Testing Environmental Kuznets Curve in the Selected Transition Economies with Panel Smooth Transition Regression Analysis

Zortuk, Mahmut; Çeken, Sinan Article Testing Environmental Kuznets Curve in the Selected Transition Economies with Panel Smooth Transition Regression Analysis Amfiteatru Economic Journal Provided in Cooperation with: The Bucharest University of Economic Studies Suggested Citation: Zortuk, Mahmut; Çeken, Sinan (2016) : Testing Environmental Kuznets Curve in the Selected Transition Economies with Panel Smooth Transition Regression Analysis, Amfiteatru Economic Journal, ISSN 2247-9104, The Bucharest University of Economic Studies, Bucharest, Vol. 18, Iss. 43, pp. 537-547 This Version is available at: https://hdl.handle.net/10419/169019 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. http://creativecommons.org/licenses/by/4.0/ The Constraints to the Economic Development in the Former Socialist EU Countries from the Central and Eastern Europe AE Vol. 18 • No. 43 • August 2016 537 TESTING ENVIRONMENTAL KUZNETS CURVE IN THE SELECTED TRANSITION ECONOMIES WITH PANEL SMOOTH TRANSITION REGRESSION ANALYSIS Mahmut Zortuk1 * and Sinan Çeken2 1)2) Dumlupınar University, Kutahya, Turkey Please cite this article as: Zortuk, M. and Çeke, S., 2016. Testing Environmental Kuznets Curve in the Selected Transition Economies with Panel Smooth Transition Regression Analysis. Amfiteatru Economic, 18(43), pp. 537-547 Article History Received: 12 February 2016 Revised: 5 May 2016 Accepted: 4 June 2016 Abstract The Environmental Kuznets Curve (EKC) introduces an inverted U-shaped relationship between environmental pollution and economic development. The inverted U-shaped curve is seen as complete pattern for developed economies. However, our study tests the EKC for developing transition economies of European Union, therefore, our results could make a significant contribution to the literature. In this paper, the relationship between carbon dioxide (CO2) emissions, gross domestic product (GDP), energy use and urban population is investigated in the Transition Economies (Bulgaria, Croatia, Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Romania, Slovakia and Slovenia). Environmental Kuznets Curve is tested by panel smooth transition regression for these economies for 1993 – 2010 periods. As a result of study, the null hypothesis of linearity was rejected and noremaining nonlinearity test showed that there is a smooth transition exists between two regimes (below $5176 GDP per capita is first one and above $5176 GDP per capita is second one) in the related period for these economies. Keywords: panel data models, panel smooth transition regression model, transition economies, environmental Kuznets curve. JEL Classification: C23, C24, P20, P28 Introduction There are several factors exist that effect the environment adversely while economic growth continues. The main factor of this degradation is Greenhouse gases (GHGs) among the others. According to IPCC report (IPCC, 2014), human-induced CO2 emissions merely holds more than 75% of the GHG emissions. Thus the main concern of this paper is the relationship of CO2 emissions and income. * Corresponding author, Mahmut Zortuk ‒ [email protected] AE Testing Environmental Kuznets Curve in the Selected Transition Economies with Panel Smooth Transition Regression Analysis 538 Amfiteatru Economic The Environmental Kuznets Curve (EKC) proposes an inverted-U shaped relationship between various indicators of environmental pollution and economic activity (Apergis and Payne, 2009). Accordingly, environmental deterioration increases in the first stage of economic growth until threshold or turning point, and then in the second stage, it begins to decrease. This pattern could be seen as the path of the developed economies. However, the findings for the developing economies’ path is still ambiguous in the literature. In our study, this unclear relationship is going to be evaluated by an unusual non-linear model. Therefore, our findings could make an important contribution to the literature. In order to assess this relationship in Transition Economies, recently developed Panel Smooth Transition Regression (PSTR) by Fok, Dijk, and Franses (2005) and Gonzalez, Terasvirta and Dijk. (2005) is implemented to data for 1993-2010 period. In the study, each countries’ path is going to be evaluated seperately. In this way, it is aimed to observe the path of the individual countries in order to obtain their curve with their threshold income level. The threshold value enables us to understand their paths better. In order to do that non-linear relationship is going to be evaluated first and then the number of transition functions is going to be defined which enables us to determine the number of regimes in the related period. The paper is organized as follows; section two outlines the related literature, section three describes our data and model specification, section four presents our empirical results, and finally section five describes conclusion. 1. Literature Review Numerous studies have been conducted in different types of techniques so far in order to shed light on CO2 – income relationship. On the basis of this relationship, in several cases, a number of empirical studies have identified a U-shaped curve; Kahuthu (2006) by Panel Data Analysis; Jalil and Mahmud (2009) by ARDL Model; Musolesi, Mazzanti and Zoboli (2009) by Panel Bayesian Estimation; Nasir and Rehman (2011) by Cointegration; Rehman, Nasir, and Kanwal (2012) by Panel Model; Kivyiro and Arminen (2014) by ARDL Model; López-Menéndez, Pérez and Moreno (2014) by Panel Model; Shahbaz, et al. (2014) by ARDL Model; Heidari, Katircioğlu, and Saeidpour (2015) by Panel Smooth Transition Regression. Main results of these studies can be thought supportive for the EKC theory. It would outline methodologies and their main results of these studies (table no.1). Table no. 1: Literature Review Authors Methodology Main Result(s) Bertinelli and Strobl (2005) Semi Parametric Regression Estimator A bell-shaped link between CO2 and GDP/capita. Azomahou, Laisney and Van (2006) Nonparametric Panel Approach There is a stable relationship between CO2 emissions per capita and GDP per capita over time during the period of the study. Galeotti, Lanza and Pauli (2006) Estimation of alternative functional forms Cointegrated relationship between per capita emissions and income. Kahuthu A. (2006) Panel Data Analysis CO2 emissions and GDP per capita seems to have changed from a linear shape to an inverted-U one. The Constraints to the Economic Development in the Former Socialist EU Countries from the Central and Eastern Europe AE Vol. 18 • No. 43 • August 2016 539 Authors Methodology Main Result(s) Huang, Lee, and Wu (2008) Quadratic Model Exhibit a quasi-L-shape or hockey-stick-curve trend. Aslanidis and Iranzo (2009) Panel Smooth Transition Regression There is no evidence of environmental Kuznets curve Halicioglu F. (2009) ARDL Model The elasticity of CO2 emissions with respect to income in the long run is found to be 12.31-1.66y. Jalil and Mahmud (2009) ARDL Model The positive sign with income and negative sign with the quadratic term of income confirms the existence of EKC for CO2 emission in the case of China. Lee and Lee (2009) Panel Cointegration Real GDP and CO2 emissions in these countries are a mixture of I(0) and I(1) processes. Musolesi, Mazzanti and Zoboli (2009) Panel Bayesian Estimation More industrialized countries show evidence of EKC in quadratic specifications. A lessdeveloped country consistently shows that CO2 emissions rise positively with income. Fodha and Zaghdoud (2010) Johansen Cointegration Approach There is a monotonically increasing linear relationship between per capita CO2 emissions and per capita GDP. He and Richard (2010) Parametric Cubic Models Per capita GDP and per capita CO2 emissions exhibit monotonically increasing but that the slope of that function changes often over time. Hossain S. Md. (2011) Johansen Panel Cointegration There is a short run casual relationship from economic growth and trade openness to CO2 emissions. Nasir and Rehman (2011) Johansen Cointegration and VECM The EKC is only a long-run phenomenon in case of Pakistan. Pao and Tsai (2011) Multivariate Granger Causality CO2 emissions appear to be GDP elastic if GDP is less than 1.174, with GDP being inelastic if GDP is greater than 1.174. Sharma S. S. (2011) Panel Model GDP per capita has positive effect on CO2 emissions. Arouri et al. (2012) Panel ECM and Cointegration Real GDP exhibits a quadratic relationship with CO2 emissions for MENA region. Borhan, Ahmed and Hitam (2012) Simultaneous Equation Models Simultaneous relationship exists between CO2 of air pollution and income. Rehman, Nasir and Kanwal (2012) Panel Model An inverted-U shaped relationship between per capita income (GDP) and the CO2 emissions. Bassetti, Benos and Karagiannis (2013) Panel Model The results do not support theoretical models predicting that a country can be caught simultaneously in a poverty and environment trap Mosheim R. (2013) Panel Model There is a support for the pattern that a greater degree of income equality leads to better environmental outcomes. AE Testing Environmental Kuznets Curve in the Selected Transition Economies with Panel Smooth Transition Regression Analysis 540 Amfiteatru Economic Authors Methodology Main Result(s) Wang C. (2013) Decomposition Technique Output growth is a major source for increases in carbon dioxide emissions while decline in energy intensity is the main contributor to the reduction of emissions. Azlina, Law and Mustapha (2014) Multivariate Cointegration There is no casual evidence of an effect of income on emission. Boutabba M. A. (2014) ARDL Model The long-run and casual relationships between per capita CO2 emissions, financial development, per capita real GDP, the square of per capita real GDP, per capita energy use and trade openness. Kivyiro and Arminen (2014) ARDL Model The results support the EKC hypothesis could be an inverted-U shaped. LopezMenendez., Pérez and Moreno, (2014) Panel Model The existence of an EKC could be assumed for these countries. Shahbaz et al. (2014) ARDL Model Economic growth is granger cause of CO2 emissions. Zeb et al. (2014) Johansen Cointegration There is a long run relationship among ERS, CO2, NRD, GDP and Poverty. Heidari, Katircioğlu and Saeidpour (2015) Panel Smooth Transition Regression Nonlinear relationship among CO2 emissions, energy consumption and GDP per capita. On the other hand, several studies do not find any supportive results for the EKC or they presented ambiguous results. Additionally, Aldy (2004) stated that, estimated EKCs are changing for panel of 48 states in the U.S. and carbon dioxide emissions – income relationship could be spurious. Moreover, Bertinelli and Strobl (2005) investigated 122 countries for 1950-1990 periods by semi-parametric regression estimator and found just a little evidence in favor of the EKC. 2. Data and Model Specification 2.1. Data With the aim of evaluation the linkage between income and carbon dioxide emissions, annual data is used on 11 Transition Economies from 1993 to 2010. All variables are collected from World Development Indicators of the World Bank. Selected Transition Economies which are the members of European Union consist of Bulgaria, Croatia, Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Romania, Slovakia and Slovenia. The descriptive statistics of variables are demonstrated in below (table no. 2). The Constraints to the Economic Development in the Former Socialist EU Countries from the Central and Eastern Europe AE Vol. 18 • No. 43 • August 2016 541 Table no. 2: Summary statistics of the variables (logarithmically) lnCOit lnGDPit lnURPOit lnENUSEit Mean 0.80 3.76 6.57 3.42 Median 0.79 3.75 6.48 3.41 Maximum 1.15 4.43 7.38 3.65 Minimum 0.42 3.03 5.96 3.17 Std.Dev. 0.18 0.33 0.42 0.12 Skewness 0.11 -0.20 0.26 0.02 Kurtosis -0.84 -0.73 -0.92 -0.99 The table summarizes descriptive statistics of the variables. Each variable has 198 observations from 1993 to 2010. Carbon dioxide emission (𝑙𝐶𝑂𝑖𝑡) variable is measured in terms of metric tons per capita as a dependent variable in the model, real GDP per capita (𝑙𝐺𝐷𝑃𝑖𝑡) is measured in current US dollars prices, urban population (𝑙𝑈𝑅𝑃𝑂𝑖𝑡) variable represents the number of people who live in urban areas, finally, kilogram of oil equivalent energy use (𝑙𝐸𝑁𝑈𝑆𝐸𝑖𝑡) is used as a transition variable in the PSTR model. 2.2. Model Specification In order to evaluate the relationship between our variables in the panel context, resolving the heterogeneity and time variability problems is vital. PSTR approach is an appropriate one for overcoming these two problems simultaneously. Following González et al. (2005), PSTR model with two extreme regimes and a single transition function can be written as shown below; 𝑦𝑖,𝑡 = 𝜇𝑖+ 𝛽0 ′𝑥𝑖,𝑡 + 𝛽1 ′𝑥𝑖,𝑡𝑔(𝑝𝑖,𝑡; 𝛾,𝑐) + 𝑢𝑖,𝑡 (1) where: i = 1, . . . ,N; t = 1, . . . , T; where: N and T represent the total number of countries and the size of related period, respectively. The dependent variable 𝑦𝑖𝑡 is a scalar, 𝑥𝑖𝑡 is a k-dimensional vector of time-varying exogenous variables, 𝜇𝑖 represents the fixed individual effect, and 𝑢𝑖𝑡 are the errors. Transition function 𝑔(𝑝𝑖𝑡; 𝛾,𝑐) is a continuous function of the transition variable 𝑝𝑖𝑡 and is normalized to be bounded between 0 and 1, and these extreme values are associated with regression coefficients 𝛽0 and 𝛽0+ 𝛽1 (Nieh and Yao, 2013). According to the study of Granger and Teräsvirta (1993) and Gonzalez, Terasvirta and Dijk (2005) the following logistic transition function is defined as follows: 𝑔(𝑝𝑖,𝑡; 𝛾,𝑐) = [1 + 𝑒𝑥𝑝 (−𝛾 ∏(𝑝𝑖,𝑡 − 𝑐𝑗 𝑚 𝑗=1 )]−1with 𝛾 > 0 𝑎𝑛𝑑 𝑐1≤ ⋯ ≤ 𝑐𝑚 (2) AE Testing Environmental Kuznets Curve in the Selected Transition Economies with Panel Smooth Transition Regression Analysis 542 Amfiteatru Economic where: 𝑐 = (𝑐1,…,𝑐𝑚)′ is an m-dimensional vector of the location parameters; 𝛾 is the slope of transition function which determines the smoothness of the transitions (Nieh and Fan, 2012). And considering the two most common cases in practice in order to capture nonlinearity, correspond to m = 1 (logistic) and m = 2 (logistic quadratic) (Coudert, Courharde and Mignon, 2014). For every value of m, when γ→∞, the PSTR becomes a panel transition regression (PTR) model. Conversely, when γ→0, the transition function is constant and the PSTR estimation becomes a panel with fixed effects (Wu, Liu and Pan, 2013). Also the three regime smoothing transition regression can be demonstrated as below; 𝑦𝑖,𝑡 = 𝜇𝑖+ 𝛽0 ′𝑥𝑖,𝑡 + 𝛽1 ′𝑥𝑖,𝑡𝑔1(𝑝𝑖,𝑡; 𝛾1,𝑐1) + 𝛽2 ′𝑥𝑖,𝑡𝑔2(𝑝𝑖,𝑡; 𝛾2,𝑐2) + 𝑢𝑖,𝑡 (3) Similarly, parameters c1 and c2 are the thresholds giving the location of the transition function and parameters 𝛾1and 𝛾2 are the slope parameters of the transition functions respectively (Giovanis, 2012). In addition, it is possible to specify the PSTR model to more than two regimes: 𝑦𝑖,𝑡 = 𝜇𝑖+ 𝛽0 ′𝑥𝑖,𝑡 +∑𝛽𝑗 ′𝑟 𝑗=1 𝑥𝑖,𝑡𝑔𝑗(𝑝𝑖,𝑡 (𝑗); 𝛾𝑗,𝑐𝑗) + 𝑢𝑖,𝑡 (4) where: 𝑟 + 1 is the number of regimes; 𝑔𝑗(𝑝𝑖,𝑡 (𝑗); 𝛾𝑗,𝑐𝑗),𝑗 = 1,…,𝑟, are the transition functions (Béreau, Villavicencio and Mignon, 2010). 3. Empirical Results The estimation of PSTR model consists of a few stages. Firstly, a linearity test is applied to the model and then if linearity is rejected, the most appropriate number of transition function is determined by no-remaining non-linearity test as a second stage. Finally, PSTR model is estimated by nonlinear least squares. As a first and second stage of PSTR model, we apply linearity and no-remaining nonlinearity tests to our model. As a result of these tests’ results (table no. 3), it is easily noticeable that we strongly reject the null hypothesis for linearity test and we cannot reject the null hypothesis of no-remaining non-linearity test and we decided that there is a one transition between two extreme regimes. Table no. 3: Linearity and no-remaining non-linearity tests m=1 m=2 LMF LMW LR LMF LMW LR 𝐻0:𝑟 = 0 vs.𝐻1:𝑟 = 1 94.059 (0.000) 119.849 (0.000) 184.067 (0.000) 55.647 (0.000) 128.396 (0.000) 206.997 (0.000) 𝐻0:𝑟 = 1 vs.𝐻1:𝑟 = 2 1.017 (0.386) 3.337 (0.343) 3.366 (0.339) 0.115 (0.995) 0.776 (0.993) 0.778 (0.993) Note: r, m represents the number of transition functions and the number of location parameters respectively. P-values are in the parentheses. The Constraints to the Economic Development in the Former Socialist EU Countries from the Central and Eastern Europe AE Vol. 18 • No. 43 • August 2016 543 After this stage, in order to determine the number of location parameters, Akaike and Schwarz information criterions are calculated based on Jude (2010). Consequently, the most suitable model consists of one transition function and one location parameter for the related period according to results (table no.4). Table no. 4: Determination of the number of location parameters PSTR Model (m, r) (1, 1) (2, 1) RSS 0.09 0.09 AIC -7.569 -7.558 SIC -7.436 -7.408 Note: r, m represents the number of transition functions and the number of location parameters respectively. The main results of the final PSTR model are reported using a specification of one smooth transition function and one location parameter (table no.5). As a result of our model, the value of slope parameter is equal to 2.8337 which indicates that a smooth and continuous transition function exists between two extreme regimes. The threshold value of our model is equal to -3.714 which its antilog is equal to 5176$. These value seperates two extreme regimes from each other. The first regime is experienced until GDP per capita of 5176$, and second regime is observed after this value. Moreover these results coincide with the literature. Grossman and Krueger (1995) stated that for different pollutants, the turning point is expected to occur until 8000$ GDP per capita. Conversely, our results show that EKC does not exist in the related period for our sample countries. The reason for that GDP per capita has negative effect on CO2 emissions per capita in the first regime while urban population and energy use variables have positive effect. In the second regime, the scenario is totally opposite. Thus, it is concluded that a U-shape EKC is valid for Transition Economies which is the member of European Union in the 1993-2010 period according to PSTR model results. Table no. 5: Estimation results of two-regime PSTR model Dependent variable COit Coefficients Regime 1 Regime 2 𝜷𝑮𝑫𝑷𝒊𝒕 -8.3030* (-445.3) 𝜷𝑼𝑹𝑷𝑶𝒊𝒕 1.6547* (10.8) 𝜷𝑬𝑵𝑼𝑺𝑬𝒊𝒕 0.1969* (1995.7) 𝜷 𝑮𝑫𝑷 𝒊𝒕 8.3030* (446.2) 𝜷 𝑼𝑹𝑷𝑶 𝒊𝒕 -1.6547* (11.7) 𝜷 𝑬𝑵𝑼𝑺𝑬 𝒊𝒕 -0.1969* (-inf.) ɣ 2.8337 c -3.714 Note: (*) 1%, significance levels. The values in parentheses are t-statistics. AE Testing Environmental Kuznets Curve in the Selected Transition Economies with Panel Smooth Transition Regression Analysis 544 Amfiteatru Economic Time series plots of CO2 and GDP per capita variables are demonstrated for 11 transition economies (figure no.1). Vertical axis refers to CO2 emissions per capita and horizontal one stands for GDP per capita. Figure no. 1: CO2 – GDP per capita Note: Vertical line represents threshold value On the one hand, we concluded that the EKC does not exist in the related period for these economies nevertheless, after or around threshold value, the majority of sample countries tend to exhibit an increasing pathway for an inverted U-shape while progressing through higher income levels such as; Bulgaria, Croatia, Estonia, Latvia, Lithuania, Poland and Slovenia. Other economies are following a decreasing trend or ambiguous period after threshold value while their economies are growing. Conclusions In this paper, the PSTR model is employed to investigate the transition dynamics of CO2 emissions per capita and GDP per capita by using eleven Transition Economies data for the period from 1993 to 2010. In order to explain the heterogeneity in time and country between CO2 emissions per capita and other variables, energy use is used as a transition variable in the model. Our PSTR model results demonstrate that an obvious non-linear relationship exists between CO2 emissions per capita and GDP per capita in the selected eleven Transition Economies. The existence of the non-linear relationship is consistent with the literature (Heidari, Katircioğlu and Saeidpour, 2015). However, the EKC is not validated by the estimation results of PSTR model, there is an ordinary U-shaped relationship in the related period. Turning point of this curve is estimated to be 5176$ GDP per capita which is the appropriate value with the literature (Grossman and Krueger, 1995). Individual country analysis showed that particularly Bulgaria, Romania, The Czech Republic, Hungary and Slovakia followed an inverted U-shaped path in the related period. They showed a decreasing trend after the threshold value, therefore, these countries paths could be seen in line with the EKC theory. On the contrary, Croatia, Latvia, Lithuania, Poland and