. International Journal of Energy Economics and Policy | Vol 5 • Issue 4 • 20151042 International Journal of Energy Economics and Policy ISSN: 2146-4553 available at http: www.econjournals.com International Journal of Energy Economics and Policy, 2015, 5(4), 1042-1049. Relationships among CO2 Emissions, Economic Growth and Foreign Direct Investment and the Environmental Kuznets Curve Hypothesis in Turkey Mesut Balıbey* Faculty of Economics and Administrative Sciences, Tunceli University, Tunceli, Turkey. *Email: mbalibey@tunceli.edu.tr ABSTRACT This study examines the causal relationships between economic growth, carbon dioxide emission and foreign direct investment (FDI) and evaluates the environmental kuznets curve (EKC) hypothesis for Turkey in 1974-2011. Firstly, the causality relationships investigated by using the Johansen Cointegration test, The Granger Causality Test, Impulse-Response and Variance Decomposition Analysis of vector autoregression model (VAR) model. The causality relationships display that FDI (LFDI) and economic growth (LGDP) have a significant effect on carbon dioxide emissions (LCO2). Moreover, impulse-response functions and variance-decompositions of VAR model support these relationships among LGDP, LCO2 and LFDI. Secondly, the study investigates the validity of the EKC hypothesis in Turkey for the period 1974-2011 by using regression model approach for the various EKC model forms such as linear, quadratic, and cubic. Consequently, economic growth leads to degradation of environment and depletion of natural resources. It must be the major aim to obtain a sustainable economic growth by less CO2 emissions and consuming less energy. Moreover, the policy makers may take account exogenous impacts such as foreign investments to plan energy policies, and to maintain economic growth against global climate warming. Keywords: CO2 Emission, Economic Growth, Environmental Kuznets Curve Hypothesis, Granger Causality, Johansen Cointeration, Impulse-Response JEL Classifications: C58, C51, Q43, Q56 1. INTRODUCTION Environmental pollution and protecting the environment have been one of the most global issues which have the priority in international political agenda. According to the Kyoto Protocol, countries have taken precautions to preserve the environment. The Kyoto protocol adapted in 1997 contains an international strategy to restrict greenhouse gas emissions. The protocol’s main aid is to succeed the reduction in the emissions of greenhouse gases by establishing quantified limitation and reduction obligations to the Organization for Economic Cooperation and Development (OECD) member States and East European countries. Increasing concentrations of greenhouse gases in the world is accepted as a significant factor affected changing of the climate conditions. Because, a small changing in the climate conditions may cause economic losses and natural disasters. Greenhouse gas emission reduction affects various sectors in the world economy, such as energy sectors, transport, production processes and industry. Increasing economic activity of countries represents the level of energy consumption and carbon dioxide (CO2) emissions (Kuo et al., 2014; Stern, 2004; Lieb, 2002). In the 1980s, environmental issues such as global warming, descending biodiversity and ozone layer depletion led to debates about the effects of environmental degradation on economic growth of the World’s countries. Therefore, there has been a need to clarify the relationships among economic growth, environmental pollution and other factors. The primary aim of this study is to reexamine the causality relationships between foreign direct investment (FDI), economic growth and carbon dioxide emission (CO2) by using Johansen Cointegration Test, the vector autoregression model (VAR) or vector error correction (VEC) model, Granger Causality Test of VAR or VEC model, Impulse-Response Functions and Variance Decompositions of the model. Secondly, the study aims to test the Balıbey: Relationships among CO2 Emissions, Economic Growth and Foreign Direct Investment and the Environmental Kuznets Curve Hypothesis in Turkey International Journal of Energy Economics and Policy | Vol 5 • Issue 4 • 2015 1043 EKC hypothesis by using OLS Regression Model approach for Turkey for the period of 1974-2011. 2. THEORY AND LITERATURE REVIEW Most empirical studies are related to testing of hypothesis namely the EKC. The EKC hypothesis was introduced by Grossman and Krueger’s (1991) study on environmental effects of the North American Free Trade Aggrement (NAFTA) in 1990s, and 1992 World Bank Report. According to the EKC hypothesis, environmental quality or emissions of pollutants are related to economic growth. The EKC hypothesis supports that at the beginning of economic development of the country, environmental degradation will increase until a specific income level “turning point” is reached, and environmental quality will begin to improve as growing income (Selden and Song, 1994). After the turning point, environmental quality indicators begin to indicate decreases in pollution and environmental degradation. This relationship in some cases means that the environmental impact indicator is an inverted U-shaped curve of income per capita (Lieb, 2002; Stern, 2004; Selden and Song, 1994; Kuo et al. 2014). A generalized EKC is plotted in Figure 1 (Yandle, 2002). From Figure 1, the EKC hypothesis actually summarizes an essentially dynamic process of change, as income of an economy increases over time, firstly, emission levels increase, reaches a peak point and then starts decreasing after a threshold level (turning point) of income (Dinda, 2004). The EKC hypothesis analysis used in the literature is identified by various forms such as linear (1) quadratic (2) and cubic (3) as the follow: Y Xit it it= + +β β ε0 1 , i=1,2,…N (1) Y X Xit it it it= + + +β β β ε0 1 2 2 , i=1,2,…N (2) Y X X Xit it it it it= + + + +β β β β ε0 1 2 2 3 3 , i=1,2,…N (3) where; i=1, 2,…N, countries t=1,…T, time Yit=CO2 emissions per capita β0=Estimated parameters Xit=GDP per capita εit=Error term The values of the parameters, if the EKC hypothesis is valid, should be, for β1 and β3 positive and for β2 negative. The squared term in model indicates the U-shape behaviour while the cubic term of model explains monotonically rising pollution (N-curves turn). If the cubic term (β3) is insignificant, it can be removed from model (3). If quadratic term (β2) is also not significant in quadratic model (2), model will returns linear form (1). In brief, cubic form produces various results such as a monotonically increasing or decreasing pollution-income relationship, an inverted U-shape (i.e., the EKC), a U-shape, a N-shape (first rising, then falling, and finally rising again), an inverted N-shape or an insignificant (i.e., flat pollution-income relationship) (Lieb, 2002; Stern, 2004). If β1 is negative and statistically significant but β2 is statistically insignificant, there are indicators that display an certain improvement with rising per capita income. If β1 is positive and statistically significant but β2 is statistically insignificant, these are indicators that indicate an certain deterioration as incomes increase. These consist per capita carbon dioxide emissions (CO2). It is possible that these indicators will show the EKC but at much higher per capita turning points. In addition, If β1 is positive and statistically significant and β2 is negative and statistically, the estimated EKC has a maximum turning point per capita income level calculated by Y* = (−β1/2β2) (Neumayer, 2003; Neumayer, 2004). On the other hand, FDI is considered as an important driving force of economic development for countries. In recent years, FDI inflows have raised questions such as if there is a causal relationship between FDI, economic growth and environmental deterioration. Therefore, several studies have implemented on the relationships among FDI, economic growth, energy intensity and CO2 emissions, and testing of the EKC hypothesis, estimate of the EKC. For example, Saidi and Hammami (2015) examined the impact of economic growth and CO2 emissions on energy consumption for a global panel of 58 countries using dynamic panel data model estimated by means of the generalized method of moments (GMM) for the period 1990-2012. They found that significant positive impact of CO2 emissions on energy consumption. Leitao (2014) investigated the correlation between economic growth, carbon dioxide emissions, renewable energy and globalization for the period 1970-2010 by using time series methods (OLS, GMM, Unit Root Test, VEC model and Granger causality) for Portuguese economy. He found that carbondioxide emissions and renewable energy are positively correlated with economic growth. Moghadam and Lotfalipour (2014) investigated the impact of financial development on environmental quality in Iran by using the auto regression model distributed lag over the period from 1970 to 2011, and they examined short-term and long- Figure 1: The environmental kuznets curve Balıbey: Relationships among CO2 Emissions, Economic Growth and Foreign Direct Investment and the Environmental Kuznets Curve Hypothesis in Turkey International Journal of Energy Economics and Policy | Vol 5 • Issue 4 • 20151044 term relationships among the variables. They found that financial development accelerated the degradation of the environment. Omri et al. (2014) investigated the causality relationships between CO2 emissions, FDI, and economic growth using dynamic simultaneous-equation panel data models for a global panel of 54 countries over the period 1990-2011. They showed that there was an evidence of bidirectional causality between FDI inflows and economic growth for all countries and between FDI and CO2 for all the panels, except Europe and North Asia. They also indicated that there was unidirectional causality relationship from CO2 emissions to economic growth, with the exception of the Middle East, North Africa, and sub-Sahara panel. Shaari et al. (2014) examined the effects of FDI and economic growth on CO2 emission by using panel data analysis by data the period of 1992 to 2012 from 15 developing countries. They showed a cointegration relationship between variables, and found that FDI didn’t has any effect on CO2 emissions. Sahinoz and Fotourehchi (2014) investigated the relationship between FDI in Turkey and CO2 emissions for the voladility of pollution haven hypothesis between 1974 and 2011. Chen and Huang (2013) examined the relationship between carbon dioxide (CO2) emission per capita and economic growth in Next Eleven (N-11) over the period 1981-2009 by using panel unit roots, cointegration in heterogeneous panels and panel causality tests. They presented that there were positive long-run relationship among CO2 emissions, electric power consumption, energy use and GDP, and there was a bi-directional causality between CO2 emission and electric power consumption. Ozturk et al. (2013) examined short-run and long-run relationship and causality between energy consumption and economic growth for the period 1960-2006 in Turkey. They employed Johansen and Juselius cointegration methods and VEC model (VECM). The findings of study indicated that there was not short-term causality relationship between energy consumption and GDP, and there was an unidirectional long-run causality from per capita GDP to per capita energy consumption. Ozturk and Acaravci (2013) investigated the causal relationships between financial development trade, economic growth, energy consumption and carbon emissions in Turkey for 1960-2007 period. They indicated that an increase in foreign trade to GDP result, an increase in per capita carbon emissions and financial development variable has no significant effect on per capita carbon emissions in the long- run. Shahbaz et al. (2013) investigated relationships between CO2 emissions, energy intensity, economic growth and globalization for the period of 1970-2010 in Turkey. They used unit root test and cointegration approach in the presence of structural breaks. They displayed that there was a cointegration relationship between the series, and energy intensity and economic growth increased CO2 emissions. Farhani and Rejeb (2012) examined the relationships between EC, GDP and CO2 emissions for 15 Mena countries by using the panel unit root tests, panel cointegration methods and panel causality test covering the annual period 1973-2008. They found that there was no causality relationship between GDP and EC; and between CO2 emissions and EC in the short run. However, in the long run, there was a unidirectional causality relationship from GDP and CO2 emissions to EC. Ozturk and Uddin (2012) investigate the long-run Granger causality relationship between energy consumption, carbon dioxide emission and economic growth in India over the period 1971-2007. The most important result is that there is feedback causal relationship between energy consumption and economic growth in India which implies that the level of economic activity and energy consumption mutually influence each other; a high level of economic growth leads to a high level of energy consumption and vice versa. The value of the error correction term confirms the expected convergence process in the long-run for carbon emissions and growth in India which implies that emission reduction policies will hurt economic growth in India if there are no supplementary policies which seek to modify this causal relationship. Kaplan et al. (2011) examined the causal relationship between energy consumption and economic growth in Turkey for the period 1971-2006. They used demand model and production model based on vector error correction model. The study indicated that energy consumption and economic growth had a cointegration relationship and there was bidirectional causality relationship between energy consumption and economic growth. Kim et al. (2010) considered the linkage between carbon dioxide emissions and economic growth in Korea. They presented that the causality relationship between carbon dioxide and growth by using Granger Causality test. Choi et al. (2010) investigated the debates over the excistence of the EKC, and used VAR/VECM models for the period 1971-2006 in China, Korea and Japan. They found that Korea, China and Japan showed very different EKC results. Ozturk and Acaravcı (2010) investigated causal relationships between economic growth, carbon emissions, energy consumption and employment ratio in Turkey. They used autoregressive distributed lag bounds testing of cointegration for the period 1968-2005, and presented an evidence of a long-term cointegration relationship between variables. Akbostancı et al. (2009) investigated the relationship between environmental quality and income for Turkey in 1968-2003 at two levels by using cointegration techniques and PM10 and SO2 measurements in Turkish provinces. They showed that the results of time series and panel data analyses do not support the EKC hypothesis. Soytas and Sarı (2009) investigated the long term granger causality relationship between economic growth, carbon dioxide emissions and energy consumption in Turkey. They found that the lack of a long term causal relationship between income and emissions could be implying that to reduce carbon emissions. Halicioglu (2008) examined empirically dynamic causality relationships between carbon emissions, energy consumption, income, and foreign trade in Turkey by using the time series data for the period 1960-2005. The study indicated that income was the most significant variable in explaining the carbon emissions in Turkey which is followed by energy consumption and foreign trade. Mazzanti et al. (2006) presented new empirical evidence on trends concerning emission-related indicators in Italy. They investigated the related EKC literature critically. They used two panel datasets concerning (a) 1990-2000 emissions at province level (b) and sectoral disaggregated NAMEA emissions sources over 1990- 2001 in analysis. The findings of study displayed mixed evidence in support of the EKC hypothesis. They found that it doesn’t exist an EKC dynamic, but many EKC dynamics, differing by period of observation, country/area, emissions/environmental pressures, Balıbey: Relationships among CO2 Emissions, Economic Growth and Foreign Direct Investment and the Environmental Kuznets Curve Hypothesis in Turkey International Journal of Energy Economics and Policy | Vol 5 • Issue 4 • 2015 1045 sectors. Lieb (2002) presented the empirical evidence about that economic growth had been promoted as a method of improving the environment. Selden and Song (1994) investigated EKCs for four emissions series: SO2, CO2 etc. They showed that the turning point for emissions was to be higher than that for ambient concentrations. Grossman and Krueger (1991) estimated EKCs for SO2 dark matter (fine smoke), and suspended particles using the GEMS data set. They found that the turning points for sulfur oxide (SO2) and dark matter were at around $4000-5000. 3. EMPIRICAL ANALYSIS 3.1. Preliminary Analysis of Data This study considers time series data set of The World Bank Database. The yearly data consists of FDI, GDP per capita used as a proxy of economic growth and CO2 emissions (metric tons per capita) for the sample period from 1974 to 2011. All variables were transformed into logarithms namely LCO2, LFDI and LGDP. All empirical tests had been carried out by using the Eviews-8. The time series of CO2, FDI and GDP are presented in Figure 2. From Figure 2, it has been seen that all variables are nonstationary. Stationary series can be described as one series with a constant mean, constant variance and constant autocovariance for each lag during time1. The augmented dickey fuller (ADF) Test, Phillips- Perron (PP) Test and Kwiatkowski-Phillips-Schmidt-Shin (KPSS) Test are used to determine the stationary of time series of LCO2, LFDI, LGDP. Table 1 presents the results of the ADF test, PP test and KPSS test. According to Table 1, the results of the stationary tests indicate that all variables are stationary at first level in ADF, PP and KPSS tests. In other words, all variables are integrated of order one I (1). 3.2. The Johansen Cointegration Test, the Granger Causality Test of the VAR/VEC Model, Impulse- Response Functions and Variance-Decompositions Because the variables are integrated with order I (1), it is tested whether there is a long term relationship among these variables by using the Johansen Cointegration test. If cointegration relationship exists among LCO2, LFDI and LGDP, VECM approach will be used to determine long term relationships among variables. Since the results of the Johansen cointegration analysis depend on the lags of the model, prior to the cointegration test, the lag order selection criteria for standard VAR are presented in Table 2. According to Table 2, one lag lenght is more appropriate for the model. In the VEC model, all variables are endogenous, and the equation of VECM system is specified as follows: Y L Y Yt t t= + +ϕ δ ε( ) / (1) where, Y=(CO2t, FDIt, GDPt), φ(L) is the coefficient matrices for lag operators L, and δ is the cointegrating vectors capturing 1 After being differentiated once is said to be integrated of order 1. It has been showed by I (1). In Table 1, variables integrated of order I (1) are presented by D (.). the long-run relationships among the variables in the system. In addition, the results of the Johansen Cointegration Analysis with one lag order are indicated in Table 3. From Table 3, the results display the rejection of null hypothesis that there isn’t any cointegration relationship between variables, and there is only a cointegration equation according to trace and maximum eigenvalue statistics at 5% significant level. According to the results of VEC (1) Granger Causality in Table 4, there does not exist any causality relationship between LCO2, LFDI and LGDP. Carbondioxide emissions, FDI and economic growth are not related in long term. Furthermore, VAR Granger Causality Test is applied also, and the results are displayed in Table 5. From Table 5; it can be said that there is bidirectional causality relationship from FDI (LFDI) to carbon emissions (CO2) at 5% Figure 2: CO2 emissions, foreign direct investment and GDP Balıbey: Relationships among CO2 Emissions, Economic Growth and Foreign Direct Investment and the Environmental Kuznets Curve Hypothesis in Turkey International Journal of Energy Economics and Policy | Vol 5 • Issue 4 • 20151046 significant level. Similarly, there is a unidirectional causality relationship from LGDP to LCO2. The causality relationships displays that FDI (LFDI) and economic growth (LGDP) have a significant effect on carbon dioxide emissions (LCO2). In addition, both LCO2 and LGDP have a significant causal effect on LFDI at 5% significant level. However, both FDI (LFDI) and carbon dioxide emissions (LCO2) do not have any causal effect on economic growth (LGDP). Accordingly, the impulse-response functions of impact of variables by one standard deviation shock on each other are plotted for ten quarter horizon in Figure 3 for VAR (1) model. The impulse-response functions of impact of variables by one standard deviation shock on each other are plotted for ten quarter horizon in Figure 3 for VAR (1) model. It can be seen from these figures that one standard deviation shock in FDI (LFDI) has a positive significant impact on carbon dioxide emission (LCO2), and that one standard deviation shock in economic growth (LGDP) has a negative minor effect on carbon dioxide emission (LCO2). Moreover, one standard deviation shock in economic growth (LGDP) and carbon dioxide emission (LCO2) have a positive significant effect on FDI (LFDI). Furthermore, the variance decomposition results of VAR (1) model are presented in Table 6. According to Table 6, the variance decomposition results indicate %100 of LCO2 variance can be expained by current LCO2 in the first period, and the percentage is continuing at the end of the tenth periods by 84.16%. At the end of the tenth periods, FDI (LFDI) and economic growth (LGDP) affect the variation in the forecast error of carbon emissions (LCO2) by 15.80% and 0.03% respectively. The variance decompositions of FDI at the end of the tenth periods display that 49.87% of LFDI variance can be explained by current LFDI. In addition LCO2 significantly contributes by 48.77% to variance of LFDI. However, the contribution of LGDP to LFDI variance is only 1.36% level. Furthermore, the variance decompositions of economic growth (LGDP) present that 44.71% of the forecast error variance of Table 1: The results of unit root tests Tests LCO2 DLCO2 LFDI DLFDI LGDP DLGDP ADF –0.623087 –5.839374* –0.755011 –8.637102* 0.093386 –5.857395* PP –0.549143 –6.080549* –0.222207 –9.399018* 0.295239 –5.854142* KPSS 0.733171 0.081123* 0.697916 0.354492* 0.734588 0.080526* *Indicates the refusal of unit root null hypothesis in the significance level at %5. (McKinnon critical value is [–2.943427], Kwiatkowski critical value is [0.463000]), ADF: Augmented dickey fuller, PP: Phillips-Perron, KPSS: Kwiatkowski-Phillips-Schmidt-Shin, LCO2: Carbon dioxide emissions, FDI: Foreign direct investment, GDP: Gross domestic product Table 2: The results of unit root tests Lag LogL LR FPE AIC SC HQ 0 27.36402 NA 4.99e-05 –1.392230 –1.258914 –1.346209 1 107.5922 142.1184* 8.54e-07* –5.462410* –4.929148* –5.278328* 2 110.8422 5.200052 1.20e-06 –5.133841 –4.200632 –4.811697 3 112.3751 2.189841 1.90e-06 –4.707149 –3.373993 –4.246943 *Indicates lag order selected by the criterion, LR: Sequential modified LR test statistic (each test at 5% level), FPE: Final prediction error, AIC: Akaike information criterion, SC: Schwarz information criterion, HQ: Hannan-Quinn information criterion Table 3: The results of Johansen cointegration test Hypothesis Variables: LCO2, LFDI, LGDP Null Alternative Eigenvalue Trace statistic Critical value 5% p-value r=0 r=1 0.435306 29.62094* 24.27596 0.0097 r≤1 r≥2 0.192720 9.047993 12.32090 0.1663 r≤2 r≥3 0.036563 1.340953 4.129906 0.2887 r=0 r=1 0.435306 20.57294* 17.79730 0.0186 r≤1 r≥2 0.192720 7.707039 11.22480 0.1941 r≤2 r≥3 0.036563 1.340953 4.129906 0.2887 *r value indicates the number of cointegrating vectors. (*) indicates rejection at the 5% critical value, LCO2: Carbon dioxide emissions, FDI: Foreign direct investment, GDP: Gross domestic product Table 4: VEC granger causality/block exogeneity wald tests Excluded Chi-square df Probability Dependent variable: D (LCO2) D (LFDI) 2.184325 1 0.1394 D (LGDP) 0.230166 1 0.6314 All 2.212405 2 0.3308 Dependent variable: D (LFDI) D (LCO2) 0.069318 1 0.7923 D (LGDP) 0.150732 1 0.6978 All 0.151145 2 0.9272 Dependent variable: D (LGDP) D (LCO2) 1.766753 1 0.1838 D (LFDI) 0.316883 1 0.5735 All 2.382403 2 0.3039 VEC: Vector error correction, LCO2: Carbon dioxide emissions, FDI: Foreign direct investment, GDP: Gross domestic product Table 5: VAR granger causality/block exogeneity wald tests Excluded Chi-square df Probability Dependent variable: LCO2 LFDI 4.223234 1 0.0399 LGDP 2.979391 1 0.0843 All 6.731664 2 0.0345 Dependent variable: LFDI LCO2 14.78222 1 0.0001 LGDP 15.67529 1 0.0001 All 17.18603 2 0.0002 Dependent variable: LGDP LCO2 0.824190 1 0.3640 LFDI 1.242881 1 0.2649 All 1.263442 2 0.5317 VAR: Vector autoregression model, LCO2: Carbon dioxide emissions, FDI: Foreign direct investment, GDP: Gross domestic product Balıbey: Relationships among CO2 Emissions, Economic Growth and Foreign Direct Investment and the Environmental Kuznets Curve Hypothesis in Turkey International Journal of Energy Economics and Policy | Vol 5 • Issue 4 • 2015 1047 economic growth (LGDP) is explained by current economic growth (LGDP) at the end of the tenth periods. In addition, LCO2 and LFDI contribute at 47.57% and 7.72% levels for variance of economic growth (LGDP) respectively. 3.3. Statistical Model Estimation for Testing of the EKC Hypothesis As mentioned earlier, the EKC hypothesis explains an inverted U-shaped relationship between economic growth (GDP) and environmental quality (CO2). In econometric analysis, this relationship could be described as quadratic form. There is also a possibility that this relationship would be a linear relationship, if economic growth (GDP) is proportional to carbon dioxide emission (CO2), or, that this relationship takes a cubic form in econometrics namely the N-shaped curve relationship. For this purpose, to evaluate if the two variables actually have the these forms of relationship described in literature, a statistical regression analysis by least square method can be performed. Finally, it will also help to determine whether the relationship between economic growth (GDP) and carbon dioxide emission (CO2) is statistically significant in different forms. The results of the three models for EKC hypothesis are presented in Table 7. All model parameters display the results to be statistically significant at 5% level. The EKC emissions reversal at higher incomes is clearly present in the data, with appropriate signs on the model coefficients. Firstly, the quadratic term (LGDP2) of quadratic model is negative and statistically significant at 5% level, and the linear term (LGDP) is positive and statistically significant also. In this case, the estimated the EKC has a maximum turning point per capita income level calculated as LGDP* = (−LGDP/2LGDP2) = (12.72614/2[–0.674319])= –4,290739 (Neumayer, 2003; Neumayer, 2004). In cubic model, the cubic term (LGDP3) is also statistically significant and positive, indicating an N-shaped curve. This would indicate that emissions would begin to rise again once a second income turning point is passed. The estimated models have R-squares (R2) values above 0.95. The results suggest a strong inverted U-shaped relation between carbon emissions (CO2) and economic growth (GDP). In the cubic model, the parameters are also statistically significant, indicating an N-shaped relation. Both an inverted-U shaped and an N-shaped EKC mean that a higher income level and a faster economic development lead to a clearer display of the trend in pollution-income relations (Yang et al., 2010). 4. CONCLUSION AND REMARKS This study aims to investigate the causal relationships between economic growth, carbon dioxide emission and FDI, and to evaluate the EKC hypothesis for Turkey in 1974-2011. Firstly, the causality relationships between the variables are examined by econometric methods. For this purpose, the methodology used in the study includes unit root tests based on ADF, PP and Kwiatkowski-Phillips-Schmidt-Shin (KPSS) Tests, the Johansen Cointegration Test, The Granger Causality Test in a VAR, Impulse- Response and Variance Decomposition Analysis of VAR model. The findings obtained display that a long term relationship exists among economic growth (GDP), carbon dioxide emission (CO2) and FDI. Table 6: The variance decompositions of VAR (1) model Period SE LCO2 LFDI LGDP Variance decomposition of LCO2 1 0.049812 100.0000 0.000000 0.000000 2 0.067372 95.06175 4.914045 0.024200 3 0.080913 91.29985 8.661361 0.038790 4 0.092300 88.92124 11.03400 0.044763 5 0.102252 87.37725 12.57668 0.046072 6 0.111156 86.31758 13.63741 0.045003 7 0.119250 85.55172 14.40553 0.042753 8 0.126694 84.97439 14.98564 0.039971 9 0.133599 84.52444 15.43853 0.037033 10 0.140049 84.16432 15.80152 0.034167 Variance decomposition of LFDI 1 0.651444 0.667665 99.33233 0.000000 2 0.736822 9.418898 90.39904 0.182067 3 0.799867 18.61078 80.98436 0.404859 4 0.856808 26.14138 73.25572 0.602897 5 0.909618 32.08342 67.14539 0.771188 6 0.959077 36.81708 62.26700 0.915917 7 1.005704 40.65849 58.29836 1.043154 8 1.049892 43.83291 55.00969 1.157402 9 1.091949 46.49817 52.23999 1.261850 10 1.132125 48.76651 49.87472 1.358776 Variance decomposition of LGDP 1 0.044006 53.02676 0.428369 46.54487 2 0.061442 50.17797 2.887877 46.93416 3 0.075042 48.69788 4.490782 46.81134 4 0.086702 47.94562 5.490560 46.56382 5 0.097135 47.57014 6.154891 46.27496 6 0.106705 47.40127 6.629428 45.96930 7 0.115628 47.35372 6.989880 45.65640 8 0.124045 47.38191 7.277282 45.34081 9 0.132055 47.45977 7.515243 45.02499 10 0.139727 47.57142 7.718161 44.71042 SE: Standard error, VAR: Vector auto regression, LCO2: Carbon dioxide emissions, FDI: Foreign direct investment, GDP: Gross domestic product Figure 3: (a and b) The impulse-response functions of vector autoregression (1) model b a Balıbey: Relationships among CO2 Emissions, Economic Growth and Foreign Direct Investment and the Environmental Kuznets Curve Hypothesis in Turkey International Journal of Energy Economics and Policy | Vol 5 • Issue 4 • 20151048 The results of the Granger Causality Test of VEC model display that any causality relationship does not exist in long run while the results of the Granger Causality Test of VAR model support that there is bidirectional causality relationship from FDI (LFDI) to carbon emissions (CO2) at 5% significant level. In addition, there is a unidirectional causality relationship from economic growth (LGDP) to carbon dioxide emissions (LCO2). In brief, the causality relationships show that FDI (LFDI) and economic growth (LGDP) have a significant effect on carbon dioxide emissions (LCO2). In addition, both LCO2 and LGDP have a significant causal effect on LFDI. Moreover, impulse-response functions and variance-decompositions of VAR model support these relationsips among LGDP, LCO2 and LFDI. Carbon dioxide emission (CO2) contributes for the variation in the forecast error of all other variables. Furthermore, FDI significantly affects variance of carbon dioxide emission (CO2). According to impulse-response functions, the shocks in foreign direct investment (LFDI) have a positive significant impact on carbon dioxide emission (LCO2) while the shocks in economic growth (LGDP) have a negative minor effect on carbon dioxide emission (LCO2). Moreover, one standard deviation shock in economic growth (LGDP) and carbon dioxide emission (LCO2) have a positive significant effect on foreign direct investment (LFDI). The findings indicate that in the long run foreign direct investment has an effect on CO2 emission. Therefore, any increase in FDI have cause any problem to the environment. An increase in economic growth is negatively related with the environment as it can contribute for decreasing of CO2 emissions. The findings are very important in the environmental policies. Therefore, the countries should find the alternative energy such as natural gas that there is no effect on the environment. Secondly, the study examined the validity of EKC hypothesis in Turkey for the period 1974-2011 by using regression model approach. For this purpose, the various EKC model forms such as linear, quadratic, and cubic are estimated. All parameters of the three models for EKC hypothesis are statistically significant at 5% level. The EKC emissions reversal at higher incomes is clearly present in the data, with appropriate signs on the model coefficients. Firstly, the quadratic term (LGDP2) of quadratic model is negative and statistically significant at 5% level, and the linear term (LGDP) is positive and statistically significant also. In this case, the estimated the EKC has a maximum turning point per capita income level. In cubic model, the cubic term (LGDP3) is also statistically significant and positive, indicating an N-shaped curve. That means that emissions would begin to rise again once a second income turning point is passed. Consequently, economic growth leads to degradation of environment and depletion of natural resources despite increasing life quality. The findings are significant in the environmental policies. Therefore, it must be the major aim to obtain a sustainable economic growth by less CO2 emissions and consuming less energy. Furthermore, the policy makers may take account exogenous impacts such as foreign investments to plan energy policies, and to maintain economic growth for global climate warming. REFERENCES Akbostancı, E., Türüt-Aşık, S., Tunç, G.İ. (2009), The relationship between income and environment in Turkey: Is there an environmental Kuznets curve?. Energy Policy, 37(3), 861-867. Chen, J.H., Huang, Y.F. (2013), The study of the relationship between carbon dioxide (CO2) emission and economic growth. Journal of International and Global Economic Studies, 6(2), 45-61. Choi, E., Heshmati, A., Cho, Y. (2010), An empirical study of the relationships between CO2 emissions, economic growth and openness. IZA Discussion Paper, 5304, 1-27. Dinda, S. (2004), Environmental kuznets curve hypothesis: A survey. Ecological Economics, 49, 431-455. Farhani, S., Rejeb, J.B. (2012), Energy consumption, economic growth and CO2 emissions: Evidence from panel data for MENA region. International Journal of Energy Economics and Policy, 2(2), 71-81. Grossman, G.M., Krueger, A.B. (1991), Environmental impacts of a North american free trade agreement. National Bureau of Economic Research Working Paper, 3914. Cambridge, MA: NBER. Halicioglu, F. (2008), An econometric study of CO2 emissions, energy consumption, income and foreign trade in Turkey. Energy Policy, 37, 1156-1164. Kaplan, M., Ozturk, I., Kalyoncu, H. (2011), Energy consumption and economic growth in Turkey: Cointegration and causality analysis. Romanian Journal of Economic Forecasting, 2(31), 31-41. Kim, S.W., Lee, K., Nam, K. (2010), The relationship between CO2 emissions and economic growth: The case of Korea with nonlinear evidence. Energy Policy, 38, 5938-5946. Kuo, C.K., Kanyasathaporn, P., Lai, S. (2014), The causal relationship between GDP, energy consumption and CO2 emissions in Hong Kong. Production Research Journal, 46-47(3), 127-138. Leitao, N.C. (2014), Economic growth, carbon dioxide emissions, renewable energy and globalization. International Journal of Energy Economics and Policy, 4(3), 391-399. Lieb, C.M. (2002), The environmental kuznets curve: A survey of the empirical evidence and of possible causes. University of Heidelberg Department of Economics Discussion Paper Series, 391, 1-60. Table 7: The estimations of regression models Dependent variable: LCO2 Independent variables Linear model coefficients Quadratic model coefficients Cubic model coefficients C (constant) –8.854106** (0.286932) [0.0000] –58.49850** (7.040325) [0.0000] –10.6747** (281.2797) [0.0163] LGDP 1.149412** (0.033515) [0.0000] 12.72614** (1.641321) [0.0000] 240.4722** (98.21036) [0.0197] LGDP2 - –0.674319** (0.095595) [0.0000] –27.17071** (11.42491) [0.0232] LGDP3 - - 1.027006 (0.442818) [0.0265] R2 0.970301 0.987736 0.989411 Adjusted R2 0.969476 0.987035 0.988477 F-statistic 1176.149 1409.432 1058.971 Prob (F-statistic) 0.000000 0.000000 0.000000 **Indicates statistically significant at level 5%, ( ) indicates standard error, [ ] indicates p-values, LCO2: Carbon dioxide emissions, FDI: Foreign direct investment, GDP: Gross domestic product Balıbey: Relationships among CO2 Emissions, Economic Growth and Foreign Direct Investment and the Environmental Kuznets Curve Hypothesis in Turkey International Journal of Energy Economics and Policy | Vol 5 • Issue 4 • 2015 1049 Mazzanti, M., Montini, A., Zoboli, R. (2006), Economic dynamics, emission trends and the EKC hypothesis New evidence using NAMEA and provincial panel data for Italy. Universita Degli Studi Di Ferrara. p1-36. Moghadam, H.E., Lotfalipour, M.R. (2014), Impact of financial development on the environmental quality in Iran. Chinese Business Review, 13(9), 537-551. Neumayer, E. (2003), Are left-wing party strength and corporatism good for the environment? A panel analysis of 21 OECD countries, 1980- 1998. Ecological Economics, 45(2), 203-220. Neumayer, E. (2004), National carbon dioxide emissions: Geography matters. Area, 36(1), 33-40. Omri, A., Nguyen, D.K., Rault, C. (2014), Causal interactions between CO2 emissions, FDI, and economic growth: Evidence from dynamic simultaneous-equation models. Economic Modelling, 42, 382-389. Ozturk, I., Acaravci, A. (2010), CO2 emissions, energy consumption and economic growth in Turkey. Renewable and Sustainable Energy Reviews, 14(9), 3220-3225. Ozturk, I., Uddin, G.S. (2012), Causality among carbon emissions, energy consumption and growth in India. Economic Research, 25(3), 752-775. Ozturk, I., Kaplan, M., Kalyoncu, H. (2013), The causal relationship between energy consumption and GDP in Turkey. Energy and Environment, 24(5), 727-734. Ozturk, I., Acaravci, A. (2013), The long-run and causal analysis of energy, growth, openness and financial development on carbon emissions in Turkey. Energy Economics, 36, 262-267. Saidi, K., Hammami, S. (2015), The impact of CO2 emissions and economic growth on energy consumption in 58 countries. Energy Reports, 1, 62-70. Selden, T.M., Song, D. (1994), Environmental quality and development: Is there a Kuznets curve for air pollution? Journal of Environmental Economics and Management, 27, 147-162. Soytas, U., Sari, R. (2009), Energy consumption, economic growth, and carbon emissions: Challenges faced by an EU candidate member. Ecological Economics, 68(6), 1667-1675. Sahinoz, A., Fotourehchi, Z. (2014), Foreign direct investments and pollution emissions: “Pollution Haven Hypothesis” test for Turkey. Socio-Economy, 1, 187-210. Shaari, M.S., Hussain, N.E., Abdullah, H., Kamil, S. (2014), Relationship among Foreign direct ınvestment, economic growth and CO2 emission: A panel data analysis. International Journal of Energy Economics and Policy, 4(4), 706-715. Shahbaz, M., Ozturk, I., Afza, T., Ali, A. (2013), Revisiting the environmental Kuznets curve in a global economy. Renewable and Sustainable Energy Reviews, 25, 494-502. Stern, D.I. (2004), The rise and fall of the environmental kuznets curve. World Development, 32(8), 1419-1439. Yandle, B., Vijayaraghavan, M., Bhattarai, M. (2002), The environmental kuznets curve. PERC Research Study, 02-1, 1-24. Yang, H., Zhou, Y., Abbaspour, K.C. (2010), An analysis of economic growth and ındustrial wastewater pollution relations in China. Consilience the Journal of Sustainable Development, 4(1), 60-79.