Determinants of CO2 Emissions in Emerging Markets: An Empirical Evidence from MINT Economies

CO2 emission is one the major contributor to climate change that the top CO2 emitting countries are always trying to mitigate. In an attempt to fill the gap in energy and environmental literature, this study explores the interaction between economic growth, energy usage, trade and urbanization on CO2 emission for MINT economies using the time coverage from 1980 to 2018, providing new perspectives into the literature by employing panel data analysis. Aiming to create robust outcomes, this paper deployed both conventional and modern econometric techniques. The panel co-integration test revealed evidence of the co-integration between CO2 and its determinants in the MINT economies. In order to explore the linkages between CO2 and its determinants, the ARDL PMG model was utilized in MINT economies. Findings based on the ARDL PMG reveals; (i) positive interconnection between CO2 emissions and energy usage; (ii) no significant link was found between CO2 and economic growth; (iii) urbanization influence CO2 positively while a negative link was found between CO2 and trade. Furthermore, the Dumitrescu-Hurlin Causality test revealed; (i) uni-directional causality from CO2 to urbanization; (ii) GDP growth cause CO2 while CO2 causes energy usage. Based on these findings, recommendations were put forward. ©2020. CBIORE-IJRED. All rights reserved


Introduction
Recently, environmental degradation, climate change and ecological distortions have been the major problems caused by the increase in the exploitation of natural resources and the production of goods and services (Ayobamiji & Kalmaz, 2020). The primary goal of developed countries is to expand their economy further, therefore there is a significant concern on the part of the environmentalists and policymakers to minimize the sideeffect of this expansion. The side effect of this growth is the accumulation of greenhouse gases (GHGs), which is generated from the production or extraction of natural resources. There has been a consensus that the significant GHGs contributing to anthropogenic climate change is CO2 emissions, which accounts for about 60% of the greenhouse effects when compared to other GHGs (Özturk and Acaravci 2010). The primary sources of CO2 emissions are fossil fuel generated from the increasing energy consumption, which accounts for about 32804.7 million tons of CO2 emissions globally (BP, 2020). The energy have a large and growing population with favorable demography; secondly, these countries are geographically placed in an advantageous position (Balsalobre-Lorente, 2019). For example, Mexico and Indonesia are firmly close to the United States of America (USA) and China, respectively (the two biggest global markets). At the same time, Turkey is a strategical place within two continents Asia and Europe. While Nigeria is situated in a favorable location where she is gifted with productive natural assets (crude oil, natural gas, coal), making her one of the highest exporters of crude oil and natural gas (Adebayo, 2020). These countries (excluding Nigeria) are members of the G20 group of countries. Due to these features, these economies are becoming a center of attraction, which provides them with an essential role in the international economic and political relations. With these opportunities, specific challenges such as political instability and corruption are being experienced by the economies.
Prior studies believed that rapid urbanization and financial development also contributes to environmental pollution (Heidari et al., 2015;Wang et al., 2018;Bakirtas & Akpolat, 2018;Usman, Akadiri, & Adeshola, 2020). According to Wang et al. (2018), from 1980 to 2011, the global urbanization rate has increased from 39.1% to 52.2%. Jedwab and Vollrath (2015) also stated that urbanization places a significant role in any nation's economic development and also improves the per capita income because most urbanized areas tend to turn into an industrialized and specialized area. Therefore, they contribute largely to the increase in the nation's economic growth; this growth tends to be induced by the energy consumed from heavy machines. Moreover, urbanization increases the consumption of industrial and residential energy, changing production structures into industrial areas, and increasing the technologically oriented production (Odugbesan & Rjoub, 2020). Examples of such urbanized cities are Lagos, Istanbul, Mexico City and Jakarta. It is expected that the trend in the movement of people from rural to urban regions will persistent in the next three decades (United Nations, 2014). The MINT countries are not also excluded from the trend, for example from the year 1995 to 2016, it was recorded that there has been an increase in percentage with regards to several people living in urban centers in Mexico, Indonesia, Nigeria, and Turkey experiencing about 47.1%, 98.4%, 160.3%, and 62.2% respectively (World Bank, 2019). MINT economies account for 4.1% of the total world GDP globally, 4.8% of the energy consumption, 4.4% leading to global CO2 emission (World Bank, 2019). The main objective of this study is to examine the effect of these macroeconomic variables: economic growth, energy consumption, trade, and urbanization on CO2 emission for MINT economies, an emerging economic block. Nevertheless, it is clear from the literature evaluated that there are numbers of weaknesses: firstly, it is evident that there are no studies examining the linkages between economic growth , energy usage, CO2 emissions and urbanization in MINT economies as a group, despite their prospective status in the global economy, combined with the difficulties developing countries are facing; Secondly, the incorporation of urbanization in this interconnection has not been thoroughly explored, particularly in the MINT nations as an emerging economic bloc. Thirdly, no past research utilized panel data to analyze the effect of energy use, trade, economic growth and urbanization on CO2 emissions for MINT economies. Therefore, the main contribution of this paper to the literature is utilizing the westerlund cointegration test proposed by Westerlund (2008) which is a second generation test to explore the long-run cointegration in the MINT economies. The structure of this research is as follows: Literature review segment contains the review of related studies done in regards to our subject matter; the data and method section showcases the data, description of data and model employed in our study; empirical methodology with findings section explains the empirical methodology utilized in this research and also discourses the outcomes or results. The concluding remark section entails the conclusion, limitation of the study, and policy implication.

Literature Review
CO2 emissions research has been conducted extensively in the literature. However, mixed results were reported concerning the relationship between CO2 emission, energy consumption, economic growth, trade, and urbanization. These mixed results are due to differences in the time range, econometric methodology, and countries or regions employed. The following studies (Dinda & Coondoo, 2006;Lee & Lee, 2009;Narayan & Narayan 2010) explored the nexus between GDP growth and CO2 emission. Dinda and Coondoo (2006) study on 88 Countries revealed a two-way causality link between GDP growth and CO2 emission. Ghosh (2010) reveals that there is a bidirectional link between GDP growth and CO2 emission in the short run, which corresponds with the findings of Govindaraju & Tang (2013) and Khoshnevis & Dariani (2019). Wang et al. (2011) study revealed a unidirectional relationship moving from GDP growth to CO2 emissions, which was corroborated in a recent study done by Farhani et al. (2014) and Ertugrul et al. (2016). However, in a recent study done by  which was concentrated on Middle East countries, a unidirectional relationship was established moving from CO2 emission to GDP growth. Zaidi et al. (2017) showed that GDP growth tends to reduce CO2 emissions while in a recent study done by Ayobamiji & Kalmaz (2020) revealed that energy and GDP growth increase CO2 emissions.
Several studies investigated the link between economic growth, energy consumption and CO2 emissions (Salahuddin & Gow, 2014;Apergis & Payne, 2009;Lean & Smyth, 2010;Zaidi et al., 2017;Gorus & Aydin, 2018;Pao & Tsai, 2010;Wang et al., 2011). Lean and Smyth's (2010) study on ASEAN confirmed a long-run relationship between energy consumption, economic growth, and CO2 emissions. Apergis & Payne, (2009) conducted a study on 6 Central American Countries and found a unidirectional moving from energy usage to CO2 emission. A study conducted on 28 provinces in China by Wang et al., (2011) also corroborated this finding. Salahuddin and Gow, (2014) findings revealed a bidirectional causality interconnection between energy usage and CO2 emissions. Pao & Tsai (2010) reveal that the link between energy consumption and GDP growth is bi-directional, which is contrary to the study done by Gorus & Aydin (2018). Akin (2014), Ertugrul et al. (2016); Ayobamiji and Kalmaz, (2020) and Farhani et al. (2014) explores the nexus between economic growth, energy consumption, Trade and CO2 emissions. Akin (2014) revealed that there is an uni-directional relationship running from CO2 emission and trade while Ertugrul et al. (2016) study shows that there is an uni-directional relationship running from trade and CO2 emission. Several studies have included urbanization in their model (Khoshnevis & Dariani, 2019;Abbas, 2020;Kasman & Duman, 2015;Odugbesan & Rjoub 2020). Khoshnevis & Dariani (2019) reveal that the link between urbanization and GDP growth is bi-directional while Abbasi et al. (2020) researched 8 Asian countries and finding revealed a bidirectional relationship between urbanization and energy consumption.
Recently, Odugbesan & Rjoub (2020) utilize the time series data set to examine the link between economic growth, energy consumption, urbanization and CO2 emissions on MINT economies. Contrary to Odugbesan & Rjoub (2020), this study employed the panel data set to examine the relationship between economic growth, energy consumption, urbanization and CO2 emissions. Also, trade was incorporated into the model, which will help in filling the gap in energy and environmental literature concerning countries with similar features such as MINT. Table 1 shows the author(s), countries, the variables used, time coverage, the techniques employed and finding.

Data
This study utilized a panel dataset of the MINT economies covering the period between 1980 and 2018. The dependent variable is CO2 emissions obtained from the OECD database, whereas its determinants are GDP growth, energy usage, trade, and urban population, which are obtained from the World Bank database. Table 2 depicts the deployed variables descriptive statistics by looking at the minimum, maximum, mean, and standard deviation. The Figure 1 illustrates the MINT in the global map while Figure 2, 3, 4, 5 and 6 respectively depicts the trends in CO2 emissions, energy consumption, economic growth, trade and urbanisation among the MINT economies.

Model
The investigators utilized the STIRPAT framework to explore the interconnection between CO2 emission and urbanization based on previous studies (Martınez-Zarzoso et al.2007;Poumanyvong & Kaneko, 2010;Khoshnevis et al. 2019). Ehrlich and Holdren (1971) created this model, which is premised on Influence, Population, Affluence, and Technology (IPAT). According to Chertow (2001), the IPAT identity illustrated in the equation is frequently utilized as the foundation for examining the different factors influencing CO2 emissions.
However, various criticism has been levied on the IPAT model such as; (i) it is seen as an equation based on mathematics which is not good for testing hypothesis; and (ii) presuming non-flexible proportionality between the indicators. As a result of the above loopholes mentioned, the stochastic version of IPAT was suggested by Dietz and Rosa (1997). Therefore, utilizing the model as a backbone for this model was suggested by Dietz & Rosa (1997).
Where the constant term is portrayed by a, and P, A and T are the same as stated in Equation 1. The elasticity of environment influences concerning P, A, and T is depicted by b, c, and d respectively, the error term is illustrated by ε_i, and i which is the country is indicated by the subscript. The impact is denoted by I, which is ideally calculated regarding the emission level of a pollutant. The size of the populace is represented by P. Society impact is denoted by A and technology index as illustrated by T. Hence, the IPAT model is utilized in examining factors influencing changes in the environment.
Several researchers, such as Wang et al. (2011), Khoshnevis et al. (2019 and Nasrollahi et al (2020) have deployed the STIRPAT framework to explore the nexus between energy usage and CO2 emission and urbanization and CO2 emissions. In Equation 2, subscript i(i=1,N) represents the country while timeframe is illustrated by i(i=1,…,T). The natural log of the variables utilizes are taken for convenient linear panel estimation. Also, the logarithm of all the variables deployed was taken in order to eliminate heteroscedastic. Therefore, equation 2 is depicted below: Where the size of the population is represented by P, GDP per capita is illustrated by A, technology index is depicted by A, and is calculated by industrial value-added share of GDP and the year is portrayed by t. Hence, to analyze the influence of these indicators on CO2 emissions, equation 3 above is re-written below as; 4"# = ∝ " + "# + "# + "# + "# (4) 4"# = ∝ " + "# + "# + "# + "# + "# In equation 5, I and t denote sub-index and different years, CO2 represents CO2 emission, Y illustrates economic growth, ENE represents energy consumption, TR depicts trade, and urbanization represents URB and e mirrors error term.

Cross Section dependence test
Data normalization is important to turn the values into similar measurement units because CO2 emissions was reported as metric tons, whereas others were reported with different measurements. The transformation into a normal log thus minimizes potential disruptions of the series' dynamic properties. Panel disturbances in data are generally believed to be cross-sectionally independent, particularly when there is a large cross-sectional dimension. Nevertheless, there is clear proof that crosssectional dependence also exists in the parameters of panel regression. The literature includes some measures for cross-section dependency. However, this study only utilized the Pesaran (2004) test for a cross-sectional dependency test. Furthermore, this study utilized Breusch & Pagan (1980), bias-corrected scaled LM, Pesaran (2004) CD, and LM, Pesaran (2004) scaled LM tests to verify the stationarity of data deployed. "# = ∝ " + " "# + "# (6) ? "# , "B C ≠ 0 The CDLM2 test is estimated as below, which is another method to analyze the cross-sectional dependency We applied this test when N and T are great ( → ∞ & → ∞) and are normally distributed asymptotically. Another test for the cross-sectional dependency is the CD LM test which is estimated using Eq. 9.

Unit root tests
There is a similarity between time series data and panel data unit root testing. The panel ADF data model can be represented as; ∆ "# = " "#UT + N " ∆ #UT + "# + "# G "RT … … … … … … … (11) Where ∆ "# denotes the variable utilized i=1,2…..,N units cross-section throughout a period t=1,2.,….T, "# Describes the exogenous variables column vector, such as fixed effects or trends of individual, coefficient of the meanreversion is portrayed by " , the autoregressive process lag length is depicted by and the error term which is presumed to be mutually dependent is illustrated by "# .
To analyze the integration order of the various variables, this research utilizes the ADF test, which was suggested by Maddala. & Wu (1999), PP test introduced by Choi (2001), Levin, Lin & Chu (2002), IPS unit root suggested by Im, Pesaran & Shin (2003). The null hypothesis was tested utilizing the above unit root tests. The outcomes of all the unit root tests in Table 4 revealed that the variables utilized are integrated at a mixed level that is I(0) and I(1). Although these outcomes of all unit root tests are alike. In addition, the research equation encompasses trend and drift. The order of variables in mixing integration allows us to use Pedroni co-integration test.

Co-Integration test
This study utilized the heterogeneous panel co-integration test suggested by Pedroni, (2004) and Westerlund cointegration test suggested by Westerlund (2008) which is a second generation cointegration test to explore the cointegration amongst the variables. The co-integration is depicted in Eq. 12 2 "# = "# + " + T" "# + 4" "# + q" "# + r" "# + "# … … (12)  The common time factors and permits for heterogeneity are taken into consideration when utilizing the panel cointegration tests. Table 5 and 6 depict Pedroni (2004) and Westerlund (2008) panel co-integration tests respectively. Pedroni (2004) tests the existence of co-integration in the long-run between CO2 emission (CO2) and economic growth (Y), energy consumption (ENE), trade (TR), and urbanization (URB). Based on the seven tests carried out, there are 11 outcomes. Out of the eleven outcomes, seven are significant, meaning that the null hypothesis can be rejected and accept that there is a co-integration between CO2 emissions and its determinants. Four tests are incorporated in the Westerlund ECM panel cointegration test. It consists of four statistics (Gt, Ga, Pt and Pa). The alternative hypothesis that the panel is cointegrated as a whole is tested by the first two tests whereas cointegration of at least one unit is tested by the other two tests (Odugbesan & Rjoub, 2019). The result obtained from Table 6 illustrates acceptance of the alternative hypothesis of cointegration in the group panel as shown by all the four tests.

Hausman Test
The Hausman test statistics is depicted in Table 7 for all the four predictor variables utilized in this research. The hypotheses for the Hausman test indicates that MG and PMG estimates are not statistically different; PMG more efficient while the alternative hypothesis shows that null hypothesis is not true.
The study utilized the PMG since the p-value > 0.05. Therefore, the null hypothesis of homogeneity cannot be rejected. Thus, the PMG estimator is supported by the model. The next thing is to conduct the pool mean group method.

Pooled Mean Group method
The research used the pooled mean group (PMG) estimator created for dynamic heterogeneous panels to explore the presence of equilibrium in the longrun between CO2 emissions and its determinants. The PMG is an intermediary method between the MG estimator and DFE. Since it includes averaging (the MG estimator) and pooling (which depicts the DFE). The PMG estimator enables for differences between the coefficients in the short-run and the error variances; however, the long-run coefficients are restricted to be similar (Khoshnevis & Dariani, 2019). Estimating the interaction in the long-run between variables is based on the cointegrating link between non-stationary variables. The maximum-likelihood PMG estimator for heterogeneous dynamic panels that fit into the ARDL model is proposed by Pesaran et al. (1999). Therefore, this can be defined as an equation for the error correction to improve economic understanding. An ARDL model for error correction (ECM) is outlined below;  Where the determinants of CO2 emission are depicted by X; the long-run dynamics is represented by ϑ ℩ ; the error correction term is denoted by θ and dynamics in the short run is depicted λij (Khoshnevis & Dariani, 2019). The next phase is calculating the long-run interaction between the CO2 emissions and its predictors. The best econometric analysis that best fits the features of our panel results is chosen. Thus, the panel data is estimated based on PMG. The result of the PMG model is depicted in Table 8.
All coefficients calculated were explained as long-run elasticity with the form of the natural logarithm of variables utilized are taken. PMG is utilized to check the relation between CO2 emissions and its determinants in the long run and short-run. The coefficient of variables is significant and negative at 1%. This indicates that the short-run adjustment speed to enter equilibrium is significant in the long run. The ECM is significant, statistically indicating a quicker return to equilibrium in the event of an imbalance. This term illustrates the speed of the adjustment process to go back to equilibrium. Furthermore, there is no proof of significant interaction between GDP growth and CO2 emissions in the MINT economies. This indicate no support for the EKC hypothesis in the MINT. The coefficient of energy usage is 2.16, suggesting a 1% increase in energy consumption will lead to 2.16% in CO2 emissions when other variables are held constant. This finding concurs with past studies (Farhani et al. 2014;Ayobamiji & Kalmaz, 2020). There is an increase in CO2 emissions due to an increase in production and consumption of energy. Though the impacts of change in technology, productivity, and energy consumption efficiency are causing a decrease, this finding corresponds to the outcomes of Wang et al. (2011), Farhani et al. (2014), and Khoshnevis & Dariani (2019. 0.46% decrease in CO2 is a result of a 1% increase in trade when other variables are kept constant. The urbanization coefficient is 0.14, which suggests that when other variables are kept constant, a 1% increase in urbanization will lead to a 0.14% increase in CO2 emissions. This finding is in support of the urban environmental transition theory. The theory claims are based on the following: one of the characteristics of urban cities is rapid industrialization, which is a significant cause of emissions. The pattern of consumption of residents in urban cities is mainly carbon intensive compared to their counterparts living in rural areas. These claims confirm the experience of MINT countries over the last two decades with massive urban growth. Major cities such as Lagos, Istanbul, Jakarta, and Mexico City are presently in the post-industrial phase. A large amount of energy has been consumed due to an increase in the use of automobiles and residential houses, public utility services such as public transport, and high electricity usage. In contrast, in small cities, industrialization's gradual development is the primary source of a large amount of energy consumption. This large amount of energy consumption will consequently lead to high emissions. This finding aligns with the findings of Ali et al. (2016), Khoshnevis & Dariani (2019), Andersson, (2019) and Wang et al. (2019). However, in the short-run, no significant relationship exists between CO2 emissions and it determinants.

Causality analysis
The Dumitrescu-Hurlin causality was also utilized to determine the path of causality between CO2 and its determinants in the MINT economies. The equation below depicts the panel causality equation.
The lag length is depicted by k, stands for the lag length, autoregressive parameter id portrayed by yi (k) , and the regression coefficient pitch depicted by βi (k) , can change between groups. Beyond these, there is no random mechanism for the tests.   Figure 7 depict the findings from the DH causality test revealed that changes in economic growth granger cause CO2 emission in MINT economies. The empirical result uncovered that GDP growth is the main contributor to CO2 emission. The outcome aligns with previous studies (Lean & Smyth, 2010;Hossain, 2011;Govindaraju & Tang, 2013;Cowan, 2014;Farhani & Ozturk 2015). No causality was found between trade and CO2 emissions. The result complies with the study of Hossain, (2011), however, it is in contrast to the studies of Halicioglu (2009) and Sebri & Ben-Salha (2014). Furthermore, CO2 emissions granger cause urbanization at a 1% significance level. The empirical finding exposed that CO2 emission is a significant contributor to urbanization. The outcome agrees with past studies (Shahbaz et al. 2018;Odugbesan & Rjoub, 2020). Lastly, a causality was found running from CO2 emissions to energy usage in the MINT economies. It indicates that CO2 emissions have predictive power over energy consumption in the MINT economies. The finding concurs with previous studies (Soytas & Sari, 2009;Wang et al. 2018;Khoshnevis & Dariani, 2019;Odugbesan & Rjoub, 2020).

Conclusion
This study empirically investigates the interconnection between CO2 and its determinants (GDP growth, trade openness, urbanization, and energy usage utilizing) in the MINTS economies as we utilized the yearly data spanning between 1980 and 2014. Various unit root tests were utilized, and findings show that the deployed variables are cointegrated at a mixed level i.e. I(0) and I(1). The cointegration test revealed that there is evidence of cointegration between CO2 and its determinants in the MINT economies. In order to explore the linkages between CO2 and its determinants, the ARDL PMG model was utilized in MINT economies. Findings based on the ARDL PMG revealed that the ECM is negative and significant statistically indicating a quicker return to equilibrium in the event of an imbalance. Furthermore, a positive ENE URB CO2 Y TO interconnection was found between energy usage and CO2 emissions while no significant connection exists between economic growth and environmental pollution. Furthermore, there is evident of negative link between trade and CO2 emissions and urbanization significantly influence environmental pollution. Also, findings from the Dumitrescu Hurlin Panel Causality test revealed that economic growth granger cause CO2 emission in MINT economies. This empirical result uncovered that GDP growth is the main contributor to CO2 emissions. Additionally, CO2 emissions granger cause urbanization and causality was found running from CO2 emissions to energy consumption in the MINT economies. Based on our findings, we recommend that policymakers in these countries should continue with their trade policies since trade has a detrimental effect on CO2 emissions. Also, it is necessary for the MINT economies to adopt energy efficiency initiatives that will boost their economic growth. This approach will be directed towards the reduction of CO2 emissions. In this respect, structural reforms are needed to enhance the quality of the environment, as well as economic growth. Additionally, the MINT economies need to improve their energy efficiency by enacting green technologies and promoting renewable energy usage. Also, strong reliance on fossil fuels should be replaced by renewable energy, as fossil fuels are the major contributor to GHGs. In addition, MINT countries need to turn their economies into a sustainable economy, which is the best way to overcome ecological issues arising from economic growth. The nations in the MINT will implement their environmental protection rules and regulations in order to put greater focus on environmental safety. Finally, in order to attain sustainable urbanization in MINT economies, efficient energy, economic and environmental measures will direct urban development growth in those nations without sacrificing economic growth and ensuring a reduction in CO2 emissions in order to accomplish a quality environment. Urban planning policy makers in the MINT states will strive to reduce the pace of urbanization by pursuing efficient land use to promote green and efficient urbanization, which will, to some degree, boost the impact of urbanization on environmental degradation. Further studies should utilize quarterly data. Although this paper allows for sound analytical outcomes and fills gaps in literature using Westerlund cointegration, PMG, and Dumitrescu Hurlin Panel Causality techniques, further research should be undertaken in the future to assess this link in the various developing countries and blocs that will enrich existing literature.