-averaged model for long-run estimates to analyse the effects of energy depletion and the human development index on natural resources for 1971–2019, while considering the roles of carbon emissions and economic growth. They found that both carbon emissions and economic growth are negatively related to natural resources, but both the human development index and energy depletion are positively related to natural resources.

From the literature survey above, it is evident that several reported findings of studies show conflicting results across different types of methodologies, the countries used (single country and cross-country), and the proxies used. Studies in South Africa failed to recognise the role of energy efficiency, energy security, and energy depletion in the energy consumption–economic growth nexus and the environmental implications. There are few studies, such as Fu et al. (2021) and Chen et al. (2023), that concentrated on energy efficiency dynamics, focusing on environmental sustainability, energy security, energy efficiency, and energy depletion on one hand and economic growth on the other.

3. Data and Methodology

The baseline model employed in this study is a modification of the one by Chen et al. (2023) for a country-specific study; it can be expressed as follows:

$$Y=f(EF,ED,ES,NRE,REN)$$

(1)

Y = Real GDP per capita growth, see Odhiambo (2023).

EF = Energy efficiency proxy (primary energy consumption/use per GDP—measured in kilowatt-hours per international-$), see Chen et al. (2023).

ED = Energy depletion (adjusted savings: energy depletion (current USD)).

ES = Energy security (total electricity generation—measured in terawatt-hours).

NRE = Non-renewable consumption (share of primary energy consumption that comes from coal—measured as a percentage of the total primary energy, using the substitution method (%)).

REN = Renewable energy consumption (share of primary energy consumption that comes from renewables—measured as a percentage of the total primary energy, using the substitution method (%)).

The modified Wald test in the Vector Auto-Regressive (VAR) approach proposed by Toda and Yamamoto (1995) is used to investigate the causality between energy efficiency, energy security, energy depletion, energy consumption, and economic growth. This approach addresses the drawbacks of the traditional Granger causality test in that it avoids the issues that may arise when variables are non-stationary or have a long-run (cointegration) relationship during the causality test.

The Toda and Yamamoto (1995) procedure of the Granger causality test starts with the determination of the optimal lag length, $k$, by applying the lag selection procedure. Subsequently, the maximal order of integration, $d_{max}$, will be established. If the stationarity test (unit root test) indicates that the variables are stationary at I (0), I (1), and I (2), $d_{max}$ will be 2. To ensure that the model is valid, $k$ should be more than or equal to $d_{max}$ ($k\ge d_{max}$). Lastly, it is of paramount significance to estimate a ($k\ge d_{max}$)th order of VAR and check the Block Exogeneity Wald test for the direction of causality.

To employ the Toda and Yamamoto (1995) technique of the Granger causality test, the base model of Equation (1) is converted into the following standard VAR presentation:

$$Y_t={\propto }_0+Z+{\epsilon }_{1t}$$

(2)

$${lnED}_t={\theta }_0+Z+{\epsilon }_{2t}$$

(3)

$${lnREN}_t={\delta }_0+Z+{\epsilon }_{3t}$$

(4)

$${lnES}_t={\phi }_0+Z+{\epsilon }_{4t}$$

(5)

$${lnEF}_t={\rho }_0+Z+{\epsilon }_{5t}$$

(6)

$${lnNRE}_t={\omega }_0+Z+{\epsilon }_{6t}$$

(7)

where

$$\begin{array}{l}\mathrm{Z}={\sum }_{i=1}^k{\propto }_{1i}{Y}_{t-1}+{\sum }_{i=k+1}^{d_{max}}{\propto }_{2i}{Y}_{t-i}+{\sum }_{i=1}^k{\theta }_{1i}{lnED}_{t-1}+\\ {\sum }_{i=k+1}^{d_{max}}{{\theta }_{2i}lnED}_{t-i}+{\sum }_{i=1}^k{\delta }_{1i}{lnREN}_{t-1}+\\ {\sum }_{i=k+1}^{d_{max}}{{\delta }_{2i}lnREN}_{t-i}+{\sum }_{i=1}^k{\phi }_{1i}{lnES}_{t-1}+{\sum }_{i=k+1}^{d_{max}}{\phi }_{2i}{lnES}_{t-i}+\\ {\sum }_{i=1}^k{\rho }_{1i}{lnEF}_{t-1}+{\sum }_{i=k+1}^{d_{max}}{{\rho }_{2i}lnEF}_{t-i}+{\sum }_{i=1}^k{\omega }_{1i}{lnNRE}_{t-1}+\\ {\sum }_{i=k+1}^{d_{max}}{{\omega }_{2i}lnNRE}_{t-i};\end{array}$$

(8)

$${\epsilon }_{it}=residual.$$

$$ln=naturallogarithm$$

where k is the optimal lag length as established by the lag selection criterion and $d_{max}$ is the maximum order of integration. Given $Y_t$ from Equation (2), ${lnED}_t$ from Equation (3) causes $Y_{t}if$ ${\theta }_{1i}\ne 0,\mathrm{f}\mathrm{o}\mathrm{r}\mathrm{e}\mathrm{v}\mathrm{e}\mathrm{r}\mathrm{y}\mathrm{i}=1,2,3,\dots ,k$. For Equation (4), ${lnREN}_t$ causes $Y_{t}if$ ${\delta }_{1i}\ne 0,\mathrm{f}\mathrm{o}\mathrm{r}\mathrm{e}\mathrm{v}\mathrm{e}\mathrm{r}\mathrm{y}\mathrm{i}=1,2,3,\dots ,k$. Furthermore, ${lnES}_t$ from Equation (5) is said to cause $Y_{t}if$ ${\phi }_{1i}\ne 0,\mathrm{f}\mathrm{o}\mathrm{r}\mathrm{e}\mathrm{v}\mathrm{e}\mathrm{r}\mathrm{y}\mathrm{i}=1,2,3,\dots ,k$. ${lnEF}_t$ from Equation (6) is said to cause $Y_{t}if$ ${\rho }_{1i}\ne 0,\mathrm{f}\mathrm{o}\mathrm{r}\mathrm{e}\mathrm{v}\mathrm{e}\mathrm{r}\mathrm{y}\mathrm{i}=1,2,3,\dots ,k$. Lastly, ${lnNRE}_t$ from Equation (7) is assumed to cause $Y_{t}if$ ${\omega }_{1i}\ne 0,\mathrm{f}\mathrm{o}\mathrm{r}\mathrm{e}\mathrm{v}\mathrm{e}\mathrm{r}\mathrm{y}\mathrm{i}=1,2,3,\dots ,k$. The Toda and Yamamoto (1995) technique pays no attention to the coefficient matrices of the last $d_{max}$ lagged vectors in the model. This further suggests that the coefficients ${\theta }_{2i}$, ${\delta }_{2i},{\phi }_{2i},{\rho }_{2i}$, and ${\omega }_{2i}$ are not included. The technique further tests the linear or non-linear constraints on the first $k$ coefficient matrices by making use of the conventional asymptotic theory (Paul, 2020).

Data

The data to be used for this study were collected from two sources, namely, World Bank Development Indicators (WDIs), and OurWorldInData.org/energy (see

Table 1. Definitions of the variables.

Variable	Definitions of Variable and Measurement	Source
Y	Economic growth (GDP was calculated as an annual percentage growth rate based on a constant prices)	World Bank Development Indicators
ED	Energy depletion (the energy depletion rate was determined by estimating savings as a percentage of GDP)	(https://ourworldindata.org/energy, accessed on 13 April 2025)
REN	Renewable energy (share of primary energy consumption that comes from renewables—measured as a percentage of the total primary energy, using the substitution method (%))	(https://ourworldindata.org/energy, accessed on 13 April 2025)
ES	Energy security (total electricity generation—measured in terawatt-hours)	(https://ourworldindata.org/energy, accessed on 13 April 2025)
EF	Energy efficiency (primary energy consumption/use per GDP—measured in kilowatt-hours per international-USD)	(https://ourworldindata.org/energy, accessed on 13 April 2025)
NRE	Non-renewable energy (share of primary energy consumption that comes from coal—measured as a percentage of the total primary energy, using the substitution method (%))	(https://ourworldindata.org/energy, accessed on 13 April 2025)

below), and covers the period from 1985 to 2021. The plot of the percentage growth rate (GDPGR) and total electricity generation (ES) (two series) is shown in Figure A1 [Appendix A below]. The scale on the left axis pertains to the economic growth rate (GDPGR), while the scale on the right axis pertains to energy security. There appears to be some co-movement between the GDPGR and ES between 1997 and 2018. Also, Figure A2 [Appendix A below] shows the strong co-movement between the percentage growth rate (GDPGR) and renewable energy consumption as a percentage of primary energy consumption (REN), particularly for the entire period. Figure A3 [Appendix A below] shows that for the periods of 1985–1991 and 2001–2018, there are co-movements between economic growth (GDPGR) and energy efficiency (EF). The scale on the left axis pertains to the economic growth rate (GDPGR), while the scale on the right axis pertains to primary energy use per GDP (EF). Figure A4 [Appendix A below] further shows possible strong co-movements between economic growth versus non-renewable energy consumption for the entire period under study. Lastly, Figure A5 [Appendix A below] shows some possible correlation between energy depletion and economic growth, particularly for the entire period. The scale on the left axis pertains to the economic growth rate (GDPGR), while the scale on the right axis pertains to the energy depletion rate, which is determined by estimating savings as a percentage of GDP. However, whether this co-movement relates to causation is verified using the Toda–Yamamoto Causality (Block Exogeneity Wald) test.

4. Empirical Analysis

4.1. Unit Root Tests

In order to apply the Toda and Yamamoto (1995) procedure, the variables included in our model should be tested for stationarity to establish the order of integration. Two techniques are used to test for the unit root: the Augmented Dickey–Fuller (ADF) test and the Phillips–Perron test (PP). The unit root test results are summarily represented in

Table 2. Unit root and stationarity tests.

Variable: Y
	Level			1st Difference
	Constant	Constant and Trend	None	Constant	Constant and Trend	None
ADF	−3.606984 **	−3.598962 **	−3.537085 ***	−6.296929 ***	−6.139641 ***	−6.400182 ***
PP	−3.571278 **	−3.577460 **	−3.537085 ***	−7.891592 ***	−9.886550 ***	−7.416417 ***
Variable: ln (ED)
ADF	−3.071143 **	−2.883738	−1.233634	−6.708545 ***	−6.778792 ***	−6.824627 ***
PP	−3.045499 **	−2.837453	−1.127161	−6.866248 ***	−6.866248 ***	−6.906979 ***
Variable: ln(REN)
ADF	−2.133017	−3.078393	−1.324630	−8.304008 ***	−8.315356 ***	−8.335274 ***
PP	−2.126549	−3.161961	−2.068486	−8.682139 ***	−10.21169 ***	−8.441800 ***
Variable: ln(EF)
ADF	−0.008020	−2.703153	−3.018536 ***	−6.330597 ***	−4.390180 ***	−4.932742 ***
PP	0.092547	−2.703106	−3.149869 ***	−6.395423 ***	−6.303031 ***	−5.015564 ***
Variable: ln(ES)
ADF	−2.909126	−0.134898	2.995655	−4.210067 ***	−5.299375 ***	−3.648703 ***
PP	−2.909126	−0.015056	2.415678	−4.243533 ***	−4.928720 ***	−3.529294 ***
Variable: ln(NRE)
ADF	−2.292455	−3.399575	−1.037296	−6.853436 ***	−6.878926 ***	−6.737369 ***
PP	−2.199682	−3.245435	−2.253214 **	−7.838867 ***	−9.525752 ***	−6.892561 ***

Note: ** and *** imply significance at 5% and 1%, respectively.

. As shown in Table 2 below, both the Augmented Dickey–Fuller (ADF) test and the Phillips–Perron test (PP) show that the series have different orders of integration, and the highest order of integration is 1. Y, lnED, lnEF, and lnNRE are integrated at order 0 (stationary variables), while lnREN and lnES are integrated at order one (differenced stationary variable). The findings for Y, lnED, lnEF, and lnNRE rejected the null hypothesis of non-stationarity for the level-scale series at the 1% level of significance, 5% level of significance, 1% level of significance, and 5% level of significance, respectively. For lnREN and lnES all stationary tests fail to reject the null of non-stationarity at the 5% level of significance for the series in the level and become stationary after differencing once at a 1% level of significance.

4.2. The Result of Johansen’s Methodology to Determine the Long Run Relationship

Before testing for the possibility of a long-run relationship between energy security, energy depletion, environmental sustainability, and sustainable economic growth, the maximum number of lags is established, i.e., the VAR lag order selection criteria. The base model is specified to the standard VAR in levels despite the prevailing circumstances of the different order of integration established from the Unit root tests. For this study, the optimal lag is one and this is based on the Akaike’s Information Criterion (AIC) and Final Prediction Error (FPE) (see

Table 3. VAR lag order selection criteria.

Lag	LogL	LR	FPE	AIC	SC	HQ
0	22.15145	NA	1.60 × 10−8	−0.922940	−0.656309	−0.830899
1	182.4637	256.4996 *	1.36 × 10−11 *	−8.026497 *	−6.160079 *	−7.382210 *
2	213.8960	39.51489	2.14 × 10−11	−7.765485	−4.299281	−6.568952

* Indicates lag order selected by criterion.

below). Liew (2004) suggested that the AIC and FPE are more powerful and reliable than the other criteria when a study uses a sample below 60 observations. Because this study has 37 observations, the AIC and FPE were used because the criteria reduce the possibility of underestimation while maximising the possibility of recovering the true lag length. Using the Johansen methodology, a cointegration test was carried out to determine the long-run relationship between Y, lnED, lnEF, lnNRE, lnREN, and lnES. From

Table 4. Johansen cointegration test results. Trend assumption: linear deterministic trend (constant without trend).

Unrestricted Cointegration Rank Test (Trace)
Hypothesized No. of CE(s)	Eigenvalue	Trace Statistic	0.05 Critical Value	Prob.**
None *	0.657376	114.5196	95.75366	0.0014
At most 1 *	0.613192	77.03033	69.81889	0.0119
At most 2	0.373813	43.78641	47.85613	0.1145
At most 3	0.319294	27.40268	29.79707	0.0922
At most 4	0.244363	13.94083	15.49471	0.0846
At most 5 *	0.111406	4.134014	3.841466	0.0420
Trace test indicates 2 cointegrating equation(s) at 0.05 level. * Denotes rejection of the hypothesis at 0.05 level. ** MacKinnon–Haug–Michelis (1999) p-values.
Unrestricted Cointegration Rank Test (Maximum Eigen Value)
Hypothesized No. of CE(s)	Eigenvalue	Max-Eigen Statistic	0.05 Critical Value	Prob.**
None	0.657376	37.48929	40.07757	0.0951
At most 1	0.613192	33.24392	33.87687	0.0593
At most 2	0.373813	16.38373	27.58434	0.6330
At most 3	0.319294	13.46185	21.13162	0.4107
At most 4	0.244363	9.806817	14.26460	0.2249
At most 5 *	0.111406	4.134014	3.841466	0.0420
Max-eigenvalue test indicates 1 cointegrating equation at 0.05 level * Denotes rejection of the hypothesis at 0.05 level. ** MacKinnon–Haug–Michelis (1999) p-values.

below, the results shown with trace statistics suggest that the null hypothesis of no cointegration should be rejected at a 5% level of significance; therefore, cointegration exists and, as such, there is a long-run relationship between/among the variables in the model (indicates two cointegrating equations). The max-eigenvalue test indicates one cointegrating equation at the 5% level. However, given that the tests have conflicting results, especially on the number of cointegration equations [which is not impossible when dealing with since we used 37 observations which is regarded as a small sample], the trace test is used (Lütkepohl et al., 2000).

4.3. Toda–Yamamoto Tests of Granger Causality

Regardless of the cointegration results, the Toda–Yamamoto technique is to be considered since the series are of different orders of integration (that is I (0) and I (1) series), as confirmed from the ADF and PP Unit root tests (Table 2). Finally, to investigate the direction of causality, in this study we estimated a VAR model using the Toda–Yamamoto procedure. For the VAR model, the optimum lag length is one (see Table 3).

The results from

Table 5. Toda–Yamamoto causality (Block Exogeneity Wald) test result.

Null Hypothesis	Chi-sq	Prob.	Direction of Causality
ED does not cause Y	11.53971	0.0031 ***	ED → Y
EF does not cause Y	5.875767	0.0530 *	EF → Y
ES does not cause Y	0.239696	0.8871	No causality
NRE does not cause Y	10.37899	0.0056 ***	NRE → Y
REN does not cause Y	1.376455	0.5025	No causality
Y does not cause ED	0.014881	0.9926	No causality
EF does not cause ED	1.339715	0.5118	No causality
ES does not cause ED	0.726559	0.6954	No causality
NRE does not cause ED	4.353396	0.1134	No causality
REN does not cause ED	2.120402	0.3464	No causality
Y does not cause EF	0.575938	0.7498	No causality
ED does not cause EF	9.382731	0.0092 ***	ED → EF
ES does not cause EF	5.824076	0.0544 *	ES → EF
NRE does not cause EF	8.100948	0.0174 **	NRE → EF
REN does not cause EF	1.637823	0.4409	No causality
Y does not cause ES	1.130843	0.5681	No causality
ED does not cause ES	5.247353	0.0725 *	ED → ES
EF does not cause ES	3.397167	0.1829	No causality
NRE does not cause ES	1.818875	0.4028	No causality
REN does not cause ES	0.964218	0.6175	No causality
Y does not cause NRE	2.784177	0.2486	No causality
ED does not cause NRE	9.625051	0.0081 ***	ED → NRE
EF does not cause NRE	4.175221	0.1240	No causality
ES does not cause NRE	0.261631	0.8774	No causality
REN does not cause NRE	3.238860	0.1980	No causality
Y does not cause REN	4.161757	0.1248	No causality
ED does not cause REN	7.159279	0.0279 **	ED → REN
EF does not cause REN	13.17416	0.0014 ***	EF → REN
ES does not cause REN	7.022401	0.0299 **	ES → REN
NRE does not cause REN	5.329011	0.1696	No causality

Note: ***, **, and * imply significance at 1%, 5%, and 10%, respectively.

below show a unidirectional causality from energy depletion (ED), energy efficiency (EF), and non-renewable energy (NRE) to the real GDP per capita (Y) but no causality between renewable energy consumption and energy security to the real GDP per capita (Y). Specifically, energy depletion (ED) causes economic growth (Y) in South Africa, and this is statistically significance at 1%. This is in contrast to the findings by Chen et al. (2023), that energy depletion is inversely related to economic growth and does not cause economic growth in the ten most energy-efficient countries.

Regarding non-renewable energy consumption (NRE), a one-way causality from non-renewable consumption (NRE) to the real GDP per capita (Y) (statistically significant at 1%) exists. This is in line with the study by Ndlovu and Inglesi-Lotz (2020), which found a one-way causality from non-renewable energy consumption to economic growth in Brazil and South Africa. The results also showed no causality between renewable energy consumption and the real GDP per capita (Y). To a greater extent, this makes sense, since, until recently, the proportion of renewable energy use in comparison with other forms of energy, such as non-renewable energy (coal), has been historically low and contributes less towards the real GDP per capita in South Africa. NRE contributes more towards economic growth, and the one-way causal relationship between non-renewable energy consumption and economic growth (NRE causes economic growth) supports the growth hypothesis. With the inclusion of energy efficiency, a one-way causality exists between energy efficiency and economic growth (statistically significant at 10%). This aligns with the study by Bayar and Gavriletea (2019), which found a unidirectional causality from energy efficiency to economic growth. The results also show no causal link between energy security and economic growth. This goes against the findings of Banna et al. (2023), who indicated that energy security and economic growth are intertwined.

Also, their study found a one-way causality from renewable energy consumption to economic growth, which is in contrast with results from this study that renewable energy consumption does not significantly cause economic growth (real GDP per capita). The neutrality hypothesis is also upheld because there is no causal relationship between renewable energy consumption and economic growth. According to the neutrality hypothesis, the decrease in energy consumption does not hurt economic growth. Consequently, energy efficiency has no impact on the causality (Menegaki & Tsani, 2018). However, this is against the study by Akram et al. (2021) in BRICS countries, which found that the panel causality test supported the feedback hypothesis between economic growth and energy efficiency. The results showed that renewable energy consumption and economic growth have a bidirectional causality. Additionally, the results demonstrated a causality from energy efficiency to renewable energy consumption.

The results further show that the null hypotheses that ED and ES do not cause EF were rejected at the 1% and 10% levels of significance, respectively, implying a unidirectional causality running from ED and ES to EF in South Africa. Energy depletion causes energy efficiency. This means that energy-depleting sources of fossil fuels in South Africa cause energy efficiency. Therefore, South Africa continues to move from conventional energy production and consumption and is restructuring its energy sector, which promotes energy efficiency, building institutional and human capacity and encouraging innovation.

Lastly, the results from Table 5 show that in South Africa the null hypotheses, ED does not cause NRE, REN, and ES, were rejected at the 1%, 5%, and 10% levels of significance, respectively, implying that there is a unidirectional causality running from ED to NRE, ED to REN, and ED to ES in South Africa. Finally, the null hypotheses, EF does not cause REN and ES does not cause REN, were rejected at the 1%, and 5% levels of significance, respectively, implying that there is a unidirectional causality running from EF to REN and ES to EF in South Africa. The results indicate no causality between NRE and REN. According to Holechek et al. (2022), this suggests that replacing fossil energy consumption with renewable energy may be challenging. Also, there is no causality between ES and NRE. While some studies, like Tugcu and Menegaki (2024) and Cergibozan (2022), suggest a unidirectional causal link from renewable energy to energy security, others, like Pang et al. (2024) and K. Khan et al. (2023), indicate that relying on non-renewable energy can lead to increased energy security risks, particularly due to the dependence on fossil fuels and geopolitical instability.

Brooks (2019) suggests the importance and the extensive use of post-estimation/diagnostic tests to ensure the statistical adequacy of the model. The most relevant post-estimation tests for multivariate models are the serial correlation test (using the LM test), normally distributed residuals (Jarque–Bera test), and conditional heteroskedasticity (White test). The results presented below (

Table 6. Robustness check.

Test	Null Hypothesis	t-Statics	Probability Value
LM test (2)	No serial correlation	1.277755	0.2205
Jarque–Bera test	Normally distributed residuals	4.440425	0.1086
White test (no cross terms)	No conditional heteroskedasticity	517.1095	0.3335

Rule of thumb: reject the null hypothesis if the probability value is less than 0.05.

) suggest that there is no serial correlation since the probability value from the LM-test (2) is 0.2205. Also, residuals are normally distributed since the probability value for the Jarque–Bera test is 0.1086. Furthermore, the results suggest that there is no conditional heteroskedasticity as the probability value for the White test (without cross terms) is 0.3335.

4.4. Further Discussion

We have, to this point, illustrated a unidirectional causality between energy efficiency, non-renewable energy consumption, and economic growth and no causality between renewable energy consumption, energy depletion, energy security, and economic growth. Here, we further consider the discussion of the results according to the real situation, in particular, regarding possible revisions to existing domestic energy efficiency strategies, changes in the energy supply, security, and economic growth strategies in South Africa.

It is argued that, regarding the unidirectional causality from energy depletion (ED) to the real GDP per capita(Y), energy depletion causes energy efficiency and less energy security and energy consumption (NRE and REN). This implies that energy depletion and the resulting energy security crisis in South Africa are significant drivers for increased energy efficiency measures and economic growth. This can offer opportunities for sustainable economic development and a reduced reliance on fossil fuels. The government may fast-track the implemented policies to promote renewable energy, such as the Renewable Energy Independent Power Producer Procurement Programme and enhance the resilience of the energy system while reducing environmental damage. Also, based on the results that energy efficiency (EF) causes economic growth and the relationship between energy efficiency and renewable energy, the calls for the government to have clear and updated energy regulations, like an Integrated Resource Plan, Integrated Energy Plan, and a Gas Utilisation Master Plan, to lower the difficulties that impede continuous investment in energy efficiency are encouraged. Also, the mandated energy efficiency institutions just cover a small portion of the economy’s total energy use.

Results show that non-renewable energy (NRE) causes economic growth in South Africa; to a greater extent, this makes sense since—until recently, the proportion of renewable energy use in comparison with other forms of energy, such as non-renewable energy (coal), has been historically low and has less of a contribution towards the economic growth in South Africa. Non-renewable energy (NRE) contributes greatly towards economic growth, and the one-way causal relationship between non-renewable energy consumption and economic growth (NRE causes economic growth) support the growth hypothesis. However, the results show no causality between NRE and REN and no causality between renewable energy consumption and energy security to the real GDP growth rate (Y). This suggests that the replacement of non-renewable energy, such as coal, with renewable energy may be difficult in South Africa. This is owing to the advantages of fossil fuels, such as coal, including a higher energy return of the energy invested, ease of transport and distribution, and ease of combustion, although these are not necessarily independent of each other. In particular, fossil fuels exert a strong influence on the transportation and energy sectors in South Africa. Coal contributes 82% of the primary energy supply according to the Department of Mineral Resources and Energy (2023). In the light of these overwhelming advantages, it seems misleading to suggest that non-renewable energy can easily be replaced by other primary energy sources without harming economic growth. Instead, more should be done to integrate technology towards a diversified energy mix.

Lastly, results show that there is no causality between energy security and economic growth, but this does not negate the importance of energy security; the energy crisis in South Africa has demonstrably impacted economic growth. We do, however, acknowledge the full complexity of any economic system and that our analysis and discussion are ultimately based on the availability of data.

5. Conclusions

This study investigated the direction of the causality between energy efficiency, energy consumption, and economic growth in South Africa. To determine if a long-run relationship between the variables exists, we relied on the Johanson cointegration test, and the results indicate that there is a long-run relationship between economic growth, energy depletion, energy efficiency, non-renewable energy consumption, renewable energy consumption, and energy security.

This study finds a unidirectional causality between energy efficiency, non-renewable energy consumption, and economic growth and no causality between renewable energy consumption, energy depletion, energy security, and economic growth. The growth hypothesis is supported, while the neutrality hypothesis is also supported but only with regard to renewable energy consumption and economic growth. Also, the unidirectional Granger causality between non-renewable consumption to economic growth and energy efficiency to economic growth exists. We relied on the Toda–Yamamoto causality (Block Exogeneity Wald) test.

The causal relationships between energy efficiency, energy consumption, and economic growth in South Africa suggest that energy efficiency improvements can drive long-term economic growth, while energy consumption remains a key determinant to industrial and economic activities. Policymakers must focus on a balanced approach that promotes efficiency while ensuring energy availability for sustained development. These results can bolster the proposal of policy implications of some significance in South Africa. The following policy implications are suggested:

• Energy efficiency is a significant tool to enhance sustainable growth and attain climate objectives. Energy efficiency helps to lower the costs of mitigating emissions, speed progress towards net-zero targets, and further advance both social and economic development.
• Energy efficiency offers room for energy savings prospects and carbon emission reductions, which remain mostly unexploited in the economy. In addition, South Africa has yet to successfully address the current energy crisis and electricity supply problems, and impressing energy efficiency can partially address the problem and lower energy usage and demand.
• Re-considering energy efficiency as a political priority would strengthen the Ministry of Electricity and Energy’s efforts.
• The South Africa’s National Energy Efficiency Strategy of 2016 acknowledges the fostering of energy efficiency as an engine to enhance the balanced, socially coherent, and ecologically sustainable growth of the economy; the findings may aid as a more widespread demand on the government to have clear, consistent, and updated energy laws and regulations, including an Integrated Resource Plan, Integrated Energy Plan, and a Gas Utilisation Master Plan, to reduce the uncertainty which hinders investment in energy efficiency. Furthermore, mandated energy efficiency laws cover just a small portion of South Africa’s total energy use.

In this current study, we conjectured that energy efficiency causes economic growth and energy consumption. Nonetheless, the causal impacts of energy efficiency on economic and energy consumption may be asymmetric due to non-linear deviations. During the course of the preparation and implementation of this study, we faced significant constraints and limitations. Firstly, the data set spans from 1985 to 2021, and the availability of data and the timeline in South Africa posed a significant constraint. Existing databases have a constrained breadth and are seldomly updated, limiting our ability to undertake a more complete and in-depth investigation. Furthermore, the rapidly and continuously changing characteristics of the energy sector in South Africa are complicated by a deficiency of consistent policies, making it impossible to identify long-term patterns or accurately anticipate future changes. This means that the impact of energy efficiency policies might not be proportional to the amount of energy saved. For example, the initial gains from energy efficiency might be larger than the gains later on, or the impact might be different at different levels of energy consumption and economic growth. Hence, the suggestion is that the forthcoming research may use asymmetrical non-linear causal links in its investigation. Lastly, the results show no causality between energy security and economic growth, which realistically does not, however, negate the importance of energy security, as the energy crisis in South Africa has demonstrably impacted economic growth. Therefore, we do, however, acknowledge the full difficulty of any economic climate and that our analysis and discussion are ultimately based on the availability of data. Forthcoming research may use asymmetrical, non-linear causal links in its investigation.