Environmental technical efficiency in EU member and candidate countries: A parametric hyperbolic distance function approach

Energy 147 (2018) 297e307

Contents lists available at ScienceDirect

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

Environmental technical efficiency in EU member and candidate countries: A parametric hyperbolic distance function approach Yavuz Selman Duman a, *, Adnan Kasman b a b

Department of Economics, Faculty of Economics and Administrative Sciences, Yalova University, 77100 Merkez, Yalova, Turkey Department of Economics, Faculty of Business, Dokuz Eylül University, 35160 Buca, Izmir, Turkey

a r t i c l e i n f o

a b s t r a c t

Article history: Received 6 February 2017 Received in revised form 18 October 2017 Accepted 5 January 2018

This paper investigates environmental technical efficiency for a panel of European Union (EU) member and candidate countries for the period 1990e2011 using parametric hyperbolic distance function. The paper also examines the environmental technical efficiency convergence among the sampled countries. The main results suggest that environmental technical efficiency scores vary among EU member and candidate countries. The EU-15 countries in contrast to the new members and the candidate countries have a greater potential for reducing CO2 emissions while increasing GDP and reducing energy use simultaneously. The results also indicate the existence of environmental technical efficiency convergence among EU member and candidate countries. © 2018 Elsevier Ltd. All rights reserved.

JEL Classification: O10 Q40 Q43 Keywords: Parametric hyperbolic distance function Environmental technical efficiency European Union Stochastic frontier analysis Efficiency convergence

United Nations Climate Change Conference, also known as Conference of the Parties (COP) was held on December 2015 in Paris with the objective to reduce the greenhouse gas (GHG) emissions. The Paris Conference holds great significance in terms of establishing a legally binding and global agreement on climate change with the aim of keeping global warming below 2
C above preindustrial levels. In this regard, it is similar to its predecessor the Kyoto Protocol that required the parties to reduce their emissions 5% by 2012. The conference is expected to be a turning point against climate change towards motivating the business sector in transitioning to a low-carbon economy. Leaders of the world addressed their concerns and reiterated the significance of climate change during the 10 days. The main aim of the “Draft Agreement” is to enhance action, cooperation and support to limit the temperature increase [1]. Stabilizing the emissions rate and transitioning into a low-

carbon economy is crucial to climate change because increasing CO2 emissions, one of the GHG with the highest concentration, affects not only human health and agriculture but also the biodiversity, and the survival of certain species. In a much broader sense, a low-carbon economy is vital for sustainable growth of environmental quality and biodiversity.1 Reducing emissions can be accomplished via clean energy and, energy is rendered clean if it does not increase the atmospheric CO2 concentrations. In that sense, reducing emissions and making energy “clean”, can be done by converting to renewable energy, adopting nuclear energy (focusing on nuclear fusion because it does not generate long lived radioactive waste), biofuels and finally, increasing environmental efficiency. Converting or diversifying energy resources with renewable energy for developed countries might not have the same amount of economic burden as in the case of developing countries, owing to the fact that renewable energy technologies require a great amount of investment. However in some countries with higher awareness

* Corresponding author. E-mail addresses: [email protected] (Y.S. Duman), adnan.kasman@deu. edu.tr (A. Kasman).

1 Sustainable growth refers to promoting growth today without harming the future.

1. Introduction

https://doi.org/10.1016/j.energy.2018.01.037 0360-5442/© 2018 Elsevier Ltd. All rights reserved.

298

Y.S. Duman, A. Kasman / Energy 147 (2018) 297e307

towards climate change, energy markets are regulated in favor of renewable energy sources with subsidies or creating incentives towards renewables by putting more tax on carbon intense sources. Such policies make renewable energy sources a more cost effective substitute. For example, renewable energy for EU is a significant factor in lowering emissions, therefore it has been regulated and promoted through policies and directorates. According to Eurostat statistics the overall share of renewables in gross final energy consumption has been consistently rising. Fig. 1 presents the share of renewable sources in the EU member states for the years 2004 and 2015. Eleven out of the 28 EU members have been able to reach their 2020 targets. Overall, the EU member states have considerably increased their share of renewable sources in their total primary energy supply Apart from the economic cost of renewable energy, there is also the availability factor that plays an important role. Despite being able to handle the economic cost and having the adequate technologies, some countries lack the natural resources to fully appreciate renewable energy. Countries without access to sea are not able to utilize waves as a source of energy, or those with no steady wind are unable to utilize wind turbines. Similar other scarcities of natural occurrences, due to a country's geographical location, prevent them from exploiting these resources. Lowering CO2 emissions, apart from transitioning to renewable energy sources, can be achieved through increasing environmental technical efficiency which is related to productive and technical efficiency in the sense that maximum number of goods and services are produced with given inputs that are used under an optimum allocation. In the environmental efficiency framework, CO2 is considered as the bad (undesirable) output. The aim is to lower the undesirable output without lowering the overall good (desirable) output. This calls for a production function that treats the good and bad output asymmetrically, therefore when we lower the bad output, the desirable will not be effected. This paper follows a hyperbolic form of the distance function because of its flexible definition of production function. Unlike the conventional distance functions which partially define production functions from either input or output

orientations, hyperbolic distance function relaxes this analytical framework and allows for simultaneous output and input characterization. In general, the hyperbolic distance function treats the desirable and undesirable output asymmetrically thus enabling us to lower the undesirable output via preserving or increasing the desirable output. It is valuable to carry out a study that investigates environmental technical efficiency for a sample of countries following the Paris conference, COP21, which introduced a new agreement on climate change. This study provides evidence on whether policies that addresses the increasing energy and environmental efficiency are successful. Hence, the main objective of this paper is to estimate environmental technical efficiency scores for a panel of European Union (EU) member and candidate countries for the period 1990e2011, employing recently developed parametric hyperbolic distance functions by Cuesta and Zofio [2] and Cuesta et al. [3]. In this study, data for capital stock, labor and energy are used to produce one desirable output GDP and one undesirable output CO2 emissions. The other objective of this paper is to examine convergence in environmental efficiency among the sampled countries. To the authors' best knowledge, this is the first paper that investigates environmental technical efficiency and its convergence for the EU member and candidate countries. Efficiency convergence produces valuable information towards assessing the success of energy and environmental regulations, particularly for countries that have common goals and unified policy agendas towards increasing environmental efficiency. By way of preview, our main empirical results suggest that environmental technical efficiency scores vary among EU member and candidate countries. EU-15 countries, in contrast to the new members and the candidate countries have a greater potential for reducing CO2 emissions while increasing GDP and reducing energy use simultaneously. The other main finding of our paper is the existence of environmental technical efficiency convergence among EU member and candidate countries. The rest of the paper is organized as follows. Section 2 reviews the previous literature on energy and environment. Section 3

Fig. 1. Share of energy from renewable sources in EU Member State.

Y.S. Duman, A. Kasman / Energy 147 (2018) 297e307

discusses the main model and econometric methodology. Section 4 discusses the data. Section 5 presents the empirical results of the estimations. The paper's concluding remarks are provided in Section 6. 2. Literature review One of the significant topics handled in the numerous directional distance function (DDF) and stochastic frontier analysis (SFA) studies is energy, and environmental efficiency. Since the use of energy in the production process results in multiple outputs, desirable and undesirable (emissions), DDF enables to measure producer performance with multiple outputs. However, the conventional distance function measure performance radially, that is, it expands all outputs and contracts all inputs at the same rate without distinguishing between desirable and undesirable outputs. What is required to increase environmental efficiency is to be able to contract undesirable outputs and expand desirable outputs. Fare et al. [4,5] introduced the hyperbolic distance function which was the first attempt to relax the restrictive assumptions on input and output, that treats desirable and undesirable outputs asymmetrically. The other study that treats outputs asymmetrically is Chamber et al. [6] that explored the relationship between Shephard's [7] distance function and Luenberger's [8] benefit functions. They introduced an alternative production function in the context of the directional distance function. The DDF introduced in [6] is one of the two major DDF's, following [7], in the study of energy and environment modeling. Chung et al. [9] is the preliminary study to adapt Chambers' DDF to environmental efficiency measurement. Following the preliminary studies and empirical applications on hyperbolic distance functions, studies that dwell on the same topic are, among the others, [10e15]. Mahlberg and Sahoo [15] and Picazo-Tadeo et al. [16] investigated eco-efficiency for EU-14 and for olive growing farms in Spain, respectively. Zhang and Choi [17] and Zhang et al. [18] investigated total factor carbon emission and pure CO2 change for China and Mahlberg and Luptacik [19] measured eco-efficiency and eco-productivity change for Austrian economy. Picazo-Tadeo et al. [20] provided a new approach for measuring intertemporal environmental performance for the EU member countries. The common methodology used in environmental performance employ Data Envelopment Analysis (DEA) which rely on mathematical, non-parametric techniques to measure directional or hyperbolic distance functions. There is no need to specify a mathematical formulation for the production function. Some major limitations of DEA that motivate towards SFA and parametric methodologies are: (i) “efficiency estimates are serially correlated” [21], (ii) inference is possible with bootstrapping, (iii) “except under constant returns to scale the program is non-linear” [3] hence the efficiencies will be under-estimated, (iv) DEA does not account for standard noise therefore on average efficiency estimates are lower than the true efficiencies compared to SFA which takes into account the standard noise. Some of the recent DEA studies that deal with energy and environmental efficiency consider China as a sample country because of its high carbon emissions. Wu et al. [22] and Wang and Wei [23] measure industrial energy efficiency using DEA for China's different provinces. A further detailed review of energy, and environmental efficiency studies using DEA and non-parametric approaches are surveyed by [24] and [25]. The major limitations of DEA motivated researchers to change course and they have recently devoted their attention on the parametric framework in estimating the distance functions. The most vital benefit of the stochastic frontier function under a

299

hyperbolic distance function framework is that unlike the conventional methodologies provides is the treatment of desirable and undesirable outputs asymmetrically. This enables us to expand desirable, and contract undesirable outputs simultaneously. In contrast to DEA, there are limited number of studies employing SFA to estimate distance functions in energy and envi€re et al. [26] estimated a quadratic directional output ronment. Fa distance function using linear programming techniques introduced by [27] to measure technical efficiency, shadow price and elasticity of substitution, and used SO2 (Sulphur dioxide) as the pollutant [28]. compared a variety of shadow pricing models and parameterization methodologies to show that the estimates are sensitive to the chosen functional form [29]. estimated the distance function for a sample of coal-fired electric power plants in the US using the model developed by Ref. [26]and further focused on the substitutability of pollutants, namely SO2 and NOx (nitrogen oxides). Matsushita and Yamane [30] derived the shadow price of CO2, substitutability of pollutants similar to [29] and technical efficiency, employing directional output distance functions. Zhang and Ye [31] extended the parametric hyperbolic distance function introduced by [3] to analyze energy and environmental efficiency for a panel of provinces in China and further analyzed the environmental total factor productivity in two components namely, environmental technical change and environmental efficiency change. Zhang and Choi [32] provided a more recent and thorough review of directional distance functions in the field of energy and environment studies and compared the two different approaches, DEA and parametric methods. 3. Methodology 3.1. Production technology Production is a process of transformation of K input vectors, xi ¼ ðx1i ; …; xKi Þ2<þ K into the following good/desirable M output vecþ tors, yi ¼ ðy1i ; …; yMi Þ2<þ M and bad/undesirable outputs b2
n o þ T ¼ ðx; y; bÞ : x2<þ K ; y2
(1)

€re and Primont which is assumed to satisfy the axioms listed inFa [33]. The axioms state that; (i) any nonnegative input corresponds to at least zero output, (ii) T is a closed set which means that there are technically efficient input and output vectors, (iii) T is bounded, finite inputs can only produce finite outputs, (iv) cðx; yÞ2<þ which is fK;Mg ; if y2TðxÞ and 0 < q  1 then qy2TðxÞ referred to as the weak disposability of outputs which allows the possibility of one output being bad. Bad output is only possible if the good output is also reduced, holding the input constant. Nevertheless this axiom has been often replaced by another; (v) 0 0 0 0 ðy; xÞ2T0ðy ; x Þ2T cðy ; x Þ  ðy; xÞ which is referred to as strong (free) disposability. Following [2] and [3] the hyperbolic þ þ þ distance function DH : Rþ K  RM  RJ /R Uf þ ∞g is defined as;

DH ðx; y; bÞ ¼ inf fq > 0 : ðx; y=q; bqÞ2Tg

(2)

The range of the distance function is, 0 < DH ðx; y; bÞ  1. Equation (2) is called the hyperbolic distance function due to its hyperbolic path generated to the production frontier. The hyperbolic distance function has the feature of treating the desirable and undesirable outputs asymmetrically thus providing an inference on the environmental performance of the production process. The economic interpretation of DH ðx; y; bÞ ¼ 1 is that the estimated

300

Y.S. Duman, A. Kasman / Energy 147 (2018) 297e307

observation is on the boundary of the production frontier and it will not be possible to reduce CO2 or increase GDP at the same time which renders the production efficient. If DH ðx; y; bÞ < 1 then the production is considered inefficient which leaves room for enhancing efficiency for this economy by increasing GDP and reducing CO2 emissions. The hyperbolic distance function satisfies the following axioms listed by [3];

  DH x; my; m1 b ¼ mDH ðx; y; bÞ; m > 0

(3)

DH ðx; ly; bÞ  DH ðx; y; bÞ; l2½0; 1

(4)

DH ðx; y; lbÞ  DH ðx; y; bÞ; l  1

(5)

DH ðlx; y; bÞ  DH ðx; y; bÞ; l  1

(6)

The axioms represent that the HDF satisfies almost homogeneity (3) it is non-decreasing in desirable outputs (4), nonincreasing in undesirable outputs (5) and, non-increasing in inputs (6).

lnDit ¼ a0 þ ak lnKit þ al lnLit þ ae lnEit þ at t þ0:5akk ðlnKit Þ2 þ 0:5all ðlnLit Þ2 þ 0:5aee ðlnEit Þ2 þ 0:5att t 2 þakl lnKit lnLit þ ake lnKit lnEit þ ale lnLit lnEit þakt t  lnKit þ alt t  lnLit þ aet t  lnEit þbGDP lnGDPit þ 0:5bGDP2 ðlnGDPit Þ2  2 þqCO2 lnCO2;it þ 0:5qCO2 lnCO2;it 2 þukGDP lnKit lnGDPit þ ulGDP lnLit lnGDPit þueGDP lnEit lnGDPit þ utGDP t  lnGDPit þjkCO2 lnKit lnCO2;it þ jlCO2 lnLit lnCO2;it þjeCO2 lnEit lnCO2;it þ jtCO2 t  lnCO2;it þ4GDP;CO2 lnGDPit lnCO2;it þ vit i ¼ 1; 2; :::; N; t ¼ 1; 2; :::; T

(9)

Dit is the distance function for country i at time t, besides t is also a time variable that captures the presence of technological change and vit is the random error term distributed, vit Nð0; s2v Þ. Translog hyperbolic distance function must satisfy the homogeneity degrees of 0, 1, 1, 1 for k1,k2,k3 and k4 respectively, that corresponds to the environmental hyperbolic distance function, therefore we depart from (7); Q R X vlnD X vlnD  ¼1 vlny vlnb q r r¼1 q¼1

(10)

Keeping in mind the logarithmic specification vD xp ¼ vlnD ; vD yq ¼ vlnD ; vD br ¼ vlnD ), the relevant partial de(vx vlnxp vyq D vlnyq vbr D vlnbr p D rivatives for the translog function derived from (8) are; 3.2. Translog hyperbolic distance function, THDF We have to specify a translog functional form for the distance function in order to estimate environmental technical efficiency. The translog functional form provides easy calculation and flexibility to impose almost homogeneity restrictions. A traditional function DH ðx; y; bÞ is almost homogenous of degrees k1,k2,k3 and k4 if, DH ðmk1 x; mk2 y; mk3 bÞ ¼ mk4 DH ðx; y; bÞ; cm > 0. From this definition the environmental hyperbolic distance function DH (2) is homogenous of degrees 0, 1, 1, 1. Assuming our distance function DH ðx; y; bÞ is continuously differentiable and satisfies the almost homogeneity condition, it takes the following functional form;

Q P X X vlnD ¼ ap þ apl lnxl þ upq lnyq vlnxp q¼1 l¼1

þ

R X

jpr lnbr ðp ¼ 1; 2; …; PÞ

(11)

r¼1 Q P X X vlnD ¼ bq þ bqm lnym þ upq lnxp vlnyq p¼1 m¼1

þ

R X

fqr lnbr ðq ¼ 1; 2; …; Q Þ

(12)

r¼1

k1

Q P R X X X vD vD vD xp þ k2 yq þ k3 br ¼ k4 D vx vy vb p q r p¼1 r¼1 q¼1

(7)

R P X X vlnD ¼ qp þ qrn lnbn þ jpr lnxp vlnbr n¼1 p¼1

Translog specification of DH ðx; y; bÞ following [3] takes the form;

þ



qrn lnbri lnbni þ

n¼1



0 1 0 Q Q P R R X X X X X @bq þ @qp bqm lnym þ upq lnxp þ fqr lnbr A 

Q X Q R R X 1X 1X bqm lnyqi lnymi þ qr lnbri þ 2 q¼1 m¼1 2 r¼1 r¼1

R X r¼1

Q P X X p¼1 q¼1

jpr lnxpi lnbri þ

Q X R X

upq lnxpi lnyqi þ

(13)

Substituting (12) and (13) to (10) we get;

l¼1

R X

fqr lnyq ðr ¼ 1; 2; …; RÞ

q¼1

Q P X P X 1X lnD ¼ a0 þ ap lnxpi þ apl lnxpi lnxli þ bq lnyqi 2 p¼1 p¼1 q¼1 P X

þ

Q X

q¼1

P X

þ

p¼1

n¼1

fqr lnyqi lnbri ði ¼ 1; 2; …; NÞ

q¼1 r¼1

p¼1

m¼1 R X

qrn lnbn þ

P X p¼1

jpr lnxp þ

r¼1 Q X

1

r¼1

fqr lnyq A

q¼1

¼1 (14)

(8) Panel data specification of the translog HDF, following (8), with inputs as capital, labor and energy and GDP as one desirable output and CO2 emissions as undesirable output will be;

Departing from (14) the restrictions required to ensure almost homogeneity of (0, 1, 1, 1) which is (1 þ P þ Q þ R), we have to modify (8) by choosing desirable output, GDP, as the normalizing variable and set m ¼ 1=y. Hence, the distance function becomes;

Y.S. Duman, A. Kasman / Energy 147 (2018) 297e307

  DH;it ðx; y; bÞ y DH;it x; ; by ¼ y y

(15)

and (8) becomes;

DH;it ln yQ ;it

! ¼ a0 þ

P X

ap lnxpi þ

p¼1

þ

Q 1 X

þ

R X

qr lnb*ri þ

r¼1



Q 1 X q¼1



R X

l¼1

bq lny*qi þ

q¼1

P X P 1X apl lnxpi lnxli 2 p¼1

1 2

Q 1 Q 1 X X

bqm lny*qi lny*mi

q¼1 m¼1

R X R P X 1X qrn lnb*ri lnb*ni þ 2 r¼1 n¼1 p¼1

upq lnxpi lny*qi þ

P X R X

jpr lnxpi lnb*ri þ

p¼1 r¼1

Q 1 X q¼1

fqr lny*qi lnb*ri ði ¼ 1; 2; …; NÞ

r¼1

(16) Following (16) the environmental hyperbolic distance function is as follows;

    lnðGDPit Þ ¼ T Kit ; Lit ; Eit ; GDPit* ; tit ; CO*2;it þ vit  ln DH;it (17) y*qi

Since ¼ yqi =yQi ¼ 1 variable involving y, GDP, is null and CO2;it ¼ CO2;it  GDPit . lnðDH;it Þ is the term representing the inefficiency in the composed error term structure of the stochastic frontier analysis, therefore we can note that uit ¼ lnðDH;it Þ and redefine (17) as;

  lnðGDPit Þ ¼ T Kit ; Lit ; Eit ; tit ; CO*2;it þ vit þ uit

(18)

For uit we follow [34] under a time-varying technical inefficiency framework and uit ¼ ui =½1 þ expðbt þ c2t Þ iid Nð0; s2t Þ where b and c are the parameters that will be estimated. Following a similar methodology but for different almost homogeneity degrees of (1, 1, 1, 1) and a slightly different definition for the distance function [3], introduces the enhanced hyperbolic distance function which enables further reductions in all of the inputs is formulated as follows;

DE ðx; y; bÞ ¼ inf fm > 0 : ðxm; y=m; bmÞ2Tg

(19)

which additional to the axioms (4,5, 6) listed for the HDF, enhanced HDF also satisfies a more inclusive almost homogeneity axiom defined as DH ðm1 x;  my; m1 bÞ ¼ mDH ðx; y; bÞ;   m > 0. The enhanced hyperbolic distance function, which we will be estimating in the following section, in the light of the previous formulation takes the form similar to (18) as;

  lnðGDPit Þ ¼ T Kit* ; L*it ; Eit* ; tit ; CO*2;it þ vit þ uit

(20)

where CO2;it ¼ CO2;it  GDPit , Kit ¼ Kit  GDPit , Lit ¼ Lit  GDPit and Eit ¼ Eit  GDPit . 3.3. Efficiency convergence (s and b convergence) As stated before, another objective of this paper is to examine environmental technical efficiency convergence among the EU

301

member and candidate countries. European Union has several prerequisites and policy adaptations regarding their candidates in their process of becoming a full member ranging from economy and health to infrastructure. One of the major topics new members have to abide by is concerning energy and environmental issues, specifically the Energy Efficiency Directive. The Directive sets binding measures to EU members in reaching the 20% energy efficiency target by 2020. Under the directive, EU members are required to increase energy efficiency at all the stages of energy use from its generation to the production of the final product. Besides, data on energy efficiency are compiled by projects such as the Odyssee-Mure which aims to monitor energy efficiency trends and measures among EU members. The literature on convergence is well established following the preliminary study of Barro and Sala-i-Martin [35] in which they introduced two types of convergence; b (beta)-convergence and s (sigma)-convergence. b-convergence implies that an economy will have a higher growth rate in contrast to another economy, given that the former has a lower initial starting point compared to the latter and, both will converge to the same steady state level in the long run. The test of b -convergence is performed using the following regression equation:

lnFit  lnFit1 ¼ a þ blnFit1 þ εit

(21)

where F denotes the efficiency scores for country i at year t and 0 < b < 1. a and b are the parameters to be estimated and the error term εit has mean zero, finite variance and independent over t and i. The regression equation is estimated using the fixed effects model. b coefficient will be observed after the estimation is undertaken by fixed effects model and, negative b coefficient suggests the existence of convergence in technical efficiency scores. It is common in efficiency studies to test for s-convergence along with b-convergence. The reason as explained by [35] is that, b-convergence does not imply s-convergence because even though the observed variables are converging, shocks to the economy can temporarily increase the dispersion. In these cases s-convergence is the relevant concept because it indicates how the variables behaved in the past and are likely to behave in the future. The test of s-convergence is performed using the following regression equation:

DDit ¼ a þ bDit1 þ εit

(22)

where D ¼ lnFit  ðlnFit Þ and DDit ¼ Dit  Dit1 . Here, Fit follows the same identification as it did in the b-convergence equation (21). ðlnFit Þ denotes the mean of lnFit . Equation (22) is also estimated following the fixed effect methodology as (21) and negative b coefficient suggests the existence of s-convergence.

4. Data To estimate the hyperbolic distance functions, this paper uses the panel of EU member and candidate countries from 1990 to 2011 including one desirable output GDP, one undesirable output CO2 and three inputs capital (K), labor (L) and energy (E). The data are compiled from different sources (see Table 1). GDP data in constant 2005 US dollars is obtained from the World Bank Development Indicators database, capital stock data calculated using perpetual inventory method (PIM) is obtained from the PENN World Tables v8.1. Summary statistics of the data is presented in Table 2.

302

Y.S. Duman, A. Kasman / Energy 147 (2018) 297e307

Table 1 Data sources. Variable

Source

Notes

GDP (constant 2005 World Bank national accounts data, and OECD National Accounts data US$) files.

CO2 Emissions (metric tons per capita) Net Capital Stock (current PPPs in mil. 2005US$) Labor (total)

Energy (kg of oil equivalent per capita)

GDP at purchaser's prices is the sum of gross value added by all resident producers in the economy plus any product taxes and minus any subsidies not included in the value of the products. It is calculated without making deductions for depreciation of fabricated assets or for depletion and degradation of natural resources. Data are in constant 2005 U.S. dollars. Carbon dioxide emissions are those stemming from the burning of fossil Carbon Dioxide Information Analysis Center, Environmental Sciences Division, Oak Ridge National Laboratory, Tennessee, United States. fuels and the manufacture of cement. They include carbon dioxide produced during consumption of solid, liquid, and gas fuels and gas flaring. Penn World Tables v8.1 (Feenstra, Robert C., Robert Inklaar and Marcel P. Capital stocks are estimated based on cumulating and depreciation past Timmer (2015), “The Next Generation of the Penn World Table” investments using the perpetual inventory method (PIM). forthcoming American Economic Review, available for download at www.ggdc.net/pwt) International Labour Organization, using World Bank population Total labor force comprises people ages 15 and older who meet the estimates. International Labour Organization definition of the economically active population: all people who supply labor for the production of goods and services during a specified period. It includes both the employed and the unemployed. While national practices vary in the treatment of such groups as the armed forces and seasonal or part-time workers, in general the labor force includes the armed forces, the unemployed, and first-time job-seekers, but excludes homemakers and other unpaid caregivers and workers in the informal sector. IEA Statistics © OECD/IEA 2012 (http://www.iea.org/stats/index.asp) Energy use refers to use of primary energy before transformation to other end-use fuels, which is equal to indigenous production plus imports and stock changes, minus exports and fuels supplied to ships and aircraft engaged in international transport.

5. Empirical results The enhanced hyperbolic distance function efficiency results and mean efficiency scores for each country are presented in Table 3.2 The first observation for the overall efficiency scores between 1990 and 2011 is that environmental technical efficiency increases, some more than others. The estimated mean environmental technical efficiency for the whole sample in 2011 is 0.682, suggesting that a European country, on average, has the potential to increase real GDP by 46.6% (1/0.682e1 ¼ 0.466) while simultaneously reducing its inputs and CO2 emissions by 31.8%, hence bridging the gap with the highest environmentally efficient country. The results further indicate that Malta, on average, has the lowest efficiency scores (0.388), suggesting that Malta could increase real GDP by 158% (1/0.388e1 ¼1.58) and while simultaneously reducing CO2 emissions, capital, labor and energy input by 61.2% (1e0.388 ¼ 0.612) using current technology. The results also indicate that Romania has the highest average environmental technical efficiency score (0.989), suggesting that it could increase real GDP by only 1.1% (1/0.989e1 ¼ 0.0107) and simultaneously reduce CO2 emissions, capital labor and energy by 1.1% (1e0.989 ¼ 0.0106). Malta and Romania share the most and least percentage change in environmental technical efficiency respectively by 27% and 0.2% over the period 1990e2011. European Environment Agency reported that emissions have risen significantly in the past 25 years in Malta and for the period between 1990 and 2007 they attribute the increase in emissions to energy, transport and waste sectors with small contributions from tourism [36]. The main reason for Malta's low environmental efficiency score is its high urbanization which is not supported with adequate environmental policies. The seasonal tourism and high transportation bring about inefficient use of resources and the lack of environmental policies hurts Malta's overall environmental efficiency. Romania's high environmental efficiency on the other hand

2

In Appendix A we have provided the MLE results for the translog function.

is mainly due to its share of agricultural land and forests with 61% and 28% respectively of the overall land cover [36]. Besides there are strict environmental policies that protect the environmentally significant and sensitive areas such as the Danube Basin and Black Sea coastal regions. Fig. 2 presents the percentage changes in environmental technical efficiency for each country from 1990 to 2011. As observed from Fig. 2, following Malta with the highest efficiency increase from 1990 to 2011 is Luxembourg and Ireland with 24.5% and 22.1%, respectively. One of the strongest economies of the EU, Germany has an efficiency score of 0.576 in 1991 and 0.648 in 2011 which experienced a 12.6% change while France increased efficiency by 13.6% and the UK increased efficiency by 14.2%. For instance, for Turkey the Intended National Determined Contribution (INDC) submitted to the COP21 indicates a greenhouse gas reduction of 21% below business as usual (BAU) in 2030. Our results indicate that Turkey for the period between 1990 and 2011 operated under an average environmental technical efficiency of 0.751, suggesting that Turkey could increase real GDP by 33.20% while simultaneously reducing CO2 by 24.9%. Our results, based on previous performance, indicate that Turkey has the potential to meet its targets set in its INDC reports. On the other hand, overall EU average environmental technical efficiency score is 0.640, indicating that EU has the potential to increase real GDP by 56.60% while simultaneously reducing CO2 by 36%. This falls in line with the EU INDC, which presents a target of 40% domestic greenhouse gas emissions reduction by 2030.3 The interesting point observed from Table 3, in general, is that new members and the candidate countries have higher scores of environmental technical efficiency than EU-15. Main reason behind such a result could be that the EU-15 countries are mostly developed countries that have transitioned into tertiary industries. Among the types of industrial structures that most countries follow

3 The emission reduction targets are retrieved directly from the INDC's submitted to UNFCCC and is available at http://www4.unfccc.int/submissions/indc/.

Y.S. Duman, A. Kasman / Energy 147 (2018) 297e307

303

Table 2 Descriptive statistics for outputs and inputs. Outputs

Inputs

GDP (2005 US$x106)

CO2 Emissions (per capita)

Capital (in mil. 2005 US$)

Labor (x102)

EU-15 Austria Belgium Denmark Finland France Germany Greece Ireland Italy Luxembourg Netherlands Portugal Spain Sweden United Kingdom

286,325 351,823 238,826 177,022 2,002,569 2,757,932 210,731 160,266 1,724,639 31,187 611,755 179,594 995,168 343,041 2,103,694

(40,239) (46,360) (29,274) (33,395) (245,727) (232,120) (38,543) (53,715) (137,719) (7,770) (95,188) (22,693) (189,330) (58,140) (335,784)

8.003 10.648 10.079 11.086 6.016 10.109 8.051 9.765 7.591 22.359 10.634 5.380 6.711 5.811 8.948

(0.538) (0.728) (1.467) (1.025) (0.341) (0.713) (0.619) (0.964) (0.445) (3.065) (0.282) (0.663) (0.879) (0.405) (0.687)

886,841 980,700 553,523 563,444 5,481,056 7,867,179 771,618 293,628 6,253,468 74,192 1,543,284 682,083 3,637,694 659,323 3,948,862

(198,801) (282,787) (139,315) (126,578) (1,741,129) (1,509,075) (196,100) (144,106) (1,524,916) (25,430) (472,240) (251,853) (1,823,363) (114,638) (1,111,599)

3,948,8 4,406,9 2,890,4 2,609,4 27,660,8 40,735,1 4,774,5 1,785,1 23,761,5 193,2 8,068,2 5,203,1 19,105,1 4,664,8 30,009,5

(242,4) (285,9) (45,5) (78,3) (1,412,8) (894,5) (335,7) (321,4) (702,0) (25,3) (711,2) (333,9) (2,800,5) (182,1) (1,239,7)

3690 5424 3615 6331 4084 4096 2418 3240 2887 8391 4671 2181 2810 5554 3610

(313.30) (261.32) (196.56) (510.65) (146.59) (127.04) (239.79) (314.40) (193.75) (745.52) (150.40) (273.28) (319.70) (227.69) (240.12)

New Members Bulgaria Croatia Cyprus Czech Republic Estonia Hungary Latvia Malta Poland Romania Slovakia Slovenia

26,195 41,322 15,626 123,045 11,880 95,599 14,442 5335 267,808 89,953 56,349 33,756

(5,1120) (6,423) (3,655) (22,501) (3,022) (15,917) (3,978) (1,062) (72,735) (17,589) (15,034) (5,741)

6.407 4.780 6.995 11.765 12.325 5.647 3.371 6.204 8.481 4.812 7.280 7.738

(0.795) (0.477) (0.501) (0.779) (1.109) (0.368) (0.408) (0.453) (0.597) (0.745) (0.522) (0.324)

173,152 213,209 60,394 689,256 50,799 445,506 80,389 22,527 1,259,746 681,373 245,345 144,462

(45,943) (66,797) (17,217) (165,528) (23,101) (166,335) (30,419) (9,337) (184,051) (211,309) (56,839) (38,007)

3,605,5 1,986,2 451,2 5,170,9 684,1 4,250,0 1,122,1 156,9 17,552,6 10,677,5 2,606,4 994,8

(225,2) (51,6) (93,0) (50,8) (13,7) (123,2) (39,1) (13,0) (311,6) (835,1) (79,3) (35,5)

2539 1870 2147 4182 3821 2547 1925 1998 2520 1903 3351 3434

(209.26) (165.99) (161.16) (189.11) (287.85) (89.18) (180.63) (135.07) (121.97) (237.22) (102.88) (204.65)

Candidate Countries Albania Macedonia Montenegro Serbia Turkey

6942 5972 2678 29,583 407,470

(2,342) (974) (239) (1,134) (104,594)

1.153 5.242 3.801 6.797 3.350

(0.453) (0.608) (0.522) (0.414) (0.515)

73,155 38,458 22,083 311,851 1,416,545

(27,466) (18,054) (3,248) (22,010) (600,090)

1,330,1 873,0 253,7 3,250,2 22,000,8

(57,9) (55,2) (4,9) (94,1) (1,885,5)

593 1357 1859 2213 1189

(136.78) (77.15) (137.14) (91.29) (174.74)

Overall

450,764

(719,968)

7.860

(3.863)

1,340,665

(2,091,622)

8,538,3

(10,499,7)

3,273

(1,646)

Energy (kgoe, per capita)

Note: The figures in the parentheses are the standard deviations.

through starting with primary heavy industries followed by secondary stage based on specialization and efficiency. The third stage of industrial development which is the tertiary industry, relative to other stages of industrialization is less energy intense which eventually translates into lower emissions. This implies that EU-15 have more potential to increase their environmental technical efficiency in the years to come compared to that of the new members and candidate countries. Another surprising result is that becoming a EU member does not seem to have a significant effect on environmental technical efficiency. Particularly looking at the new members and comparing the pre and post membership trends from before and after 2004, we cannot observe a distinctive change in the trend due to EU accession. However, overall results suggest that there is a catching-up process in environmental technical efficiency among the sampled countries since the countries that experience the highest efficiency increase are the ones with the lowest efficiency scores in 1990. Energys systems or in particular the energy conversion, transmission, and distribution systems in various countries play a significant role in terms of energy efficiency. In a wider perspective, improving energy efficiency under the assumption of “no reboundeffect” might lead to improving the environmental efficiency as well. According to the European Environmental Agency report [37], EU-28 is still dependent on fossil fuels which accounts for %73 of total gross energy consumption and the efficiency of energy

systems is different across member states. Both the availability of energy for final consumption and the source it is generated from differs as well, for example while %96 is available for end-use in Luxembourg, only %45 is available for Estonia. Apart from the availability, the source of electricity for Luxembourg is mostly other states while Estonia domestically uses low-efficiency energy systems which in return increases its loss in transmission, conversion and distribution. The losses faced for Estonia in this regard is higher relative to Luxembourg. Various energy systems and their dynamics in differents countries indicates that it directly effects energy efficiency in end use due to losses faced through conversion, transmission and distribution. However, although the EU is decarbonising its energy systems the current dependency to fossil fuel does not help to improve environmental efficiency. Fig. 3 shows the overview of EU28 energy systems and when the sectors that use energy are compared, transportation and industry have the highest share which are abundantly using fossil fuels. Energys systems should be altered in a way that clean energy should the primary source for sectors that demand the highest amount of energy and release the most CO2 emissions. Because this is the first study to derive the environmental technical efficiency scores and study the relationship between environmental technical efficiency, emissions and GDP for EU candidate and member countries, we do not have the chance to compare our results with the previous findings. However, a

304

Table 3 EHDF efficiency results.

Average EU15 New Members Bulgaria Croatia Cyprus Czech Republic Estonia Hungary Latvia Malta Poland Romania Slovakia Slovenia Average New Members

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

Mean

0.512 0.539 0.452 0.605 0.565 0.544 0.411 0.527 0.375 0.534 0.580 0.585 0.625 0.551

0.516 0.543 0.457 0.609 0.569 0.576 0.548 0.415 0.531 0.380 0.538 0.584 0.589 0.628 0.555

0.520 0.547 0.461 0.613 0.573 0.579 0.552 0.420 0.535 0.384 0.542 0.588 0.593 0.632 0.559

0.524 0.551 0.465 0.616 0.577 0.583 0.556 0.424 0.539 0.389 0.546 0.591 0.596 0.635 0.563

0.528 0.555 0.470 0.620 0.581 0.587 0.559 0.428 0.543 0.393 0.550 0.595 0.600 0.639 0.567

0.532 0.559 0.474 0.623 0.585 0.591 0.563 0.433 0.547 0.398 0.554 0.599 0.604 0.642 0.571

0.536 0.563 0.478 0.627 0.589 0.595 0.567 0.437 0.551 0.402 0.558 0.603 0.607 0.646 0.575

0.540 0.567 0.482 0.631 0.592 0.598 0.571 0.441 0.555 0.406 0.562 0.606 0.611 0.649 0.579

0.544 0.571 0.487 0.634 0.596 0.602 0.575 0.446 0.559 0.411 0.565 0.610 0.615 0.652 0.583

0.548 0.575 0.491 0.638 0.600 0.606 0.579 0.450 0.563 0.415 0.569 0.614 0.618 0.656 0.586

0.552 0.578 0.495 0.641 0.603 0.609 0.583 0.455 0.567 0.420 0.573 0.617 0.622 0.659 0.590

0.556 0.582 0.499 0.644 0.607 0.613 0.586 0.459 0.571 0.424 0.577 0.621 0.625 0.662 0.594

0.560 0.586 0.504 0.648 0.611 0.617 0.590 0.463 0.574 0.428 0.581 0.624 0.629 0.666 0.598

0.564 0.590 0.508 0.651 0.614 0.620 0.594 0.467 0.578 0.433 0.585 0.628 0.633 0.669 0.601

0.568 0.594 0.512 0.655 0.618 0.624 0.598 0.472 0.582 0.437 0.588 0.631 0.636 0.672 0.605

0.572 0.597 0.516 0.658 0.622 0.627 0.601 0.476 0.586 0.442 0.592 0.635 0.639 0.675 0.609

0.576 0.601 0.520 0.661 0.625 0.631 0.605 0.480 0.590 0.446 0.596 0.638 0.643 0.679 0.612

0.579 0.605 0.524 0.665 0.629 0.634 0.609 0.485 0.593 0.450 0.600 0.642 0.646 0.682 0.616

0.583 0.608 0.528 0.668 0.632 0.638 0.612 0.489 0.597 0.455 0.603 0.645 0.650 0.685 0.620

0.587 0.612 0.532 0.671 0.636 0.641 0.616 0.493 0.601 0.459 0.607 0.649 0.653 0.688 0.623

0.591 0.616 0.536 0.674 0.639 0.645 0.620 0.497 0.605 0.463 0.611 0.652 0.656 0.691 0.627

0.595 0.619 0.540 0.677 0.643 0.648 0.623 0.501 0.608 0.468 0.614 0.655 0.660 0.694 0.630

0.554 0.580 0.497 0.642 0.605 0.613 0.584 0.456 0.568 0.422 0.575 0.619 0.623 0.660 0.592

0.529

0.536

0.540

0.544

0.548

0.552

0.556

0.559

0.563

0.567

0.571

0.575

0.579

0.582

0.586

0.590

0.594

0.597

0.601

0.605

0.608

0.612

0.981

0.981

0.982

0.982

0.982

0.415

0.419

0.423 0.810

0.428 0.812

0.432 0.814

0.703

0.706

0.709

0.712

0.341 0.954 0.988

0.345 0.955 0.988

0.350 0.955 0.988 0.753

0.354 0.956 0.988 0.755

0.359 0.956 0.989 0.758

0.982 0.609 0.437 0.816 0.702 0.715 0.677 0.363 0.957 0.989 0.760 0.598

0.982 0.613 0.441 0.818 0.705 0.718 0.680 0.368 0.957 0.989 0.763 0.602

0.983 0.617 0.445 0.820 0.708 0.721 0.683 0.372 0.958 0.989 0.765 0.606

0.983 0.620 0.450 0.822 0.710 0.724 0.686 0.377 0.958 0.989 0.768 0.609

0.983 0.624 0.454 0.824 0.713 0.726 0.689 0.381 0.959 0.989 0.770 0.613

0.983 0.627 0.458 0.826 0.716 0.729 0.692 0.385 0.959 0.989 0.773 0.617

0.984 0.631 0.463 0.828 0.719 0.732 0.695 0.390 0.960 0.989 0.775 0.620

0.984 0.634 0.467 0.830 0.722 0.735 0.698 0.394 0.960 0.990 0.777 0.624

0.984 0.638 0.471 0.832 0.725 0.737 0.701 0.399 0.961 0.990 0.780 0.627

0.984 0.641 0.476 0.834 0.728 0.740 0.704 0.403 0.961 0.990 0.782 0.631

0.984 0.645 0.480 0.835 0.730 0.743 0.707 0.408 0.962 0.990 0.784 0.634

0.984 0.648 0.484 0.837 0.733 0.745 0.710 0.412 0.962 0.990 0.787 0.638

0.985 0.651 0.488 0.839 0.736 0.748 0.713 0.416 0.962 0.990 0.789 0.641

0.985 0.655 0.493 0.841 0.739 0.751 0.716 0.421 0.963 0.990 0.791 0.645

0.985 0.658 0.497 0.843 0.741 0.753 0.719 0.425 0.963 0.990 0.793 0.648

0.985 0.661 0.501 0.844 0.744 0.756 0.722 0.430 0.964 0.991 0.796 0.651

0.985 0.665 0.505 0.846 0.747 0.758 0.725 0.434 0.964 0.991 0.798 0.655

0.736

0.732

0.746

0.748

0.750

0.717

0.720

0.722

0.725

0.727

0.730

0.732

0.735

0.737

0.739

0.742

0.744

0.747

0.749

0.751

0.754

0.756

0.591 0.643 0.589 0.929 0.768

0.595 0.647 0.592 0.930 0.770

0.599 0.650 0.596 0.931 0.772

0.602 0.653 0.600 0.931 0.775

0.606 0.657 0.604 0.932 0.777

0.704

0.707

0.710

0.712

0.715

Candidate Countries Albania Macedonia Montenegro Serbia Turkey

0.524

0.528

0.532 0.589

0.537 0.593

0.541 0.597

0.545 0.600

0.549 0.604

0.553 0.608

0.557 0.611

0.560 0.615

0.564 0.619

0.568 0.622

0.572 0.626

0.576 0.629

0.580 0.633

0.584 0.636 0.581

0.723

0.725

0.728

0.731

0.734

0.737

0.739

0.742

0.745

0.747

0.750

0.752

0.755

0.758

0.760

0.763

0.587 0.640 0.585 0.928 0.765

Average Candidates

0.623

0.627

0.617

0.620

0.624

0.627

0.631

0.634

0.638

0.641

0.644

0.648

0.651

0.654

0.658

0.641

0.701

0.983 0.637 0.460 0.829 0.725 0.732 0.701 0.388 0.959 0.989 0.776 0.627

0.566 0.624 0.592 0.930 0.751

Y.S. Duman, A. Kasman / Energy 147 (2018) 297e307

EU-15 Austria Belgium Denmark Finland France Germany Greece Ireland Italy Luxembourg Netherlands Portugal Spain Sweden United Kingdom

Y.S. Duman, A. Kasman / Energy 147 (2018) 297e307

305

Table 4 show that the coefficients of lnFit1 and Dit1 are negative and statistically significant at conventional levels, indicating that both b-convergence and s-convergence exist among the EU and candidate countries. The existence of b-convergence implies that although the efficiency levels of the sample countries differs in terms of their initial levels, based on the comparative differences in their efficiency growth rates, they will converge to a steady state. The existence of b-convergence is a necessary condition to test for s-convergence. From Table 4, the existence of s-convergence implies that the dispersion of efficiency scores among the sampled countries decreases overtime.

6. Conclusion

Fig. 2. % change in environmental technical efficiency from 1990 to 2011.

research fellow at the World Resource Institute published a recent article [38], which utilizes data from BP Statistical Review of World Energy 2015 and World Bank Development Indicators to direct attention to 21 countries that have reduced annual GHG emissions while experiencing GDP growth. Through analyzing the GHG emissions and GDP statistics of different countries [38], draws attention to the fact that emission reduction is statistically possible without compromising GDP growth. The article does not mention energy or environmental technical efficiency and ties the simultaneous decrease in emissions with GDP growth to environmental policies, switching to renewable energy, carbon taxes, reducing industrial sector share of their economies and transitioning from high-carbon production to low-carbon production. In that sense, our findings not only support but also present a different approach for providing evidence on simultaneous GDP growth and emissions reduction possible. One of the primary objectives of this paper is to check for efficiency convergence. The results of the b-convergence and s-convergence are presented in Table 4. The environmental technical efficiency scores are used in the convergence analysis. We have tested the whole sample for both the b-convergence and s-convergence over the period 1990e2011. The results presented in

This paper introduces a new perspective on the analysis of environmental technical efficiency by employing parametric hyperbolic distance functions and further looking into efficiency convergence among the EU member and candidate countries for the period 1990e2011. Departing from a recent study of [3] we have estimated the environmental technical efficiency using a translog production function which allowed for increases in desirable output GDP, and decreases in the inputs capital, labor and energy as well as the undesirable output CO2 emissions. We have found that the EU-15 countries in contrast to the new members and the candidate countries have a greater potential for reducing CO2 emissions while increasing GDP and reducing energy use simultaneously. Although the environmental technical efficiency scores vary among the EU members and candidates, countries with higher GDP levels tend to have higher potential in reducing carbon emissions. One common feature among the EU members and candidates is that for the period between 1990 and 2011 environmental technical efficiency follows an increasing trend which led us to check for any convergence exists or not in terms of efficiency. Departing from the traditional output production function, this paper, using hyperbolic distance function, has enabled the policy makers to take into consideration the environmental aspect of the production process by allowing the reduction of undesirable outputs and inputs while simultaneously increase production. However, one of the most important contribution of this study is that it provides evidence on the existence of environmental technical

Fig. 3. Overview of the EU28 energy systems.

306

Y.S. Duman, A. Kasman / Energy 147 (2018) 297e307

Table 4 Regression results for environmental technical efficiency convergence.

b- convergence a b (lnFit1 ) s- convergence a b (Dit1 )

0.0002* (0.0007) 0.0119* (0.0001) 0.0020* (0.0001) 0.161** (0.0284)

Note: The figures in the parentheses are the standard-errors. *, **, and *** denote significance level at 1%, 5% and 10%, respectively.

efficiency convergence among the EU member and candidate countries. The results of this study have at least two policy implications. First, based on the estimated environmental technical efficiency scores, new policies can be designed specific to each member and candidate to improve efficiency and production. Transitioning from high carbon intense production to a low carbon economy, increasing the share of renewable energy in the energy profile and improving production technology towards higher energy efficiency are some of the policies that can be adopted towards increasing environmental technical efficiency. Our results could provide insights towards matching policies to the right country that needs it and fits best. Furthermore, our results suggest that policy makers should follow the Odyssee-Mure project, which provides energy consumption and efficiency trends as well as an evaluation of energy efficiency policy measures. The second policy implication is based on the existence of convergence which is perhaps the most significant finding and contribution of this study. EU members are bound to a common and consistent set of environmental policies and candidates are required to satisfy these policies during their accession process. Therefore, the existence of environmental technical efficiency convergence among EU member and candidates provides a good example of collaborative action that facilitates coordination and better dialog among countries. This form of collective action help raise efficiency in the countries that lag and catch up with a standard that benefits

all parties. Eco-design Directive is one of the common set of EUwide rules that target the improvement of environmental performance while reducing energy and resource consumption. The results of this paper also support and prove that such Directives are economically sound and applicable with real life results. A similar agreement (The Paris Agreement) was ratified on April 22, 2016 by 174 countries including the EU member and candidates, that will help reduce carbon emissions provided that countries fulfill their goals stated in their INDC's. The common feature of these Directives and Agreements is that besides being a binding roadmap for policymakers, they also provide financial or technological support to those countries that lack the sufficient resources and help them converge to a common and higher efficiency and emissions standards. There are several ways to build on this study by further research. First, to expand the sample countries to a wider range for example using the UNFCCC members and observing convergence. Second, to include different greenhouse gasses as undesirable outputs, since there are many air pollutants besides CO2, such as carbon monoxide, nitrogen oxide or relatively a more dangerous output SO2 emissions. Including more pollutants may give a wider perspective on environmental technical efficiency for our future study. Appendix The maximum likelihood estimates of the enhanced hyperbolic distance function are summarized in Table A1. The estimated parameters enable us to determine how the desirable and undesirable input and outputs effect the distance function. Although the results are mostly significant, contrary to our expectations qCO2 has a negative sign which contradicts the notion that any increase in CO2 would increase the value of the distance function. The same inference can be made for ak,evalues, since as expected they have a negative sign, with the exception of al, which means that any increase in the capital and energy will also increase the distance from the frontier.

Table A.1 Maximum Likelihood Estimates of the distance function Parameter

Enhanced Hyperbolic Distance Function Estimated Value

a0 ak al ae at akk all aee att akl ake ale akt alt aet qCO2 qCO22 jkCO2 jlCO2 jeCO2 jtCO2 s2 g m h

t-statistic

8.6413 1.6480 0.8365 0.4942 0.0848 0.1846 0.1254 0.1710 0.0001 0.1497 0.0139 0.0262 0.0107 0.0074 0.0039 0.1240 0.1536

(3.4659) (0.2940) (0.2860) (0.6317) (0.0164) (0.0249) (0.0200) (0.0969) (0.0001) (0.0207) (0.0303) (0.0332) (0.0013) (0.0013) (0.0013) (0.5343) (0.0723)

2.4932 5.6055 2.9242 0.7823 5.1623 7.4127 6.2723 1.7643 0.8418 7.2343 0.4592 0.7886 8.0854 5.8711 3.0045 0.2320 2.1239

0.0744 0.0975 0.1581 0.0026 0.0488

(0.0259) (0.0279) (0.0813) (0.0012) (0.0166)

2.8693 3.4914 1.9440 2.1430 2.9422

0.9874 0.3830 0.0121

(0.0044) (0.0507) (0.0015)

224.1644 7.5524 7.9844

Note: The figures in the parentheses are the standard-errors. s2 ¼ s2v þ s2u and g ¼ s2u =ðs2u þ s2v Þ.

Y.S. Duman, A. Kasman / Energy 147 (2018) 297e307

References [1] UNFCCC. Draft Paris outcome. Proposal by the President. Paris. 2015. [2] Cuesta RA, Zofío JL. Hyperbolic efficiency and parametric distance functions: with application to Spanish savings banks. J Prod Anal 2005;24:31e48. [3] Cuesta RA, Lovell CAK, Zofío JL. Environmental efficiency measurement with translog distance functions: a parametric approach. Ecol Econ 2009;68: 2232e42. [4] F€ are R, Grosskopf S, Lovell CAK, Pasurka C. Multilateral productivity comparisons when some outputs are undesirable: a nonparametric approach. Rev Econ Stat 1989;71:90e8. [5] F€ are R, Grosskopf S, Lovell CAK. The measurement of efficiency of production. Boston. MA: Kluwer-Nijhoff Publishing; 1985. [6] Chambers RG, Chung Y, F€ are R. Benefit and distance functions. J Econ Theor 1996;70:407e19. [7] Shephard RW. Cost and production functions. Princeton University Press; 1953. [8] Luenberger DG. Benefit functions and duality. J Math Econ 1992;21:461e81. €re R, Grosskopf S. Productivity and undesirable outputs: a [9] Chung YH, Fa directional distance function approach. J Environ Manag 1997;51:229e40. €re R, Grosskopf S, Pasurka CA. Accounting for air pollution emissions in [10] Fa measures of state manufacturing productivity growth. J Reg Sci 2001;41: 381e409. [11] Weber WL, Domazlicky B. Productivity growth and pollution in state manufacturing. Rev Econ Stat 2001;83:195e9. [12] Arocena P, Waddams Price C. Generating efficiency: economic and environmental regulation of public and private electricity generators in Spain. Int J Ind Organ 2002;20:41e69. [13] Boyd GA, Tolley G, Pang J. Plant level productivity, efficiency, and environmental performance of the container glass. Environ Resour Econ 2002;23: 29e43. €rük BK, Zaim O. Productivity growth in OECD countries: a comparison with [14] Yo Malmquist indices. J Comp Econ 2005;33:401e20. [15] Mahlberg B, Sahoo BK. Radial and non-radial decompositions of Luenberger productivity indicator with an illustrative application. Int J Prod Econ 2011;131:721e6. https://doi.org/10.1016/j.ijpe.2011.02.021. n-Esteve M, Go  mez-Limo  n JA. Assessing eco-efficiency [16] Picazo-Tadeo AJ, Beltra with directional distance functions. Eur J Oper Res 2012;220:798e809. [17] Zhang N, Choi Y. A comparative study of dynamic changes in CO2 emission performance of fossil fuel power plants in China and Korea. Energy Pol 2013;62:324e32. [18] Zhang N, Zhou P, Choi Y. Energy efficiency, CO2 emission performance and technology gaps in fossil fuel electricity generation in Korea: a meta-frontier non-radial directional distance function analysis. Energy Pol 2013;56:653e62. https://doi.org/10.1016/j.enpol.2013.01.033. [19] Mahlberg B, Luptacik M. Eco-efficiency and eco-productivity change over time in a multisectoral economic system. Eur J Oper Res 2014;234:885e97. nez J, Beltr [20] Picazo-Tadeo AJ, Castillo-Gime an-Esteve M. An intertemporal

[21] [22]

[23] [24] [25] [26] [27] [28]

[29] [30] [31]

[32]

[33] [34] [35] [36] [37]

[38]

307

approach to measuring environmental performance with directional distance functions: greenhouse gas emissions in the European Union. Ecol Econ 2014;100:173e82. Simar L, Wilson PW. Estimation and inference in two-stage, semi-parametric models of production processes. J Econom 2007;136:31e64. Wu F, Fan LW, Zhou P, Zhou DQ. Industrial energy efficiency with CO2 emissions in China: a nonparametric analysis. Energy Pol 2012;49:164e72. https://doi.org/10.1016/j.enpol.2012.05.035. Wang K, Wei YM. China's regional industrial energy efficiency and carbon emissions abatement costs. Appl Energy 2014;130:617e31. Ang BW, Zhang FQ. A survey of index decomposition analysis in energy and environmental studies. Energy 2000;25:1149e76. Zhou P, Ang BW, Poh KL. A survey of data envelopment analysis in energy and environmental studies. Eur J Oper Res 2008;189:1e18. €re R, Grosskopf S, Noh D-W, Weber W. Characteristics of a polluting techFa nology: theory and practice. J Econom 2005;126:469e92. Aigner DJ, Chu SF. On estimating the industry production function. Am Econ Rev 1968;58:826e39. Vardanyan M, Noh D-W. Approximating pollution abatement costs via alternative specifications of a multi-output production technology: a case of the US electric utility industry. J Environ Manag 2006;80:177e90. €re R, Grosskopf S, Pasurka CA, Weber WL. Substitutability among undesirFa able outputs. Appl Econ 2012;44:39e47. Matsushita K, Yamane F. Pollution from the electric power sector in Japan and efficient pollution reduction. Energy Econ 2012;34:1124e30. Zhang Z, Ye J. Decomposition of environmental total factor productivity growth using hyperbolic distance functions: a panel data analysis for China. Energy Econ 2015;47:87e97. Zhang N, Choi Y. A note on the evolution of directional distance function and its development in energy and environmental studies 1997e2013. Renew Sustain Energy Rev 2014;33:50e9. €re R, Primont D. Multi-output production and duality: theory and applicaFa tions. 1995. Kumbhakar SC. Production frontiers, panel data, and time-varying technical inefficiency. J Econom 1990;46:201e11. Barro RJ, Sala-i-Martin X. Convergence across states and regions. Brookings Pap Econ Activ 1991;22:107e82. European Environment Agency (EEA). The European environment: state and outlook 2015: synthesis. Eur Environ 2015:42. https://doi.org/10.2800/45773. European Environment Agency (EEA). Overview of the European energy system (ENER 036). Indic assess. 2017. https://www.eea.europa.eu/data-andmaps/indicators/overview-of-theeuropean-energy-system-3/assessment. [Accessed 18 October 2017]. Aden N. The roads to decoupling: 21 countries are reducing carbon emissions while growing GDP. World Resour Inst; 2016. http://www.wri.org/blog/2016/ 04/roads-decoupling-21-countries-are-reducing-carbon-emissions-whilegrowing-gdp. [Accessed 5 April 2016].