THE IMPACT OF RENEWABLE ENERGY CONSUMPTION ON CARBON DIOXIDE EMISSIONS: EMPIRICAL EVIDENCE FROM DEVELOPING COUNTRIES IN ASIA

This empirical study examines the relationship between environmental quality, economic development, renewable and non-renewable energy (RE) consumption in 13 developing countries in Asia. We use panel data in the period from 1980 to 2014 and panel cointegration, fully modified ordinary least squares (OLS) and dynamics OLS estimators are employed to test for cointegration in the long-run. The study confirms the existence of the inverted U-shape Environmental Kuznets Curve hypothesis in 13 Asia countries for both estimators with the increment of GDP per capita and conventional energy consumption decreasing the environmental quality. However, the empirical finding suggests that RE consumption is insignificant in contributing to less pollution regarding CO2 emissions. This study concludes that to comprehend better the potential factors affecting the CO2 emissions, the sampled countries can design a strategic plan to mitigate the rate of global warming and climate change, while at the same time stimulating economic development and promoting energy from eco-friendly resources.


INTRODUCTION
The linkages between environmental quality and economic development have been discussed thoroughly. In particular, developing countries in Asia (i.e., Bangladesh, China, India, Iran, Iraq, Indonesia, Malaysia, Myanmar, Nepal, Pakistan, Philippines, Sri Lanka and Thailand) have recorded excellent economic development in the most recent three decades as illustrated in Table 1. However, this impressive performance has also led to environmental issues such as increasing the level of CO 2 emissions. The release of CO 2 emissions is one of the contributors to the greenhouse effect which is largely due to human activities such as burning fossil fuels (FOSS) and deforestation (NASA 2018). As in Figure 1, among the selected 13 Asian countries, in 2014, Iran was the largest emitter of CO 2 emissions with 8.28 metric tonnes per capita followed by Malaysia and China. In the same year, Iran was also the largest country in Asia that consumed energy from FOSS based with 98.98%, closely followed by Iraq (97.27%), Malaysia (96.63%) and China (87.48%). In addition, Nepal was the largest country in Asia that uses energy from renewable sources with 84.37%, followed by Myanmar (66.13%) and Sri Lanka (57.59%) as in Figure 1.
Nevertheless, excellent economic performance requires a lot of energy. The high demand, dependency and consumption of energy from conventional sources mainly generated from natural gas, oil and coal have contributed to environmental problems such as increased level of CO 2 emissions which lead to climate change. Apart from that, the empirical evidence remains scarce in the case of Asian countries. In a recent study by Apergis and Ozturk (2015) which incorporating only CO 2 emissions and gross domestic product in their study found the existence of This Journal is licensed under a Creative Commons Attribution 4.0 International License the Environmental Kuznets Curve (EKC) hypothesis 14 Asian countries. The existence of the EKC hypothesis indicates that decreasing environmental quality due to pollution is an inevitable consequence at the beginning of economic improvement, but as income grows, the environmental quality improves over time.
As of July 2018, Bangladesh, China, India, Indonesia, Malaysia, Myanmar, Nepal, Pakistan, Philippines, Sri Lanka and Thailand have ratified the Paris Agreement which committed to reducing pollution to mitigate climate change by keeping a global temperature rise below 2 degrees Celsius (UNFCC 2018). In order to reach this goal, every party involved in developed and developing countries should be ready to take action in terms of financial support, capacity and advanced innovation to achieve their national objectives.
In line with that, one of the targets in the Sustainable Development Goals (SDGs) is to provide affordable and clean energy by enhancing and promoting the contribution of renewable energy (RE) in the total energy mix, enhance international cooperation for better research on environmentally friendly energy with technologies, innovation and investment by 2030 (SDG, 2018). Enhancing RE consumption in the global energy mix will provide a positive impact on environmental quality. As reported by the Energy Information Administration (2018), increasing contribution of electricity generation from renewable and nuclear sources by 35% has reduced the carbon intensity of the electricity supply in 2016. In addition, the electricity consumption from hydroelectric source shows encouraging improvement mostly in China and India as well as other countries over time (Environmental Impact Assessment, 2018).
Hence, this study examines the linkages among CO 2 emissions, economic growth, renewable and non-RE consumption in 13 Asian countries using Fully Modified Ordinary Least Squares (OLS) (FMOLS) and dynamic OLS (DOLS) from 1980 to 2014. It intends to contribute to the research regarding the EKC hypothesis by providing new evidence based on empirical analysis of electricity consumption from FOSS based on a proxy for non-RE consumption. Other than that, this empirical study also examines the potential of electricity consumption from hydroelectric resources as a proxy for RE consumption in 13 Asian countries.
Given the above background, it is crucial to investigate the role of RE consumption within the EKC framework in Asia, which positively impacts on environmental quality by reducing pollution and enhancing economic performance. This study additionally contributes to the literature by using the panel data technique to explore the linkages between environmental quality, economic performance, renewable and non-RE used in 13 Asian countries. The selected 13 countries are developing countries in Asia. They have high consumption levels of energy especially non-renewable sources and are large emitters of CO 2 compared with other Asian members. Fully modified OLS (FMOLS) and Dynamic OLS (DOLS) approaches are employed to support a better understanding of the relationship between CO 2 emissions, real per capita GDP, electricity consumption from hydroelectric and electricity consumption from FOSS. The study also examines the existence of the EKC hypothesis. The outcomes would have policy implications for the region.
The paper is organised as follows. Section 2 reviews the recent literature, Section 3 details the selected data, model and used methodology, Section 4 and 5 reports the empirical results and discussions based on the analysis. It also concludes the study while offering policy recommendations.

LITERATURE REVIEW
Studies of environmental quality and economic performance hypothesised that environmental pollution is an inverted U-shape curved of income per capita. This inverted U-shape curve is widely known as the EKC hypothesis. The EKC hypothesis indicates that  pollution worsens in the early stage of economic growth until it reaches a turning point after which pollution reduces with higher income per capita (Grossman and Krueger 1991). According to Selden and Song (1994), a high level of pollution is inevitable since many technological innovations in various economic sectors is required in the early phase of economic growth.
Along with the rapid economic activities, the abundant demand for energy, especially from conventional resources such as coal, oil and natural gas, also increases every year. A close relationship between energy consumption and per capita income has contributed to the decline in environmental quality. Hence, numerous studies explored the linkages between non-RE consumption, economic growth and environmental quality (Chandran and Tang, 2013;Heidari et al., 2015;Niu et al., 2011;Zhu et al., 2016). Regarding the EKC hypothesis, the empirical studies from Chandran and Tang (2013), Heidari et al. (2015) and Zhu et al. (2016) investigated the existence of this hypothesis in ASEAN-5 and failed to validate its presence in those countries. Summary of selected previous literature on the renewable energy consumption, CO2 emissions, GDP per capita and energy consumption published between 2007 till 2018 as presented in Table 2.
With increasing demand for energy in various industrial sectors, energy production from conventional sources is unable to cope with such high demand. In this regard, the emergence of energy from renewable sources is an alternative to reduce dependency and reliance on traditional sources as well as improving economic performance. Hence, the empirical studies by Bölük and Mert (2014) Liu et al. (2017) found that the EKC hypothesis is invalidated in 16 European Union, 25 OECD and ASEAN-4 countries, respectively. Meanwhile, Seker (2016a, 2016b) found that the EKC hypothesis is validated in 15 EU and 40 top RE countries. According to Seker (2016a, 2016b), Jebli et al. (2016) and Liu et al. (2017), high consumption from RE improves the environmental quality by reducing CO 2 emissions, and extensive use of energy from conventional sources increases the emissions. Meanwhile, Bölük and Mert (2014) found that energy used from renewable sources contributes to a 50% reduction in emissions compared to energy from conventional sources.
Several studies examined the presence of the inverted U-shape between environmental pollution and per capita income for individual countries without taking into account RE in their modelling framework (Ang, 2007;Lau et al., 2014). The empirical findings from Ang (2007) (2015) and Sugiawan and Managi (2016) provides evidence that the inverted U-shape of the EKC hypothesis is validated and energy production from renewable sources can mitigate pollution by reducing CO 2 emissions in the long-run for the case of Malaysia, Turkey and Indonesia. In contrast, previous study by Azlina et al. (2014) found that the inverted U-shape of the EKC hypothesis is invalid, but energy consumption from renewable sources still provides a positive impact on environmental quality in Malaysia.
Meanwhile, Aung et al. (2017) found that EKC hypothesis is failed to validate for CO2 emissions but the existence of inverted U-shaped can be observed for methane (CH4) and nitrous oxide (N2O) in Myanmar.
Apart from that, the literature reveals the presence of bidirectional causality between RE consumption and economic growth in the European Union, OECD, Eurasia and G7 countries (Al-Mulali et al., 2013;Alper and Oguz, 2016;Payne, 2010a, 2010b;Chang et al., 2015). According to Al-Mulali et al. (2013), higher income countries are more likely to have bidirectional causality between RE consumption and economic growth in the long-run. The findings also reported that the feedback hypothesis also exists in ASEAN countries, namely Malaysia, Indonesia, Philippines and Thailand. However, Alper and Oguz (2016) failed to confirm the presence of the feedback hypothesis in most new EU member countries (Cyprus, Estonia, Hungary, Poland and Slovenia) due to less energy production from renewable sources. Meanwhile, in the case of ASEAN countries, evidence from Chandran and Tang (2013) reported unidirectional causality running from CO 2 emissions to conventional energy consumption in the short-run for Indonesia and Malaysia and bidirectional causality between CO 2 emissions and energy consumption in the Philippines and Thailand. In another study conducted by Azam et al. (2015) and Wang et al. (2016), the empirical results reveal the existence of unidirectional causality running from conventional energy used to economic growth and CO 2 emissions in Malaysia and ASEAN-8 countries.   The significance of conventional energy used for economic development and environmental quality creates an argument whether consumption from RE is significant in the EKC hypothesis. Thus, the present paper examines the linkages between per capita CO 2 emissions, economic growth, electricity consumption from FOSS and electricity consumption from renewable sources to validate the existence of the EKC curve hypothesis for a panel of 13 Asian countries.

Specification Models
Numerous studies by Ang (2007) To investigate the long-run relationships between per capita CO 2 emissions (CO 2 ), economic growth (GDP) and squared economic growth (GDP 2 ), the multivariate framework is established following the model from Seker (2016a, 2016b) and Jebli et al. (2016). Each selected variable is converted into a natural logarithm. The standard EKC model can be shown as: By introducing electricity consumption from FOSS and electricity consumption from renewable sources as another independent variable into the Equation.
(1), the modified EKC model is as follows: Where i and t stands for country and the time; while denotes normally distributed error term; β 1 ,β 2 β 3 and β 4 are the coefficient estimates on the selected variables, respectively. Equation.
(2) employs both GDP and the quadratic of GDP as exogenous variables. Thus, the expected coefficients of β 1 > 0, and β 2 < 0 validated the existence of the EKC hypothesis. A threshold of EKC is when per capita income is low risk on the pollution, and the relationship between these two indicators resembles an inverted U-shaped curve. According to Stern (2017), the value of the turning point (τ) of the EKC can be formulated as τ=exp f()(−(0.5β 1 )/β 2 ). First, as the economy grows in the early phase of economic development, lots of energy is required to fulfil the energy demand from various sectors and industries. Accordingly, heavy dependence on conventional energy sources from natural gas, oil and coal is inevitable because of its reliability and cheaper compared to RE. As consequences, the level of environmental quality reduced as CO 2 emissions increases which contribute to pollution. The positive impact of environmentally friendly energy on environmental quality can be achieved by promoting its consumption and contribution in the total energy mix.
Thus, by applying the panel cointegration approach, this empirical study examines the relationship between environmental quality, economic development, RE and non-RE consumption for a panel of 13 Asian countries during the 1980-2014 period.
The empirical analysis is started by examining the stationarity of each variables using common and individual unit root tests. This study applies panel cointegration tests; namely Pedroni cointegration test Pedroni (1999Pedroni ( , 2004 and Kao cointegration test Kao (1999) to test the existence of a long-run relationship among variables.
After confirming the presence of cointegration, this study uses the fully modified OLS (FMOLS) and dynamic OLS (DOLS) estimators to show long-run coefficient estimates of the real income per capita, the square of real income per capita, renewable and non-RE consumption for CO 2 emissions. The FMOLS approach and the DOLS estimators have been proposed by Pedroni (2001Pedroni ( , 2004 and Kao and Chiang (2000) and Mark and Sul (2003).
This study uses the FMOLS and the DOLS estimators in order to show long-run coefficient estimates of the CO 2 emissions, GDP per capita, squared of GDP per capita, squared of GDP per capita, energy consumption from conventional sources and energy consumption from renewables.

Data and Variables
CO 2 emissions denoted as CO 2 as a proxy for environmental quality in metric tonnes per capita concerning the independent variables, while independent variables, GDP per capita (constant 2010 US$) as a proxy for low economic growth and squared of GDP per capita as a proxy for high economic growth. FOSS is electricity consumption from FOSS measured in million Kw based as a proxy for energy consumption from conventional sources, and RE is electricity consumption from renewable sources measured in million Kw as proxy for energy consumption from renewable sources. The dataset was obtained from 1980 to 2014. The panel data on CO 2 emissions and GDP per capita are taken from the World Development Indicators (WDI), and the data on FOSS and RE retrieved from the U.S. Energy Information Administration. This study selected 13 countries in Asia based on the availability of data extracted from WDI and Energy Information Administration including Bangladesh, China, India, Iran, Iraq, Indonesia, Malaysia, Myanmar, Nepal, Pakistan, Philippines, Sri Lanka and Thailand. The data used in this study are transformed into the natural logarithm in order to interpret the coefficient estimates as the elasticities of the response variable.
The magnitude of coefficients for EC and RE, β 3 and β 4 are predicted to be positive and negative respectively as shown by Dogan and Seker (2016b). Table 3 reports a summary of descriptive statistics with means, standard deviation, maximum and minimum of each selected variables before transforming into the logarithm form from 1980-2014.

Panel Unit Root Tests
This empirical work uses four different panel unit root tests developed by Levin-Lin-Chu (LLC) Levin et al. (2002), Breitung (2000), Im-Pesaran-Shin (IPS) Im et al. (2003) and Fisher Phillips-Perron (PP) Phillips and Perron (1988) tests to examine the stationarity of each series. As in Table 4, the panel unit root tests can be classified into common and individual unit root. The first group of common unit root consists of the t-statistics of Breitung (2001) and LLC's test by Levin et al. (2002). Next, another group of individual unit root tests includes IPS-W-statistic (Im et al., 2003) and PP-Fisher Chi-square (Phillips and Perron, 1988).
The assumption for the null hypothesis is that the series is not stationary with a unit root, while the alternative hypothesis is that the series is stationary with no unit root. All panel unit root tests are estimated with intercept and deterministic trends.

Panel Co-integration Test
First and foremost, this study employs the Pedroni panel cointegration test (Pedroni, 1999) to find a possible cointegration relationship between the analysed variables in Equation (2). According to Pedroni (1999), there are seven tests statistics as shown in Table 5. The results from the Pedroni panel cointegration test indicate that majority of the test statistics provide evidence with the presence of cointegrating relationship between the analysed variables in Equation (2).
The second panel cointegration test is the Kao panel cointegration test (Kao 1999). According to the results in Table 5, the analysed variables in Equation (2) are cointegrated and have long-run relationships since we have enough evidence to reject the null hypothesis of no cointegration in favour of the alternative hypothesis of cointegration at 5% level of significance.
As reported in Table 5, results from Pedroni cointegration tests (Pedroni, 2001(Pedroni, , 2004 can be categorised into within-dimension and between-dimension parts. Two out of four panel statistics reject the null hypothesis of no cointegrating relationship among selected variables at the 1% and 5% significance level in the within-dimension part. Meanwhile, for the between-dimension part, two out of three group statistics reject the null hypothesis. Thus, this result indicates a long-run cointegrating relationship between these selected variables in 13 Asian countries. The Kao cointegration test developed by Kao (1999) is applied to confirm the accuracy and reliability of the previous result.

Long-run Estimates
The FMOLS and the DOLS approaches are effective in eliminating the endogeneity problems and serial correlation. Results from panel both estimators Table 6 indicate that GDP is highly positive and statistically significant on carbon dioxide emissions. Meanwhile, GDP 2 is negatively statistically significant at the 1% level in explaining carbon dioxide emissions. Thus, the connection between environmental quality and economic performance resembles an inverted U-shaped curve and thus validates the EKC hypothesis in 13 Asian countries. These empirical findings align with previous studies by Apergis and Ozturk (2015) and Heidari et al. (2015) that found that the EKC hypothesis is validated in 14 Asian and ASEAN-5 countries. In contradict, findings from Bölük and Mert (2014), Chandran and Tang (2013) and Liu et al. (2017) found that EKC hypothesis is invalid in 16 European Union, ASEAN-5 and ASEAN-4 countries. The implication of this findings indicates that carbon emissions increases in the early phase of economic development, but when it reaches a turning point or threshold value, the emissions reduces without jeopardising the economic growth.
In relation to that, Tamazian et al. (2009) found that environmental pollution declines during high levels of economic and financial development in BRIC countries.
As expected, FOSS is positively statistically significant in explaining carbon dioxide emissions for FMOLS and DOLS estimators. Therefore, a 1% increase in energy consumption from electricity consumption from FOSS contributes to carbon dioxide emissions by 38.7% and 36.4% in the long-run for 13 Asian countries. Our result is similar to those of Seker (2016a, 2016b) and Liu et al. (2017) for the European Union, top RE and ASEAN-4 countries. Economic development requires extensive consumption of energy, but its impact on environmental quality is inevitable. With the high demand for energy, the level of pollution also increases due to most energy supply coming from non-renewable sources. Energy from conventional sources is favourable due to its reliability and is cheaper compared to RE.
The result from FMOLS and DOLS long-run estimates indicates that RE is insignificant in explaining environmental quality. The empirical finding is contradicted with the studies by Dogan and Seker (2016b), Dong et al. (2017) and Liu et al. (2017) in top RE, BRICS and ASEAN-4 countries, respectively. This indicates that consumption of energy from renewable sources is insufficient in reducing the impact of climate change in selected 13 Asian countries. Meanwhile, high dependency on energy from FOSS compared to RE in the total energy mix is dominant. According to a report from the Energy Information Administration (2018) issued in September 2017, energy consumption by OECD countries exceeds the non-OECD countries by 2015.
Nevertheless, by 2030, non-OECD countries are expected to utilise energy beyond OECD countries due to excellent and promising economic development. Additionally, world energy consumption from petroleum and other liquids resources is expected to continue to grow until 2040. However, RE consumption is predicted to grow rapidly between 2015 and 2040 and reduce consumption and dependency on consumption from coal sources. Serious actions taken by promoting RE consumption to contribute as much as possible to the total energy mix and reduce dependency on energy from FOSS will be beneficial for the quality of the environment in the future.

CONCLUSION
This study investigates the linkages of environmental quality, economic development, renewable and non-RE consumption in selected 13 developing countries in Asia from 1980 to 2014. Previously, empirical evidence concerning the relationship between environmental quality, economic development and energy consumption did not include energy from renewable sources in Asia countries. Hence, this study contributes to the literature by examining the relationship between carbon dioxide emissions, economic growth, hydroelectricity consumption and electricity consumption from FOSS within the EKC framework. This study used the panel unit root cointegration, fully modified (OLS) and dynamic (OLS) approach to analyse the linkages between carbon dioxide emissions and its determinants in Asia countries.
The main empirical findings suggest that the EKC hypothesis (inverted U-shaped) is validated in both FMOLS and DOLS estimators in the long-run. The positive relationship between the release of carbon dioxide and GDP per capita postulates that environmental quality is decreasing in the early stage of economic development. However, as income grows with advanced technology and innovation as well as better understanding and awareness, the level of pollution decreases over time.
From empirical findings, increased energy consumption from conventional sources has resulted in an increment in the level of pollution as expected. The extensive consumption of energy from conventional sources such as FOSS has negatively impacted on environmental quality by increasing the level of carbon emissions. Unfortunately, consumption from RE which is environmentally friendly is insignificant in explaining carbon dioxide emissions in Asian countries. This may happen because the share of energy consumption from renewable sources is relatively low and thus insufficient contribution in total energy supply for countries like Iran, Iraq, Malaysia and China failed to have a beneficial effect on environmental quality. Although this empirical study has failed to prove that RE is capable of contributing positively to the environment in 13 Asia's developing countries, it is undeniable that the energy efficiency of this resource is more environmentally friendly and has minimal risk compared to conventional resources.
Last but not least, the findings are critical to enable the Asia countries to design and structure suitable policies regarding RE including enhancing its consumption and reducing heavy reliance on non-renewable sources without risking economic development. In line with the aim of the parties ratified in the Paris Agreement to mitigate climate change and to support the SDGs, future studies can investigate the importance of technology and innovation in RE development to enhance and promote the consumption of this alternative energy in the total primary energy supply.