Energy consumption and economic growth nexus in Somalia: an empirical evidence from nonlinear ARDL technique

ABSTRACT

The objective of this study is to ascertain the asymmetric impact of energy consumption on economic growth in Somalia for the period 1985–2017. A novelty econometric method of the Nonlinear Autoregressive distributed lag (NARDL) technique and the Granger causality test are employed to confirm the aim of the study. The empirical findings revealed the presence of asymmetric co-integration between energy consumption and economic growth in the long run. Both positive and negative shocks in energy consumption enhance economic growth in Somalia in the long run, even though, the positive shock has a stronger significant effect on economic growth in the long run. In contrast, a negative shock in energy consumption causes economic growth, thus, confirming the energy-led growth hypothesis. Hence, this calls for the policymakers to devise investment policies that are aimed at the improvement of foreign and local investments in energy sector.

KEYWORDS:

• NARDL; economic growth; energy consumption; Somalia

1. Introduction

Energy plays a key role in sustainable economic growth for both developing and developed nations. It is considered a key input for the production process, particularly in manufacturing industries which are heavily dependent on energy for production purposes. Thus, the relationship between energy use and economic growth has been extensively investigated since the 1950s (Kraft and Kraft 1978; Mason 1955).

Somalia has been enduring a stateless period for more than two decades. Food insecurity, vulnerability, and extreme poverty are constant threats to most Somalis. Around 2.3 million people do not receive enough food – thus facing food insecurity. Moreover, the population who lives below extreme poverty and poverty lines are estimated at 43% and 73% of the total population, respectively (World Bank 2018). Despite the importance of energy to economic growth and poverty reduction, majority of Sub-Saharan African countries including Somalia face energy shortages. There is a limited energy supply in Somalia where only 15% of the total population have access to energy services. The accessibility of energy services varies based on the demographics. A negligible percentage of the population can access energy in rural areas (4%) compared to urban areas (33%) (World Bank 2018).

The availability of sufficient energy supply ensures lift millions of people out of poverty, and food insecurity (World Bank 2018). It will lead to make sure major progress on the sustainable development goal (SDG) 7: ‘ensure access to affordable, reliable, sustainable, and modern energy for all’. I It also touches with SDG 15: resilient infrastructure, as reliable electric power will assist other infrastructure. Greater access to energy tends to make people more productive, and generate more income which will ultimately lead to reduced poverty (SDG: 1). The emergence of conflicts in Somalia in 1991 did not only hamper the structure of governance but also impeded the services intended that the government provides. Following the collapse of the central government, private sectors filled the void left by the government by providing more than 90% of power to the urban and peri-urban areas using diesel generation and mini-grids. But, notably, the cost of electricity in Somalia is one of the most expensive in the world which is maximally around $1/kWh (World Bank 2018). Thus, traditional biomass energy – typically charcoal and firewood – emerged as the dominant energy source corresponding with 82% of the total energy consumption (African Development Bank 2015). The dominance of biomass energy is attributed to the limited ability of large-scale imports of energy due to the low income of the population. This type of energy is not only unsustainable but also undermines environmental quality.

Indeed, the link between energy-growth nexus is extensively studied in the literature. Because it is essential for energy conservation policy. If causality is established between energy and growth, introducing an energy conservation policy will affect growth. But, if there is no causality detected among them, using an energy conservation policy would not have any harm on growth. The energy and growth relationship is based on four hypotheses, namely, the energy-led growth hypothesis; feedback hypothesis; conservative hypothesis; and neutral hypothesis. Accordingly, several studies have documented the energy-led growth hypothesis – attributing energy use as a determinant of economic development (Chontanawat, Hunt, and Pierse 2008). A neutrality hypothesis – which considers that energy and growth are not related – is confirmed by Kourtzidis, Tzeremes, and Tzeremes (2018) in the United States of America (USA) using threshold analysis at sectoral levels of energy. This is congruent with Śmiech and Papiez (2014); and Stern (1993) who arrived at the same conclusion. Moreover, a study conducted in Turkey assessed the association between CO₂ emissions, energy consumption, FDI, and economic growth (Gökmenoğlu and Taspinar 2016). The findings signify a unidirectional causality from economic growth to energy use, hence, confirming the conservative hypothesis. Several other studies support this hypothesis, such as Fatai, Oxley, and Scrimgeour (2004) for New Zealand and Pao and Fu (2013) for Brazil. Moreover, Ebohon (1996) and Khobai and Roux (2017) have confirmed the feedback hypothesis – which implies that energy and growth affect each other.

Furthermore, the empirical literature highlighted the importance of energy use to growth. Using static and dynamic panel analysis, Salari et al. (2021) assessed the impact of renewable and nonrenewable energy on growth in USA states. The results revealed that renewable and nonrenewable energy improve economic growth.. Using nonparametric methods, Ivanovski, Hailemariam, and Smyth (2021) modelled the role of renewable and nonrenewable energy use in economic growth in cross-country analysis – OECD and non-OECD countries. The result signifies that nonrenewable energy exerts a positive effect on growth in both sampled countries. A recent study conducted in three income group countries; namely, lower, middle, and upper-middle income levels, modelled the impact of total energy and renewable energy consumption on economic growth (Namahoro, Nzabanita, and Wu 2021). It is observed that total energy consumption exerts a positive effect on growth in all income group countries. A similar result is observed in a panel of South Asian countries (Rahman and Velayutham 2020).

Even though the empirical literature demonstrates that energy usage positively contributes to the economies of energy-producing countries, it has also a favourable effect on the economies of energy-importing countries. Using pooled mean group and mean group estimation methods, Esen and Bayrak (2017) revealed that energy utilisation has a favourable impact on economic growth in a sample of net energy-importing countries. In a recent study, Zhang, Nuruzzaman, and Su (2021) observed that both gas and residential electrical energies enhance economic growth in Bangladesh in the long run using the ARDL technique.

Energy shocks and financial crises exert energy use to have asymmetric effects on economic growth. Therefore, recent studies have tested the asymmetric effect of energy use on growth. Using nonlinear ARDL (NARDL), Awodumi and Adewuyi (2020) assessed the asymmetric impact of nonrenewable energy on growth in African oil-producing countries. The results signify that the positive and negative changes in petroleum energy retards economic growth in Egypt and Angola. On the contrary, it exerts a positive effect on Nigerian economic growth. Shastri, Mohapatra, and Giri (2020) examined the impact of renewable and nonrenewable energy consumption on economic growth in India using the NARDL. They revealed that a positive shock in nonrenewable energy consumption exerts a positive effect on growth, whereas a negative shock in nonrenewable energy utilisation significantly hampers economic growth in the long run. In the same vein, nonrenewable energy and growth are observed to be asymmetrically co-integrated in the United States, Italy, and France in the long run (Tugcu and Topcu 2018). The heterogeneous results of asymmetric co-integration between growth-energy nexus were also found by several previous studies (Adekoya, Ogunnusi, and Oliyide 2021; Lee and Chang 2007). The inconclusive asymmetric results of the energy use – growth nexus are due to the discrepancies in the methodology applied, and the measurement of energy consumption employed. Thus, the hypothesis of energy and growth nexus is:

H₀: Energy consumption does not have a significant positive effect on economic growth.

H₁: Energy consumption has a significant positive effect on economic growth.

The overarching research inquiry of this study is: how much does energy utilisation contribute to economic growth in Somalia? This question is critical because the empirical studies in the literature on economic growth and energy consumption nexus have produced blended results. Some plausible explanation for this contradictory result in previous undertakings is that most of the previous studies consider energy use and economic growth nexus symmetrically. Indeed, the symmetric association only produces an average relationship over time. The nexus of energy and growth are asymmetrically co-integrated (Adekoya, Ogunnusi, and Oliyide 2021; Lee and Chang 2007).

Hence, given the role of energy consumption in economic growth in Somalia, there are very scanty studies in this regard.. Very limited studies have been examined in this respect in Sub-Saharan African economies (Awodumi and Adewuyi 2020; Bildirici and Ozaksoy 2017; Ekeocha, Penzin, and Ogbuabor 2020), but most of them have been conducted in developed and developing countries (Kourtzidis, Tzeremes, and Tzeremes 2018; Oryani et al. 2021; Tuna and Tuna 2019). The very few studies on the energy-growth nexus in Sub-Saharan African countries have not only excluded Somalia in their sample but also failed to consider the asymmetric relationship between energy and growth (Ebohon 1996; Al-mulali and Binti Che Sab 2012; Khobai and Roux 2017; Wolde-Rufael 2009). This study contributes to the literature in several ways. First, this is the first study that examines the impact of energy consumption on economic growth in Somalia. Second, it is notable that the variation in energy use does not exert the same effect on growth due to the existence of political instability and energy shocks that could exert an adverse effect on the energy-growth nexus. To bridge this gap, this undertaking ascertains the asymmetric effect of energy consumption on economic growth in Somalia utilising time series data stretching from 1985 to 2017 and the nonlinear autoregressive distributed lag (NARDL) model postulated by Shin, Yu, and Greenwood-nimmo (2014). To avoid model misspecification, we incorporate labour and capital as control variables to find out robust results.

The rest of the paper is organised as follows; theoretical literature and stylised facts are presented in section two; section three reports data sources and descriptions, and econometric methodology; empirical analysis and discussion are presented in section four. Finally, the conclusion and policy implications are reported in section five.

2. Theoretical literature and stylized facts about economic growth and energy consumption

2.1. Theoretical literature about energy and growth

Energy and growth nexus is based on two competing views in the literature. The first theory is the standard neoclassical economic growth that considers two important inputs for economic production which are capital and labour. Technological development also plays an essential role in stimulating economic growth. They argue that energy has a negligible or no effect on growth. The second view underscores that natural resources – energy – is a key input for production and a source of long-term sustainable economic growth. It is considered an intermediate input, so is assumed to affect economic growth (Stern 2004). Ecological and natural resource economists have emphasised the importance of energy consumption to economic growth (Cleveland et al. 1984). It is, therefore, needed for energy to use with capital and labour (Stern 2011). In light of the above, the seminal work of energy and growth nexus conducted observed that energy is a critical determinant of economic growth in the USA (Kraft and Kraft 1978). Since then, many empirical examinations related to the energy and growth nexus have been performed in other nations. Indeed, the scope has extended beyond growth, as the impact of energy consumption on other economic and socio-economic variables has been investigated such as poverty, human capital, employment, and the environment (Nagatomo et al. 2021; Ogbeide-Osaretin 2021; Warsame et al. 2022a). Notably, given the importance of energy to growth, a concise conclusion about the energy and growth nexus has not been reached yet. Some studies have produced that energy causes economic growth; while others established that growth leads to high energy utilisation (Magazzino 2012; Magazzino et al. 2021). Other studies have confirmed that energy and growth cause each other. Finally, some others have concluded that energy and growth are not related (Pao and Fu 2013; Ebohon 1996; Khobai and Roux 2017).

2.2. Stylised facts about energy and growth

Somalia is an agrarian-based economy and agriculture production is sensitive to climate shocks – floods and droughts. In the drought periods, agriculture production dramatically declines, even though it constitutes 65% of the gross domestic product (GDP) in Somalia (Hussein et al. 2023). Real gross domestic product per capita (RGDPC) – measured for economic growth – has been decreasing as shown in Figure 1. From 1990 to 1997, the RGDPC substantially plummeted. But from 1998 to 2017, RGDPC became stagnant, and this could be attributed to several factors. First, Somalia endured statelessness period in 1991 which played a key role in the reduction of the GDP. Because it led to the destruction of the infrastructure, the infant industries that the nation had which were booming, and caused massive migration of people. Moreover, stagnant economic growth and increasing population growth have also contributed to the reduction of the real gross domestic product per capita.

Figure 1. Real GDP per capita in Somalia. Source: SESRIC database (2020).

Furthermore, energy consumption per capita in Somalia is reported in Figure 2. Likewise, It has shown a decreasing trend. In 1990, energy consumption per capita in Somalia was 560 kilowatts. But this number decreased to 309 and 277 kilowatts in 1992 and 1999, respectively. Nevertheless, it started to recover and reached 335 kilowatts in 2000. It continued to increase to 343 kilowatts in 2002, even though it is less than the value of 1990. From 2003 and onward, energy consumption per capita dramatically plummeted. Several reasons are attributed to its reduction. First, biomass energy – firewood and fossil fuel – constitutes 82% of Somalia final energy consumption (Warsame 2022). Firewood is not a sustainable option for energy, because it impedes environmental quality. Moreover, Somalia is a poor country that could not import enough fossil fuel energy from abroad. Hence, Somalia’s declining economic growth could be attributed to the limited energy supply.

Figure 2. Energy consumption per capita in Somalia. Source: Ourworldindata (2021).

3. Material and methods

3.1. Data sources and descriptions

This study has chosen Somalia as a case study for several reasons. First, Somalia is a rich country in terms of both renewable and nonrenewable energy sources, but the country faces energy crises where only 36% of the population has access to electricity. Energy resources in the country are still untapped. Second, the country depends on traditional biomass energy which constitutes 82% of the total energy consumption. Currently, the country encounters energy crises that impede economic productivities (African Development Bank 2015).

The data used in the study contains four variables, namely, energy consumption, labour, capital, and real gross domestic product per capita. All the variables were extracted from the Organization of Islamic Cooperation (OIC) countries database, except energy consumption which was taken from the OurWorld inData website. The time series dataset used stretches from 1985 to 2017. Real gross domestic product per capita is measured for economic growth, population growth is taken as a proxy of labour, gross fixed capital formation is taken as a proxy of capital, and energy consumption is measured for per capita energy consumption. All the variables are based on annual frequency. Notably, all the series were converted into natural logarithms to find precise coefficient results and more importantly to eliminate the heteroskedasticity problem. A detailed explanation of the data is displayed in Table 1.

Table 1. Variable descriptions and sources.

Variable	Code	Proxy	Source
Real Gross Domestic Product Per Capita	RGDPC	Real GDPC in the US dollar (constant 2010)	OIC database
Labour	L	Population growth rate	OIC database
Capital	K	Gross fixed capital formation (constant in the US million)	OIC database
Energy Consumption	EC	Energy consumption per capita (kWh)	Ourworldindata

3.2. Econometric methodology

In the previous studies, the linear methodologies such as; ARDL, Johansen co-integration, Ordinary Least Square (OLS), and Engle & Granger methods were inadequate to investigate the nexus between energy use and economic growth for both in short-run and long-run. However, in recent years the nexus between these variables has been tested as a nonlinear form (Oryani et al. 2021; Tuna and Tuna 2019). The nonlinear models are good at capturing the effects of financial crises occurred and political instability in the case of Somalia which could result in structural breaks in time series data. To capture these hidden effects, the study adopts the nonlinear ARDL method postulated by Shin, Yu, and Greenwood-nimmo (2014). This method decomposes the variable into positive and negative shocks which absorbs the asymmetric effects in the long- and short run. Besides this advantage, the NARDL is appealing and addresses several econometric issues. First, NARDL is a better method for addressing nonlinear co-integration as it accounts for both short- and long-run asymmetric association compared to threshold co-integration method which ascertains only for long run asymmetries. Second, traditional co-integration methods – Engle & Granger, and Johansen and Juselius co-integration methods – require variables to be integrated at the same order of first difference I(1). But NARDL could regress variables which are integrated at level I (0), first difference I(1), or the combination of both. Third, smooth transition methods and threshold vector error correction modelling suffer from convergence problems due to proliferation of the number of parameters, but NARDL addresses this issue emphatically. Finally, nonlinear ARDL decomposes the variable into positive and negative shocks by assessing the asymmetric effect of the regressors on dependent variable which is essential for policy implications.

In this study, the NARDL method is adopted to estimate the impact of energy consumption on economic growth. To achieve this objective, the study used the framework proposed by (Apergis and Payne 2009) which is based on the neoclassical economic growth and Cobb Douglas production function. This theory employs labour, capital, and energy as key determinants of economic growth. Labour and capital are incorporated in the model as control variables to capture the exact effect of energy on growth. More importantly, we control them to address the omission of variables. By employing these parameters, an empirical model is specified as follows: (1) $lnRGDP{C}_t={\alpha }_0+{\text{β}}_{1}lnE{C}_t+{\text{β}}_{2}ln{L}_t+{\text{β}}_{3}ln{K}_t+{ϵ}_t$(1) $lnRGDP{C}_t$ represent gross domestic product per capita measured for economic growth, $lnE{C}_t$is the energy consumption, $ln{L}_t$ stands for labour, $ln{K}_t$ stands for capital, t is the time period, ln represents natural logarithm, and ${ϵ}_t$ is the disturbance term.

According to Shin, Yu, and Greenwood-nimmo (2014). The decomposition of energy use into positive and negative shocks on economic growth can be specified as: (2) $E{C}_t=E{C}_0+E{C}_t^{+}+E{C}_t^{-}$(2) Where $E{C}_t^{+}andE{C}_t^{-}$are the decomposition of positive and negative changes in $E{C}_t$: (3) $E{C}_t^{+}=\sum _{j=1}^t\mathrm{\Delta }E{C}_j^{+}=\sum _{j=1}^t\mathrm{m}\mathrm{a}\mathrm{x}\left(\mathrm{\Delta }E{C}_j,0\right)$(3) $E{C}_t^{-}=\sum _{j=1}^t\mathrm{\Delta }E{C}_j^{-}=\sum _{j=1}^t\mathrm{m}\mathrm{i}\mathrm{n}\left(\mathrm{\Delta }E{C}_j,0\right)$The study’s final model of the interested parameters – lnRGDPC, lnEC⁺, lnEC⁻, lnL, and lnK – in the NARDL model can be expressed using Cobb–Douglas production function and the empirical previous studies of Wolde-Rufael (2009) and Oryani et al. (2021) as follows: (4) $\begin{array}{rl}& \mathrm{\Delta }lnRGDP{C}_t={\text{β}}_0+\mathrm{\Delta }lnRGDP{C}_{t-1}+{\text{β}}_1^{+}lnE{C}_{t-1}^{+}+{\text{β}}_1^{-}lnE{C}_{t-1}^{-}+{\text{β}}_{3}ln{L}_{t-1}+{\text{β}}_{4}ln{K}_{t-1}+\\ & \sum _{j=1}^{c-1}{\alpha }_j\mathrm{\Delta }lnRGDP{C}_{t-j}+\sum _{j=0}^{d-1}{\alpha }_{1j}^{+}\mathrm{\Delta }lnE{C}_{t-j}^{+}+\sum _{j=0}^{d-1}{\alpha }_{1j}^{-}\mathrm{\Delta }lnE{C}_{t-j}^{-}+\sum _{i=0}^d\mathrm{\Delta }{\alpha }_{3}ln{L}_{t-k}+\\ & \sum _{i=0}^d\mathrm{\Delta }{\alpha }_{4}ln{K}_{t-k}+{ϵ}_t\end{array}$(4) Where lnRGDPC is the natural logarithm of per capita real gross domestic product, $lnE{C}_{t-1}^{+}$is the positive change in energy consumption, $lnE{C}_{t-1}^{-}$ is the negative change in energy consumption, lnL stands for labour, lnK signifies capital, c is the lag length of the dependent variable and d denotes the optimal lag length of the explanatory variables. ${\text{β}}_{1-4}$ and $\alpha$ represent long-run and short-run coefficients of the parameters respectively. Moreover, the null hypothesis states that energy use and economic growth are not co-integrated in the long run, while the alternative hypothesis presents that they are co-integrated in the long run.

4: Empirical result and discussion

4.1. Descriptive statistics

The characteristics of the series are presented in Table 2. Our outcome indicates that the population – which is a proxy for labour – is negatively skewed, whereas economic growth, capital, and energy use have shown positive skewness. The Jarque-Bera P-values demonstrate that the null hypothesis of normality of the series such as; economic growth, labour, and energy use could be rejected. In contrast, the correlation matrix is also reported in Table 2. Economic growth is strongly and positively correlated with capital and energy use and is negatively related to labour. In addition, labour is found to be moderately and negatively associated with capital and energy use. Finally, capital and energy use have a positive correlation between them.

Table 2. Descriptive statistics.

	lnRGDPC	lnL	lnK	lnEC
Mean	4.649785	1.158153	19.20464	5.853653
Median	4.523417	1.317473	19.18837	5.745077
Maximum	5.064555	1.567599	19.69285	6.778529
Minimum	4.498364	0.247130	18.92412	5.496287
Std. Dev.	0.211245	0.383316	0.199413	0.349871
Skewness	1.054855	−1.194581	0.566580	1.516613
Kurtosis	2.281085	3.236095	2.558876	4.035579
Jarque-Bera	6.623622	7.685120	1.971525	13.69717
Probability	0.036450	0.021439	0.373155	0.001061
lnRGDPC	1
lnL	−0.6975	1
lnK	0.7423	−0.4552	1
lnEC	0.8568	−0.4246	0.6231	1

4.2. Unit root test.

Time series data often contain a trend that violates the assumption of stationary. Hence, to avoid this fallacy, a unit root test is conducted using Augmented-Dickey Fuller (ADF) and Philips Perron (PP) tests to verify that none of the variables are stationary at second order I (2). Regarding NARDL, it is critical to know that all the variables are integrated at level I (0), the first difference I (1), or the combination of both levels. Table 3 displays that economic growth and energy use contain a unit root problem at level 1 (0) in both tests. Whereas labour is non-stationary at order one I (1) in both tests. However, the result shows that none of the variables are stationary at the second difference I (2) which implies that they are stationary at the combination of order I (0) and order I (1). Besides, the utilisation of ADF and PP is not sufficient to address the structural breaks in the series. Therefore, we employ Zivot and Andrews (1992) to determine the structural breaks in the data. This method is good at capturing the stationary and a single break in the series. Its result reported in Table 3 revealed that the stationary of the parameters are mixed orders I (0) and I (1). Hence, we could proceed to examine the long run cointegration of the sampled variables.

Table 3. Unit root tests.

Variable	ADF	PP	ZA structural break unit root
			T-statistics	Time break
LnRGDPC	−3.1825	−1.1392	−8.721***	1994
$\mathrm{\Delta }$LnRGDPC	−2.7325	−5.9296***	−17.921***	1996
LnEC	−2.2325	−2.1970	−4.639	2012
$\mathrm{\Delta }$LnEC	−5.3904***	−5.3908***	−7.224***	1993
LnL	−35.4002***	−2.2718	−9.291***	1996
$\mathrm{\Delta }$LnL	−1.6992	−2.9030	−7.959***	1997
LnK	−8.6845***	−1.1149	−8.165***	1994
$\mathrm{\Delta }$LnK	−4.5229***	−8.5201***	−8.219***	1996

Notes: *** and ** represent significance levels at 1% and 5%, respectively. The t-statistics reported are intercept and trend only. ZA stands for Zivot and Andrews.

To detect the presence of the variables’ nonlinear co-integration, we conduct the Wald bound test. Its result reported in Table 4 demonstrates that the Wald F-statistics – 15.7 – is above the upper bound critical value (6.97) at a 1% significance level which implies that economic growth, energy consumption, labour, and capital are co-integrated in the long run. In addition, it also shows the presence of long-run asymmetric co-integration among the variables. Hence, the presence of long-run co-integration depicts the dismissal of the null hypothesis of no long-run co-integration.

Table 4. F-bounds tests.

Model		Significance level	Critical values
lnRGDPC = f (lnEC^+, lnEC^–,lnL, lnK)			K (3)
	F-statistics		I (0)	I (1)
	15.7065	1%	5.333	6.975
		5%	3.653	4.965
		10%	2.985	4.133

Note: K is the number of explanatory variables.

The coefficients of the parameters are reported in Table 5. It reveals a positive shock in energy usage stimulates economic growth by about 0.68% in the long run if it is increased by a 1%. Similarly, a negative shock in energy utilisation raises economic growth by 0.08% in the long run for a 1% increase in negative shock energy use. Moreover, since the positive and negative shocks of energy consumption have significant positive effects on economic growth; we test whether there is an asymmetry between the two coefficients. Its result confirmed that the two coefficients are different and have an asymmetric impact on economic growth as shown by the statistical significance level of the coefficient F-statistics in Table 5. Besides, labour hampers economic growth in Somalia. Labour decreases economic growth by about 0.10% in the long run, if labour is increased by a 1%. On the contrary, capital is a crucial driving force of economic growth in Somalia. It is statistically significant and has a positive coefficient. This implies that a 1% in the capital will lead to the economic growth increasing by about 0.093% in the long run.

Table 5. Long- and short-run results.

Variable	Coefficient
lnEC^+	0.685358*** (7.5830)
lnEC^–	0.080904*** (5.7228)
lnL	−0.104789*** (−7.3340)
lnK	0.092442*** (4.9020)
Δ (lnRGDPC (−1))	−0.3987*** (−3.5897)
Δ (ln EC^+ (−1))	−0.5389*** (−4.6801)
Δ (ln EC^— (−1))	0.0626 (1.220)
Δ (lnL)	−0.3106*** (−3.6808)
Δ (lnL (−1))	0.1567** (2.6069)
Δ (lnK)	0.2660*** (2.8860)
Δ (lnK (−1))	0.0667 (1.0037)
ECT (−1)	−0.2164*** (−3.8848)
lnEC^+=  lnEC^-	18.785***

Notes: *** and ** indicate significance levels at 1% and 5%, respectively. T-statistics are reported in parenthesis. Δ is the short-run parameter.

Further, the short-run results of the study are presented in Table 5. It demonstrates that the previous year's economic growth undermines the current year's economic growth. A positive change in energy consumption has a negative coefficient and is statistically significant. This is interpreted as a 1% increase in positive shock in energy use will result in a 0.54% reduction in economic growth in the short run. On the contrary, a negative shock in energy use has a positive coefficient, albeit, it is insignificant. The result of labour in the short run is similar to its negative result in the long run. Labour inhibits economic growth by 0.31% in the short run for a 1% increase in labour. In the same vein, capital stimulates growth by 0.26% if it is increased by a 1% in the short run. More importantly, Table 5 also presents the error correction term (ECT) which are also known as the speed of adjustment.

To make robust that the series are integrated into the long run, the ECT should be statistically significant and has a negative coefficient. The negative coefficient denotes the convergence of the variables to the long-run equilibrium while the positive coefficient implies the divergence of the series. Hence, the value of ECT is 0.21 which is significant and accompanied by a negative sign. This presents that the model makes a convergence rather than a divergence. It is interpreted as 21% of the disequilibrium that occurs in economic growth in the short run is adjusted by the scrutinised explanatory variables in the long run annually.

Besides, the diagnostic tests (Table 6) and model stability display a shred of conclusive evidence which indicates that the NARDL model result is correctly estimated and can be applied as policy implications. The diagnostic results indicated that the model is free from model misspecification bias, heteroskedasticity, non-normality, serial correlation, and model instability as presented in Table 6 and Figure 3. The goodness fit of the model is good as shown by the adjusted R-squared. The sampled independent variables explain 95% of the variation that happens in economic growth.

Figure 3. Model stability test. (a) Cusum test; (b) Cusum square test.

Table 6. Diagnostic tests.

$X_{SC}^2$	0.1807 (0.6747)
$X_{HET}^2$	0.5306 (0.7583)
Normality Test	1.3457 (0.5103)
Reset Test	1.4717 (0.2467)
Adjusted R^2	0.9576

Note: $X_{SC}^2$ and $X_{HET}^2$ are serial correlation and heteroskedasticity respectively.

4.3. Granger causality test

Determining the causation of energy and economic growth is crucial in the context of energy economics. Hence, we estimate the Pairwise Granger causality of the variables to find out the causation of the interesting parameters. The results of causality – presented in Table 7 – demonstrate bidirectional causality between labour and economic growth, and capital and labour. In addition, a negative shock in energy consumption Granger causes economic growth which confirms the energy-led growth hypothesis. Energy consumption is an essential determinant of economic growth. This is congruent with several previous studies that revealed energy use causes economic growth (Chontanawat, Hunt, and Pierse 2008). In the same vein, a negative change in energy use causes labour whereas a unidirectional causality is established from a positive change in energy use to capital.

Table 7. Granger causality tests.

Null Hypothesis:	F-Statistic	Prob.
lnL ⇏ lnRGDPC	10.7024	0.0004
lnRGDPC ⇏ lnL	34.1129	5.E-08
lnK ⇏ lnRGDPC	0.79559	0.4620
lnRGDPC ⇏ lnK	1.52725	0.2360
lnEC^+ ⇏ lnRGDPC	0.69234	0.5097
lnRGDPC ⇏ lnEC^+	1.02486	0.3734
lnEC^– ⇏ lnRGDPC	9.33803	0.0009
lnRGDPC ⇏ lnEC^–	0.33487	0.7186
lnK ⇏ lnL	8.35754	0.0016
lnL ⇏ lnK	4.56563	0.0200
lnEC^+ ⇏ lnL	2.54059	0.0990
lnL ⇏ lnEC^+	0.27975	0.7583
lnEC^– ⇏ lnL	48.6573	2.E-09
lnL ⇏ lnEC^–	0.57392	0.5706
lnEC^+ ⇏ lnK	10.6937	0.0004
LK ⇏ lnEC^+	0.19614	0.8231
lnEC^– ⇏ lnK	0.41748	0.6632
lnK ⇏ lnEC^–	0.17876	0.8374
lnEC^– ⇏ lnEC^+	0.52664	0.5970
lnEC^+ ⇏ lnEC^–	1.93177	0.1659

Note: ⇏ indicates that variables do not cause each other.

4.4. Robust analysis

To find robust results about the interested parameters, we utilise fully modified ordinary least squares (FMOLS) and dynamic ordinary least squares (DOLS). Their results are reported in Table 8. Both models uncovered the same results which are also similar to the nonlinear ARDL cointegration method in the long run. Energy consumption and capital enhance economic growth whereas labour has a detrimental impact on economic growth in Somalia in the long run.

Table 8. Robust analysis.

Method:	(FMOLS)	(DOLS)
Variable	Coefficient	Coefficient
	(T-Statistic)	(T-Statistic)
LEC	0.394*** (7.709)	0.259*** (3.961)
LK	0.169* (1.981)	0.153*** (3.1447)
LPG	−0.202*** (−5.177)	−0.249*** (−3.828)
Constant	−0.672 (−0.443)	0.481 (0.524)
R-squared	0.90	0.98
Adjusted R-squared	0.89	0.98

4.5. Discussion about the result

Energy consumption is the greasing wheel of economic growth since it is a key input for production activities (Warsame et al. 2022a). The fractionalisation of positive and negative changes in energy use both enhance economic growth by about 0.68% and 0.08%, respectively, in the long run in Somalia. The positive shock in energy use has a significantly stronger effect on growth than the negative energy shock. Higher energy use will lead to high economic growth if the efficiency and production of energy are held constant. This is consistent with the previous studies that documented the positive shock effect of energy on economic growth (Kouton 2019). Moreover, using nonlinear ARDL technique and annual time series data from Iran, Oryani et al. (2021) reported that a positive shock in energy consumption increases economic growth in Iran. Moreover, Shastri, Mohapatra, and Giri (2020) reported that a positive shock in nonrenewable energy use enhances economic growth in India.

The negative shock in energy use has a smaller positive effect on economic growth. This could be attributed to the utilisation of modern energy technologies and the implementation of energy-saving policies. Notably, traditional biomass energy – typically firewood and charcoal – constitutes 82% of the total final energy consumption in Somalia (African Development Bank 2015). Any reduction in these types of energy and substitute for clean energy supports sustainable economic growth which is also a favour to the environment. This is congruent with several previous studies. For instance, Baz et al. (2021) noted that a negative shock in fossil fuel energy consumption tends to increase the economic growth in Pakistan. Similar results were found by Jiang and Chen (2020) in China who found the stimulating effects of negative shocks in energy use to economic growth.

Further, labour and capital were observed to have significant positive and negative effects on economic growth in Somalia respectively. Labour, which is measured for population growth, harms economic growth. This could be explained to that population growth is substantially increasing in Somalia while domestic economic sectors are malfunctioning. It is estimated that 70% of the Somali population are youth, but unfortunately, the youth unemployment rate is estimated at 67% (Warsame et al. 2022b). The adverse effect of labour corroborates with the previous study of Intisar et al. (2020) in a sample of panel Asian countries. On the contrary, capital formation substantially contributes to the economic growth in Somalia. It significantly supports economic productivity. This is in line with the previous study of Warsame et al. (2023) which revealed that capital enhances economic growth in Somalia.

5. Conclusion and policy implications

Harnessing energy resources leads to economic development and prosperity. The ability of the nations to access energy resources is different. Countries that can access a larger share of energy often reach economic development and reduce poverty. In least developed countries such as; Somalia, there is an energy supply shortage that can be a bottleneck to eradicating poverty and reaching sustainable economic growth. Hence, uncovering the role of energy use in growth is essential to be addressed in Somalia. To this end, this undertaking aims to model the effect of energy use on economic growth in Somalia for the period 1985–2017. Moreover, labour and capital were incorporated in the study as control variables. This study employed a NARDL technique, a recently developed econometric methodology. The contribution of this study is to address the energy-growth nexus asymmetrically, whereas the majority of the previous undertakings examined this nexus symmetrically which can lead to biased inferences. Moreover, Granger causality is conducted to determine the causality of the interested variables.

The empirical results demonstrate the presence of asymmetric co-integration between economic growth, labour, capital, and energy use in the long run. Furthermore, a positive shock in energy use has a stronger impact on economic growth than a negative shock in energy in the long run. This implies that energy use is critical for the economic growth in Somalia. Likewise, capital enhances economic growth in Somalia both in the short and long-run. On the contrary, labour hampers economic growth in the long run. Besides, the result of causality indicates bidirectional causality between labour and economic growth; and capital and labour. In addition, a negative shock in energy consumption Granger causes economic growth which confirms the energy-led growth hypothesis. Energy consumption is an important driver of economic growth in Somalia. In the same vein, a negative change in energy use causes labour, whereas, a unidirectional causality is established from a positive change in energy use to capital.

In light of the empirical results, energy use is critical for sustainable economic development in Somalia. However, the available energy supply is not sufficient to cover the increasing energy demand in the country. This study recommends policymakers devise and install energy policies that attract foreign and local investments in the energy sector. Somalia is endowed with both renewable and nonrenewable energy sources. Extracting these energy sources, specifically clean energy, will cover the domestic energy needs and also make Somalia to export energy to the neighbouring countries. Clean energy stimulates sustainable economic growth without harming environmental quality. Furthermore, Somalia is simultaneously experiencing water crises in some seasons of the year and an energy crisis. These issues could be fixed simultaneously by implementing hydropower dams. This will provide cheap energy and will also offer water reservoirs for agriculture production – crop and livestock production.

The limitation of the study is that it provides an avenue for future studies to examine the asymmetric impact of disaggregate energy utilisation in single and cross-country studies. This study is only limited to Somalia and used aggregate energy consumption.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The datasets used and/or analyzed during the current study are available at these links: https://www.sesric.org/query.php. https://ourworldindata.org/