Does improvement in capital intensity facilitate the transition to renewable energies? Evidence from Tunisia

Abstract

The objective of this study is to investigate the potential role of capital deepening in promoting the transition to renewable energy in Tunisia. To this end, the long and short run effects of capital deepening on the renewable energy transition were explored using the vector error correction model (VECM) and the Johansen cointegration technique, along with a linear and nonlinear causality test in the context of Tunisia for the period 1990 to 2018. In particular, we found that capital deepening contributes positively to the transition to clean energy resources. In fact, the results of the linear and nonlinear causality tests confirm a unidirectional causal relationship between capital intensity and the transition to renewable energy. This explains that the increase in capital intensity ratio conducts technical change towards renewable energy, which constitutes a capital-intensive technology. Moreover, these results enable us to draw a conclusion about the energy policies in Tunisia and the developing countries in general. In fact, the renewable energy substitution depends on capital intensity, through the development of specific energy policies, such as renewable energy policies. Gradual substitution of fossil fuel subsidies with renewable energy subsidies is essential to faster the transition to renewable energy and promote capital-intensive production methods.

Introduction

The renewable energy transition or the so-called the decarbonization of the energy sector has become an important debate among the policymakers and economic players since it is considered an urgent action to ensure the environmental and energy sustainability (Belaid and Zrelli 2019; Awijen et al. 2022; Omri et al. 2022; Samour et al. 2022a).

In this vein, Sadorsky (2021) illustrates that the substitution of the fossil fuels by clean energy resources helps to achieve the goals of the climate change agreement, boosts energy security, improves access to electricity, and also internalizes the negative impacts of fossil fuel consumption.

However, this transition to clean energy system will require significant efforts, especially for developing countries (Saadaoui and Chtourou 2022a). In general, several barriers to the diffusion of renewable resources in the emerging countries can be identified, especially in relation to the financing of new technologies. Therefore, they need to accentuate their efforts and make the strategies feasible, in terms of dissemination of renewable energy.

Moreover, Tunisia is one of the developing countries that have a strong plan to diversify its energy mix. For instance, policymakers in Tunisia have set out to achieve a 30% renewable energy share by 2030. Basically, Tunisia has a huge potential in renewable energy. More specifically, the country has a high rate of sunshine per year (more than 3000 h) and an important gross wind potential [International Renewable Energy Agency (IRENA 2021)].

However, the Tunisian energy mix is dominated by a non-renewable energy use. As reported by the International Energy Agency (IEA 2020), natural gas and oil dominate the Tunisian energy mix (production and consumption). In fact, natural gas and petroleum continued to increase in 2018, bringing the amount of energy to 5498 kilotons of oil equivalent (Ktoe) and 4664 Ktoe, respectively.

In contrast, total renewable resource production, excluding hydropower, gained 11% of the total energy production mix. Thus, the above target of the Tunisian government seems ambitious in terms of efforts made (Omri et al. 2019). However, Tunisia is still very far from optimally exploiting its solar and wind energy potential, while the developing countries have made great progress in recent years (Omri and Saadaoui 2022).

Otherwise, the Tunisian economy is disturbed by the continuous deterioration of its energy balance, which plummeted to −5678 ktoe in 2019 (the Ministry of Industry, Energy and Mines statistics 2020). The increase in total energy demand coupled with the depletion of hydrocarbon production in Tunisia may worsen the country’s energy balance.

This could have disastrous consequences for the economic situation. On the other hand, the rapid rise of energy imports underlines the country’s social and economic vulnerability to the volatility of international energy prices and the devaluation of the national currency (IRENA 2021). Therefore, the shift towards the use of accessible energy sources such as renewable energy will reduce imports of fossil fuels. This essentially leads to an improvement in the energy balance, as well as protecting the economy from energy price volatility, reducing energy poverty and confronting the adverse effects of climate change (Saadaoui and Chtourou 2022b).

Indeed, in this context, interest in accelerating the instruments of the transition to renewable technology has increased in recent years (Saadaoui 2022). Undoubtedly, a comprehensive dissection of the drivers of green technology can benefit policy makers to design efficient policies to accelerate the clean transition process. On the other hand, the literature links the mutation to renewable technology to environmental and economic factors (see Sadorsky 2009; Chang et al. 2009; Salim and Rafiq 2012; Silva et al. 2018; Ankrah and Lin 2020; Shahbaz et al. 2021; Samour et al. 2022b, etc.).

For their part, Wang et al. (2019) argue that capital deepening is an important factor influencing the modern energy transition. Moreover, the conversion from one type of energy to another demands a change in the shares of capital and labor (Kander et al. 2013; Wang et al. 2019). In this vein, Best (2017) pointed out that capital intensity differs significantly per energy type. The author indicated that there are different capital intensities of the sources of electricity generation in the USA. In fact, natural gas and coal have the lowest capital intensity, while renewable energy sources (wind and solar) have much higher capital intensity. In this context, labor demand and capital intensity differ by energy type. Therefore, an expansion in capital deepening will cause a technical shift towards modern capital-intensive technology (Antweiler et al. 2001; Wang et al. 2019). In contrast, Hou et al. (2021) found that technical change at the Portuguese firm level favors fossil fuels over green energy. For his part, Hötte (2020) found that at the microeconomic level, lower productivity of the provided capital is one of the barriers to the diffusion of clean energy.

In addition, recent empirical work has focused on how capital stock and labor affect renewable energy consumption. Namely, they have scrutinized the causality links between energy, capital, and labor in the context of an economic growth model (Stern 2010; Apergis and Payne 2011, 2012; Pao and Fu 2013; Ocal and Aslan 2013; Alper and Oguz 2016; Amri 2017; Kahia et al. 2017). Although, most studies have concentrated on the interaction between energy resources and economic performance, no alternative causal channels and potentially contradictory results have been found. In this context, Lien and Lee and Chien (2010) argued that the principal obstacle is that many studies disregard the important role of the capital stock. In fact, there are many other studies that evaluate the potential relationship between energy resources and capital.

However, Ma et al. (2009), Lin and Xie (2014), Lin and Ahmad (2016), Lin et al. (2017), and Saadaoui and Chtourou (2022b) agree with the substitution of energy and capital in the production process, while Berndt and wood (1975) and Griffin and Gregory (1976) recommend a complementary hypothesis between energy and capital.

Nevertheless, these previous studies have overlooked the direct link between the energy structure mutation and capital deepening, especially in the context of the developing countries.

Furthermore, our research will contribute to the existing literature by emphasizing the driving role of capital deepening in the diffusion of clean energies. Therefore, to achieve this goal, we used energy mix and capital intensity data for the period from 1990 to 2018 to analyze the long- and short-term relationship using the VECM and Johansen’s cointegration model. Moreover, this study examines the various linear and nonlinear causal links between these two factors.

This study is the first investigation to examine this dynamic relationship in the context of a developing country, such as Tunisia.

In fact, our choice of Tunisia as a case study is motivated by the indispensability of integrating the transition to renewable energies into the economic model of this country, since this type of energy is an accessible source. On the other hand, Tunisia is intended to achieve the “SDG” Sustainable Development Goals through the diffusion of renewable energy.

Therefore, it is useful for policy makers to analyze the factors that can affect this transition, including the contribution of capital deepening in order to achieve reliable recommendations.

Similarly, recent empirical studies focused on the context of Tunisia as an important case study to analyze sustainability factors related to energy and environmental quality (see Amri et al. 2018; Ben Lahouel et al. 2021; Saadaoui and Chtourou 2022a, b; Omri and Saadaoui 2022).

Indeed, sustainability is a strategic measure for Tunisia, which is considered an energy importing country and endowed with a huge energy potential.

The rest of this paper is structured as follows: “A brief overview of the Tunisian context” section highlights the economic and energy context in Tunisia. “Data specification and methodology” section shows the methodology and data used; results and main discussion are presented in the “Results and discussion” section; finally, conclusions and policy implications are exhibited in the “Conclusion and policy implications” section.

A brief overview of the Tunisian context

Economic context

Tunisia is a North African country bordered to the north and the east by the Mediterranean Sea (1148 km of coastline), Libya to the south (459 km), and Algeria to the west (965 km of common border). The Tunisian economy has been hit by structural shocks that threaten the country’s economic, political, and geopolitical stability. Indeed, the economic situation has changed significantly.

Figure 1 shows annual economic performance in Tunisia for the period from 1990 to 2020. Prior to 2011, average annual growth in Tunisia was about 5%. Despite high growth rates at one point, Tunisia experienced political changes in 2011 that revealed the inability of its economic model to withstand high youth unemployment and large regional inequalities. Subsequently, economic growth declined to −1.9% in 2011 due to political uncertainties and social unrest that affected tourism and foreign direct investment. In addition, the Libyan conflict negatively impacted the Tunisian economy through trade. Therefore, the Tunisian economy remained fragile during this period following the revolution and the global economic crisis. However, economic growth had recovered slightly and increased to 2% in 2017. Certainly, the improvement in security conditions contributed to the recovery of the tourism sector, which registered a 23% increase in 2017, and also to a resumption of GDP growth. Nevertheless, the rate of the current account deficit increased for the first time to 10% of GDP (10.1 billion Tunisian dinars or 10.3% of GDP). Moreover, the deterioration between 2013 and 2017 was glaring compared to the rate of 3.1% already reached between 2006 and 2010. However, in 2020, the growth rate will reach a very low rate of −8.6%. In 2020, the whole world, including Tunisia, will experience serious changes due to the pandemic COVID-19, which will greatly affect the Tunisian economy.

Fig. 1.

The annual GDP growth in Tunisia from 1990 to 2020.

Source: data collected from the World Bank (2021)

As for the trade balance, it recorded a deficit of 19,408.70 million dinars in 2019, mainly due to the increase in imports and the decrease in exports in some sectors. Figure 2 shows the situation of the trade balance by indicating the changes in imports, exports, and trade balance in the period from 1975 to 2019. Likewise, in recent years, the fiscal balance has recorded a permanent deficit, due to the decline in national production, the evolution of national consumption, and the decline in investment activity in the exploration and development of hydrocarbon sectors. In addition, since 2011, there has been a sharp decline in the volume of exports of phosphate and its derivatives, due to social tensions in the production sites of phosphate and its transportation.

Fig. 2.

The trade balance in Tunisia in million dinars from 1975 to 2019.

Source: data collected from INS (2020)

On the other hand, in recent years, we have seen an improvement in manufacturing exports, especially machinery and electrical, as demand from several European Union countries has increased. So, if we focus on the services balance, we notice significant fluctuations since 2011, due to the decline in tourism activities. In fact, after 2015, there was a slight progress due to the improvement of security conditions in the last 2 years. Moreover, the public debt reached 69.5% of GDP at the end of 2017. Moreover, the depreciation of the Tunisian dinar against the US dollar and the euro reached levels of 17.3% and 16.1% in 2017 and 2016, respectively. Foreign exchange reserves fell to US $5 billion, equivalent to 3 months of imports of goods and services. This level is considered the lowest since 2006. In addition, the inflation rate remained high at 6.7% in 2019. On the other hand, the unemployment rate remained high in 2019, as it stood at 14.9% in the fourth quarter of 2019, being higher among women (21.7%) than among men and young people. In particular, the unemployment rate among young graduates was twice the average rate (27.8%). According to the National Institute of Statistics (2020), the unemployment rate in the first quarter of 2020 was 15.1%.

The energy context in Tunisia

The energy sector helps drive Tunisia’s various economic sectors. Indeed, the favorable situation of the Tunisian energy sector has played a fundamental role in the Tunisian economy. In fact, until 1980, this sector contributed about 13% to GDP and 16% to national exports [Agence Nationale pour la Maîtrise de l'Énergie (ANME) 2004; Omri et al. 2015]. However, from the late 1980s, the situation in Tunisia changed due to a growing deficiency in the energy balance and, in particular, due to the dependence on fossil fuels (natural gas and oil) to meet the increased energy resource demand (IRENA 2021). On the other hand, according to the IEA statistical studies in 2020 (see Fig. 3), the imports of petroleum and natural gas fluctuated in the period from 1990 to 2017. In fact, the imports of natural gas and petroleum in 1994 amounted to 69,374 terajoules (Tj) and 664 Ktoe, respectively. Moreover, the imported amount of natural gas has been continuously increasing since 1999 until it reached 153,343 Tj in 2017. As for the import of petroleum, it was the lowest in 2010 (184 Ktoe), while it reached a value of 635 Ktoe in 2017. Therefore, the increase in energy imports contributed to a deficiency in the total energy balance that reached 5672 Ktoe, noting that 49% of the total primary energy consumed in Tunisia is imported (IRENA 2021).

Fig. 3.

The oil and natural gas imports.

Source: data collected from IEA (2020)

On the other hand, the electricity generation mix shown in Fig. 4 indicates that Tunisia is on the verge of being fully electrified by natural gas. In fact, electricity generated from gas increased from 3700 gigawatt-hours (GWH) to 19,667 GWH between 1990 and 2019, demonstrating the massive use of natural gas in electricity generation. Therefore, the high dependence on natural gas may have some serious consequences, especially for energy security, as natural gas production in Tunisia has stagnated or decreased in recent years. According to IRENA (2021), national production of oil and natural gas has decreased by 54% and 47%, respectively, since 2010. On the other hand, electricity generation from oil has deteriorated over the period. In 2018, it even decreased to 46 GWH. Moreover, the share of renewable energy in the electricity mix is still very low, reaching 740 GWH in 2019 (of which 66 GWH from hydropower, 500 from wind energy, and 174 from solar energy). Moreover, Tunisia has made a gradual net inclusion of wind and solar energy in its electricity mix since 2010.

Fig. 4.

Electricity generation mix in Tunisia for the period 1990–2019.

Source: data collected from IEA (2020)

In fact, Tunisia has more than 3000 h of sunshine per year. ANME estimates the exploitable potential of photovoltaics in Tunisia at several hundred gigawatts (GW). The average global horizontal radiation (GHI) is equivalent to 1850 kWh/m², which can generate about of 1650 kWh/kWp annual production from solar “PV” systems [Gesellschaft fuer Internationale Zusammenarbeit (GIZ) (2019). We can therefore consider solar energy as a key element to consolidate energy balance and environmental protection.

Moreover, some regions of Tunisia have favorable wind conditions with 7 m/s at 60 m (GIZ 2019). This renewable energy potential, especially solar and wind energy, is expected to meet 30% of electricity demand by 2030 through clean resources. To accelerate these processes, Tunisia has actively participated in the international momentum to fight climate change and has therefore fulfilled all its commitments to the United Nations Framework Convention on Climate Change (UNFCCC) (Omri et al. 2015).

Tunisia is indeed considered one of the few developing countries to have adopted incentive policies such as the voluntary energy control policy in the mid-1980s. This initiative anticipated the energy deficit identified in the mid-1990s. The energy strategy then gained momentum in the mid-2000s after several increases in oil prices and their derivatives, in addition to a deterioration in the energy balance. Moreover, energy efficiency is considered one of the strategic pillars. Certainly, the rise in oil prices in the mid-1970s increased the importance of this strategy. Indeed, the period from 1960 to 1980 was marked by the development of supply and the establishment of an institutional structure of the energy sector through the creation of the Tunisian Electricity and Gas Company [Société tunisienne de l'électricité et du gaz (STEG)] and the Tunisian Company for Petroleum Activities [Entreprise tunisienne d'activités pétrolières (ETAP)]. Then, in 1985, Tunisia created ANME to support the implementation of energy control policies. Some institutional reforms were then carried out in the 1990s. In addition, the 2000 period was characterized by private investment in electricity generation, mainly from natural gas. This period was characterized by the incentive to use renewable energy as part of the energy management policy in conjunction with energy efficiency. In addition, Tunisia has enacted a number of laws and decrees to develop renewable energy, the most important of which are listed in Table 1, the 2004–72 Law on Energy Management and the Use of Renewable Energy, which prioritizes solar and wind energies.

Table 1.

The legislative background

	The legislative background
2004	Law 2004–72 on energy efficiency
2005	Law 2005–82 on the construction of the Energy Efficiency Fund, supporting clean development mechanisms
2009	Law 2009_7 on renewable energy and energy efficiency: introduction of a regime of self-consumption of renewable energy (this law completes the law of 2004, with a right to sell the surplus to the STEG within the limit of 30%)
2015	Law 2015–12 describes the legal framework for electricity generation from renewable energy sources. An incentive to independents for renewable energy production and liberalization of the export of electricity from renewable sources using 3 schemes: self-consumption, independent production of electricity for domestic consumption and export
2016	- Decree 2016–1123 defines the conditions and procedures regarding the implementation of projects and sale of electricity from renewable sources
2017	Order of February 9, 2017: Order supplementing the 2015 law establishing: -The specifications for connection The contract for self-generation in low voltage (net- metering) The contract for self-generation in high voltage/medium voltage -The PPA (Purchase Power Agreement) for the authorization regime
2018	Order of August 30, 2018: Carrying approval of the revision of the standard contract of sale to STEG of electrical energy produced from renewable energy subject to authorization

Source: collected from GIZ (2019)

Furthermore, this law was amended and supplemented by Law 2009–7, which now allows STEG to buy back the surpluses of self-generators of electricity (up to 30%). In addition, the 2005 Law created the National Fund for Energy Management, which is a financial support system for the development of renewable energy resources. In addition, Law No. 2015–12 on the total electricity production from renewable sources corresponds to the main test in the field of clean energy in Tunisia. This law was promulgated on May 11, 2015 and regulates the implementation of renewable electricity generation projects. It describes the national plan for the electricity generation from renewable energy sources and establishes the framework for the development of projects. This law also describes the role of the technical commission for private electricity generation from decommissioning obligations of the facilities, the procedures of control and violations, as well as the role of the specialized authority in charge of examining the problems related to the projects of electricity generation from renewable energy sources.

Data specification and methodology

Data specification

The data set selected in this study involves the renewable energy transition (in Eq. 1) and the capital intensity (in Eq. 2) for the Tunisian economy for the period spending between 1990 and 2018.

RET is used to express the renewable energy transition. In fact, Tunisia has endorsed various strategies in order to maximize the use of its huge potential from renewable energies (prosocial, prosocial elec, etc.). Accordingly, these plainings have yielded to promoting the part of renewable energies’ resources, while the energy mix in Tunisia is still mainly dominated by fossil fuel consumption (oil and natural gas). In our case, we measure the renewable energy transition (RET) by the relative share of renewable energy and fossil energy (Eq. 1).

$$RET=\frac{{\mathit{share}}_{\mathit{ER}}}{{\mathit{share}}_{\mathit{ENR}}}=\frac{{}^{{\mathit{Q}}_{\mathit{E}\mathit{R}}}\left./,{}_{\mathit{Q}}\right)}{{}^{{\mathit{Q}}_{\mathit{E}\mathit{N}\mathit{R}}}\left./,{}_{\mathit{Q}}\right)}$$ 1

From Eq. (1), ${\mathrm{share}}_{\mathit{ER}}$ is part of renewable energy consumption from the total energy mix in Tunisia. The share of clean energy presents the quantity of clean energy $(Q_{\mathit{ER}})$ divided by the entire energy quantity $(Q$) in Tunisia. Similarly, ${\mathrm{share}}_{\mathit{ENR}}$ represents the quantity of fossil fuel ($Q_{\mathit{ENR}})$ divided by the total energy quantity $\left(Q\right).$ These previous variables are extracted from IEA.

The explanatory variable in our model is the capital intensity (CI), which is determined by the increasing of the ratio (capital/labor) of the economy (Antony 2009; Chen et al. 2020). This ratio designs the continuous elevation of capital factors linked to a unit labor factors (Eq. 2):

$${\mathit{CI}}_{\mathrm{t}}=\frac{{\mathrm{K}}_{\mathrm{t}}}{{\mathrm{L}}_{\mathrm{t}}}$$ 2

where ${\mathrm{K}}_{\mathrm{t}}$ is the capital stock and L is the labor. The capital intensity measurement requires data on labor and capital stock (K). The labor (L) is the active population extracted from the database of the World Bank.

For the capital stock, it is measured using the perpetual inventory technique, which is presented in the following equation:

$$K_t=K_{t-1}\left(1,-,\delta \right)+I_t$$ 3

In this equation, $K_t$ represents the current capital stock at the constant prices (2010 = 100), then $K_{t-1}$ refers to the previews year of capital stock “ t-1,” δ is the rate of the depreciation of the capital, and $I_t$ exposes the gross fixed capital formation of the current year “t.” Concerning the depreciation rate, it is equivalent to 5%.

The stock of initial capital ${}^{\prime \prime }{K}_0^{\prime \prime }$ is determined in the following equation:

$$K_0=\frac{I_0}{g+\delta }$$ 4

$K_0$ is the initial capital stock, δ represents the rate depreciation of capital, $I_0$ is the initial capital investment, and g refers to the average growth rate of capital investment for the period 1990/2018.

Table 2 reports the main descriptive statistics for capital intensity (lnCI) and energy transition (lnRET) in natural logarithm form of and in Tunisia from 1990 to 2018.

Table 2.

Stochastic properties of the variables

	lnCI	lnRET
Mean	9.970	−1.692
Maximum	10.347	−1.494
Minimum	9.580	−1.890
Std. dev	0.269	0.086
Skewness	0.046	−0.569
Kurtosis	1.555	3.496
Jarque–Bera	2.531	1.864
	[0.282]	[0.393]

p-values are in []

Moreover, Fig. 5 illustrates the distribution of renewable capital intensity (lnCI) and renewable energy transition (lnRET) in Tunisia from 1990 to 2018.

Fig. 5.

Trends of variables

VECM specification

This empirical study explores the long run and short run links among energy transition and capital intensity based on the case of Tunisia over the period 1990–2018. To do so, the VECM model and the Johansen test for the cointegration technique are adopted. The VECM technique is usually used to examine the sense of causality between the components of the model while giving estimates for the long- and short-term. In the case of the existence of cointegration, a VECM model specification is represented in Eq. 5:

$$\mathrm{\Delta }{Y}_t=C+{\sum }_{k=1}^{n-1}{\omega }_i\mathrm{\Delta }{Y}_{t-1}+{\sigma }_i{Y}_{t-1}+{\rho }_i{X}_t+{\mu }_t$$ 5

The $\mathrm{\Delta }$ represents the difference operator, $Y_t$ is the vector of endogenous variable. ${\omega }_i$ and ${\rho }_i$ are the coefficient matrices of both endogenous and exogenous components, respectively. Moreover, ${\omega }_i$ is the matrices that represent the dynamics of the the short run of the model, and ${\sigma }_i$ captures the cointegration in the model, which show the long run dynamics. And ${\mu }_t$ designates the residuals in our model.

Linear and nonlinear causality

The last step of our empirical study concerns the identification of the existing causality between the different variables. In this context, the Toda and Yamamoto (1995) test and the Diks and Panchenko (2006) test are applied.

Concerning the linear causality test, it is based on an adjustment in the Wald test in an augmented VAR model. Toda and Yamamoto (1995) reported that the distribution of the modification in Wald statistic rides to a chi-square random variable (stationary or not).

The unrestricted VAR model is as follows:

$Y_t=\mathrm{\beta }+{\theta }_1{Y}_{t-1}+\cdots \cdots +{\theta }_p{Y}_{t-p}+{\mu }_t$ are n-dimensional vectors, and ${\theta }_p$ is an n × n matrix of parameters for the lag p. The adjustment presented in Toda and Yamamoto’s method attempts to extend the real lag dimension of the unrestricted bivariate VAR model by a maximum integration order.

However, the linear Granger causality test does not take into account the non-linearity observed in the dynamics time series. It is obvious that macroeconomic and financial variables exhibit non-linear behavior over time (Omri and Saadaoui 2022). Neglecting these non-linearity dynamics can lead to misidentification of the causal relationship between two variables or otherwise reduce the estimation effectiveness of the causality test. Although these types of linear Granger tests (Toda and Yamamoto causality) are widely used in empirical work on the energy and economic growth relationship, they are considered insufficient, especially in the case of the increasing likelihood of complex links, such as those between energy resources production or consumption determinants. Baek and Brock (1992) indicated that these linear tests have lower effect compared to some alternative non-linear tests. Some of Granger’s best common nonparametric causality tests are those featured in Hiemstra and Jones (1994), Diks and Panchenko (2006), and Su and White (2008).

Although the Hiemstra-Jones test of causality is extensively applied (Kumar 2017; Mishra et al. 2019), Diks and Panchenko (2005) noted that this test has some limitations which can contribute to serious over-rejection of the null assumption. Therefore, the causality test suggested by Diks and Panchenko (2006) is approved (Massa and Rosellón 2020).

Diks and Panchenko (2006) developed a nonparametric, nonlinear method of Granger causality, which provides more vigorous information about the causal links between variables (Rahimi et al. 2016).

The null hypotheses of the approach of Dicks and Pancheko (2006) examine that past information of $X_t^{\mathit{ix}}$ encompasses any additional observations about $Y_{t+1}$ (beyond that in $Y_t^{\mathit{iy}}$):

6

The following equation represents the test statistic:

$$T_n\left({\epsilon }_n\right)=\frac{n-1}{n\left(n,-,2\right)}{\sum }_i(f_{X,Z,Y}(X,{\mathrm{Y}}_{\mathrm{i}},{\mathrm{Z}}_{\mathrm{i}})/{\mathrm{f}}_{\mathrm{y}}(Y_{\mathrm{i}})-{\mathrm{f}}_{\mathrm{x},\mathrm{y}}({\mathrm{X}}_{\mathrm{i}},{\mathrm{Y}}_{\mathrm{i}}){\mathrm{f}}_{\mathrm{y},\mathrm{x}}({\mathrm{Y}}_{\mathrm{i}},{\mathrm{Z}}_{\mathrm{i}})$$ 7

where f _x,y,z (X,Y_, Z) is joint probability density function for l_x= l_y=1 and if ${\epsilon }_n$=Cn ^–β (C > 0, $\frac{1}{4}$ <β < $\frac{1}{3}$), Diks and Panchenko (2006) prove that the test statistic in Eq. (7) satisfies the following:

$$\surd n\frac{(T_n\left({\epsilon }_n\right)-q)}{S_n}{\to }^D\mathrm{N}(0,1)$$ 8

where ${\to }^D$ signifies convergence in distribution, and $S_n$ is an estimator of the asymptotic variance of $T_n$ (.)

Results and discussion

This section presents the various results of this empirical study.

Test results for unit roots

As the first step, the statistical tests of Dickey–Fuller augmented (ADF) (Dickey et Fuller 1979) and Zivot–Andrews (ZA) (Zivot and Andrews 1992) unit root tests are applied to examine the order of integration. The null hypothesis indicates the presence of a unit root in the series. Table 3 illustrates the results for the static tests (both ADF and ZA). The ADF unit root test results confirm the non-stationarity of all series at the level. Furthermore, they are found to be stationary at I (1).

Table 3.

Order of integration

Variables	ADF	ZA
lnRET	−1.619	−4.281 (2010)
lnCI	−1.619	−2.269 (2008)
$\mathrm{\Delta }$ nRET	−8.101*	−10.700* (2010)
$\mathrm{\Delta }$ lnCI	−8.101*	−5.950* (2008)

^* Indicates the significativity at the level of 1%, and breaks years are in ()

The exceptional result of the traditional stationarity test, such as ADF in our case, does not take into account a structural break in the time series data. This omission of structural breaks can have a significant effect on the order of integration (Saadaoui and Chtourou 2022a). Indeed, the development of renewable energy deployment in Tunisia has changed since establishing national programs (energy management). Similarly, the global financial crisis of 2008 and the period of the revolution since 2011 could also cause a discontinuity in increasing capital intensity. Furthermore, the ZA unit root test is applied to capture the possible structural break in our model. The ZA test confirms our above result concerning the order of integration of lnCI and lnRET, and approves the structural point existing in 2010 and 2008 for both lnRET and lnCI, respectively.

The unique order of integration of the variables (ADF and ZA tests) helps us apply the Johansen cointegration and VECM model to examine the short and long run relationships between the two variables.

Test results for cointegration test

After obtaining variables of the same order, the next step is the determination of the number of cointegrating relationships between the different variables using the approach of Johansen (1991). This step requires determining the optimal number of lags in estimating the VAR model using the information criteria, namely the Akaike information criterion “AIC” (Akaike 1974), the Hannan and Quinn information criterion “HQIC” (Hannan and Quinn 1979), and Schwarz’s Bayesian information criterion “SBIC” (Schwarz 1978). The results provided by these criteria are illustrated in Table 4. The lag equivalent to 2 is selected as an optimal lag order based on the information given by the AIC and HQIC.

Table 4.

Lag selection order

Lag	0	1	2
AIC	−2.815740	−9.794913	−9.889621*
SBIC	−2.717569	−9.500400*	−9.398765
HQIC	−2.789696	−9.716779	−9.759397*

To check the number of cointegration vectors, we use the Johansen trace tests. Table 5 shows that at the 5% level of significance, the trace test declines the null assumption of absence of cointegration, although it does not reject the null assumption of the existence of at most one cointegrating equation. We, therefore, accept the null hypothesis that there is a single cointegration equation between the energy transition and capital intensity.

Table 5.

Johansen cointegration rank test

Maximum rank	Eigenvalue	Trace	5% critical value	λ-max	5% critical value
0	1	29.3711	15.41	26.2510	14.07
1	0.64363	3.1200	3.76	3.1200	3.76

The trace test shows 1 cointegrating relationship at 5% threshold. The maximum eigenvalue test highlights one cointegrating result at the 5% threshold

Test results for VECM model

The results in Table 6 describe the interaction between the variables from the VECM estimation. Moreover, the error correction term “ECT” is negative and significant, which supports the existence of a long-term relationship between energy transition and capital intensity for the lnRET = lnCI model. The “ECT” implies the existence of a speedy adjustment towards equilibrium. The speed at which “lnRET = lnCI” adjusts towards the equilibrium is 0.672. Therefore, when there is an exogenous shock in the system, “lnRET = lnCI” corrects its state of disequilibrium by 0.672 speed of adjustment per year.

Table 6.

The VECM results

	Coefficients	Std. err	P-value
Short run analysis
Dependent variable: Δ lnRET
ECT	−0.672**	0.323	0.037
Δ lnRET(-1)	−0.068	0.248	0.783
Δ lnCI(-1)	2.436***	1.481	0.100
C	0.001	0.031	0.963
Dependent variable: ΔCI
ECT	0.154*	0.043	0,000
Δ lnRET(-1)	−0.054***	0.033	0.104
Δ lnCI(-1)	0.098	1.983	0.619
C	0.006	0.004	0.136
Long run analysis
Dependent variable: lnTE
LnCI	0.136*	0.023	0.000

^* and *** indicate the significant level at 1% and 10% respectively

On the other, in the short run, the results revealed that a 1% augmentation in capital intensity increases the clean energy transition by 2.436%.

Therefore, from the long run Johansen’s cointegration equations, we realize that the rise in capital intensity enhances the renewable transition by 0.136%, at 1% level of significance, in the long run. Moreover, our findings suggest that any increase of capital intensity in Tunisia boosts renewable energy transition in both the long and short run, whereas its effect is greater in the short-term than in the long run. Moreover, this finding recommends that the response of the total energy structure to the adjustment in the ratio of capital intensity is more reactive in the short than in the long run.

In fact, this positive relationship between capital intensity and the transition to renewable energies underlines that renewable energies are capital-intensive. This implies that a deepening of the capital factor, i.e., favoring the capital factor over the labor one, leads to directing technical change towards capital-intensive energies, which are the clean energies in the long- and short-term. Therefore, this finding is in line with that of Wang et al. (2019), who found that capital intensity stimulates modern energy transition in the long run but not in the short run for China from 1978 to 2015. Moreover, Savina et al. (2021) argue that capital intensity influences the solar energy market.

In the same sense, Abbas et al. (2020) pointed out that the augmentation in fixed capital formation, as a consequence of technological innovation, can generate economic profits without damaging the environment in the “Belt and Road Initiative” selected countries. The authors also indicate that in the long-term, it is necessary to invest in technological development in order to identify alternative types of energies, such as the nuclear, hydroelectricity, the solar energy, and geothermal. Likewise, our results can be compared with those of previous papers that dealt with the effect of productivity and the economic factors on the sustainability factors. In fact, the total factor productivity is considered a proxy of technical change and economic growth. In this direction, Amri (2018) found that productivity factors increase the CO₂ emission in Tunisia both in the long and short run between 1975 and 2014. Moreover, the same result is proved by Amri et al. (2019) for the same case study. In addition, Sadorsky (2009) found a positive link between the economic income and renewable energy transition for a sample of G7 countries between 1980 and 2005. For their part, Apergis and Payne (2010) argue that fixed capital formation, GDP, and labor positively impact renewable energy consumption for 20 OECD countries between 1985 and 2005. Moreover, Samour et al. (2022a, b) showed a positive impact of the economic factor on renewable consumption in the UAE. Recently, Omri and Saadaoui (2022) have demonstrated that the total factor production drives the total consumption on renewable energy in Tunisia.

Nerveless, these results are contrasted with those of Shahbaz et al. (2021), which indicate that the economic income hampers the renewable energy consumption for 34 upper middle income developing countries. The same result appeared in the research of Uzar (2020) for 38 countries.

Furthermore, we apply the impulse response analysis for the VECM model to better understand the relationship between the renewable energy transition and capital intensity.

Figure 6 represents the response of the renewable energy transition to a one-time shock in capital intensity at the time. This function shows that the response will be stable after 5 years, so an adjustment period does not exceed 5 years.

Fig. 6.

The impulse response functions for VECM model

Diagnostic tests

To evaluate the stability and effectiveness of our model, we use the Lagrange-multiplier test, the normality test, and the stability of the model in Table 7 and Fig. 7.

Table 7.

Lagrange multiplier test for serial correlation and normality check

	Normality check
	Skewness/kurtosis	Chi2	df	Prob > Chi2
Jarque–Bera test	–	1.818	2	0.40295
Skewness test	−0.39011	0.685	1	0.40793
Kurtosis test	4.0036	1.133	1	0.28713
Lag	Lagrange multiplier test for serial correlation
	Chi2	df	Prob > Chi2
1	4.1275	4	0.38902
2	2.7886	4	0.59380

Fig. 7.

Residual stability test of ECM model

The serial correlation test demonstrates that the errors are serially uncorrelated due to the probability is greater than 1%, 5%, and 10% (level of significance). The Jarque Bera test for normality shows evidence of normality for the model. The residues follow a normal distribution, which is favorable for our model.

Moreover, the stability condition test reveals that all the points are included in the circle, and none of them are outside. This indication means that our model is stable.

Causality results

Our empirical study’s final phase consists in determining whether there is any causal relationship between the various variables. For this purpose, we use Diks and Panchenko’s (2006) test as well as the Toda and Yamamoto’s (1995) test.

In fact, Table 8 displays the results of the linear Granger causality test. More precisely, it illustrates the causality running from capital intensity to the shift to renewable energy at the level of 5%. Furthermore, this unidirectional relationship between capital intensity and energy transition shows how improving capital intensity can advance the transition to renewable energy. Therefore, investment can facilitate the transition to renewable energy by encouraging technological advances in the sector as well as promoting capital deepening. Therefore, this result is comparable to that of Wang et al. (2019), who found a unidirectional causal relationship, in the case of China, between capital intensity and the modern energy transition.

Table 8.

Results of the Toda and Yamamoto (1995) causality test

H0	F-statistics	P-value
lnRET cannot Granger cause lnCI	4.269	0.118
lnCI cannot nonlinearly Granger cause lnRET	6.263**	0.043

^** Indicates rejection of the null hypothesis of no causality at the level of 5%

This finding is contrast with Apergis and Payne (2010), who discovered a bidirectional causal relationship between capital and renewable energy.

In fact, this finding is similar to others that investigate the relationship between renewable energy and economic growth. In this vein, Omri (2014) demonstrated that the empirical literature on the causal relationship between renewable energy and economic growth produces a wide range of results. In fact, the feedback hypothesis is confirmed by 33%. Moreover, 11% of the data support the growth concept, while the neutrality supposition accounts for 22%, which implies that the conservation assumption is around 34% in the end.

As a next step, we perform further the non-linear Granger causality test “Diks and Panchenko (2006)” to examine the nonlinear causal relationship between energy transition and capital intensity. The bandwidth was suggested at “0.5,” “1,” and “1.5,” and the embedding dimensions were selected at “1” and “2.” It can be seen in Table 9 that the non-linear Granger causality test result suggests that there is unidirectional causality running from capital intensity to renewable energy transition. The non-linear granger causality supports the result of the Toda and Yamamoto causality.

Table 9.

Results of the Diks and Panchenko (2006) causality test

ED	m = 1			m = 2			Conclusion
B	0.5	1	1.5	0.5	1	1.5
lnRET =  > lnCI	0.321	0.159	1.703	0.278	0.479	−1.097	No causality
lnCI =  > lnRET	0.000	1.368***	1.416***	−0.564	1.607**	1.647**	Causality

^** and *** indicate rejection of the null hypothesis of no causality at the 5% and 10% significance level, respectively. ED and B indicate embedding dimensions and bandwidth respectively

Conclusion and policy implications

This study scrutinizes the short run and long run relationship between the energy structure and capital intensity for the case of Tunisia and investigates the potential causality between the previous variables for the 1990–2018 period. For this purpose, the VECM model and the Johansen cointegration test are employed to examine the cointegration relationship among the variables. Besides, this study utilizes Toda and Yamamoto (1995), and the Diks and Panchenko’s (2006) tests to identify the linear and non-linear causality links, respectively. The results of the Johansen cointegration test show evidence of a long run cointegration relationship between the renewable energy transition and capital intensity. Likewise, the result of the VECM implies that the substitution in the energy mix in Tunisia from fossil to renewable energy is consistent with capital deepening in the long run and short run, while this effect is more serious in the short run. An increase in capital intensity promotes technical changes in the energy sector, favoring capital intensive energy, which is renewable energy. Using the most expensive factor, the capital, instead of labor factor, allows directing the technical change towards renewable energy, which forms a capital-intensive technology. The impulse response function demonstrates that a shock on the capital deepening will positively impact the renewable energy mutation when the adjustment period is 5 years.

Finally, both linear and non-linear Granger causality test results show that one-way unidirectional causality runs from capital deepening to the renewable energy transition.

On the other hand, from a policy perspective, our findings suggest that Tunisian policymakers may apply initiative policies to enhance renewable energy substitution in both the long and the short run by boosting capital deepening. In other words, they can use incentive policies to enhance capital intensity, which may ameliorate the investment in renewable energy. Therefore, capital deepening policies help to directly stimulate investment, which can lead to direct biased technical change toward more capital-intensive renewable energy. In this case, the industrial policy significantly stimulates investment in renewable energy and encourages investors to create projects and invest in clean sources, which is well demonstrated by Pegels and Lütkenhorst (2014).

Furthermore, Tunisia can orient the investment to solar energy, which represents a promising source in the future. Therefore, this sector can stimulate economic growth and ameliorate the energy balance situation by satisfying the domestic demand, and also enables exporting electricity from renewable sources to the European countries (Omri et al. 2015). It is worth mentioning that Tunisia has not completed its achievements in this sector, compared with its great potential. In practice, the installed cumulative capacity of solar photovoltaic represented about 47.1 megawatts (MWe) of electricity in 2017 (Omri et al. 2019). Therefore, policymakers must continue promoting investment in the solar sector to better take advantage of its potential long-term benefits. It should be noted that Tunisia has made strategies in this sector. For example, Tunisia adopted the Law 2015–12 to encourage private investors to invest in clean technologies (Saadaoui and Chtourou 2022a, b). Indeed, in 2005, Tunisia launched first an ambitious program called the solar program to enhance the investment and diffusion of solar energy, then, the strategy to 2030 to stimulate energy efficiency and investment in renewable energy. These incentive policies can encourage investment and stimulate technological innovation to support the solar sector. Moreover, the government can adopt a price adjustment policy, which reflects the internal and external energy costs to ensure that energy prices are rationally determined. On the other hand, the economic theory explains that price intervention internalizes the positive externalities of green energy against traditional energy. In addition, price adjustment can generate efficient basic research and development levels by private firms (Morris et al. 2012).

In fact, the price change can encourage investment in renewable energy. For Tunisia, it has put in place the first political and financial support instruments by granting premiums and subsidies acquire solar energy production equipment in the residential sector (Giz 2015). These policies can affect the choice of investors to finance renewable energy projects when the capital intensity of the types of energy changes due to price changes. Moreover, Tunisia has a competitive advantage in adjusting its energy policy by adopting a feed-in tariff mechanism instead of fossil fuel subsidies. One of the efficient mechanisms in subsidizing electricity production is the feed-in tariffs instrument. This mechanism encourages investment and the diffusion of renewable energies instead of fossil energies. However, this change must be progressive in order not to penalize the purchasing power of the consumers.

Therefore, future research can contribute to the existing literature by analyzing the effect of green capital on renewable energy transition, similarly to recent studies using, for example, the green total factor productivity (see Xie and Zhang 2021) or sustainable factor productivity (see Shen et al. 2020).

Author contribution

Haifa Saadaoui: data collection, software, investigation, conceptualization, methodology, and language editing. Nouri Chtourou: conceptualization, methodology, investigation, language editing, and revision.

Data availability

All data is provided in full in the “Data specification and methodology” section of this manuscript.

Declarations

Ethics approval

Not applicable

Consent to participate

Not applicable

Consent for publication

Not applicable

Conflict of interest

The authors declare no competing interests.