The role of renewable and non-renewable energy consumption in CO₂ emissions: a disaggregate analysis of Pakistan

Abstract

The energy sector has become the largest contributor to greenhouse gas (GHG) emissions. Among these GHG emissions, most threatening is CO₂ emission which comes from the consumption of fossil fuels. This empirical work analyzes the roles of renewable energy consumption and non-renewable energy consumption in CO₂ emissions in Pakistan. The empirical evidence is based on an auto-regressive distributive lag (ARDL) model of data from 1970 to 2016. The disaggregate analysis reveals that renewable energy consumption has an insignificant impact on CO₂ emission in Pakistan and that, in the non-renewable energy model, natural gas and coal are the main contributors to the level of pollution in Pakistan. Economic growth positively contributes to CO₂ emission in the renewable energy model but not in the non-renewable energy model. Policies that emphasize the contribution of renewable energy to economic growth and that add more clean energy into the energy mix are suggested.

Introduction

Energy sector is responsible for 75% of the global GHG emissions (International Energy Association 2015). Carbon dioxide (CO₂) emissions have increased over the years due to continuous rise in the global energy demand and have severe implications for the environment and a significant contributor to global climate change. In an effort to address the increasing concerns about climate change, the Conference of Parties (COP) of the United Nations Framework Convention for Climate Change (UNFCCC) agreed to limit the increase in the global temperature to 2 °C above pre-industrial levels by 2020 in 2015 (UNFCCC 2015). Since this target cannot be achieved until the pattern of energy consumption is changed, therefore, combating climate change with sustainable development has become an essential global agenda in planning for energy production and consumption. An economy may turn to a sustainable track if it uses a mixture of renewable and non-renewable energy resources (Dogan 2016). Therefore, policymakers must know the individual contributions of energy sources (renewable and non-renewable) on economic growth and CO₂ emissions.

Numerous studies concerned with energy consumption, economic growth, and environmental degradation such as Shahbaz et al. (2012); Alkhathlan and Javid (2015); and Ibrahiem (2015) concluded that high levels of energy consumption are central to economic growth while at the same time, they have a tendency to deteriorate the environment in developing and developed economies (Azad et al. 2015).

There has been a continuous increase in energy use for developing countries during recent years to achieve higher levels of living standards and economic development (Shahbaz et al. 2012). Attaining higher ladders of economic development at the cost of natural environment is never desirable. Therefore, examining the role of renewable and non-renewable energy consumption in CO₂ emissions has remained debatable in empirical literature due to differences in data sets, regions, and research methodologies employed (Mirza and Kanwal 2017). Our study attempts to find more evidence on causal relationships between renewable and non-renewable energy consumption, CO₂ emissions, and economic growth concurrently in a single study for the case of Pakistan.

Pakistan is an interesting case study for this empirical work as its share of energy-led emissions is increasing. As per the Global Climate Risk Index (2018), Pakistan has been one of the most affected countries due to climate during the last two decades. The impact of climate change compared to the country’s diminutive per capita GHG emissions has been very high in Pakistan (Abas et al. 2017). Moreover, Pakistan has faced acute energy shortages from 2007 and onwards that have adversely affected its economic growth (Komal and Abbas 2015). To address these energy shortages, Pakistan has resorted primarily to non-renewable energy-producing sources, which are the main contributors to the country’s CO₂ emissions¹ (Nasir and Ur Rehman 2011). During the last three years, Pakistan has initiated seven energy projects that are based on coal consumption and will further add to GHG emissions in the country. Although the country has set numerous goals and strategies to encourage consumption from renewable resources, the energy sector is ill-managed and the share of renewable energy consumption is very small. Despite the initiation of renewable energy policy in 2006 and the existence of huge potential² for renewable energy production, no time path is available to achieve sustainable energy development in Pakistan.

Therefore, having a large population and being one of the major contributors to GHG emissions among the developing countries, Pakistan is an ideal candidate for an exclusive study that examines the environmental and growth effects of any possible fuel substitutions in the coming years. In this paper, we attempt to carry out a disaggregated analysis to test for the existence of long-run and short-run relationship between individual energy consumption sources, CO₂ emissions, and economic growth. We also implement causality tests to study the direction of causality between these variables to suggest optimal policies. Analysis of renewable and non-renewable energy consumption by source at disaggregate levels facilitates the examination of the relationships among each source of energy consumption, economic growth, and CO₂ emissions. In addition, research at disaggregate level is essential for examining the barriers to replacing traditional energy resources with newer ones, along the lines of Greiner et al. (2018) who investigated whether natural gas consumption can mitigate CO₂ emissions produced from coal consumption.

Novelty of our study relative to the existing literature lies mainly in the difference in analytical perspective. Previous studies on Pakistan have investigated aggregate relationships among selected variables; however, this study examines the role of different renewable and non-renewable energy sources in CO₂ emissions. With disaggregated level analysis, we are able to compare the individual impact of renewable and non-renewable energy consumption on CO₂ emissions and economic growth. Our contribution also includes a comparative assessment of renewable and non-renewable consumption in a holistic manner to suggest a comprehensive policy framework towards CO₂ emission reduction. The analysis provides valuable information for policymakers to construct an optimal combination of renewable and non-renewable sources in order to meet the national demand.

The remainder of the paper is distributed as follows. The “Literature review” section reviews related work in the literature. The “Methodology and data” section describes the study’s data collection and econometric approach. The results and discussion are presented in the “Empirical analysis and discussion of results” section, while the “Conclusion” section provides a conclusion.

Literature review

The relationships among renewable and non-renewable energy sources, economic growth, and CO₂ emissions have been investigated in many studies. Many have used panel country data to investigate these relationships. For example, Apergis and Payne (2011a) conducted a study of 80 countries and found bidirectional causality between renewable energy consumption and economic growth and between non-renewable energy consumption and economic growth. The same results were reported by Tugcu et al. (2012), who used the auto-regressive distributive lag (ARDL) approach to assess the classical production function in the G-7 countries. Using fully modified ordinary least square (FMOLS), Sadorsky (2009) found a positive relationship between renewable energy consumption and real income per capita in 18 countries. By employing similar methods to a group of 69 countries, Ben Jebli and Ben Youssef (2015) validated the growth hypothesis for both renewable and non-renewable energy consumption. These results were later supported by Wesseh and Lin (2016a), who estimated a translog model for 34 African countries. Kahia et al. (2016) explored an energy-growth nexus for Middle East and North African (MENA) countries using data from 1980 to 2012. Applying a panel cointegration technique, they reported bidirectional causality between renewable and non-renewable energy consumption, found significant but negative short-run coefficients, and specified substitutability of renewable and non-renewable energy resources. Similarly, Bhattacharya et al. (2016) applied a heterogeneous panel Granger causality test to multiple countries and found no causality between renewable energy consumption and GDP. Ito (2017) applied a GMM model to 42 developing countries and argued that renewable consumption reduces CO₂ emissions and has a positive influence on economic growth in the long run.

Other studies have used single-country data sets. For instance, Dogan (2016) investigated the case of Turkey by applying a vector error correction model (VECM) Granger causality test with a structural break and found bidirectional long-run and short-run causality between non-renewable energy consumption and GDP. They also reported one-way short-run causality from GDP towards renewable energy consumption, while in long run they found bidirectional causality. Using the Toda-Yamamoto causality method for Italy, Vaona (2012) supported the feedback hypothesis for non-renewable energy consumption and growth while neutrality hypothesis for the relationship between renewable energy consumption and economic growth. Apergis and Payne (2014) used the Toda-Yamamoto causality technique and found no relation among renewable, non-renewable energy consumption and economic growth in the USA. Table 6 shown in the Annexure lists the literature on renewable and non-renewable energy consumption, CO₂ emissions, and economic growth.

Some studies on Pakistan have assessed the role of renewable and non-renewable energy consumption on economic growth and CO₂ emissions by applying various econometric models. Studies on Pakistan’s economy have focused only on aggregate analysis, while the relationships at the disaggregated level have not been explored for the case of Pakistan. For example, Mirza and Kanwal (2017) carried out an analysis on aggregate data of total energy consumption and found bidirectional long-run causalities between total energy consumption, economic growth, and CO₂ emissions. Danish et al. (2017) conducted an aggregate study on renewable and non-renewable energy consumption and observed bidirectional causality between renewable energy consumption and CO₂ emissions in Pakistan. Shahzad et al. (2017) used Granger causality to report that energy consumption is positively related to CO₂ emissions. Muhammad et al. (2014) examined the nexus between renewable and non-renewable energy consumption, real GDP, and CO₂ emissions for Pakistan by applying structural VAR technique. But they also used aggregate data for analysis. One common conclusion from the studies in this area is the support for renewable energy resources. Table 7 shown in the Annexure lists the literature on disaggregated studies that have been conducted in various countries. We cannot ignore the analysis at disaggregated levels because Pakistan’s energy is a mixture of renewable and non-renewable energy resources. The current study addresses this research gap by considering each source of energy with its related CO₂ emissions and economic growth.

Methodology and data

Methodology

This study explores the relationships among renewable and non-renewable energy consumption, economic growth, and CO₂ emissions in Pakistan. The disaggregated analysis examines the corresponding effects of energy consumption on economic growth and CO₂ emissions. We use a standard linear-log function to test the per capita relationship between CO₂ emissions, renewable energy consumption, non-renewable energy consumption, and GDP. To discuss disaggregate contributions to CO₂ emissions, we employ the following linear-log model for analytical purposes:

$$ {\mathrm{CO}}_{2t}={\alpha}_0+{\beta}_1{\mathrm{lnGDP}}_t+{\beta}_2{\mathrm{lnHydro}}_t+{\beta}_3{\mathrm{lnNuc}}_t+{\beta}_4{\mathrm{lnOil}}_t+{\beta}_5{\mathrm{lnCoal}}_t+{\beta}_6{\mathrm{lnGas}}_t+{\varepsilon}_t $$

(1)

where CO_2t reflects carbon emissions, GDP denotes gross domestic products, Hydro reflects hydroelectricity, Nuc implies Nuclear energy, and ε_t is the disturbance term. In order to measure different contribution of renewable energy consumption and non-renewable energy consumption to CO₂ emissions, we have divided the above-mentioned model into two sub-models which can be described as below:

Model 1: renewable energy consumption, GDP, and CO₂ emissions

$$ {\mathrm{CO}}_{2t}={\alpha}_0+{\beta}_1{\mathrm{lnGDP}}_t+{\beta}_2{\mathrm{lnHydro}}_t+{\beta}_3{\mathrm{lnNuc}}_t+{\varepsilon}_t $$

(2)

Model 2: non-renewable energy consumption, GDP, and CO₂ emissions

$$ {\mathrm{CO}}_{2t}={\alpha}_0+{\beta}_1{\mathrm{lnGDP}}_t+{\beta}_2\ln {\mathrm{Oil}}_t+{\beta}_3\ln {\mathrm{Coal}}_t++{\beta}_4{\mathrm{lnGas}}_t+{\varepsilon}_t $$

(3)

Model 1 shows the relationship between renewable energy consumption (hydroelectricity and nuclear energy), GDP, and CO₂ emissions while Model 2 shows non-renewable energy consumption (through oil, coal, and natural gas) GDP, and CO₂ emissions. For this study, we collected time series data from 1970 to 2016 from two major sources: World Development Indicators (WDI) (World Bank 2017) and BP Statistics (2017). The data for GDP per capita was obtained from WDI, while data on CO₂ emissions, oil, coal, natural gas, hydroelectricity, and nuclear consumption were collected from BP Statistics.

Estimation technique

This study applies the ARDL bound testing technique to capture the short-run and long-run dynamics at disaggregate levels. The ARDL methodology was introduced by Pesaran et al. (2001) to test for cointegration among variables. This methodology has several benefits and can be applied if variables are integrated at level I(0) or first difference I(1). It provides robust results regardless of sample size, adjusts the lags in the model, and delivers unbiased estimates with valid t-statistics of the long-run model (Harris and Sollis 2003). Moreover, with the help of a simple linear transformation, ARDL derives a dynamic unrestricted error correction model (UECM). The UECM joins long-run equilibrium with short-run dynamics while keeping long-run information intact. ARDL is a suitable model in the presence of endogeneity and serial correlation in time series data (Pesaran et al. 2001).

Based on the objective of this study, we employed ARDL twice for two different models, i.e., Model 1 for renewable energy consumption, GDP, and CO₂ emissions and Model 2 for non-renewable energy consumption, GDP, and CO₂ emissions. Both models are specified as follows:

ARDL Model 1: renewable energy consumption, GDP, and CO₂ emissions

$$ {\displaystyle \begin{array}{c}\varDelta \mathrm{CO}2t={c}_0+\sum \limits_{i=1}^p{\beta}_1\varDelta {\mathrm{CO}}_{2,t-r}+\sum \limits_{i=0}^p{\beta}_{2i}\varDelta \ln {\mathrm{GDP}}_{t-r}+\sum \limits_{i=0}^p{\beta}_{3i}\varDelta \ln {\mathrm{Hydro}}_{t-r}+\sum \limits_{i=0}^p{\beta}_{4i}\varDelta \ln {\mathrm{Nuc}}_{t-r}\\ {}+{\lambda}_1\ln {\mathrm{CO}}_{2,t-1}+{\lambda}_2\ln {\mathrm{GDP}}_{t-1}+{\lambda}_3\ln {\mathrm{Hydro}}_{t-1}+{\lambda}_4\ln {\mathrm{Nuc}}_{t-1}+{\varepsilon}_t\end{array}} $$

(4)

where Δ is the first difference operator and p denotes the lag length. We derived two hypotheses from Eq. (4) for the long relationships. The first is null hypothesis of no cointegration (H₀:λ₁ = λ₂ = λ₃ = λ₄ = 0) which tested against the second one, i.e., the alternative hypothesis (H₁:λ₁ ≠ λ₂ ≠ λ₃ ≠ λ₄ ≠ 0).

ARDL Model 2: non-renewable energy consumption, GDP, and CO₂ emissions

$$ {\displaystyle \begin{array}{c}\varDelta \mathrm{CO}2t={c}_0+\sum \limits_{i=1}^q{\beta}_1\varDelta {\mathrm{CO}}_{2,t-r}+\sum \limits_{i=0}^q{\beta}_{2i}\varDelta {\mathrm{lnGDP}}_{t-r}+\sum \limits_{i=0}^q{\beta}_{3i}\varDelta \ln {\mathrm{Coal}}_{t-r}+\sum \limits_{i=0}^q{\beta}_{4i}\varDelta \ln {\mathrm{Oil}}_{t-r}+\sum \limits_{i=0}^q{\beta}_{5i}\varDelta \ln {\mathrm{Gas}}_{t-r}\\ {}+{\gamma}_1\ln {\mathrm{CO}}_{2,t-1}+{\gamma}_2\ln {\mathrm{GDP}}_{t-1}+{\gamma}_3\ln {\mathrm{Coal}}_{t-1}+{\gamma}_4\ln {\mathrm{Oil}}_{t-1}+{\gamma}_5\ln {\mathrm{Gas}}_{t-1}+{\varepsilon}_t\end{array}} $$

(5)

where Δ is the first difference operator and q denotes the lag length. The null hypothesis of no cointegration is (H₀:γ₁ = γ₂ = γ₃ = γ₄ = γ₅ = 0) tested against the alternative hypothesis of cointegration (H₁:γ₁ ≠ γ₂ ≠ γ₃ ≠ γ₄ ≠ γ₅ ≠ 0).

Empirical analysis and discussion of results

As a first step, we check the unit root in each series using the Augmented Dicky Fuller (ADF) and Phillips Pearson (PP) tests. Table 1 presents the results of the unit root tests, which reveal that the variables are stationary at I(1). As no variable is integrated at 2nd difference, we reject the null hypotheses of no stationary at I(1) for all of the series.

Table 1 Result of ADF and PP unit root tests

In order to check structural break in the data, we also employed Zivot and Andrews structural break unit root test. Table 2 reveals that structural breaks exist and all variables are integrated at first difference except Coal which is integrated at level. These breaks may occur due to changes in government and economic condition or due to the introduction of new regulations. For example, structure break of 2008 recalls the global financial crises when most of countries were affected economically. GDP rates in many countries declined in the comparison of CO₂ emissions. The problem of structural break can be overcome by adding additional variable or dummy variable from the period of structural change in dependence variable. We have also considered dummy variable to improve the long-run stability of the results. All the unit root tests allowed us to use ARDL technique as all variables are integrated at I(0) and I(I).

Table 2 Results of Zivot and Andrews structural break unit root tests

After checking the unit root, we move to ARDL bound testing to test for cointegration among variables. Table 3 depicts the results of bound testing for Model 1 and Model 2. The results show a long-run relationship among all of the selected variables. The calculated F-statistics are greater than the appropriate critical values of upper-bound. Hence, the null hypothesis of no cointegration is rejected. Diagnostic tests for serial correlation (the Breusch-Godfrey test) and heteroscedasticity (the Arch Test) support the conclusion that the error term is white noise. The Ramsey RESET test also indicates that the model is well-specified.

Table 3 Bound testing cointegration results

We also used the Johansen Cointegration technique to check the robustness of our findings. This technique provides two types of values: trace statistics and maximum eigenvalue statistics. Table 8, presented in the Annexure, indicates that there is at least one cointegration relationship present between renewable and non-renewable energy consumption, economic growth, and CO₂ emissions.

After confirmation of cointegration through the ARDL technique, we check for the long-run and short-run dynamics of Model 1 and Model 2. Results are shown in Table 4 for both models.

Table 4 Long-run and short-run dynamics

In Model 1, the coefficient of GDP is positive and significant (β = 1.665411), which means that economic growth accelerates the CO₂ emissions in the long-run path when energy is consumed from renewable resources. This result is similar to those of Danish et al. (2017) and Mirza and Kanwal (2017) for Pakistan and Zoundi (2017) for 25 African countries. This relationship indicates that an increase in economic growth enhances the demand for energy and so indirectly adds to CO₂ emissions (Shahbaz et al. 2012). The results reflect the increasing population, urbanization, and industrialization in Pakistan, where energy demand is increasing in parallel. Energy use in Pakistan is based primarily on a combination of fossil fuels, so the share of renewable resources is minor. Therefore, energy consumption increases the GDP parallel to an increase in CO₂ emissions, as in the case of Algeria that Bélaïd and Youssef (2017) discussed. Some authors have found that renewable energy consumption causes a decline in CO₂ emissions when GDP increases, as Dogan and Ozturk (2017) found for the USA. However, this finding does not hold for Pakistan because of the seismic differences between the two economies. Shahzad et al. (2017) argued that Pakistan is operating below the threshold level of economic activities, so until it achieves the threshold level, CO₂ emissions are likely to rise in Pakistan. If energy consumption is below the threshold level, then the technology effect remains meager and the effects of scale and composition dominate. In the short run, there is a negative—albeit insignificant—relationship between GDP and CO₂ emissions. The relationship between GDP and CO₂ emissions has been studied many times for different data sets. Many authors (for example, Danish et al. 2017; Sinha and Shahbaz 2018) formulated their studies using the hypothesis of environmental Kuznets curve (EKC). The EKC hypothesis suggests that there might be an inverted U-shape or a U-type nexus between environmental quality and GDP per capita which implies that in the early stages of economic development, economic growth will sooner or later undo the environmental impact. Thus, we can say that a positive or negative relationship between GDP and CO₂ emissions is not obvious.

Our results show a positive but insignificant relationship between hydroelectricity use and CO₂ emissions. A 1% change in the use of hydroelectricity leads to a unit change in CO₂ emissions of only 0.11. Nuclear consumption and CO₂ emissions also have a positive but insignificant relationship, as a 1% change in nuclear consumption brings a unit change of 0.013 in CO₂ emissions. In the short run, then, hydroelectricity and nuclear energy have positive but insignificant relationships with CO₂ emissions.

In the nexus among non-renewable energy consumption, GDP, and emissions (Model 2) is a negative but insignificant long-run relationship between GDP and CO₂ emissions such that GDP is negatively related to CO₂ emissions. However, such is not always the case; for example, Martínez-Zarzoso and Bengochea-Morancho (2004) stated that CO₂ emissions’ declining as a result of increased income sustains to a certain level, after which CO₂ emissions increase with additional increases in income. This finding is an indication for Pakistan’s economy that, as income rises, at some point will come an increase in CO₂ emissions such that this relationship becomes positive. The mediator in this relationship is non-renewable energy consumption. The use of fossil fuels increases the CO₂ emissions in Pakistan, which reduces the economy’s energy efficiency and deteriorates the environment (Muhammad et al. 2014; Danish et al. 2018). Moreover, excessive CO₂ emissions result in the long run in economic benefits’ being outweighed by the economic cost associated with the use of non-renewable resources (Apergis et al. 2010). Pakistan must pursue smart policies on the use of fossil fuels to prevent GDP’s decreasing as a result of inefficient and excessive use of non-renewable energy resources (Soytas et al. 2007).

In the long run, coal consumption bears a positive and statistically significant relationship with CO₂ emissions. Our results match Shahbaz et al.’s (2015) results for China, Ahmad et al.’s (2016) results for India, and Mohiuddin et al.’s (2016) results for Pakistan that coal consumption can increase economic development, but its environmental cost is high. Chandran Govindaraju and Tang (2013) also examined a disaggregate link between coal consumption and CO₂ for India and China and found a strong long-run influence of coal consumption on growth and CO₂ emissions for China. China’s policy of reducing coal consumption could cut CO₂ emissions but at the cost of economic growth, as is the case for Pakistan. Although coal consumption increases CO₂ emissions, technology can reduce its environmental effects. Policymakers must plan to decrease the share of coal consumption in Pakistan’s overall energy mix.

In the long run, natural gas consumption has a positive relationship with CO₂ emissions. Natural gas has the largest share of Pakistan’s total energy mix. Alkhathlan and Javid (2015) concluded that natural gas in Saudi Arabia is friendlier to the environment than other energy sources are, but in Pakistan, natural gas is the least environmentally friendly source of energy, so Alkhathlan and Javid’s (2015) results contrast ours. The reason for difference may be the difference in economies and the level of reliance on natural gas for energy. This outcome is consistent with Shahzad et al.’s (2017) findings for Pakistan. Where there is heavy dependence on natural gas in the overall energy mix, the sustainability of native sources becomes questionable. Per the estimates of the Planning Commission of Pakistan (2017), at the current speed of consumption, the country’s native natural gas resources will be depleted within seventeen years. Therefore, Pakistan must shift its consumption from natural gas to coal or, preferably, more renewable energy sources.

Oil, as the second-largest source of energy consumption in Pakistan, also contributes to the country’s CO₂ emissions, although the positive relationship between oil and CO₂ in the country is insignificant. Dependence on oil in Pakistan has increased because of a severe energy crises and a reduction in natural gas resources but besides oil supply, risk has also increased (Mohsin et al. 2018). Exploration of sustainable ways to produce energy in Pakistan is vital. In the short run, as Model 2 indicates, GDP is negatively associated with CO₂ emissions, while all non-renewable energy sources have positive and significant relationships with CO₂ emissions. Therefore, per capita CO₂ emissions are largely affected by the non-renewable energy consumption. This result makes sense since Pakistan’s primary sources of energy are mainly non-renewable (natural gas, oil, and coal).

We also performed diagnostic tests to examine the models’ stability. Table 4 shows that, on disaggregate levels, there is no serial correlation, heteroscedasticity, or model misspecification. The cumulative sum (CUSUM) and CUSUM of squared recursive residual (CUSUMSQ) plots are executed to ratify that long-run and short-run links are stable. The results are shown in Figs. 1 and 2 for renewable and non-renewable energy sources, respectively.

Fig. 1

CUSUM and CUSUM of squares plots of recursive residuals for Model 1

Fig. 2

CUSUM and CUSUM of squares plots of recursive residuals for Model 2

Model I: renewable energy consumption, GDP, and CO₂ emissions

Model 2: non-renewable energy consumption, GDP, and CO₂ emissions

The CUSUM and CUSUMQ values fall between the upper and lower critical bounds at the 5% levels, indicating the stability and reliability of long-run and short-run dynamics.

Cointegration results indicate the presence of long-run relationships among the variables. We apply the VECM to find the direction of causal relationships. Toda and Philips (1993) indicated that if a long-term relationship exists, then an error correction model can be applied to determine the direction of causality. An error correction model also allows us to differentiate between long-term and short-term Granger causality. VECM Granger causality is determined by using the Wald statistic for all independent variables to determine the difference and lag difference coefficients. Table 5 depicts the causality results for Models 1 and 2, along with the growth and CO₂ emissions.

Table 5 Results of VECM Granger causality

Short-run causality is determined based on the F-statistic calculated through the Wald test, while long-run causality is calculated with the help of the error correction term (ECT). An ECT_t−1 that is statistically significant and that has a negative sign is sign of long-run causality (Danish et al. 2018). The econometric equations for Models 1 and 2 are as follows:

$$ {\displaystyle \begin{array}{l}\mathrm{Model}\;1:\left[\begin{array}{l}\varDelta {\mathrm{LCO}}_{2 it}\\ {}\varDelta {\mathrm{LGDP}}_{it}\\ {}\varDelta {\mathrm{LHyd}}_{it}\\ {}\varDelta {\mathrm{LNuc}}_{it}\end{array}\right]=\left[\begin{array}{l}{\delta}_1\\ {}{\delta}_2\\ {}{\delta}_3\\ {}{\delta}_4\end{array}\right]+\sum \limits_{p-1}^q\left[\begin{array}{cccc}{\theta}_{11p}& {\theta}_{12p}& {\theta}_{13p}& {\theta}_{14p}\\ {}{\theta}_{21p}& {\theta}_{22p}& {\theta}_{23p}& {\theta}_{24p}\\ {}{\theta}_{31p}& {\theta}_{32p}& {\theta}_{33p}& {\theta}_{34p}\\ {}{\theta}_{41p}& {\theta}_{42p}& {\theta}_{43p}& {\theta}_{44p}\end{array}\right]\times \left[\begin{array}{l}\varDelta {\mathrm{LCO}}_{2 it-1}\\ {}\varDelta {\mathrm{LGDP}}_{it-1}\\ {}\varDelta {\mathrm{LHyd}}_{it-1}\\ {}\varDelta {\mathrm{LNuc}}_{it-1}\end{array}\right]+\left[\begin{array}{l}{\alpha}_1\\ {}{\alpha}_2\\ {}{\alpha}_3\\ {}{\alpha}_4\end{array}\right]{\mathrm{ECT}}_{it-1}+\left[\begin{array}{l}{\mu}_{1 it}\\ {}{\mu}_{2 it}\\ {}{\mu}_{3 it}\\ {}{\mu}_{4 it}\end{array}\right]\\ {}\mathrm{Model}\;2:\left[\begin{array}{l}\varDelta {\mathrm{LCO}}_{2 it}\\ {}\varDelta {\mathrm{LGDP}}_{it}\\ {}\varDelta {\mathrm{LCoal}}_{it}\\ {}\varDelta {\mathrm{LOil}}_{it}\\ {}\varDelta {\mathrm{LGas}}_{it}\end{array}\right]=\left[\begin{array}{l}{\delta}_1\\ {}{\delta}_2\\ {}{\delta}_3\\ {}{\delta}_4\\ {}{\delta}_5\end{array}\right]+\sum \limits_{p-1}^q\left[\begin{array}{ccccc}{\theta}_{11p}& {\theta}_{12p}& {\theta}_{13p}& {\theta}_{14p}& {\theta}_{15p}\\ {}{\theta}_{21p}& {\theta}_{22p}& {\theta}_{23p}& {\theta}_{24p}& {\theta}_{25p}\\ {}{\theta}_{31p}& {\theta}_{32p}& {\theta}_{33p}& {\theta}_{34p}& {\theta}_{35p}\\ {}{\theta}_{41p}& {\theta}_{42p}& {\theta}_{43p}& {\theta}_{44p}& {\theta}_{45p}\\ {}{\theta}_{51p}& {\theta}_{52p}& {\theta}_{53p}& {\theta}_{54p}& {\theta}_{55p}\end{array}\right]\times \left[\begin{array}{l}\varDelta {\mathrm{LCO}}_{2 it-1}\\ {}\varDelta {\mathrm{LGDP}}_{it-1}\\ {}\varDelta {\mathrm{LCoal}}_{it-1}\\ {}\varDelta {\mathrm{LOil}}_{it-1}\\ {}\varDelta {\mathrm{LGas}}_{it-1}\end{array}\right]+\left[\begin{array}{l}{\alpha}_1\\ {}{\alpha}_2\\ {}{\alpha}_3\\ {}{\alpha}_4\\ {}{\alpha}_5\end{array}\right]{\mathrm{ECT}}_{it-1}+\left[\begin{array}{l}{\mu}_{1 it}\\ {}{\mu}_{2 it}\\ {}{\mu}_{3 it}\\ {}{\mu}_{4 it}\\ {}{\mu}_{5 it}\end{array}\right]\end{array}} $$

The results ratify the presence of long-run causality among hydroelectricity, nuclear, GDP, and CO₂ emissions. ECT_t−1 is significant in the long run for CO₂ emissions, hydroelectricity, and nuclear. Bidirectional causality is present between, nuclear and CO₂ emissions, which suggests that any change in hydroelectricity or nuclear will cause a change in CO₂ emissions and vice versa. Moreover, we observe unidirectional causality from GDP to hydroelectricity, GDP to nuclear, and GDP to CO₂ emissions. However, the short-run results of Model 1 reveal that causality runs in one direction, from GDP to CO₂ emissions, from CO₂ emissions to nuclear, from GDP to nuclear, and from hydroelectricity to nuclear. In the short run, there is no causality between hydroelectricity and CO₂ emissions, between hydroelectricity and GDP, or between hydroelectricity and nuclear. Numerous studies, such as Al-Mulali et al. (2015), Apergis and Payne (2014), Farhani and Shahbaz (2014), Ohler and Fetters (2014), and Yuan et al. (2008), have also confirmed the relationships among renewable energy consumption, economic growth, and CO₂ emissions. Short-run values are shown in the chi-square coefficient and p values, while long-run values are shown through t-statistics and p values.

For Model 2, the long-run results reveal a bidirectional relationship between GDP and coal and between GDP and oil. This result is consistent with Zhang and Yang (2013) and Lim et al. (2014), who examined the disaggregated nexus of energy-emissions growth for China and the Philippines, respectively. There is evidence that CO₂ Granger-causes GDP, as Lim et al. (2014) found that growth can continue without increasing CO₂ emissions. A unidirectional causality runs from CO₂ to coal, so growing CO₂ emissions per capital increase coal consumption. Neutral causality is observed between oil and CO₂ emissions, between natural gas and CO₂ emissions, between natural gas and GDP, and between natural gas and oil. We endorse the policy of Shahbaz and Lean (2012) that government can protect its GDP rate if it explores alternate energy sources to cater the energy needs of Pakistan. In the short run, CO₂ emissions and oil Granger-cause GDP. CO₂ and natural gas consumption Granger-cause coal consumption in the short run.

Conclusion

This study examines the roles of renewable and non-renewable energy consumption in economic growth and CO₂ emissions at disaggregate levels for Pakistan. The study confirms that energy consumption is central to a country’s economic development, but some energy resources are harmful to the environment. For Pakistan, our results indicate that consumption of renewable energy (hydroelectricity and nuclear) produces less CO₂ emissions than non-renewable energy consumption (oil, coal, and natural gas) does. At the disaggregated level, natural gas consumption is a major source of energy production and the main driver of CO₂ emissions, followed by oil and coal consumption.

Pakistan’s economy is growing fast, but its growth depends heavily on energy consumption. Increased amounts of energy are needed to cater to the increasing demand from the production, household, and transport sectors, but more energy consumption will add more CO₂ emissions to the air if Pakistan’s existing energy mix remains as it is. To achieve the desired growth rate without harming the environmental quality, policymakers should analyze the country’s energy mix at disaggregated levels. A polluted environment not only has a negative effect on human health but also deteriorates water quality and agricultural production. Pakistan’s government can limit CO₂ emissions by shifting from natural gas energy to other alternatives to lower the environmental burden. As Solarin et al. (2018) suggested, the government should encourage hydropower activities and more projects should be started to expand the hydropower production. Our results show that consumption of natural gas generates more CO₂ pollution than the other energy resources in the country’s energy mix. Even so, natural gas reserves are inadequate, whereas coal reserves are ample, so coal is expected to remain the primary source of energy in Pakistan in the future.

The results of this study provide valuable information for policymakers to construct an optimal combination of renewable and non-renewable sources in order to meet the national demand. We put forward a policy such as there is need to plan a strategic mix of all available energy resources in Pakistan to meet the growing economy’s energy demands while also reducing CO₂ emissions. The government should also encourage the industrial infrastructure to use high-level technologies for energy conversion. For example, natural gas-to-liquid technologies and coal-bed methane techniques are useful in converting energy to increase efficiency. Similarly, the government should construct more hydroelectricity plants, as hydroelectricity is more environmentally friendly and economical than coal-fired electricity or natural gas.

Moreover, Pakistan’s government should use awareness campaigns to encourage and motivate consumers and producers to use energy-efficient technologies to improve the environmental quality. Nuclear power and hydroelectricity are the best alternatives to fossil fuels for helping economic development and reducing CO₂ emissions. Therefore, there is a strong need to increase investment in renewable energy sources like solar power, hydroelectricity, wind, and biofuels to stimulate sustainable development in Pakistan.

Annexure

Table 6 ᅟ

Table 7 Literature on disaggregated renewable and non-renewable energy consumption, economic growth, and CO2 emissions

Table 8 Johansen cointegration test results