(greenhouse gases) and economic growth

Energy 74 (2014) 439e446

Contents lists available at ScienceDirect

Energy journal homepage: www.elsevier.com/locate/energy

Fossil & renewable energy consumption, GHGs (greenhouse gases) and economic growth: Evidence from a panel of EU (European Union) countries € lük a, *, Mehmet Mert b Gülden Bo a b

Akdeniz University, Department of Economics, Antalya, Turkey Akdeniz University, Department of Econometrics, Antalya, Turkey

a r t i c l e i n f o

a b s t r a c t

Article history: Received 13 March 2014 Received in revised form 2 July 2014 Accepted 3 July 2014 Available online 28 July 2014

Recently a great number of empirical research studies have been conducted on the relationship between certain indicators of environmental degradation and income. The EKC (Environmental Kuznets Curve) hypothesis has been tested for various types of environmental degradation. The EKC hypothesis states that the relationship between environmental degradation and income per capita takes the form of an inverted U shape. In this paper the EKC hypothesis was investigated with regards to the relationship between carbon emissions, income and energy consumption in 16 EU (European Union) countries. We conducted panel data analysis for the period of 1990e2008 by fixing the multicollinearity problem between the explanatory variables using their centered values. The main contribution of this paper is that the EKC hypothesis has been investigated by separating final energy consumption into renewable and fossil fuel energy consumption. Unfortunately, the inverted U-shape relationship (EKC) does not hold for carbon emissions in the 16 EU countries. The other important finding is that renewable energy consumption contributes around 1/2 less per unit of energy consumed than fossil energy consumption in terms of GHG (greenhouse gas) emissions in EU countries. This implies that a shift in energy consumption mix towards alternative renewable energy technologies might decrease the GHG emissions. © 2014 Elsevier Ltd. All rights reserved.

Keywords: Environmental Kuznets Curve Carbon dioxide emissions Renewable energy

1. Introduction The increasing threat of global warming and climate change has been a major on-going concern since the 1990s. Because GHG (greenhouse gas) emissions result primarily from the combustion of fossil fuels, energy consumption and production are at the centre of climate change debates. According to the latest report of the EUJRC (Joint Research Centre), (see Ref. [54], fossil fuel combustion accounts for about 90% of total global CO2 emissions. The relationship between energy consumption and economic development, as well as economic development and environmental pollution has been studied intensively for the last three decades [27,39,41]. There are three main research branches discussed in the literature that consider economic growth, energy consumption and

* Corresponding author. Department of Economics, Akdeniz University, Dumlupinar Bulvari, Kampus, 07058 Antalya, Turkey. Tel.: þ90 2423106407. €lük), [email protected]. E-mail addresses: [email protected] (G. Bo tr (M. Mert). http://dx.doi.org/10.1016/j.energy.2014.07.008 0360-5442/© 2014 Elsevier Ltd. All rights reserved.

environmental pollution [79]. The first branch focuses on the relationship environmental pollution and economic growth, testing the validity of the so-called EKC (Environmental Kuznets Curve) hypothesis. Within the first branch of studies, the EKC was interpreted as reflecting the relative strength of scale versus technique effect [16]: 1768). However, as suggested by Panayotou in Ref. [57]; the shape of the EKC reflects some mixture of scale, composition and technique effect. When a country at the early phase of industrialization, due to the setting up rudimentary, inefficient industries, scale effect takes place and pollution emerges. Generally, however, as per capita income increases, both the output mix and production techniques change. This composition effect states that the movement from an agrarian to an industrial and finally to a service economy shifts gradually the economic growth to sectors that pollute less [40]. Technique effect allows for the possibility that as countries grow, “cleaner” technologies substitute for “dirtier” ones in the production process [15]. In many empirical studies a U-shaped relationship appears as follows: at a relatively low level of income per capita, growth leads to greater environmental damage, until it levels off at an intermediate level of income after which further growth leads to

440

€lük, M. Mert / Energy 74 (2014) 439e446 G. Bo

improvements in the environment [24]. This relationship (EKC) has been explored for a variety of pollutants such as nitrous oxide, sulphur dioxide, suspended particulate matter, carbon monoxide, lead, deforestation, biological oxygen demand, etc. [25]. If EKC were true, this hypothesis would suggest that countries do not need to struggle to reduce the CO2 emissions envisaged by the Kyoto Protocol, since economic development would eventually lead to environmental improvement [80]. The EKC hypothesis assumes emissions to be a function of income which indicates unidirectional causality running from income to GHG emissions [1]. The EKC hypothesis argues that the pollution level increases as a country develops but begins to decrease as rising income passes beyond a turning point. In other words, environmental quality will get worse first and then improve with economic growth [14] (See Fig. 1). Therefore, the EKC hypothesis states that the initial phases of economic development are associated with greater production of garbage and pollution emissions, though at some given income level (and/or per capita income) there is a turning point where greater GDP growth implies lesser environmental degradation [28]; p.9). The results of early EKC studies showed that some important indicators of environmental quality (for example, sulphur dioxide and particulates in the air) improved as income and consumption increased [75]; p. 1). The EKC concept first emerged in 1991 with Grossman and Krueger's pioneering study of the potential impacts of the NAFTA (North American Free Trade Agreement). Within the first wave of EKC studies, basic EKC models were used and income growth and its environmental impacts were estimated in the model without any explanatory variables. Refs. [30,31,70,12,57,37,69,68,53,67,34] tested the economic growth and environmental pollution relationship and EKC hypothesis. Refs. [73,21,45,14] provided review surveys of empirical EKC studies. The second branch of EKC studies concentrates on the energy consumption and economic growth relationship. The main argument of these studies is that economic growth and output are closely related to growth in energy consumption. Energy is the engine of economic growth, since all production and consumption activities need some form of energy as a basic input. These studies test the causal relationship between economic output and energy consumption and intensively use time series models such as causality (Granger, Toda Yamamoto, DoladoLütkepohl), the VECM (Vector Error Correction Model), the VAR (Vector Autoregressive Model) and the ARDL (Autoregressive Distributed Lag) bound test etc. A paper by Kraft and Kraft (1978) found unidirectional causality from GNP growth to energy consumption in the USA for the period from 1947 to 1974. A number of empirical studies tested the relationship between energy and economic development (See Refs.

Fig. 1. The Environmental Kuznets Curve: a developmenteenvironment relationship. Source: Ref. [57]; p. 46.

[2,77,23,78,51,38,71,76,42,74,26] etc.). Payne [59] provided an extensive review of the studies on the empirical results of the energy consumption and economic growth relationship. Due to the omitted variables in previous studies, a third group of EKC studies has emerged. The third wave of EKC studies combines the two approaches used in previous research groups. Moreover, within the third group of EKC studies, the relationship between environmental pollution, economic growth, energy consumption and some other variables such as urbanization and trade openness has been investigated. Studies by Refs. [61,4,79,32,5,72,3,47,1,55,56,46,64] focused on the relationship between economic growth, energy consumption and pollution. According to the [22]; renewables are the fastest growing source of world energy and the share of renewables in total energy use will increase from 10% in 2008 to 14% in 2035 [22]. In many countries, considerable attention has been focused on renewable energy because of concerns over the volatility of oil prices, dependency on foreign energy sources (the energy security problem) and the environmental consequences of GHG emissions. Renewable energy market is supported by various incentive mechanisms to ensure sufficient investments in renewable energy sector. More than 100 countries both defined specific goals and developed focused policies regarding renewable energy. To develop renewable energy all over the world, market-based and non-market based promotion mechanisms such as feed-in tariffs, premiums, quota based green certificates, bidding incentives, incentives for investment, tax exemptions and discounts have been put in place by the governments. Worldwide governmental support for renewable energy rose from 41 billion US dollars in 2007 to 88 billion US dollars by 2011. Worldwide support for renewable energy is expected to be around 115 thousand million US Dollars in 2015 [20]. Recently, a number of energy consumption-economic growth studies have focused on renewable energy consumption Refs. [17,18,65,66,6e8,48] examined the relationship between renewable energy consumption and economic growth. The existing literature indicates that a large part of EKC studies focuses on the nexus of energy-output or pollution-output. Although recently a few studies have “combined” the approaches of these strands and investigated the inter-temporal linkage in the energy-environment-income nexus, in EKC studies generally other explanatory variables have been excluded. One important variable which is generally omitted in these relationships is the energy [50]. Including the fuels and/or energy and or splitting into energy types in EKC studies helps policy makers' understanding of the factors that may affect energy use and/or carbon emissions in the long term. To the best of our knowledge, there has been no study so far that tests the EKC hypothesis including renewable energy consumption as a variable affecting the environment. In a research paper by Ref. [50]; however, panel data analysis was conducted for 24 EU countries and it was assumed that the impacts of energy consumption on emissions were dependent on the primary energy mix. Despite this, we directly used renewable energy consumption as an important variable having affects on GHG emissions, since it was expected that greater use of renewables in final energy consumption would eventually lower GHG emissions in the world. In this study, we aim to examine the relationship between economic growth, GHG emissions and energy consumption taking into account renewable energy for EU countries using panel data analysis. Moreover, we also use centered values of explanatory variables to fix the multicollinerity problem which has been generally ignored in empirical EKC studies. The rest of the paper is organized as follows. In the second section, material and methods are presented. The third section discusses the empirical results. The fourth section discusses the results of the model. The final


section concludes with important results and presents policy implications. 2. Material and methods The annual time series data for EU countries were taken from the WDI (World Development Indicators) online database for 16 EU countries. The time period examined did not vary for the 16 EU countries however, some countries (Bulgaria, Czech Republic, Estonia, Ireland, Latvia, Lithuania, Romania, Slovakia and Slovenia) were excluded from the data set because their greenhouse emissions data was not available over the entire time period. The countries taken into the analysis and some descriptive statistics for them can be seen in Table 1. In Table 1, CO2 stands for carbon dioxide emissions per capita (measured in metric tons of CO2 per capita), GDPPC stands for per capita Gross Domestic Product (at constant 2000 US Dollars), REN stands for the renewable energy consumption (measured in kt (kilotons) of oil equivalent) per capita, and FOSS stands for fossil fuel energy consumption per capita (measured in kt of oil equivalent). According to Table 1, the minimum average value for per capita carbon emissions is evidently in Portugal, and Luxembourg has the maximum per capita carbon emissions for the period of 1990e2008. Additionally, Luxembourg also has the maximum average GDP and average fossil fuel energy consumption per capita and Portugal has the minimum average fossil fuel energy consumption per capita. Furthermore, the minimum renewable energy consumption per capita is in the UK (United Kingdom) and the maximum renewable energy consumption is in Finland for the analysed period of the current study. Three types of empirical specifications are typically used in the analysis of the EKC hypothesis: linear, a quadratic (inverted-U) and cubic specifications (N-shaped) or sideways-mirrored (S-shaped) [25]. There is a general functional form which may also consist of other relevant factors such as; time, regional characteristics and technical factors as external variables. These general forms are given in the following equations [39]; p. 242):

Qt ¼ a0 þ a1 ln Yt þ Gt þ εt :

(1)

Qt ¼ a0 þ a1 Yt þ a2 Yt2 þ Gt þ εt :

(2)

ln Qt ¼ a0 þ a1 ln Yt þ a2 ðln Yt Þ2 þ Gt þ εt : Qt ¼ a0 þ a1 Yt þ a2 Yt2 þ a3 Y 3 þ Gt þ εt :

(3)

ln Qt ¼ a0 þ a1 ln Yt þ a2 ðln Yt Þ2 þ a3 ðln Yt Þ3 þ Gt þ εt : In these specifications, Q is the per capita GHG emissions, Y is the per capita annual GDP, t is time and G is the external variable, ε is the stochastic error term and ai are the coefficients of the models (also called marginal propensity to emit) [39]; p. 242). In the linear specifications, if a1 > 0, the relationship between income and GHG emissions is linearly increasing and any increase in income leads a proportional increase in GHG emissions. This linear relationship reflects the scale effect that was discussed earlier. If a1 < 0, the relationship would be monotonically decreasing. In both cases the link between emissions and income only exists if a1 is significant [35]. In the quadratic case, if a1 > 0, a2 < 0 and a3 ¼ 0, emissions exhibit an inverted-U relationship to per capita income. This means that environmental degradation (pollution level) firstly increases with growing GDP, then eventually decreases when the economy develops (so-called EKC). In the cubic case, if a1 > 0, a2 < 0 and a3 > 0, an N-shaped relationship between emissions and income

441

can result. If the coefficients are reversed in terms of sign (a1 < 0, a2 > 0 and a3 < 0), a sideways-mirrored-S shape can be identified [25]; p. 5). To test the EKC hypothesis, an econometric framework has been specified to examine the relationship between per capita GHG emissions, per capita real GDP, squared of per capita real GDP, per capita renewable energy consumption and per capita fossil energy consumption in EU countries. Based upon the conceptional model, to conduct the empirical analysis, the following equation has been employed;

CO2it ¼ a þ b1 GDPPCit þ b1 GDPPC2it þ b3 RENit þ b4 FOSSit þ εit (4) t ¼ 1990, 1992,…,2008, i ¼ 1,2,…,16. In the above equation, as mentioned before, CO2 is per capita carbon dioxide emission (CO2) (measured in metric tons per capita), GDPPC is the per capita Gross Domestic Product (in constant 2000 US Dollars), GDPPC2 is the square of per capita real GDP, REN is the renewable energy consumption (measured in kt of oil equivalent) per capita, FOSS is fossil fuel energy consumption per capita (measured in kt of oil equivalent) and ε is the error term. Moreover, i is the subscript of countries and t is the subscript of the time dimension. In Eq. (4), a is a country specific term that captures all fixed factors inherent to each country that are not considered in the model, such as geographical, social and local policy aspects of the countries and the initial pollution level. Because of the relationship between economic development and energy utilization, we expected multicollinearity among the independent variables. The dependence between energy use and output is generally and unintentionally neglected due to the model concept of EKC Hypothesis [41]; p.277). To solve this problem, we used centered values of the independent variables. The centered values of the variables were obtained by recalculating the variables by subtracting their means. That is, we obtained the independent variables GDPPC-mean (GDPPC), REN-mean (REN) and FOSS-mean (FOSS) for each country. After centering, we took the variable GDPPC2 as the square of the centered GDPPC in the model. To check the multicollinearity, we got correlations for the variables, VIF (variance inflation factors) and coefficients of determination after running fixed effect regression for the centered and original values of the independent variables. We obtained R2 as the coefficient of determination for the general regression when the dependent variable CO2 was regressed on all the explanatory variables (GDPPC, GDPPC2, REN and FOSS). When the dependent variable GDPPC was regressed on GDPPC2, REN and FOSS, we obtained R21 ; when the dependent variable GDPPC2 was regressed on GDPPC, REN and FOSS, we obtained R22 ; when the dependent variable REN was regressed on GDPPC, GDPPC2 and FOSS, we obtained R23 and finally, when the dependent variable FOSS was regressed on GDPPC, GDPPC2 and REN, we obtained R24 . We obtained the coefficients of determination for centered and original values. As seen in Table 2, without centering, we had a serious multicollinearity problem. The correlation matrix had high correlations, VIF values were very high and R21 > R2 . However, after centering, the correlations and VIF values became acceptably smaller and R2 was higher than the other coefficients of determination R21 , R22 , R23 and R24 . As a result of these diagnostics, we used the centered values of the independent variables in the model. The expected sign of fossil energy consumption is positive since a higher level of energy consumption should result in greater economic activity and CO2 emissions. Renewable energy in our model consists of all types of renewables such as hydro, wind, solar, biofuels, biomass and waste. Fossil fuel is used in the production process of these renewable energy types. Therefore, a higher level


442

Table 1 Descriptive statistics of the variables for the countries analyzed.

BELGIUM

DENMARK

GERMANY

GREECE

SPAIN

FRANCE

HUNGARY

ITALY

LUXEMBOURG

NETHERLANDS

AUSTRIA

POLAND

PORTUGAL

FINLAND

SWEDEN

UK

TOTAL

CO2 GDPPC REN FOSS CO2 GDPPC REN FOSS CO2 GDPPC REN FOSS CO2 GDPPC REN FOSS CO2 GDPPC REN FOSS CO2 GDPPC REN FOSS CO2 GDPPC REN FOSS CO2 GDPPC REN FOSS CO2 GDPPC REN FOSS CO2 GDPPC REN FOSS CO2 GDPPC REN FOSS CO2 GDPPC REN FOSS CO2 GDPPC REN FOSS CO2 GDPPC REN FOSS CO2 GDPPC REN FOSS CO2 GDPPC REN FOSS CO2 GDPPC REN FOSS

Mean

Median

St. Deviation

Min.

Max.

10.94 21,775.06 115.54 5312.56 10.17 28,576.35 354.33 3300.08 10.58 22,350.24 116.57 4047.13 8.03 11,570.97 90.79 2312.63 7.00 13,746.97 103.95 2725.31 6.38 20,973.73 192.49 3916.38 5.77 4627.83 86.15 2476.74 7.72 18,612.43 43.32 2842.26 22.57 43,684.26 115.93 8318.01 10.91 22,814.90 109.83 4542.39 8.06 22,921.55 423.19 3221.72 8.56 4247.74 109.21 2398.31 5.53 10,614.95 264.36 1907.19 11.21 22,594.13 1177.17 5120.51 6.02 26,980.61 865.81 4744.30 9.41 24,051.33 38.17 3656.75 9.30 20,008.94 262.93 3802.64

11.15 21,946.73 99.15 5342.52 10.01 29,055.27 313.48 3289.22 10.39 22,295.10 93.23 4030.66 8.10 10,942.83 89.61 2276.42 7.07 13,836.70 102.04 2809.57 6.23 21,146.40 188.51 3939.51 5.74 4358.70 75.89 2451.81 7.70 18,713.35 38.20 2880.02 21.96 43,420.31 101.53 8356.11 10.87 23,429.80 104.67 4534.48 7.94 23,181.69 391.12 3174.92 8.37 4249.28 116.16 2402.80 5.58 11,097.16 260.00 2004.03 10.95 22,386.63 1217.23 5103.58 6.07 26,724.35 866.70 4813.58 9.31 24,098.55 30.42 3694.07 8.51 20,317.72 117.66 3548.13

0.64 2197.69 43.95 261.20 1.21 2888.69 107.30 222.69 0.73 1721.39 68.03 139.92 0.70 1844.94 4.08 249.25 0.77 1824.47 10.91 341.50 0.39 1717.02 14.00 143.33 0.20 803.16 24.23 93.79 0.24 1296.47 23.64 178.86 3.57 7817.49 50.74 803.50 0.48 2763.88 40.15 108.87 0.52 2635.17 96.35 242.48 0.61 1037.52 26.30 134.46 0.60 1160.40 18.30 281.04 1.00 3763.67 227.43 326.27 0.43 3808.87 129.83 208.54 0.51 3814.82 20.19 129.25 4.06 9886.25 319.96 1572.52

9.65 18,738.03 74.67 4768.49 8.38 24,097.91 221.68 2950.95 9.57 19,600.84 59.83 3757.54 6.59 9635.29 87.45 2000.92 5.79 11,335.55 89.86 2214.15 5.85 18,731.66 173.29 3652.17 5.44 3707.13 63.27 2338.31 7.25 16,610.79 16.59 2554.48 17.32 32,476.86 59.66 6968.41 10.21 18,858.00 63.77 4329.44 7.26 19,192.07 322.72 2880.72 7.77 2872.97 54.17 2208.04 4.38 8771.93 244.15 1428.69 9.74 17,643.55 810.68 4570.32 5.25 22,065.24 643.34 4302.70 8.52 19,008.94 10.96 3313.17 4.38 2872.97 10.96 1428.69

11.78 25,100.47 221.06 5606.41 13.05 32,767.40 547.03 3879.93 12.01 25,620.08 278.15 4363.52 8.89 14,728.86 105.53 2632.66 8.14 16,351.11 126.19 3149.95 7.30 23,516.22 226.99 4116.01 6.12 5947.16 151.79 2699.15 8.13 20,291.23 89.30 3066.16 30.10 56,388.99 246.06 9351.93 12.00 27,348.47 191.53 4720.49 9.04 27,295.13 663.20 3606.99 9.62 6235.76 153.58 2646.90 6.44 11,966.00 300.38 2227.68 13.21 28,839.22 1475.93 5671.95 6.85 33,259.26 1079.09 4982.08 10.34 29,627.91 77.14 3850.64 30.10 56,388.99 1475.93 9351.93

Note: This table was prepared by authors using the data taken from the WDI (World Development Indicators) online database.


of renewable energy consumption should result in greater economic activity and GHG emissions as well. The expected sign of renewable energy consumption is also positive since renewable energy should result in greater economic activity and consequently, stimulate CO2 emissions. Despite this, consumption of renewable energy should stimulate emissions at a lower level than fossil fuels. According to the EKC hypothesis, the signs of GDP per capita and GDP per capita square are expected to be both positive and negative respectively, to reflect the inverted U-shape pattern.

3. Results The annual data from EU countries for the period 1990e2008 has been used to evaluate the EKC Hypothesis (Eq. (4)). This study utilized panel fixed effect analysis to examine the relationship between greenhouse gases, energy consumption (fossil energy and renewable energy) and GDP. The panel data method increases the power of empirical analysis, since it combines information from both the time and cross-section dimensions of the data set, and therefore, allows the researcher greater flexibility in modelling, especially with regard to differences in behaviours of individuals [29]; p. 345). There are three models that are used for analysis of panel data. The first model simply combines or pools all the time series and cross section data and estimates the model using a pooled OLS (pooled ordinary least squares) method. The intercept term is assumed to be common [60].

ait ¼ a

(5)

The difficulty of pooled least squares is that the assumption of a constant intercept and slope is unreasonable. The fixed effects model is estimated to allow for different intercepts for the different cross section units, hence:

ait ¼ ai

where Eðai εit Þs0

(6)

Despite this, the random effects model treats intercepts as a random variable across pooled member countries, therefore:

ait ¼ a þ ui

where Eðui εit Þ ¼ 0

(7)

The main concern here was that the random effects model could not be estimated as it required the number of cross-sections to be greater than the number of regressors [11]. In order to test which model was better between the pooled and fixed effect models, we

Table 2 Multicollinearity analysis of independent variables. Without centering Correlation matrix

GDPPC GDPPC2 REN FOSS

GDPPC 1 0.9376 0.2508 0.8030

VIF GDPPC2 1 0.1609 0.7988

REN

1 0.2198

FOSS 148.17 62.23 34.07 32.98

1

Coefficients of determination R2 R21 R22 R23 R24

¼ ¼ ¼ ¼ ¼

0.98 0.99 0.98 0.97 0.97

conducted the following F-test, where the null and alternative hypothesis were;

H0 ¼ ait ¼ a HA ¼ ait ¼ ai : According to the F test, the test statistic was F15, 284 ¼ 950.51 and Prob > F ¼ 0.000. Since this is evidence against the null hypothesis being true, it means that the country factors are not equal to a constant. As a result of this test, we concluded that the fixed effects model was better than the pooled model for our data set. After the F test, we performed the Hausman test to decide between the fixed or random effects model for the data set. As a result of the Hausman test, we got the statistic chi2(1) ¼ 32.96, P ¼ 0.000 and concluded that the FE (fixed effects) estimation was necessary to best fit the data. Hausman test results can be seen in Table 3. Test: Ho: difference in coefficients not systematic

h i chi2ð1Þ ¼ ðb BÞ0 ðV_b V_BÞ^ð1Þ ðb BÞ ¼ 32:96 Prob > chi2 ¼ 0:0000 We ran FE panel estimation and then checked for the presence of cross-sectional dependency, heteroskedasticity and autocorrelation problems. Testing for cross-sectional dependency, we performed Pesaran's test of cross-sectional independence and obtained the test statistic 3.488 and P-value ¼ 0.000. Therefore, considering this test result, we concluded that our data set had a cross-sectional dependency problem. Additionally, we performed a modified Wald test for groupwise heteroskedasticity in the fixed effects regression models and got Chi2(16) ¼ 4763.83 and Prob > Chi2 ¼ 0.000 and rejected the null of constant variance. Finally, we calculated Bhargava, Franzini and Narendranathan's DurbineWatson test statistic to be 0.883, and Baltagi-Wu's LBI (Locally Best Invariant) statistic to be 1.061. These two statistics indicated that we had a serious autocorrelation problem since they were fairly smaller than number 2 [11]. As a result of these diagnostics, our fixed effect model was cross-sectional dependent, heteroskedastic and autocorrelated. Due to this result, we used a fixed effects estimator with Driscoll and Kraay standard errors to estimate Eq. (4) [36]. 4. Empirical results This study estimated Eq. (4) using the panel fixed effect model for the 16 EU countries. The empirical findings provided no evidence of an inverted U-shaped relationship between real GDP per capita and GHG emissions. The results of the panel analysis are given in Table 4. As seen in Table 4, the estimated panel model is significant (F(4, 18) ¼ 141.62, Prob > F ¼ 0.000). The estimated regression of the CO2 emission functions appears to fit the data well with more than 85 Table 3 Hausman test. Coefficients

With centering Correlation matrix

GDPPC GDPPC2 REN FOSS

GDPPC 1 0.1296 0.6568 0.1758

VIF GDPPC2 1 0.1136 0.1881

REN

1 0.1530

FOSS

1

1.79 2.31 1.77 1.10

Coefficients of determination R2 R21 R22 R23 R24

¼ ¼ ¼ ¼ ¼

0.99 0.44 0.57 0.43 0.09

443

GDPPC GDPPC2 REN FOSS a b

(b)a Fixed

(B)b Random

(b B) Difference

sqrt(diag(V_b V_B)) S.E.

0.0001698 1.53e-08 0.0014468 0.0031594

0.0001701 1.59e-08 0.0014415 0.0031529

2.69e-07 5.80e-10 5.36e-06 6.42e-06

4.69e-08 1.01e-10 9.33e-07 1.12e-06

b ¼ consistent under Ho and Ha; obtained from xtreg. B ¼ inconsistent under Ha, efficient under Ho; obtained from xtreg.


444 Table 4 Panel results for CO2.

Cons. GDPPC GDPPC2 REN FOSS N F(4, 18) Prob > F Within R-squared a b

Coef.

Drisc/Kraay Std. Err.

P-value

9.166838a 0.0001698a 1.53e-08a 0.0014468b 0.0031594a 304 141.62 0.000 0.8569

0.0402996 0.0000243 2.49E-09 0.0005974 0.0001514

0.000 0.000 0.000 0.026 0.000

Stands for the 1% significance level. Stands for the 5% significance level.

percent of the variation in CO2 emissions explained by the model (R-squared ¼ 0.8569). All the coefficients are statistically significant at the 0.05 level of significance, except for constant term. All country constants (intercepts) have positive values and they are significant. The signs of the coefficients of the GDP per capita and its squared form indicate the shape of a Kuznets Curve. Furthermore, the renewable energy quantity per capita has positive effects on carbon emissions (coeff. ¼ 0.0014468, P ¼ 0.026) and also fossil fuel energy consumption per capita has positive effects on carbon emissions (coeff. ¼ 0.0031594, P ¼ 0.000). The statistical significance of the square of per capita real income rules out the suggestion that output increases monotonically with the level of CO2 emissions. However, the results do not seem to provide support for the EKC hypothesis that the level of environmental pollution initially increases with income until it reaches its stabilization point then decreases. In other words, our results show that the invertedU shape is not applicable for EU countries. Hence, first environmental pollution decreases with income instead of increasing since the statistically significant coefficients of GDPPC and GDPPC2 are negative and positive, respectively (See Table 4). The coefficients of fossil energy and renewable energy consumption are also statistically significant at the 0.05 significance level and they have the expected signs. Estimated coefficients state that pollution increases with both fossil and renewable energy consumption. As expected, fossil fuel consumption leads to a greater increase in pollution levels than renewable energy consumption. Moreover, the results state that renewable energy consumption contributes less than around 1/2 that of fossil energy consumption to CO2 emissions in EU counties. The EKC hypothesis has been studied empirically in numerous studies and most of these studies used cross-sectional data. The empirical evidence for the existence of an EKC can be found in various research papers: Refs. [19,37,53,55,49,10,63,43,33]. Despite this, the results of many research studies concluded that there was no EKC relationship: Refs. [30,31,13,50,58,25,52,3,62,44,47]. Our results showed similar findings in line with the second group. Therefore, it could be inferred that empirical evidence for the existence of the EKC hypothesis was not conclusive. Since there was no statistical evidence in favour of the existence of an EKC (inverted-U) for CO2 emissions per capita in the 16 EU countries that between 1990 and 2008, economic development itself ultimately would not lead to the reduction of emissions in the European Union. 5. Conclusions and policy implications This paper proposed and estimated a panel model for EU-16 over the period of 1990e2008, which linked greenhouse emissions (CO2) per capita with real GDP per capita and aggregate

energy consumption. Since the emissions data were obtained for the 1990e2008 period, the data availability on EU countries during this period could constitute a limitation for EKC hypothesis testing. However, a limited data set of countries can still provide an initial inference on the validity of the EKC hypothesis. The main contribution of this paper was two-fold: to investigate the EKC hypothesis by splitting aggregate energy consumption into renewable and fossil fuel energy components for EU countries and to take into account the multicollinearity among the income and energy variables which has been generally ignored in the EKC studies. The main findings of this paper can be summarized as follows. Our results indicated that between 1990 and 2008, there was no statistical evidence in favour of the existence of an EKC (inverted-U) for CO2 emissions per capita in the 16 EU countries. Our results not confirm that GHG emissions increase with economic growth then reach a turning point and then start to decline with higher levels of economic growth. We were surprised to see that the EKC hypothesis for the EU countries was rejected, since the EU was determined to decrease the GHG emissions and additionally had ambitious renewable energy targets (Kyoto commitment and EU 20/20/20 targets). On the contrary, we found the existence of a U shaped relationship between GHG emissions and real GDP in EU countries. This result has two important implications. Firstly, regulations in EU countries have not actually improved efforts to reduce their CO2 emissions. As a result, we conclude that policies to reduce fossil energy use and/or carbon emissions must go beyond promoting economic growth. This means that economic development itself cannot be expected to control CO2 emissions and/or environmental pollution. In other words, policies that move towards environmental remediation should not wait for a rise in income levels, but be implemented immediately. Secondly, some other factors not included in this model should be taken into account such as technology, taxes, trade, and so on. In our opinion, the main result of this study strengthens the opposite view of the EKC hypothesis. There is an important criticism of the EKC hypothesis (see Ref. [25], as the inverted U-shape is only a statistical result, and not a common development path since it is derived from cross-section data. Therefore, testing the EKC hypothesis country by country and comparing the obtained results might give more consistent results than that of panel data analysis. Last but not least, based on the model we have estimated, it is found that renewable energy consumption significantly lowers GHG emissions (around 1/2) that of fossil energy consumption in the EU countries. This implies that much more improvement in energy efficiency and/or, a shift in the energy mix towards less polluting energies (renewable energy technologies) could be very important in achieving environmental targets. Since the inputs of finite energy resources and the green house emissions are considerably low compared to the conventional energy sources in renewable energy chain, deploying the renewable energy instead of the fossil fuels might contribute to climate mitigation efforts. Our findings also highlight that regulations to support renewable sources would yield significant reductions in per capita carbon dioxide emissions. Therefore, we conclude that the findings of our research provide highly important information for policy makers, both in the EU and in the world.

References [1] Acaravci A, Ozturk I. On the relationship between energy consumption, CO2 emissions and economic growth in Europe. Energy 2010;35:5412e20. [2] Akarca AT, Long TV. On relationship between energy consumption and GNP: a re-examination. J Energy Dev 1980;5:326e31. [3] Akbostancı E, Turut-Asık S, Tunc GI. The relationship between income and environment in Turkey: is there an environmental Kuznets curve? Energy Policy 2009;37(3):861e7.

€lük, M. Mert / Energy 74 (2014) 439e446 G. Bo [4] Ang JB. CO2 emissions, energy consumption, and output in France. Energy Policy 2007;35:4772e8. [5] Apergis N, Payne J. CO2 emissions, energy usage, and output in Central America. Energy Policy 2009;37:3282e6. [6] Apergis N, Payne J. The emissions, Energy consumption and growth nexus: evidence from the commonwealth of independent states. Energy Policy 2010a;38:650e5. [7] Apergis N, Payne J. Renewable energy consumption and economic growth: evidence from a panel of OECD countries. Energy Policy 2010b;38:656e60. [8] Apergis N, Payne J. Renewable energy consumption, growth, and economic development: an exploratory panel analysis. Adv Energy Res 2011;9:359e72. [10] Auci S, Travato G. The environmental Kuznets curve within European Countries and sectors: greenhouse emission, production function and technology; 2011. www.webmeets.com/files/papers/EAERE/2011/549/auci_trovato% 20june%202011.pdf [accessed 25.10.12]. [11] Baltagi BH. Econometric analysis of panel data. 4th ed. UK: John Wiley & Sons ltd.; 2009. [12] Beckerman W. Economic growth and the environment: whose growth? Whose environment? World Dev 1992;20(4):481e96. [13] Bengochea-Morancho A, Tamarit FH, Zarzoso IM. Economic growth and CO2 emissions in the European Union. Environ Resour Econ 2011;19:165e72. [14] Bo S. A literature survey on environmental Kuznets curve. Energy Procedia 2011;5:1322e5. [15] Bouvier R. Working Paper Series. Air pollution and per capita income, 84. Political Economy Research Institute (PERI); 2004., www.peri.umass.edu/ fileadmin/pdf/working_papers/working_papers_51-100/WP84.pdf [accessed 25.04.14]. [16] Brock WA, Taylor MS. Economic growth and the environment: a review of theory and empirics [Chapter 28]. In: Aghion Philippe, Durlaof Steven N, editors. Handbook of economic growth, vol. 1B. Amsterdam, Netherland: Elsevier B.V.; 2005. [17] Chien T, Hu JL. Renewable energy and macroeconomic efficiency of OECD and non-OECD economies. Energy Policy 2007;35:3606e15. [18] Chien T, Hu JL. Renewable energy: an efficient mechanism to improve GDP. Energy Policy 2008;36:3045e52. [19] de Bruyn SM, van den Bergh JCJM, Opschoor JB. Economic growth and emissions: reconsidering the empirical basis of environmental Kuznets curves. Ecol Econ 1998;25(2):161e75. [20] Deloitte. New life for renewable energy resources e renewable energy policies and expectations; 2011. www.deloitte.com/assets/Dcom-Turkey/Local% 20Assets/Documents/turkey_tr_energy_renewableenergy_090611.pdf [accessed 06.09.12]. [21] Dinda S. Environmental Kuznets curve hypothesis: a survey. Ecol Econ 2004;49:431e55. [22] EIA. International energy outlook. U.S. Energy Information Administration; 2011. www.eia.gov/forecasts/ieo/pdf/0484(2011).pdf [accessed 12.10.12]. [23] Erol U, Yu ESH. Time series analysis of the causal relationships between US energy and employment. Resour Energy 1987;9:75e89. [24] Frankel J. Global environmental policy and global trade policy. Discussion Paper 08-14. Harward Kennedy School, John F.Kenedy School of Government; 2008., belfercenter.ksg.harvard.edu/files/Frankel2Web.pdf [accessed 10.10.12]. [25] Friedl B, Getzner M. Environment and growth in a small open economy: an EKC case-study for Austrian CO2 Emissions. Discussion Paper, No:2002/2. College of Business Administration, University of Klagenfurt; 2002., wiwi.uniklu.ac.at/Forschung/2002_02.pdf [accessed 10.11.12]. [26] Ghosh S, Kanjilal K. Long-term equilibrium relationship between urbanization, energy consumption and economic activity: emprical evidence from India. Energy 2014;66:324e31. [27] Giovanis E. Environmental Kuznets curve: evidence from British Household Panel Survey. Econ Model 2013;30:602e11. [28] Granados JAT, Carpintero O. Dispelling the smoke: CO2 emissions and economic growth from a global perspective; 2009. sitemaker.umich.edu/tapia_ granados/files/co2_emissions_and_gdp_growth_-_dec_2009_-_f2.pdf [accessed 14.01.13]. [29] Greene W. Models for panel data; 2010. pages.stern.nyu.edu/~wgreene/ DiscreteChoice/Readings/Greene-Chapter-9.pdf [accessed 02.11.13]. [30] Grossman GM, Krueger AB. Environmental impacts of a North American free trade agreement. NBER Working Paper no. 3914; 1991., www.nber.org/ papers/w3914.pdf [accessed 10.10.12]. [31] Grossman GM, Krueger AB. Economic growth and the environment. Q J Econ 1995;110:353e77. [32] Halicioglu F. An econometric study of CO2 emissions, energy consumption, income and foreign trade in Turkey. Energy Policy 2009;37:1156e64. [33] Hamit-Haggar M. Greenhouse gas emissions, energy consumption and economic growth: a panel cointegration analysis from Canadian industrial sector perspective. Energy Econ 2012;34:358e64. [34] Heil MT, Selden TM. International trade intensity and carbon emissions: a cross-country econometric analysis. J Environ Dev 2001;10:35e49. [35] Hervieax MS, Darne O. Environmental Kuznets curve and ecological footprint: A time series analysis. Working Paper. University of Nantes; 2013., www. touteconomie.org/afse/index.php/congres/2013/paper/viewFile/51/19 [accessed 26.04.16]. [36] Hoechle D. Robust standard errors for panel regressions with cross-sectional dependence. Stata J 2007;7(3):281e312.

445

[37] Holtz-Eakin D, Selden TM. Stoking the fires? CO2 emissions and economic growth. J Public Econ 1995;57:85e101. [38] Hondroyiannis G, Lolos S, Papapetrou E. Energy consumption and economic growth: assessing the evidence from Greece. Energy Econ 2002;24: 319e36. [39] Huang WM, Lee GWM, Wu CC. GHG emissions, GDP growth and the Kyoto protocol: a revisit of environmental Kuznets curve hypothesis. Energy Policy 2008;36:239e47. [40] Hussen AM. Principles of environmental economics and sustainability: an integrated and ecological approach. New York: Routledge; 2005. [41] Itkonen JVA. Problems estimating the carbon Kuznets curve. Energy 2012;39: 274e80. [42] Lee C, Lee J. A panel data analysis of the demand for total energy and electricity in OECD countries. Energy J 2010;31(1):1e23. [43] Lee CC, Chiu Y-B, Sun CB. Does one size fit all? a reexamination of the environmental Kuznets curve using the dynamic panel data approach. Rev Agric Econ 2009;31(4):751e78. [44] Lekakis JN. Environment and development in a Southern European Country: which environmental Kuznets curves? J Environ Plan Manag 2000;43(1): 139e53. [45] Lieb CM. The environmental Kuznets curve e a survey of the empirical evidence and of possible causes. Discussion Paper Series, No:391. University of Heidelberg, Department of Economics; 2003., www.uni-heidelberg.de/md/ awi/forschung/dp391.pdf [accessed 06.10.12]. [46] Lotfalipour MR, Falahi MA, Ashena M. Economic growth, CO2 emissions and fossil fuels consumption in Iran. Energy 2010;35:5115e20. [47] Luzzati T, Orsini M. Investigating the energyeenvironmental Kuznets curve. Energy 2009;34(3):291e300. [48] Mahmoodi M, Mahmodi E. Renewable energy consumption and economic growth: the case of 7 Asian developing countries. Am J Sci Res 2011;35: 146e52. [49] Markandya A, Golub A, Pedroso-Galinato S. Empirical analysis of national income and SO2 emissions in selected European Countries. Environ Resour Econ 2006;35:221e57. [50] Marrero GA. Greenhouse gases emissions, growth and the energy mix in Europe. Energy Econ 2010;32:1356e63. [51] Masih A, Masih R. Energy consumption, real income and temporal causality: results from a multicounty study based on cointegration and error-correction modelling techniques. Energy Econ 1996;18:165e83. [52] Mazzanti M, Zoboli R. Delinking and environmental Kuznets curves for waste indicators in Europe. Environ Sci 2005;2(4):409e25. [53] Moomaw WR, Unruh GC. Are environmental Kuznets curves misleading us? The case of CO2 emissions. Environ Dev Econ 1997;2(4):451e63. [54] Oliver JGJ, Jansens G, Peters JAHW. Trends in global CO2 emissions. 2012 Report, JRC Background Studies; 2012., edgar.jrc.ec.europa.eu/ CO2REPORT2012.pdf [accessed 03.12.12]. [55] Pao HT, Tsa CM. Modelling and forecasting the CO2 emissions, energy consumption, and economic growth in Brazil. Energy 2011;36:2450e8. [56] Pao HT, Yu HC, Yang YH. Modeling the CO2 emissions, energy use, and economic growth in Russia. Energy 2011;36:5094e100. [57] Panayotou T. Chapter 2: economic growth and the environment. Economic survey of Europe, No: 2; 2003. www.unece.org/fileadmin/DAM/ead/pub/032/ 032_c2.pdf [accessed 15.06.13]. [58] Poudel BW, Paudel KP, Bhattarai K. Searching for an environmental Kuznets curve in carbon dioxide pollutant in Latin American countries. J Agric Appl Econ 2009;41(1):13e27. [59] Payne JE. A survey of the electricity consumption-growth literature. Appl Energy 2010;87:723e31. [60] Paul RK. Econometric analysis using panel data; 2011. www.iasri.res.in/ sscnars/socialsci/12-Panel%20data.pdf [accessed 02.12.12]. [61] Richmond AK, Kaufmann RK. Is there a turning Point in the relationship between income and energy use and/or carbon emissions? Ecol Econ 2006;56: 176e89. M, Galletto V. Economic growth and atmospheric [62] Roca J, Padilla E, Farre pollution in Spain: discussing the environmental Kuznets curve hypothesis; 2001. dep-economia-aplicada.uab.cat/repec/doc/wp0101.pdf [accessed 20.10.12]. [63] Saatci M, Dumrul Y. The relationship between economic growth and environmental pollution: the estimation of environmental Kuznets curve with a cointegration analysis of Structural Breaks for Turkish economy (in Turkish). _ _ Erciyes Üniversitesi Iktisadi ve Idari Bilimler Fakültesi Dergisi 2011;66(37): 65e86. [64] Saboori B, Sapri M, bin Baba M. Economic growth, energy consumption and CO2 emissions in OECD (Organization for Economic Co-operation and Development)'s transport sector: a fully modified bi-directional approach. Energy 2014;66:150e61. [65] Sadorsky P. Renewable energy consumption and income in emerging economies. Energy Policy 2009a;37(10):4021e8. [66] Sadorsky P. Renewable energy consumption, CO2 emissions and oil prices in G7 countries. Energy Econ 2009b;31(3):456e62. [67] Schmalensee R, Stoker TM, Judson RA. World carbon dioxide emissions: 1950e2050. Rev Econ Stat 1998;80(1):15e27. [68] Selden TM, Song D. Environmental quality and development: is there a Kuznets Curve for air pollution emissions? J Environ Econ Manag 1994;27: 147e62.

446


[69] Shafik N. Economic development and environmental quality: an econometric analysis. Oxf Econ Pap 1994;46:757e73. [70] Shafik N, Bandyopadhyay S. Economic growth and environmental quality: time series and cross-country evidence (Background paper for World Development Report 1992). Washington D.C.: World Bank; 1992. [71] Soytas U, Sari R. Energy consumption and GDP: causality relationship in G-7 countries and emerging markets. Energy Econ 2003;25:33e7. [72] Soytas U, Sari U, Ewing BT. Energy consumption, income and carbon emissions in the United States. Ecol Econ 2007;62:482e9. [73] Stern DI. The rise and fall of the environmental Kuznets curve. World Dev 2004;32(8):1419e39. [74] Warr BS, Ayres RU. Evidence of causality between the quantity and quality of energy consumption and economic growth. Energy 2010;35(4):1688e93. [75] Yandlee B, Vijayaraghavan M, Bhattarai M. The environmental Kuznets curve a premier. PERC Research Study, 02-1; 2002., www.macalester.edu/~wests/ econ231/yandleetal.pdf [accessed 25.11.12].

[76] Yoo SH. Electricity consumption and economic growth:evidence from Korea. Energy Policy 2005;33(12):1627e32. [77] Yu ESH, Choi JY. The causal relationship between energy and GNP: an international comparison. J Energy Dev 1985;10:249e72. [78] Yu ESH, Jin JC. Cointegration tests of energy consumption, income, and employment. Resour Energy 1992;14:259e66. [79] Zhang XP, Cheng XM. Energy consumption, carbon emissions, and economic growth in China. Ecol Econ 2009;68(10):2706e12. [80] Zhao X. The impact of CO2 emission cuts on income, Agricultural & Applied Economics Association's 2011 AAEA & NAREA joint annual Meeting, July 24e26, 2011. Pittsburgh, Pennsylvania, USA; 2011. ageconsearch.umn.edu/ bitstream/103412/2/Zhao-1-the%20impact%20of%20The%20cost%20of% 20CO2%20emission%20cuts%20on%20income.pdf [accessed 05.12.12].