Exploring the Macroeconomic Determinants of Carbon Emissions in Transitional Economies: A Panel Data Analysis Approach

The study’s main objective was to investigate the macroeconomic determinants of carbon emissions in transitional economies using panel methods with data ranging from 1996 to 2014. The main data analysis was done using econometric estimation methods such as fixed effects, random effects, pooled ordinary least squares (OLS) and the dynamic generalized methods of moments (GMM) approach whilst robustness tests were done under the umbrella term, the lagged independent variable approach. To a larger extent, infrastructural development, economic growth, trade openness, financial development and natural resources were found to have had a significant positive effect on carbon emissions, in line with major theoretical predictions. On the other hand, renewable energy consumption, foreign direct investment, information and communication technology and human capital development were mainly found to have reduced carbon emissions in transitional economies. The results are firmly supported by literature. Transitional economies are therefore urged to increase their use of renewable energy and information and communication technology (ICT) infrastructure, attract more foreign direct investment (FDI) and implement policies aimed at enhancing human capital development to reduce carbon emissions. Given data availability, future studies must investigate whether other macroeconomic variables mentioned in the empirical literature that they determine carbon emissions are relevant in transitional economies.


INTRODUCTION
The relationship between the growth of the economy and carbon emissions has for the past few decades been the subject of discussion among environmental economists, researchers and in finance (Hooi and Smyth, 2010;Lee and Lee. 2009;Narayan and Narayan. 2010;Ahmed et al., 2017). Central to such a discussion is The Environmental Kuznets Curve (EKC) which has been described by some researchers as U-shaped (Hooi and Smyth, 2010) and has been used to explain the relationship between emission of pollutants and economic activities. Over time, the determinants of carbon emissions have included a whole lot of other factors explained in Table 1 from a theoretical point of view. These include trade openness, foreign direct investment (FDI), financial development, renewable energy consumption, human capital development, tourism, infrastructural development, information and communication technology, among others. However, the empirical findings on the determinants of carbon emissions are mixed and include quite divergent and convergence views in some instances, evidence that there is no yet consensus on the list of factors which affect carbon emissions.
The economic growth-led carbon emissions nexus is the side of the relationship between carbon emissions and economic activities that has been well researched, and the findings are no longer contestable. What is still not yet quite known is an agreeable list of determinants of carbon emissions determinants hence the reason why this author attempts to contribute to literature by investigating the determinants of carbon emissions in transitional economies. Not a single study that has so far explored the determinants of carbon emissions has exclusively focused on transitional economies as a bloc of countries. The current study therefore seeks to tell a story on the determinants of carbon emissions in transitional economies to fill such a gap.
Most of the empirical studies that have been done so far on carbon emissions determinants on transitional economies focused on single country studies. For example, Rasool et al. (2019), Zhang et al. (2016), Khan  However, closest empirical studies have been done by Magazzino and Cerulli (2019), Zakarya et al. (2015) and Nordin et al. (2015) which respectively focused on Middle East and North African (MENA) countries, BRICS (Brazil, Russia, India, China, South Africa) and ASEAN countries. Although majority of countries in these groups are emerging economies, these groupings of countries are far from being a true representation of transitional economies bloc of countries. Still, these closest empirical studies suffer from

GDP per capita
Economic growth is associated with high level of economic activities associated with a lot of energy usage which produces more pollution and carbon emissions, following Aye and Edoja (2017). A study done by Aye and Edoja (2017) in developing nations also revealed that economic growth increased carbon emissions + Natural resources (NAT) Total natural resources rents (% of GDP) Kwakwa et al. (2018) argued that the extraction of natural resources is done using heavy equipment and machinery which emit carbon dioxide and uses a of energy + Trade openness (OPEN) Total trade (% of GDP) According to Grossman and Krueger (1991), increased trade openness means that more production activities in the economy can happen as domestic companies can now easily source inputs from other countries wherever they are found. The same authors argued that with high levels of trade openness, the country can now import clean energy using machinery and or efficient machinery from advanced economies hence contributing to lower carbon emissions Following Zhang and Liu (2015), the continuous use of ICT technology is associated with more energy consumption which consequently emit more carbon emissions. Lee and Brahmasrene (2014) also noted that ICT brings in a lot of energy usage efficiency thus triggering less carbon emissions and pollution In the case of China, FDI reduced carbon emissions only up to a certain level (Cheng and Yang, 2016). Foreign direct investment into the host country is associated with increased manufacturing activities which produces more pollution and carbon emissions (Blanco et al., 2013).
Individuals using the internet (% of population) Contrary to theoretical expectation, clean infrastructure such as internet infrastructure had an insignificant positive effect on carbon emissions in the OECD countries (Salahuddin et al., 2016). Consistent with Kwakwa et al. (2018), the use of heavy infrastructure in the process of extracting natural resources involve emission of carbon dioxide + FIN Domestic credit provided by the financial sector (% of GDP) Financial development increases carbon emissions in the following three ways. Firstly, availing credit to the consumers can increase the purchase of equipment (automobiles and machinery) which uses more energy (Xing et al., 2017). According to Aye and Edoja (2017:10), financial development attracts FDI into more energy usage activities + Source: Author compilation methodological deficiencies such as failure (1) to capture the dynamic characteristics of carbon emissions data, (2) to address the endogeneity problem and (3) to accept reality that the impact of one macroeconomic variable on another is not immediate. The current study addresses all these methodological concerns.
The remaining section of the paper is structured into five main headings: Section 2 is the theoretical literature on the determinant of carbon emissions whilst Section 3 presents the determinants of carbon emissions from an empirical point of view. Section 4 is the broad research methodology which encompasses data and its description, pre-estimation diagnostics, diagnostic tests (panel unit root tests and co-integration), main data analysis, discussion and interpretation of results. Section 5 concludes the study whilst Section 6 is the bibliography. Table 1 is a summary of determinants of carbon emissions, their relevant proxies and how each of them is related to carbon emissions. Table 2 below is a discussion of the empirical literature on the determinants of carbon emissions.

DETERMİNANTS OF CARBON EMİSSİONS (CO 2 ) -EMPİRİCAL LİTERATURE REVİEW
A variety of variables have been mentioned in Table 2 as having conflicting influence on carbon emissions, itself one of the reasons why the current study is pursuing further empirical tests to address that problem.

Pre-estimation Diagnostics, Panel Root and Co-integration Tests
The following variables were found to be positively and significantly related with carbon emissions, in line with theoretical predictions (Table 4), namely economic growth, trade openness, foreign direct investment, financial development, infrastructural development, information and communication technology and human capital development. As expected, renewable energy consumption and carbon emissions are negatively and significantly related. A non-significant positive relationship between natural resources and carbon emissions was detected. The maximum size of the relationship was found to be between FDI and trade openness (79%), hence there is no multi-collinearity problem in the data set used, consistent with Stead (1996).
Economic growth data has got abnormal values since standard deviation is far more than 100. The range value of economic growth also supports this argument. All the probabilities of the Jarque-Bera criteria are equal to zero, an indication that the data for all the variables is not normally distributed. This is the main reason why the author had to transform all the data sets into natural logarithms before using it for main data analysis in order to effectively address such a statistical problem.
The data was integrated of order 1 (Table 5) whilst a long run relationship between and among the variables was also detected (

General Model Description
In line with theoretical literature and some of the most recent empirical literature (Zakarya et al., 2015;Kongo, 2018;Gianmoena and Ibanez. 2018;Faisal et al., 2018) on the determinants of carbon emissions, equation 1 is the general model specification used in the study.
Where the description of GROWTH, NAT, OPEN, HCAP, RENEW, ICT, FDI, INFR, FIN and CO 2 is shown in Table 7.
Equation 2 shows an econometric equation on the relationship between carbon emissions and its determinants in transitional economies (a transformation of equation 1).
Equation 2 variables are explained in Table 8.
The current study estimated equation 2 using panel data analysis methods such as fixed effects, random effects and pooled OLS approaches, whose main strengths have already been elucidated.
The findings are included in Table 9.
According to Table 9, economic growth had a significant positive impact on carbon emissions across all the three panel data analysis methods (fixed effects, random effects, pooled OLS), a finding which resonates with Aye and Edoja (2017) whose argument is that economic growth is associated with high level of economic activities associated with a lot of energy usage which produces more pollution and carbon emissions. Under    Source: Author compilation from E-Views. ***/**/* denotes statistical significance at the 1%/5%/10% level respectively Source: Author compilation from E-Views. ***/**/* denotes statistical significance at the 1%/5%/10% level respectively  fixed effects, natural resources had a non-significant positive effect on carbon emissions whereas random and pooled OLS shows that natural resources had a significant positive influence on carbon emissions, results which are in line with Kwakwa et al. (2018) whose study noted that the extraction of natural resources is done using heavy equipment and machinery which emit carbon dioxide.
A significant positive relationship running from trade openness towards carbon emissions was also observed across all the three panel data analysis methods, a finding which agrees Grossman and Krueger's (1991) view that high level of trade openness promote more industrial activities in the local economy which are associated with more carbon emissions. FDI was found to have had a significant negative impact on carbon emissions, results which are in line with Cheng and Yang (2016) finding that FDI had a deleterious effect on carbon emissions up to a certain level in China.
A significant positive relationship running from financial development towards carbon emissions across all the three panel data analysis methods resonates with Aye and Edoja (2017:10), that financial development attracts FDI into more energy usage activities which consequently emit more carbon dioxide. Consistent with Kwakwa et al. (2018) whose view is that the use of heavy infrastructure to extract natural resources produces more carbon dioxide, the study found out that infrastructural development had a significant positive impact on carbon emissions under fixed effects, random effects and pooled OLS.
Across all the three panel methods, ICT had a significant negative influence on carbon emissions in line with Lee and Brahmasrene (2014) whose view is that ICT enables more energy efficiency usage hence lowering the amount of carbon emissions emitted. According to the fixed and random effects, human capital development had a non-significant negative effect on carbon emissions yet a significant negative relationship running from human capital development towards carbon emissions was observed under the pooled OLS approach. These results resonate with Saleem et al.'s (2019) argument that human capital development enhances efficiency in the use of energy thereby lowering the quantity of carbon emissions.
To capture Morales-Lage et al. (2016) argument that carbon emissions trigger more carbon emissions as the situation gets out of hand, the current study captured the dynamic characteristic of carbon emissions data (see equation 3).
CO 2it = β 0 + β 1 CO 2 LAG it + β 2 ROWTH it + β 3 NAT it + β 4 OPEN it + β 5 HCAP it + β 6 RENEW it + β 7 ICT it +β 8 FDI it +β 9 FIN it +Ɛ it Where β 1 CO 2 LAG it captures the dynamic feature of the carbon emissions data. Equation 3 was estimated using the dynamic GMM approach, whose results are presented in Table 10.     -Lage et al. (2016) argument that carbon emissions trigger more carbon emissions, the dynamic GMM approach shows that carbon emissions were positively and significantly affected by its own lag (see results in Table 10). Economic growth and natural resources had a separate nonsignificant positive influence on carbon emissions, findings which are supported by existing literature (Table 1). A non-significant negative relationship from trade openness to carbon emissions was observed, in line with Grossman and Krueger (1991) whose study noted that high trade openness means that local companies are now able to import clean energy from international markets.
A significant negative relationship running from renewable energy consumption towards carbon emissions was observed under the dynamic GMM method, a finding which is consistent with Dogan and Seker (2016) that renewable energy usage is associated with less pollution and carbon emissions. The dynamic GMM approach produced results which show a non-significant negative influence of FDI on carbon emissions, in line with Cheng and Yang's (2016) findings. In line with Xing et al. (2017) and Aye and Edoja (2017), the study noted that financial development had a non-significant positive influence on carbon emissions.
The dynamic GMM approach also observed that infrastructural development had an insignificant positive effect on carbon emissions, findings which are aligned with those of Salahuddin et al. (2016) and Kwakwa et al. (2018). ICT reduced carbon emissions, in line with Lee and Brahmasrene (2014)'s view that ICT is associated with energy usage efficiency. Last but not least, the dynamic GMM method noted that human capital development had a significant negative influence on carbon emissions, a finding which is in line with Saleem et al. (2019) which says that high level of human capital development enhances efficiency in the use of energy.

Robustness Tests Using the Lagged Panel Data
Analysis Framework Matthew and Johnson's (2014) approach is that it takes time for one macroeconomic variable to affect another (the current study assumed it takes 1 year) -see equation 4.
The results are quite robust because they mirror the main results of the study. According to the lagged independent variable approach (robustness checks approach), the variables which were found to have had a significant positive influence on carbon emissions include economic growth, natural resources, trade openness, financial development and infrastructural development. These results mirror the main findings presented in Table 11. The lagged independent variable approach also indicates that renewable energy consumption reduced carbon emissions in a significant way, findings which are firmly rooted in literature. The robustness approach also shows that separately, both FDI and ICT had a deleterious effect on carbon emissions, in line with majority theoretical predictions.
The notable difference is that the robustness approach shows that fixed and random effects had a non-significant positive influence on carbon emissions, a finding which is contrary to available literature.

CONCLUSİON
The study's main objective was to investigate the macroeconomic determinants of carbon emissions in transitional economies using panel methods with data ranging from 1996 to 2014. The main data analysis was done using econometric estimation methods such as fixed effects, random effects, pooled OLS and the dynamic GMM approach whilst robustness tests were done under the umbrella term, the lagged independent variable approach. To a larger extent, infrastructural development, economic growth, trade openness, financial development and natural resources were found to have had a significant positive effect on carbon emissions, in line with major theoretical predictions (Table 1). On the other hand, renewable energy consumption, foreign direct investment, information and communication technology and human capital development were mainly found to have reduced carbon emissions in transitional economies.
The results are firmly supported by literature (Table 1). Transitional economies are therefore urged to increase their use of renewable  Source: Author's compilation from E-Views. ***, ** and * denote 1%, 5% and 10% levels of significance, respectively energy and ICT infrastructure, attract more FDI and implement policies aimed at enhancing human capital development in order to reduce carbon emissions. Given data availability, future studies must investigate whether other macroeconomic variables mentioned in the empirical literature that they determine carbon emissions are relevant in transitional economies.