Renewable and Non-Renewable Energy Consumption on Economic Growth: Evidence from Asymmetric Analysis across Countries Connected to Eastern Africa Power Pool

Abstract

Though various studies have examined the energy-growth nexus, the non-linear asymmetry relationship between economic growth and energy use has received little attention. In order to investigate the stratified asymmetric relationship between total, renewable, and nonrenewable energy consumption and economic growth in nine Eastern African nations connected via the Eastern Africa power pool (EAPP) and power trade (EAPT), this study used common correlated effect means group (CCEMG), nonlinear Autoregressive Distributed Lags (NARDL) approaches, and asymmetric causality tests from both a country and regional perspective. The time span is 1980 to 2017. The results from cross-sectional dependence confirms the existence of cross-sectionally dependence, findings from unit root and westerlund cointegration agreed the presence of long-run relations between variables and support the use of NARDL. CCEMG results reveal that energy in total contributes to growth, while nonrenewable energy reduces economic growth across the panel of selected countries. Nonlinear results from positive and negative shocks to energy as total, renewable and nonrenewable energy use have mixed nonlinear effect (positive and negative) on economic growth in long-run across the selected countries, while symmetric effect was unavailable in several countries. Bi-directional causation was noted between growth and all considered energy types at the panel of seven countries, and between energy as total and nonrenewable and growth in Ethiopia and Sudan, while all of the examined nations showed a strong one-way causal relationship between growth and renewable and nonrenewable energy, Rwanda showed a neutral relationship between growth and energy sources. For sustainable economic growth, policymakers, investors, and government officials may use this information to help them develop energy policies that promote renewable energy output while reducing reliance on nonrenewable energy in the region.

1. Introduction

Energy consumption not only delivered sustainable development but also establishing environmental relief globally and regionally through its role in economic growth during the previous two decades. Renewable energy projects in Africa commit to speeding energy access with the target of eradicating energy poverty and securing substantial financial support from cooperative development banks by 2020 [1]. African countries engaged in cross-border power trade, which received significant attention for more than two decades and led to the creation of regional power pools including Southern, Western, Central, and Eastern Africa power pools [2]. By the World Bank report, the existence of these power pool not only decreased the share of energy demand at the regional markets but also reduced the dependence on nonrenewable energy use. As a result, the Southern Africa power pool (SAPP) is currently the most liquid market and met 3.5% of energy share demand, while the Eastern Africa power pool (EAPP) is the least and met about 1% or 1.5% at best for the year [2]. Thus, there is a need to increase the energy and income generation from EAPP and the associated energy market for inter-connected. However, for establishing energy policies or enhancing the structure of income creation, an examination of the influence of energy consumption and economic growth in terms of country membership is critical.

Lack of knowledge regarding the historical effects of energy produced by renewable energy projects and nonrenewable energy on economic growth may cause irregular changes in the amount invested in the energy sector and the amount of money made from energy consumption in the countries linked by the Eastern African Power Trade (EAPT) and the Eastern African Power Pool (EAPP), which are the subject of this study. This fact led some research to claim that the cumulative trend of energy use, including both renewable and nonrenewable sources, will cause environmental deterioration and hence slow down economic growth [3,4,5,6]. In this context, the negative side effect of environmental externalities may lead to a decrease in growth through energy consumption. Other research has found that the impact of energy use on growth varies depending on structural economic variations [7,8,9]. Thus, examining the impact of renewable and nonrenewable energies, separately on economic growth can add contribution to the literature and assist policymakers to establish policies that can reflect on how to each energy types in the interest of economic growth and environmental quality [10,11,12,13]. Squalli [8] argued that due to an economy that is growing, and extreme energy directed to the nonproductive sectors joined with an insufficient energy supply; the economy’s production structure is projected to shift to less energy-intensive industries. In this respect, energy consumption may either negatively or positively affect economic growth, as argued by [14]. Due to the influence of external factors, there is a case where energy generation can be increased but negatively affect growth, and lead to nonlinear linkage between energy use and economic growth. Research on the nonlinear link between energy use and economic growth, however, is required at the national and regional levels [15,16,17,18,19,20].

The establishment of EAPP, in conjunction with hydropower and hydroelectricity facilities built in Ethiopia, the Democratic Republic of Congo (DRC), and Tanzania as regional renewable energy projects, addressed the initial confusion in energy delivery and economic growth [21,22,23]. On the other hand, Ozturk and Bilgili [24] argued that the historical dependence on biomass increases income generation, 1% increase in biomass leads to a 1.818% increase in the gross domestic product (GDP). Fortunately, about 68% of total energy consumption in the country members of EAPP and EAPT are wood biomass, see [25,26] in Rwanda; in Ethiopia [27]; in Tanzania [28]; in Egypt [29]; and others, see [30]; and about 20% is petroleum, and 10% is renewable energy in some countries, such as Kenya. But other energy sources, including those that aren’t renewable and those that come from wood biomass, harm the environment by creating deforestation, which has a detrimental impact on economic growth [31].

According to The World Bank, countries connected at EAPP and EAPT are grouped into low- and lower and middle-income categories. However, beyond the main input of GDP (labor, and capital); energy (renewable and nonrenewable) consumption received a considerable share to affect economic growth [32,33]. In this respect, the countries connected at EAPP and EAPT are increasingly gaining in energy production, such as renewable energy development through new inter-connected projects and committed connections [34]. By this effort, ref. [35] argued that the investment in renewable energy projects will lead to significant economic growth and reduce the menaces from emissions in the region.

This study aims to examine the asymmetric (positive or negative) effects of nonrenewable and renewable energy sources, as well as total energy consumption, on economic growth in the countries linked by the EAPP and EAPT. This study aims to examine the asymmetric (positive or negative) effects of nonrenewable and renewable energy sources, as well as total energy consumption, on economic growth in the countries linked by the EAPP and EAPT.

Due to the unavailable historical data of energy and income gained from EAPP and EAPT, and the fact that energy generated from EAPP and other projects combined with existing nonrenewable energy to contribute to the growth of the economy, as argued by Remy and Chattopadhyay [21]; this study uses the available data from U.S energy information administration database [36]. Several studies have examined the relationship between energy and growth in certain nations using this data., although they used estimators, such as ARDL, Granger causality, and others see [32,37,38], which assume the asymmetry among variables and lead to methodological omission. To address this methodological flaw, particularly when the impression of the energy-growth nexus is altered by unobserved factors like environmental factors or structural economic changes [3,7]; nonlinear Autoregressive distributed lags (NARDL) have suggested to offer the thorough study and assist in identifying the covert impact of energy on economic growth [39].

Therefore, our study uses the data available in EIA data (the total data, the renewable data, and nonrenewable energy data) and The World Bank database (GDP) and contributes to filling the following literature gaps: First, very few country-specific studies examined the energy-growth nexus by assuming asymmetry in some country connected at EAPP and EAPT. This is the first study to measure the effects of probable nonlinear asymmetry in total, renewable, and nonrenewable energy consumption on economic development in nine nations using NARDL from 1980 to 2017. The nonlinear (NARDL) proposed by [40] is employed to investigate nonlinear asymmetric interactions between variables, as well as to evaluate and contrast the asymmetric structure’s dimension in terms of positive and negative changes in country-specific variables. Additionally, the asymmetric causality test suggested by [41] is further used.

Second, this study provides a regional perspective on the stratified effects of renewable, nonrenewable, and total energy on economic growth in EAPP and EAPT nations. To make that possible, the most current estimator, panel common correlated effect of means group (CCEMG), was used. It was developed by [42] and modified by [43], and it permits the influence of variability across variables and cross-sectional dependence on the relevant variable.

Few studies on the asymmetric relationship between energy consumption and economic growth have been done [44,45,46,47,48], including [49] in G7 nations, Ndoricimpa [50] in South Africa, and [51] in India. Refs. [52,53] looked at the asymmetry of the wood biomass consumption-growth nexus in many nations, including few countries connected at EAPP and EAPT (Kenya, Uganda, and Djibouti). Despite employing the NARDL estimator, compared with existing studies, the study has two differences. To begin, this research is based on nations that are classed as low, lower, and middle-income, and are linked through EAPP and EAPT initiatives. Furthermore, these countries share the common targets to satisfy the energy demand, energy security, and income generation through a common market, which are working on regional integrated policy planning, development, and energy access projects. Their renewable energy projects started operating from 1980, and various database recorded energies data from 1980 to 2017 across all sampled countries. Second, as proposed by [14], the predicted coefficient on energy consumption might change considerably across two categories of energy usage, such as renewable and nonrenewable. In this respect, the paper analyses evidence at the aggregate level, assesses the stratified impact of renewable and non-renewable energy sources on economic growth, and identifies new factors that might change investment in and income generation in the energy industry.

The study’s second section is about related works. The technique in Section 3 introduces NARDL. The empirical results and debate are presented in Section 4, and some empirical policies are concluded and supplied in Section 5.

2. Relative Studies in the Countries Connected at EAPP and EAPT

2.1. Review on the Total, Renewable, and Nonrenewable Energy-Growth Nexus

Due to the historical use of generated energy before and after the creation of EAPP and EAPT projects, various studies examined the energy-growth nexus in the countries connected to these projects, and findings have briefed within hypotheses (bidirectional, one-way directional, and neutral), see

Table 1. Selected empirical studies in East Africa on the relationship between total, renewable, and nonrenewable energy use and economic growth.

Authors	Country	Variables	Period	Methods	Findings
Odhiambo [23]	Kenya, DRC, and South Africa	EC, Price, and Y	1972–2006	ARDL	$EC\to Y$ in SA and Kenya $Y\to EC$ in DRC
Adams et al. [30]	30 African countries	REC, NREC, and Y	1980–2012	Panel co-integration and panel error correlation test	$REC\to Y$ and $NREC\to Y$
Onuonga [31]	Kenya	EC and Y	1970–2005	Granger causality error correlation model	$Y\to EC$
Kebede [32]	Ethiopia	EC, and Y	1970–2014	ARDL co-integration	$Y\leftrightarrow EC$
Sekantsi and Okot [33]	Uganda	Electricity consumption and Y	1981–2013	ARDL and Granger causality	$Y\leftrightarrow EC$
Atinafu [37]		EC and Y	1970–2017	ARDL and Granger causality	$Y\to EC$
Elfaki et al. [38]	Sudan	EC and Y	1984–2014	ARDL	$Y\to EC$
Bildirici and Ozaksoy [52]	Kenya, Uganda, Djibouti, and others	Wood biomass EC and Y	1980–2012	ARDL, NARDL, Error correlation model	$Y\leftrightarrow EC$ in Kenya, $Y\to EC$ Uganda, and Djibouti
Odhiambo [54]	Tanzania	EC and Y	1971–2006	ARDL bound test	$EC\to Y$
Aïssa et al. [55]	Kenya, Sudan, Egypt, and 9 other countries	REC, Trade, and Y	1980–2008	Error correlation model	Link from REC, Trade to Y
Nyasha et al. [56]		EC and Y	1971–2013	ARDL bound test	$Y\to EC$
Mustapha and Fagge [57]		EC, and Y with trade and Urbanization	1981–2011	Johansen Co-integration	$EC\to Y$

EC: energy consumption, REC: renewable energy consumption, NREC: nonrenewable energy consumption, and Y: economic growth.

. Ref. [54] examined the energy-growth nexus in Tanzania, moreover, the one-directional causal between these variables was found and was moving from energy to growth. In Ethiopia, Atinafu et al. [32,37] proposed that energy and growth had bi-directional causal linkages. The energy-growth nexus has been studied in Kenya [31], Uganda [33], Sudan [38], and elsewhere, with findings revealing the presence of correlations among the variables.

The relationship between growth and renewable energy has not been extensively studied. Ref. [55] conducted a multi-country examination of the impact of using renewable energy on economic growth, including Kenya, Sudan, and Egypt. According to the research, there is no connection between the use of renewable energy and economic growth. Additionally, Adams’s investigation on the nonlinear asymmetric causal relationship between the usage of renewable and nonrenewable energy sources and economic growth in many African nations reveals an asymmetric causative relationship [30]. Overall, the impact of the asymmetrical relationship between the use of renewable and nonrenewable energy on economic growth has not only been studied among EAPP and EAPT national members but has also not been researched at an aggregate level. To illustrate the comprehensive implications to the causal hidden characteristics, it makes appropriate to highlight the likely influence of nonlinear changes in energy consumption on economic development.

2.2. Review on Existing Estimation Approaches

Various studies that explored the energy-growth nexus in the countries connected to EAPP and EAPT have used the conventional estimators, which assume the asymmetric linkage, such as Autoregressive distributed lags (ARDL), panel-cointegrations analysis, Granger causality, and others. In the case of energy-growth nexus, several studies have used ARDL, see [32,37,38] and others. However, Hajko [39] highlighted that employing these approaches has a significant risk of methodological omissions as economic variables may be impacted by both structural economic changes and harmful hidden qualities (environmental externalities). In the context of this study, nonlinear ARDL proposed by [40] accepted a high consideration for detecting the possible nonlinear relationship between variables. This approach is not only used to examine nonlinear asymmetric relationships among variables but also can assess and compare the dimension of the asymmetric structure through dynamic multiplier plots. Only one assumption for NARDL is that all variables should be integrated at zero or one order and a combination of these two orders. As a result, Ndoricimpa [50] used NARDL to investigate the asymmetries in South Africa’s energy consumption and economic growth.

Ref. [30] have also used panel cointegrations (DOLS, FMOLS) and ECM study the relationship between growth and renewable and non-renewable energy in 30 countries of Africa, including some countries connected at EAPP and EAPT; and Aïssa et al. [55] has used ECM to assess the renewable energy-growth in multi-countries, include three countries connected at EAPP and EAPT. Furthermore, using the panel cointegrations approach, such as DOLS and FMOLS belong to the first-generation estimators, can sometimes provide inconsistency results or misleading information, because they are resistant to potential cross-country cross-sectional reliance [58]. To our knowledge, only a few researches have employed NARDL to investigate the asymmetric nexus of energy (total, renewable, and nonrenewable) consumption on economic development., more specifically, this is the first study that investigates the nonlinear link between the selected variables in the countries connected at EAPP and EAPT. Furthermore, for the case of our study, the second-generation estimators, such as the common correlation effect of means group (CCEMG), developed by Chudik [43] and Pesaran after Pesaran’s first proposal [42], which considers the impact of cross-sectional dependency and heterogeneity, have not been used yet.

3. Materials and Methods

3.1. Methodology

3.1.1. Mathematical Model

For nine countries connected with the EAPP and EAPT, we conducted country-specific and panel studies to examine the impact of total, renewable, and non-renewable energy consumption on economic development. To properly investigate the contribution of total, renewable, and non-renewable energy to economic growth while maintaining the generality of the production function, labor and capital are utilized as control variables, resulting in the mathematical model:

$$y_{it}=f(L_{it}, K_{it},E{C}_{it})$$

(1)

For $i=1,2,\dots N$ represent the country, $t=1,2,\dots T$ time, $y_{it}$ is economic growth, $E{C}_{it}$ is energy consumption (total, renewable and non-renewable), $L_{it}$ is Labor, and $K_{it}$ is the Capital. Thus, the multivariate regression model for the sampled countries can be written as follow:

$$y_{it}={\alpha }_{0i}+{\alpha }_{1i}E{C}_{it}+{\alpha }_{2i}RE{C}_{it}+{\alpha }_{3i}NRE{C}_{it}+{\alpha }_{4i}{L}_{it}+{\alpha }_{5i}{K}_{it}+u_{it}$$

(2)

For ${\alpha }_{0i}$ is an unobserved country fixed effect, ${\alpha }_1-{\alpha }_5$ are the equilibrium coefficients in the long term, and $u_{it}$ the error term. For the country-specific, Equation (2) can be written as follow:

$$y_t={\alpha }_0+{\alpha }_{1}E{C}_t+{\alpha }_{2}RE{C}_t+{\alpha }_{3}NRE{C}_t+{\alpha }_4{L}_t+{\alpha }_5{K}_t+u_t$$

(3)

3.1.2. Cross-Sectional Dependence Test

Ref. [59] claimed that disregarding cross-sectional dependency in panel data analysis would result in erroneous estimates and misleading information. For identifying cross-sectional dependency, Pesaran [60] introduced Pesaran CD and standardized Lagrange Multiplier (LM) tests, while Breusch and Pagan [61] proposed the Breusch-Pagan LM test. Pasaran’s suggested standardized test has the capacity to handle huge panel data sizes N and time T, and may be approximated as follows:

$$LM=\sqrt{\frac{1}{N\left(N-1\right)}}{{\displaystyle \sum }}_{i=1}^{N-1}{{\displaystyle \sum }}_{j=i+1}^N\left(T_{ij}{\mu }_{ij}^2-1\right)\to N\left(0,1\right)$$

(4)

$$CD=\sqrt{\raisebox{1ex}{$2$}\!\left/ \!\raisebox{-1ex}{$N\left(N-1\right)$}\right.}{{\displaystyle \sum }}_{i=1}^{N-1}{{\displaystyle \sum }}_{j=i+1}^N{T}_{ij}{\mu }_{ij}^2\to N\left(0,1\right)$$

(5)

The Breusch-pagan LM test is efficient for small size and T, and may be approximated as follows:

$$LM={{\displaystyle \sum }}_{i=1}^{N-1}{{\displaystyle \sum }}_{j=i+1}^N{T}_{ij}{\mu }_{ij}^2\to {\chi }^2\left(\raisebox{1ex}{$N\left(N-1\right)$}\!\left/ \!\raisebox{-1ex}{$2$}\right.\right)$$

(6)

where Equation (4) is used for big size and variable time T, while Equation (5) is used for large N and constant T.

For ${\mu }_{ij}^2$ is the correlation coefficients obtained from the residuals of the Equation (3), can be determined in the following way for:

$${\mu }_{ij}={\mu }_{ji}=\frac{{\sum }_{t-1}^T{\epsilon }_{ij}{\epsilon }_{ji}}{{({\sum }_{t-1}^T{\epsilon }_{ij}^2)}^{\frac{1}{2}}{({\sum }_{t-1}^T{\epsilon }_{jt}^2)}^{1/2}}$$

(7)

where ${\epsilon }_{ij} \mathrm{and} {\epsilon }_{ji}$ are standard errors.

3.1.3. Panel Pesaran CIPS Unit Root Test

By evaluating at the averages of lagged levels and differences for each unit, the Pesaran CIPS panel unit root test suggested by [62] allows for cross-sectional dependency. Cross-sectionally augmented Dickey-Fuller is the name for this method, which can be represented as follows:

$$\Delta y_{it}={\psi }_i+{\alpha }_i{y}_{i,t-1}+{\beta }_i{\overline{y}}_{t-1}+{{\displaystyle \sum }}_{j=0}^p{d}_{ij}\Delta {\overline{y}}_{t-j}+{{\displaystyle \sum }}_{j=1}^p{\xi }_{ij}\Delta y_{i,t-j}+u_{it}$$

(8)

For ${\overline{y}}_{t-1}$ and $\Delta {\overline{y}}_{t-j}$ are the lagged levels’ cross-sectional averages and the initial difference, respectively. The CIPS statistic was estimated using cross-sectionally augmented Dickey-Fuller (CADF) statistics in the following expression:

$$CIPS=\frac{1}{N}{{\displaystyle \sum }}_{i=1}^{N}CAD{F}_i$$

(9)

3.1.4. Unit Root Tests

In this study, the impact of total, renewable, and non-renewable energy consumption on growth has been examined, considering country-specific factors, therefore, it is important to detect the unit root in the selected variables. Under the null hypothesis that the variable has a unit root, Dickey and Fuller (ADF) [63] introduced the unit root test, and (pp) [64] are used. The ADF and PP test results are frequently compared to the (KPSS) [65] unit root test to check if the conclusions are the same. The following is the model:

$$\Delta y_t=\psi y_{t-1}+{\displaystyle \sum _{i=1}^p{\varphi }_i}\Delta y_{t-i}+u_t$$

(10)

For $\psi =1$ (null hypothesis by using DF test), ${\varphi }_i=1$, $i=1,2,\dots ,p$ unit root at maximum lags (p) by using ADF and PP test, $\Delta$ indicates the differencing operator, and $u$ the error term.

3.1.5. Panel Cointegration Test

The error correction panel cointegration test, which is effective for cross-sectional dependency by inserting an error correction term, was utilized in this work (ECT). Westerlund proposed the error correction panel cointegration test [66]. It is expressed as follows:

$$\Delta z_{it}={\acute{\alpha }}_i{d}_i+{\vartheta }_i\left(z_{i\left(t-1\right)}+{\acute{\pi }}_i{y}_{i\left(t-1\right)}\right)+{{\displaystyle \sum }}_{j=1}^m{\varphi }_{ij}\Delta z_{i\left(t-1\right)}+{{\displaystyle \sum }}_{j=0}^m{\phi }_{ij}\Delta y_{i\left(t-1\right)}+{\omega }_{it}$$

(11)

For ${\vartheta }_i$ is the adjustment term, $d_i$ is a vector of deterministic components, while other parameters introduce the nuisance in the variable of interest. Therefore, referred to the estimates of ${\vartheta }_i$, the statistics of Westerlund ECT based panel cointegration tests can be determined as follows:

$$G_{\tau }=\frac{1}{N}{{\displaystyle \sum }}_{i=1}^N\frac{{\vartheta }_i}{SE({\acute{\vartheta }}_i)}$$

(12)

$$G_{\alpha }=\frac{1}{N}{{\displaystyle \sum }}_{i=1}^N\frac{T{\vartheta }_i}{{\acute{\vartheta }}_i\left(1\right)}$$

(13)

where $G_{\tau }$ and $G_{\alpha }$ are group mean statistics and judges the null hypothesis of no presence of cointegration in the cross-sectional panel. The presence of cointegration for at least one cross-sectional unit in the panel shows rejection of this hypothesis; the panel statistic may be computed as follows:

$$P_{\tau }=\frac{{\widehat{\vartheta }}_i}{SE({\widehat{\vartheta }}_i)}$$

(14)

$$P_{\alpha }=T{\widehat{\vartheta }}_i$$

(15)

The null hypothesis is rejected, which shows that the entire panel does not exhibit cointegration.

3.1.6. Panel Common Correlated Effects Mean Group Estimator

The panel CCEMG developed by [42] and extended by [43] was used in the study since the panel data is more likely to exhibit cross-sectional dependency across nations. This estimator considers the possibility of cross-sectional dependency and heterogeneity, and respond to multicollinearity by using partialled strategy out during the estimation procedures and estimates their impact on the variables of interest, as follows:

$$y_{it}={\alpha }_i+{\displaystyle \sum }_{l=0}^p{\beta }_{il}{y}_{it-l}+{\displaystyle \sum }_{l=0}^q{\delta }_{il}{x}_{it-l}+{\displaystyle \sum }_{l=0}^Z{\mu }_{il}{\overline{z}}_{it-l}+u_{it}$$

(16)

where ${\overline{z}}_t={\left({\overline{y}}_t,{\overline{x}}_t\right)}^{\prime }$, ${\overline{y}}_t=n^{-1}\sum _i^N{y}_t$ and ${\overline{x}}_t=n^{-1}\sum _i^N{x}_t$, for (p, q, z) are the lags (l).

For $i=1,2,\dots .,N$, ${\alpha }_i$ and $u_{it}$ are intercept and error terms, respectively. The linear combinations of the cross-sectional averages of the variable of interest and regressors, which common consequences can be seen, are used in the CCEMG estimator, using coefficients from [67].

3.1.7. The NARDL Estimator

One of the objectives of the study is to determine if growth and overall, renewable and nonrenewable energy consumption have a nonlinear connection (short and long-run asymmetry). Nonlinear ARDL is expanded by [68] from the ARDL used in this situation. The NARDL identifies both short- and long-run nonlinearities effect of explanatory variables to the variable of interest and describe the asymmetric dynamic multipliers that geometrically show the cross between the long- and the short-run. Furthermore, for small sample size data, this technique efficiently illustrates short- and long-run patterns, and able to be used for variables co-integrated at one or zero [I(1) or I(0)], and combination of these orders [68]. To identify the link between the variable of interest and explanatory variables, the input variables of Equation (3) checked whether they are co-integrated either at zero-order or one, see (Section 3.1.4). Hatemi-j [41] proposed the technique for decomposing explanatory variables into positive and negative changes. Thus, employing the negative and positive changes as new variables and using the error-correction term, lead to the visible nonlinear effects and their dynamic multipliers (asymmetrical presentation), which is checked by using the test based on statistical inference in vector autoregressive [69]. Therefore, Equation (3) can be written as follow:

$$\begin{gathered} \Delta \ln \left(y_t\right)={\delta }_0+{\sum }_{l=1}^p{\rho }_1\Delta ln{y}_{t-l}+{\sum }_{l=0}^q{\beta }_1^{+}\Delta ln{x}_{t-l}^{+}+{\sum }_{l=0}^q{\beta }_1^{-}\Delta ln{x}_{t-l}^{-}+{\sum }_{l=0}^q{\beta }_2^{+}ln{x}_{t-l}^{+}+{\sum }_{l=0}^q{\beta }_2^{-}ln{x}_{t-l}^{-}+ \\ {\rho }_2\ln y_{t-1}+{\theta }_1^{+}\ln x_{t-1}^{+}+{\theta }_2^{-}\ln x_{t-1}^{-}+m_t \end{gathered}$$

(17)

where the coefficients of this equation are the combination of long- and short-run coefficients, and long-run adjustment rates; $y_t$ is the GDP per capita, $x_t$ indicates the labor, capital, and energy consumption (total, renewable, and non-renewable) in Equation (3).

3.1.8. Causality Test

In this study, the asymmetric causality test was applied, which takes into account any asymmetries by summing up positive and negative changes in underlying variables. This test, which may be stated as follows, is used in this study to look for asymmetric causation between economic development, total energy consumption, and the usage of renewable energy at the regional and global levels across income groups.

$$Wald={\left(cb\right)}^{\prime }\left[c\left(z^{\prime }z{)}^{-1}\otimes S_u\right)c{)}^{-1}\right]\left(cb\right)$$

(18)

$b$ is the vector representation of matrix coefficients of vector autoregressive (VAR) model, c is a $p\times n\left(1+np\right)$ indicator matrix with elements ones for constrained parameters and zeroes for the remaining parameters, $\otimes$ indicates the Kronecker product, and $S_u$ is the variance-covariance matrix of the VAR model, for detail, see [41] for more information.

3.2. Data

The relationship between total, renewable, and nonrenewable energy use and national growth and panel data of countries linked to EAPP and EAPT were examined employing time-varying data extracted from the databases of the World Bank and the US Energy Information Administration time-varying data extracted from the databases of the World Bank and the US Energy Information Administration [36]. (see the list in the Appendix A). Total renewable and nonrenewable energy consumption measured in quadrillion Btu and converted to kilogram(kg) of oil equivalent mined from the EIA, GDP (in constant 2010 US dollars) was used to measure economic growth, and the control variables employed are labor and capital. By dividing the yearly total population and applying the natural logarithm, the data were transformed to per capita values in order to produce a valid study. After all transformation, the selected variables were noted as follows: Total energy denoted as lnEN, renewable energy denoted as lnREN, nonrenewable energy denoted as lnNREN, capital denoted as lnK, and labor denoted as lnL.

Table 2. Descriptive statistics.

Country	Variables	Mean	Median	Maximum	Minimum	Observations
Burundi	lnGDP	2.414911	2.3799	2.527089	2.330696	38
	lnEN	1.242604	1.203814	1.453165	0.959908	38
	lnREN	0.602462	0.661007	0.752427	−0.07195	38
	lnREN	1.120452	1.048054	1.356696	0.856756	38
	lnL	−0.25723	−0.24889	−0.20959	−0.31042	38
Rwanda	lnGDP	0.427906	0.409156	0.72953	0.025087	38
	lnEN	−0.70058	−0.6724	−0.55515	−0.99402	38
	lnREN	−1.52286	−1.4769	−1.23288	−2.00983	38
	lnNREN	−0.7754	−0.7462	−0.62926	−1.19035	38
	lnK	−0.58278	−0.54923	−0.2334	−1.33501	38
	lnL	−2.46679	−2.46359	−2.41668	−2.53735	38
Uganda	lnGDP	2.743079	2.728778	2.95917	2.555669	38
	lnEN	1.558126	1.513388	1.824908	1.346607	38
	lnREN	1.123968	1.103224	1.367156	0.934682	38
	lnNREN	1.352566	1.296169	1.643233	1.110025	38
	lnK	2.003425	2.038926	2.296768	1.758725	38
	lnL	−0.30522	−0.30427	−0.28533	−0.31931	38
Kenya	lnGDP	2.953017	2.941103	3.063118	2.908464	38
	lnEN	2.114068	2.108665	2.214393	2.044065	38
	lnREN	1.478224	1.485208	1.620047	1.23363	38
	lnNREN	1.997074	1.992483	2.125412	1.89411	38
	lnK	2.010442	2.003707	2.19729	1.835265	38
	lnL	−0.27611	−0.27188	−0.23751	−0.31809	38
Tanzania	lnGDP	2.767678	2.742504	2.958976	2.665763	38
	lnEN	1.760297	1.728407	1.991711	1.588567	38
	lnREN	1.1398	1.14451	1.295736	0.940289	38
	lnNREN	1.628477	1.579389	1.943476	1.428206	38
	lnL	−0.19567	−0.1979	−0.1834	−0.20835	38
Ethiopia	lnGDP	2.393282	2.34166	2.739113	2.215734	38
	lnEN	1.487247	1.445094	1.884188	1.221362	38
	lnREN	0.912055	0.816717	1.475397	0.547298	38
	lnNREN	1.344155	1.338781	1.669426	1.003063	38
	lnL	−0.22748	−0.22881	−0.20947	−0.24426	38
Egypt	lnGDP	3.2709	3.269952	3.449836	3.042303	38
	lnEN	2.880352	2.859502	3.042183	2.618884	38
	lnREN	1.641184	1.622005	1.775799	1.571949	38
	lnNREN	2.852591	2.82866	3.026105	2.553172	38
	lnL	−0.4292	−0.436	−0.39349	−0.46237	38
	lnK	2.507552	2.489465	2.757632	2.259573	38
Sudan	lnGDP	3.108232	3.068222	3.278775	2.959064	38
	lnEN	2.068141	1.993109	2.371552	1.819885	38
	lnREN	1.21556	1.098499	1.752485	0.953638	38
	lnNREN	1.996725	1.934184	2.27393	1.746952	38
	lnL	−0.42659	−0.4293	−0.39503	−0.44962	38
	lnK	2.305357	2.321774	2.711999	1.851259	38
DRC	lnGDP	2.655823	2.601255	2.926601	2.440997	38
	lnEN	1.742462	1.720585	1.93306	1.566422	38
	lnREN	1.526374	1.492315	1.673456	1.415994	38
	lnNREN	1.312114	1.385171	1.635607	0.878733	38
	lnL	−0.29376	−0.29087	−0.26645	−0.32292	38
Regional level	lnGDP	2.526092	2.71466	3.449836	0.025087	342
	lnEN	1.572524	1.704322	3.042183	−0.99402	342
	lnREN	0.901863	1.132654	1.775799	−2.00983	342
	lnNREN	1.425418	1.52303	3.026105	−1.19035	342
	lnL	−0.54201	−0.29678	−0.1834	−2.53735	342
	lnK	1.740628	2.063848	4.009741	−1.33501	190

displays the descriptive statistics for each variable in terms of mean, median, maximum and minimum, and show the observation of variable across the countries.

4. Results and Discussion

4.1. Cross-Sectional Dependence and Panel Unit Root Test Results

Table 3. Cross-sectional dependence test results.

	Breusch-LM	Pesaran LM	Pesaran CD
lnGDP	726.492 *	81.375 *	11.405 *
lnEN	455.679 *	49.459 *	49.338 *
lnREN	265.370 *	27.031 *	−1.130
lnNREN	421.190 *	45.395 *	4.296 *
lnL	369.283 *	39.062 *	3.631 *
lnK	239.021 *	26.736 *	2.329

* indicates 1% significant level.

displays the results of the cross-sectional dependence tests performed by Breusch LM [60,61]. For all selected variables, the null hypothesis of no cross-sectional independence is rejected at a 1% significance level, showing that cross-sectional dependency exists in the panel of nine nations connected at EAPP and EAPT. This implies that methods can handle cross-sectional dependence should be used to examine the link between variables.

Table 4. Panel Pesaran CIPS unit root test results.

	Levels		1st Difference		Order
Variable	C	C-T	C	C-T
lnGDP	−0.525	−2.195	−4.949 *	−4.86 3 *	I(1)
lnEN	−1.431	−2.878 **	−9.391 *	−8.874 *	I(0)
lnREN	−1.687	−2.207	−9.913 *	−9.071 *	I(1)
lnNREN	−1.482	−2.841 **	−8.759 *	−7.634 *	I(0)
lnL	−1.428	−2.182	−5.724 *	−5.487 *	I(1)
lnK	−2.172	−3.834 **	−7.478 *	−6.831 *	I(0)

C: constant, C-T: constant, and trends, * and ** indicate significant levels at 1%, 5%, and 10%.

displays the outcomes of the CIPS panel unit root test after accounting for and ignoring cross-sectional panel dependency. Based on this Table 4, the unit root null hypothesis is rejected at the first difference. It denotes that the variables are integrated at zero and one order, I(0) or I(1), respectively, and implies that there is possibility of having equilibrium relationships between variables.

Table 5. Westerlund ECT Panel cointegration test results.

Dependent	Test	Gt	Gα	Pt	Pα
lnGDP	Statistic	−1.694	−4.575	−8.761 *	−11.215 *
	z-value	0.044	1.554	−3.670	−3.299
	Variance ratio	−2.182 **

Ho: no cointegration, and H1: All panel are cointegrated, * indicates significant levels at 1%, ** indicates significant levels at 5%.

displays the outcomes of the Westerlund panel cointegration test [66] for the panel of chosen countries. The null hypothesis of no cointegration in the panel was rejected by this Westerlund test statistic, confirming the long-run cointegration connections among the chosen variables by adopting the premise that all panels are cointegrated. This implies existence of equilibrium relationship between variables across the panel of selected countries.

4.2. Unit Root Test Results

The objective of this research is to determine if growth and total, renewable, and nonrenewable energy in the countries associated with the EAPP and EAPT have a nonlinear asymmetry relationship. By this, the NARDL approach is used and PP, ADF, and KPSS are used to check whether the variables exhibit the assumption.

Table 6. Unit root test results.

Country	Test	lnGDP		lnEN		lnREN		lnNREN		lnLab		lnK
		C	C-T	C	C-T	C	C-T	C	C-T	C	C-T	C	C-T
Ethiopia	PP	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(0)	I(1)	I(1)	NA	NA
	ADF	I(1)	I(1)	I(1)	I(0)	I(1)	I(1)	I(1)	I(0)	I(0)	I(1)	NA	NA
	KPSS	I(0)	I(0)	I(0)	I(0)	I(0)	I(1)	I(0)	I(1)	I(1)	I(0)	NA	NA
Tanzania	PP	I(1)	I(0)	I(0)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(0)	NA	NA
	ADF	I(1)	I(0)	I(0)	I(0)	I(1)	I(0)	I(1)	I(0)	I(1)	I(0)	NA	NA
	KPSS	I(0)	I(0)	I(0)	I(1)	I(1)	I(1)	I(0)	I(0)	I(1)	I(1)	NA	NA
Sudan	PP	I(1)	I(1)	I(0)	I(0)	I(1)	I(1)	I(1)	I(1)	I(1)	I(0)	I(1)	I(1)
	ADF	I(1)	I(1)	I(0)	I(0)	I(1)	I(1)	I(1)	I(1)	I(0)	I(1)	I(1)	I(1)
	KPSS	I(0)	I(1)	I(0)	I(0)	I(0)	I(0)	I(0)	I(0)	I(0)	I(1)	I(0)	I(0)
Egypt	PP	I(1)	I(1)	I(1)	I(1)	I(0)	I(1)	I(1)	I(1)	I(0)	I(0)	I(0)	I(1)
	ADF	I(1)	I(0)	I(1)	I(1)	I(0)	I(1)	I(1)	I(1)	I(1)	I(0)	I(0)	I(1)
	KPSS	I(1)	I(0)	I(0)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)
DRC	PP	I(0)	I(0)	I(0)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(0)	NA	NA
	ADF	I(0)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	NA	NA
	KPSS	I(0)	I(0)	I(0)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	NA	NA
Kenya	PP	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(0)	I(1)	I(1)
	ADF	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(0)	I(1)	I(1)	I(1)
	KPSS	I(0)	I(0)	I(1)	I(1)	I(1)	I(1)	I(0)	I(1)	I(1)	I(1)	I(1)	I(0)
Burundi	PP	I(0)	I(0)	I(0)	I(1)	I(1)	I(1)	I(1)	I(1)	I(0)	I(0)	NA	NA
	ADF	I(0)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	NA	NA
	KPSS	I(0)	I(0)	I(0)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	NA	NA
Rwanda	PP	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)
	ADF	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)
	KPSS	I(0)	I(0)	I(1)	I(1)	I(1)	I(1)	I(0)	I(1)	I(1)	I(1)	I(1)	I(0)
Uganda	PP	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)
	ADF	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)	I(1)
	KPSS	I(0)	I(0)	I(1)	I(1)	I(1)	I(1)	I(0)	I(1)	I(1)	I(1)	I(1)	I(0)

I(0) represents level stationarity, I(1) indicates first-difference stationarity, unknown structural breaks occurred in the brackets; NA: non observations.

displays the stationarity results from PP, ADF, and KPSS, which demonstrate that economic growth, total, renewable, and nonrenewable energy consumption, and the first difference, are all stable at current levels. As a result, all variables are stationary and integrated in the I(0) and I(1). This support the convenience of using NARDL methods for estimating the nonlinear relationships between economic growth and different forms of energy, since one of the critical issues is that NARDL used for variables are cointegrated at zero or 1st order.

4.3. NARDL and CCEMG Estimates

Table 7. Panel CCEMG estimates.

Dependent: lnGDP
Variables	LnEN	LnREN	LnNREN	LnGDP.avg	lnEN.avg	lnREN.avg	LnNREN.avg
Estimates	2.154 **	0.115	−1.576 **	0.727 **	−0.549	0.042	0.381
p-value	0.031	0.138	0.044	0.041	0.674	0.906	0.686
Wald test	8.11 **

** indicates significant levels at 5%.

and

Table 8. NARDL model estimates.

Dependent: lnGDP
		Burundi	Rwanda	Uganda	Kenya	Tanzania	Ethiopia	DRC	Egypt	Sudan
lnEN	C	0.586 **	0.187	−0.167	0.893 *	0.492 **	0.662 **	0.708 *	1.058 *	1.233 *
	${\mathrm{y}}_{-1}$	−0.255 **	−0.247 **	0.036	−0.302 *	−0.181 **	−0.277 **	−0.237 *	−0.338 *	−0.411 *
	${\mathrm{E}}^{+}$	−1.252 **	−0.297	−0.195	0.239 **	0.004 **	2.326	−0.282	0.839	0.635
	${\mathrm{E}}^{-}$	0.076	−0.135 **	−0.399	−0.416	−0.132	2.740	−0.228 **	0.788	0.440
	${\Delta \mathrm{E}}^{+}$	−1.709 *	-	-	−0.066 *	-	1.148	-	0.451	0.515
	${\Delta \mathrm{E}}^{-}$	0.243	1.152 **	-	-	-	-	−1.071 **	1.649	-
lnREN	${\mathrm{E}}^{+}$	0.327	0.168	−0.041 **	0.094	0.050 **	−0.699	−0.309 **	0.071	−0.083
	${\mathrm{E}}^{-}$	0.071	−0.002	0.080	−0.076	−0.005	2.519 **	0.351	0.030	−0.130
	${\Delta \mathrm{E}}^{+}$	−0.295 **	-	-	−0.123 *	-	-	−0.087	-	-
	${\Delta \mathrm{E}}^{-}$	-	-	-	-	-	0.599	1.006 *	-	-
lnNREN	${\mathrm{E}}^{+}$	0.878	−0.049	0.177 ***	−0.234 **	0.037	−1.525	0.172	−0.571	−0.413
	${\mathrm{E}}^{-}$	−0.001	−0.241 **	0.007	0.407 **	0.163 ***	−1.836	0.019	−0.952	−0.346
	${\Delta \mathrm{E}}^{+}$	−1.169 **	-	-	-	-	−0.957	-	-	-
	${\Delta \mathrm{E}}^{-}$	-	1.039 **	-	-	-	-	0.327 **	-	0.402
lnL	${\mathrm{L}}^{+}$	−1.607 *	2.305 *	2.695 **	0.068	0.932	1.487	−0.059	0.713 **	1.057
	${\mathrm{L}}^{-}$	−0.676 *	13.480 **	−1.126 **	−0.046	−0.370	−1.410	0.934	−0.189	−1.036 *
	${\Delta \mathrm{L}}^{+}$	−3.842 *	6.292 *	10.055 *	0.810 *	-	−21.114 **	-	-	-
	${\Delta \mathrm{L}}^{-}$	-	9.218 *	7.061 ***	-	-	25.761	4.664 **	−0.882 **	−13.694 **
lnK	${\mathrm{K}}^{+}$	NA	0.139 **	0.138 *	0.039 **	NA	NA	NA	0.034	0.052 *
	${\mathrm{K}}^{-}$	NA	0.124 **	0.009	0.048 **	NA	NA	NA	0.014	0.004
	${\Delta \mathrm{K}}^{+}$	NA	0.057	−0.011	0.094 ***	NA	NA	NA	0.220 *	-
	${\Delta \mathrm{K}}^{-}$	NA	0.443 *	−0.136	-	NA	NA	NA	−0.076 *	0.234 *
R-square		0.739	0.647	0.754	0.842	0.856	0.743	0.876	0.730	0.761

* indicates significant levels at 1%, ** indicates significant levels at 5%, *** indicates significant levels at 10%, -: no estimate, NA: non observations.

show the estimates from the CCEMG and NARDL estimators. Table 7 presents that overall energy has substantial and positive influence on economic growth; renewable energy has a negative impact on growth, and non-renewable energy has a considerable negative impact on economic growth in the panel of all chosen nations. The cross-sectional dependence averages have also a positive impact (renewable, GDP, and nonrenewable), while total energy has a negative effect on growth. According to these findings, a 5% increase in total and renewable energy causes growth to increase by 2.154 percent and 0.115 percent, respectively, whereas a 5% increase in nonrenewable energy causes growth to fall by 1.576 percent. This implies that increment in nonrenewable dependence reasonably reduce the increment rate of growth and disturb the green growth policy which is highly depends on green energy utilization. These results suggest that intensive investment in the energy sector, especially renewable energies, such as EAPP and EAPT projects positively contribute to economic growth and reduce negative effect from nonrenewable energy to not only economic growth but also environmental degradation [70,71]. These findings are in line with [30], which found that renewable and nonrenewable energy both contribute to economic growth in 30 Sub-Saharan African nations, although environmental issue was noted taken into consideration.

From Table 8, both the labor and capital improve the economic growth in terms of negative and positive shocks. In terms of energy consumption, in the long and short-term, the positive shocks to total have a substantial detrimental influence on economic growth; while positive shocks to renewable and non-renewable energy have a major negative influence on Burundi’s growth in the short term. In Rwanda, negative shocks to renewable and total energy have a large and detrimental long-term and short-term impact on growth. Positive shocks to renewable and nonrenewable energy, on the other hand, have a considerable negative and positive impact on long-term growth in Uganda. Beneficial shocks to total energy have a large positive influence on Kenyan growth. In Ethiopia, negative shocks to renewable energy have a big and favorable long-term impact on growth. Favorable shocks to total energy and negative shocks to renewable and nonrenewable energy have a large and positive impact on long-term growth in Tanzania. Negative shocks to total energy have a large positive and negative impact on long- and short-term growth in the DRC; negative shocks to renewable energy have a strong positive and negative impact on long- and short-term growth in the DRC. These results imply that the impacted countries need to establish effective policies to monitor energy sector investment and energy use in the projects related to economic development. The existence of a nonlinear asymmetric and symmetric link between growth and energy consumption (total, renewable, and nonrenewable) in the selected nations is highlighted by these findings. This information is positive to the policymakers and government officials for establishing new policies to avoid a negative nonlinear effect, which may be caused by environmental fluctuations. These findings are consistent with the findings of [52,55], demonstrating an uneven relationship between total and renewable energy and growth in certain of the analysed countries.

4.4. Long-Run and Short-Run Asymmetry and Symmetry Restrictions

Because the mentioned nonlinear link between growth and energy usage is considerable in some nations in the long and short term, it’s necessary to explain how they differ and are dispersed among a few countries from 1980 to 2017. Therefore,

Table 9. Long- and short-run asymmetry and symmetry restrictions.

Dependent: lnGDP
Country	Long-and Short-Run	lnEN		lnREN		lnNREN
Burundi	LR	2.441	S	1.024	S	2.472	S
	SR	4.526 **	A	4.718 **	A	4.061 **	A
Rwanda	LR	0.6331	S	0.283	S	0.701	S
	SR	4.685 **	A	-	-	-	-
Uganda	LR	0.356	S	3.710 **	A	0.748	S
	SR	-	-	-	-	-	-
Ethiopia	LR	1.799	S	3.155 **	A	1.622	S
	SR	0.826	S	0.589	S	3.987 **	A
Tanzania	LR	1.524	S	1.597	S	2.898 ***	A
	SR	-	-	-	-	-	-
Kenya	LR	4.535 **	A	8.410 *	A	6.373 *	A
	SR	0.768	S	3.291 **	A	10.443 *	A
Egypt	LR	0.689	S	0.997	S	0.457	S
	SR	1.185	S	-	-	-	-
DRC	LR	0.252	S	3.753 **	A	2.469	S
	SR	5.665 **	A	7.343 *	A	2.174 **	A
Sudan	LR	0.572	S	0.565	S	0.312	S
	SR	0.816	S	-	-	4.144 **	A

* indicates significance level at 1%, ** indicates significance level at 5%, *** indicates significance level at and 10%, A: Asymmetry and S: Symmetry, LR: Long-run, SR: Short-run.

and dynamic multipliers (Figure 1, Figure 2 and Figure 3) show findings for asymmetric and symmetric nonlinear long-run and short-run causations from the total, renewable, and nonrenewable energy consumption to the growth. From Table 9 (A and S columns), total energy consumption has a long-run nonlinear asymmetric causation to growth in Kenya; a short-run asymmetry in Burundi, Rwanda, and DRC; and symmetry in other countries. In renewable energy, long-run asymmetry is noted in four countries; and three countries for the short-run; and symmetric causation is noted in six countries for both long- and short-run. Asymmetric long- and short-run connections between the usage of nonrenewable energy and growth are also seen in five countries.

Finally, Figure 1, Figure 2 and Figure 3 show the dynamic multipliers adjustment findings for positive and negative shocks to total, renewable, and nonrenewable energy consumption, showing that the economic growth adjustment is moving towards the long-run equilibrium. Nonlinear asymmetric and symmetric interactions between variables have differing dimensions and distributions, as seen in the figures. Figure 1 displays the results from Burundi, Uganda, and Rwanda. This implies that in Burundi and Rwanda, renewable energy and growth have a positive nonlinear asymmetric connection, but in Uganda, there is a negative asymmetry. Total energy and nonrenewable have positive nonlinear asymmetry with growth in Rwanda and Uganda, and in these three countries, respectively. Figure 2 shows findings from Kenya, Tanzania, and Ethiopia, which show a positive nonlinear asymmetric relationship between renewable energy consumption and growth in these countries, as well as a positive asymmetric nexus between total energy and growth in two countries and a negative nonlinear asymmetric relationship between nonrenewable energy and growth in these countries. These findings are consistent with those obtained in Tanzania [54] and Ethiopia [32,56], which confirmed that renewable and nonrenewable affect growth in both positive and negative sides, however, based on the NRDL approach, the results of this study emphasize that the effect can be linear or nonlinear, and differ in dimension and distributions, which indicate the different use of energy and economic growth. According to the findings from DRC, Egypt, and Sudan, which are represented in Figure 3, there is a negative nonlinear asymmetric causality between renewable energy and economic growth in DRC, while the positive asymmetric causation is noted in Egypt and Sudan. The asymmetric nexus from total and nonrenewable on growth is positive and more nonlinear in DRC than that from Egypt and Sudan. The results are similar to [30].

4.5. Causality

Table 10. Asymmetric and symmetric causalities.

	Variables	Burundi	Rwanda	Uganda	Ethiopia	Tanzania	Kenya	Egypt	Sudan	DRC
lnEN	${\mathrm{y}}_{\mathrm{t}}\Rightarrow {\mathrm{EC}}_{\mathrm{t}}^{+}$	0.320	5.595 *	1.268	2.001 ***	1.308	0.016	3.643 **	0.821	0.321
	${\mathrm{EC}}_{\mathrm{t}}^{+}\Rightarrow {\mathrm{y}}_{\mathrm{t}}$	2.530 ***	0.555	2.446	7.500 *	5.397 *	2.446	2.595 ***	8.537 *	0.213
	${\mathrm{y}}_{\mathrm{t}}\Rightarrow {\mathrm{EC}}_{\mathrm{t}}^{-}$	1.805	0.877	1.458	2.231 ***	0.462	0.327	4.985 **	0.647	1.160
	${\mathrm{EC}}_{\mathrm{t}}^{-}\Rightarrow {\mathrm{y}}_{\mathrm{t}}$	0.295	5.433 *	2.356	2.810 ***	6.739 *	1.424	1.446	11.444 *	0.028
lnREN	${\mathrm{y}}_{\mathrm{t}}\Rightarrow {\mathrm{EC}}_{\mathrm{t}}^{-}$	3.441 **	0.782	2.497 ***	0.133	2.409	0.507	4.814 **	0.023	0.965
	${\mathrm{EC}}_{\mathrm{t}}^{-}\Rightarrow {\mathrm{y}}_{\mathrm{t}}$	0.901	3.290 **	0.059	6.875 *	4.751 **	2.992 ***	9.331	10.293 *	0.624
	${\mathrm{y}}_{\mathrm{t}}\Rightarrow {\mathrm{EC}}_{\mathrm{t}}^{+}$	4.178 **	0.313	6.447 *	0.544	0.976	0.167	2.279	7.103 *	1.909
	${\mathrm{EC}}_{\mathrm{t}}^{+}\Rightarrow {\mathrm{y}}_{\mathrm{t}}$	1.513	2.326	0.585	3.495 **	4.568 **	2.755 ***	0.675	0.024	0.210
lnNREN	${\mathrm{y}}_{\mathrm{t}}\Rightarrow {\mathrm{EC}}_{\mathrm{t}}^{+}$	0.426	5.179*	0.532	6.248 *	2.786 ***	0.321	5.207 **	0.661	0.440
	${\mathrm{EC}}_{\mathrm{t}}^{+}\Rightarrow {\mathrm{y}}_{\mathrm{t}}$	2.058	0.380	2.044	1.249	2.950 ***	2.953 ***	2.107	9.733 *	0.035
	${\mathrm{y}}_{\mathrm{t}}\Rightarrow {\mathrm{EC}}_{\mathrm{t}}^{-}$	1.558	0.544	0.372	2.175	0.457	0.067	5.074 **	0.596	0.285
	${\mathrm{EC}}_{\mathrm{t}}^{-}\Rightarrow {\mathrm{y}}_{\mathrm{t}}$	0.296	4.303 **	2.135	2.730 ***	4.710 **	1.701	0.873	11.208 *	0.085

* indicates significance level at 1%, ** indicates significance level at 5%, *** indicates significance level at and 10%.

illustrates the results of the causality tests in terms of nonlinear asymmetry and symmetry, as well as validated hypotheses about the effects of positive and negative shocks to total, renewable, and nonrenewable energy usage on growth in a number of nations. The bidirectional causality between a positive shock to total energy and growth in Egypt and a negative shock to total energy and growth in Ethiopia can be shown in this table, and a positive shock to nonrenewable energy to growth in Tanzania. Among the selected variables in nine countries, the neutral hypothesis get highly support. The next hypothesis is a growth hypothesis, which spans energy consumption to economic expansion. This means that energy consumption has a direct impact on economic growth, with renewable energy usage having a greater positive impact on growth than nonrenewable and total energy use in many countries (see Figure 1, Figure 2 and Figure 3). The conservative hypothesis indicates the relationship which is between growth and energy is running from growth to energy. That is only supported in few countries among the all selected variables.

The linear causations between energy (total, renewable, and nonrenewable) use are illustrated in Figure 4.

In Sudan and Ethiopia, there is a bidirectional causation between total and nonrenewable energy and economic growth; however, in Ethiopia, there is a one-way directional causation moving from renewable energy to economic growth; and in Burundi, Tanzania, DRC, Kenya, and Egypt, there is a one-way directional causation running from economic growth to total and nonrenewable energy; and neutral causation is noted in Rwanda for all selected variables. These findings are consistent with country-specific studies [23,32,56]. As for the panel data from all selected countries connected to EAPP and EAPT, we can see that there is a bidirectional causal relationship between total and nonrenewable energy and economic growth, and the conservative hypothesis is that growth leads to the usage of renewable energy.

5. Conclusions and Policy Implications

The objective of this study is to examine the potential for a nonlinear imbalance between energy consumption (total, renewable, and nonrenewable) and economic growth in the Eastern Africa Power Pool (EAPP) and Eastern Africa Power trade members (EAPT). By assuming that the energy consumption is proportional to energy generated, the data mined from the EIA database and The World Bank from nine countries (the time span is 1980–2017) were used. For this reason, we used the NARDL and CCEMG estimators were used to provide the main results. The results reveal that long-run and short-run nonlinear asymmetric and symmetric relationships between energy and growth are noted in the selected counties, and differently distributed in direction and dimension in terms of positive and negatives changes. Total and renewable energy consumption have a favorable impact on regional growth, but nonrenewable energy has a negative impact. Furthermore, bidirectional, ono-way directional, and neutral hypotheses are noted among the selected variables in some countries, more specifically the neutral hypothesis is highly supported, followed by the growth hypothesis; A bidirectional hypothesis exists between total and nonrenewable energy and economic growth at the regional level, as well as a one-way directional hypothesis between growth and renewable energy, which runs from economic growth to renewable energy.

Based on our research results, we suggest that policy makers could follow: First, policymakers, investors, and government officials are urged to look into the hidden characteristics, which affect energy sector investment and trade, such as environmental externalities that lead to the decrease in income generation from energy use in the regional level and country levels. Second, sustainable energy generation can be established through policy leadership to encourage existing and dedicated cross-border connections to optimize East African electricity trade and intensive investment in EAPPs, which can positively affect economic growth in the region. Building an additional integrated power system in the region to assist EAPP would also support energy access and reduce the dependence on nonrenewable energy within the inter-connected countries.