Using the LMDI method to analyze the change in greenhouse gas emissions in Turkish sectors

In this study, CO2 emissions of the Turkish economy are decomposed for the 1998–2017 period for four sectors; agriculture, forestry and fishery, manufacturing industries and construction, public electricity and heat production, transport, and residential. The analyses are conducted for five fuel types; liquid, solid, gaseous fuels, biomass, and other fuels. In decomposition analysis, Log Mean Divisia Index (LMDI) method is used. The analysis results point out that energy intensity is one of the determining factors behind the change in CO2 emissions, aside from economic activity. The fuel mix component, especially for the manufacturing industries and construction sector, lowers CO2 emissions during the crisis periods when the economic activity declines. Mainly, it is found that changes in total industrial activity and energy intensity are the primary factors determining the changes in CO2 emissions during the study period. Among GDP sectors, manufacturing industries and construction and public electricity and heat production are the two sectors that dominate the change in CO2 emissions. Additionally, the residential and transport sectors’ contributions have gained importance during recent years. Among the manufacturing industries and construction, the non-metallic minerals sector contributes to CO2 emissions, followed by the chemicals sector.

excluding LULUCF, the energy sector has the largest share with 67.8%, followed by the industrial processes and product use with 15.7%, the agriculture with 10.8% and the waste with 5.7% [2]. CO2 emissions per capita were 6.04 t in 2013, while it was 3.96 t for the year 1990 [3]. Per capita electricity consumption and per capita greenhouse gas emission levels in Turkey are about one third of those in other OECD countries. On the other hand, the energy intensity of the Turkish economy is higher than that of other OECD countries nearly by one third [4].
The major source of greenhouse gases is fossil fuel combustion; and energy sector is the main responsible sector in terms of emission of greenhouse gases in Turkey. Considering greenhouse gas emissions (excluding LULUCF) the energy sector's share is 70.2% in 2017. This number increases to 82.2% in terms of CO2 emissions [2]. The highest proportion of CO2 emissions from fuel combustion is from energy industries (public electricity and heat production, petroleum refining and manufacture of solid fuels) both in 1990 (27%) and 2013 (37%) [2].
Considering Turkey's ambitious growth perspective that does not take environmental impact of emissions into account, analyzing the sources of emissions is an important issue for developing alternative policy proposals.
Energy intensity (TPES divided by GDP) is around the OECD average. Turkey's target to reduce energy intensity by 20% by 2023 (from 2011) requires additional efforts to be reached (İEA,2016) (Chapter 4). TPES intensity has decreased to 96% of the 2005 value, whereas total final consumption intensity has decreased more (88%). The decrease in energy intensity is, however, nor steady. It remains dependent on external economic conditions, as the effect of the 2008-09 global financial crisis [5].

Literature review
The decomposition Analysis (IDA) which utilizes index number computations is used to analyze the effects of CO2 emissions on the economy both at the aggregate level and at the sectoral level, specifically on the manufacturing and construction sector. Log Mean Divisia Index (LMDI) method developed by Ang [6] is employed to investigate the leading factors that cause the change in Turki In this study, among available methods the LMDI method developed by Ang [1] is used to figure out the leading factors that contribute to the change in CO2 emissions in the Turkish main four combustion sectors. Researchers and policy-makers use various decomposition methods to quantify the impact of different factors on the change of CO2 emissions. However, there is no consensus among them as to which is the ''best'' decomposition method. The first model was proposed in 1997 and the term "LMDI" was introduced a year later in 1998. The two methods, LMDI-I and LMDI-II, were only formally introduced in 2001. The LMDI decomposition approach comprises two different methods, LMDI-I and LMDI-II. The difference between them lies in the weights formulae used. In each case, several decomposition models have been reported. The popularity of the LMDI approach stems from a number of desirable properties it possesses (Ang, 2004). A practical guide to LMDI-I, based on the change of Canada's industrial energy consumption and CO2 emission was reported in Ang (2005). Ang, has underlined that decomposition analysis is a subject area that has gained in importance in policymaking in the energy field in the last 40 years. He compared various index decomposition analyses, and recommended that the multiplicative and additive LMDI I method, due to its theoretical foundation, adaptability, ease of use and result interpretation and some other desirable properties in the context of decomposition analysis.
With LMDI now firmly established as the preferred approach in IDA, it is timely to conduct stocktaking by providing a precise and definitive documentation of the various LMDI models, including their origin, basic formulae, and key features. This will help potential users to make sensible choices and decisions when implementing it in their studies.sh economy's CO2 emissions.
In a previous study by Akbostancı et al. [7] evaluated GHG from production sector of Turkey that covers fifty seven sectors of Turkey by application the LMDI method and defined that industrial development varieties and intensity of energy are the primary key drivers of alterations in greenhouse gases. Coal is the crucial determinant in fuels, while steel and iron industry sectors are the most polluting industry-dominating CO2 emissions in the production sector of Turkey.
In the earlier studies, the computation of CO2 emissions was based on available energy statistics and emission factors from the Intergovernmental Panel on Climate Change (IPCC). The underlying energy statistics, however, did not include enough sectoral and technological detail to allow for a precise calculation of emissions. Turkey's First National Communicationwas published in 2007 and GHG inventory submissions to the UNFCCC have been annually reported since then. The national GHG inventory reports provide the official emission database for Turkey, consistent over time and consistent with international standards. This accurate and reliable data source for CO2 emissions has been employed in the current analysis.
Lise [8] explores development trajectories of four sectors but decomposes emissions at the aggregate level; Tunç et al. [9], the other hand, differentiate three sectors and decompose at sectoral level but aggregate transportation, housing and services under the service sector. The subsectors, however, differ significantly in terms of their technological infrastructure and have diverging intensity/scale/composition effects on emission growth which cannot be captured in an aggregate analysis. This study is done at a more disaggregate level; five sectors are distinguished and the decomposition analysis is employed on each sector separately.
Gonzalez and Martinez [10] performed a decomposition study to describe the forces that affected the alteration in the greenhouse gas in entire Mexico's industry and its 16 critical industrial sub-divisions throughout the period 1965-2003. They found the impacts of activity, composition, and fuel mix for electricity generation led to increasing CO2 emissions from 1965 to 2003, respectively, they were mitigated by energy intensity and end-use fuel mix. Zhang et al. [11], Emodi et al. [12], and Sumabat et al. [13] used the LMDI decomposition technique to analyses CO2 emissions from power production in China, Nigeria and Philippine. They noticed that the most crucial contribution to rising CO2 emissions from power production is the effect of economic activity. Economic activities and energy intensity are the primary but adverse effects of energy-related carbon emissions. Thus, while financial activities increase energyinduced carbon emissions, energy intensity has a decreasing effect. The Industry and its subsectors are energy-intensive sectors and hence produce significant carbon emissions. Therefore, scholars have done much scientific research on the decomposition analysis of the industry.
Lise [8] also found that relatively rapidly growing economies, the most significant key driver in raising CO2 emissions is economic development. In contrast, the decreasing energy intensifies of the economy is accounted for a small decline in CO2 emissions in Turkey during the period 1980-2003.
Akbostancı et al. [14] have analyzed CO2 emissions of 5 industrials of the Turkish economy between 1990 and 2013. These sectors are; manufacturing, electricity and heat production, transportation, and residential industries. They found that the intensity of energy and wealth generation are the major factors of CO2 emission change. The first two sectors (MC and electricity) are the main crucial sectors that prevail in the alteration in GHG emissions. concluded that the primary increase in greenhouse gases is economic growth. In contrast, energy intensity brings down CO2 in 1980-1990 and 1995-2008 periods, and the industrial system is not a significant factor in minimizing carbon dioxide. Moreover, taking into account the contributions of sectors to CO2 emission changes, it was also found key drivers in CO2 emissions; the industry and services industries as predicted. The industry's contribution has increased in the latest years.
Rüstemoğlu [15] aims to identify and analysis by means of a technique of LMDI decomposition the factors which increase or decrease CO2 emissions for Turkey and Iran from 1990 to 2011.
Economic growth and workforce are the major sources of CO2 emissions both in nations.. Surprisingly, the impact of energy intensity could be the third significant determining factor in Iranian carbon emissions. In contrast, it has a minimal lower impact on Turkey.
Ediger and Havuz [16] utilized the LMDI approach to Turkish sectoral power use evaluation beetween 1980-2000. Although there is a positive correlation among energy demand and the economic output, estimation that the sector specific energy usage differs significantly over the periods 1982, 1988-1989, 1994, and 1998-2000. Such policies would include improvements in the composition of final energy needs, improved material, and energy quality, and the replacement of more suitable products and oils.
Chontanawat et al. [17] also used the LMDI method to decompose the Thai manufacturing sector's source of changes in the level of CO2 emissions and the rate of CO2 emissions for the period 2005-2017. They found that the level of CO2 emissions and the intensity of CO2 emissions increased yearly on average during this timeframe. The effect of the systemic change led to alleviate both the sum of CO2 pollution and the emission intensity. However, the rising energy production of each enterprise increased the sum of CO2 emissions and the rate of CO2 emissions. The study, therefore, argues for strategies to curb the energy density of enterprises in industries to Thailand can profit through growing without having to incur pollutants furthermore.
Trotta [18] has isolated and quantified energy savings generated by improvements in energy Shao et al. [19] have used the LMDI model. Tapio decoupling model and an emission estimation technique to predict to analyze the related decoupling and its impact influence the growth of China's commercial aviation and pollution, along with estimate predicted CO2 pollution. They found that cumulative greenhouse gases change over time on a generally ascending tendency, but there is consistent downward tendency oil consumption per tonnekilometer revenue. The transportation quantities growth impact is contributing effect to increased CO2 pollution among the four main drivers; accompanied by changes in the transport structure and alternate fuel effects. The "pace of energy consumption" aspect plays an essential part in hindering CO2 emissions. They also ended up estimating that China's commercial aviation would be accountable for 0.13 Gt of CO2 emissions by 2020 depending on eight simulations. A factor of 1.6 to 3.9 could raise CO2 emissions between 2020 and 2050.
Shao et al. [20] have used LMDI to factors influencing the usage of natural gas in every province in China. Generally, the primary critical factors of natural gas usage are the financial impact and the impact of the fossil fuel energy sector. The analyzes show that the effect on energy intensity is one of the critical restricting drivers for natural gas consumption; the most significant key drivers for natural gas usage in the net spillover block is the economic impact, followed by spatial expansion. They also found that low energy performance is a significant impediment to the growth of the Xinjiang and Ningxia natural gas sectors. The effect of population density is a significant driver leading to a discrepancy between Beijing and Shanghai in the market competition for natural gas. They concluded that it is particularly significant to prepare reasons for the growth of the natural gas sector to attain the energy transition targets and propose that the Chinese government increasing tax transfer payments, help net spillover growth.
Zhang et al. [21] analyze the influence of four drivers on overall CO2 emission increases, namely the effect of carbon density, the effect of energy intensity, the GDP per capita, and the economic impact of the population. The study reveals that the effect of GDP per capita is the primary factor behind increased CO2 emissions. The carbon concentration and the impact of the population also play a role here. The intensity of carbon has strongly significantly contributed to the decline in CO2 emissions in almost all the nations studied. In C, carbon policies were aimed to decomposition financial development from ecological pressure. Energy policies need to increase the proportion of renewable energy sources in China and the ASEAN countries, raising energy efficiency, and introduce ecological growth as lengthy-term goals in the countries to decouple financial development from ecological repression.
Zhang et al. [22] have used the LMDI to decompose the Chinese manufacturing. The key drivers illustrated various effects on different manufacturing categories (or subsectors) and rates due to the variations in the industrialization stage, growth type, manufacturing investment, and research and development spending. Thus this analysis both discussed the general level of the manufacturing industry and explored the key drivers in each stage of economic development from the perspectives of the market segment and industry category.
Qian et al. [23] has applied the LMDI decomposition approach is utilized to evaluate the key drivers behind industrial SO2 pollution. The research phase is categorized through four phases, Fang et al. [25] has showed that the conomic development has been found to have a substantial impact on energy consumption, while technological development can efficiently mitigate it.
Furthermore, the outcomes of analyzing electricity usage from an industrial and regional context demonstrate that the economic system and intensity of usage have different effects in the eastern, central, and western regions. Jiang et al. [27] have considered that economic development is the key factor affecting the increasing energy consumption in non-housing regions. Besides, the increased energy consumption rate as a second key factor reduces the growth of non-housing electricity usage.
Finally, growth in the population as the third factor has a low impact on rising electricity usage.
Jiang et al. [28] used the LMDI method to describe the main factors of energy consumption.
They found that foreign trade development, economic growth, and inhabitants are due to increased use of embodied energy. Conversely, the modernization and optimization of industrial systems and structures would support to reduce the growth of embodied energy usage.
A specific Investigation is under way on the Turkish high -intensity combustion sector in 4 categories with the use of LMDI in order to achieve emission reductions goals in order to achieve a low-carbon transition in Turkey.

The Log-mean Divisia Index Method
In this study Ang [6] used the LMDI technique to disintegrate the key drivers of CO2 emissions from four fuel types in the Turkey four major ignition industries.
Where C is the energy-related total CO2 emissions (Unit: 103 tons) of Turkish four high-energy intensive industries; i denotes the i-th energy intensive sector; j denotes the j-th type of energy; C_ij is the CO2 emissions contributed by the j-th type of energy consumed by i-th sector (Unit: 103 tons), Q(=∑▒Q_i ) is the total economic activity level, Q_i is the value added of i-th sector, E_i (=∑▒E_ij ) is the energy consumption of i-th sector, and the unit of this variable is TJ (109 kj), E_ij is the consumption of fuel j in sector i.
Where Si (=Qi/Q) is the share of sector and represent the industrial structure, Ii (=Ei/Qi) is the energy intensity of sector i; the fuel-mix variable is given by Mij (=Eij/Ei) and Uij (=Cij/Eij) denotes the CO2 emissions coefficient of energy j consumed by i-th industry.
Let V be an energy-related aggregate. Assume that there are n factors contributing to changes in V over time and each is associated with a quantifiable variable whereby there are n variables, x1; x2;y; xn: Let subscript i be a sub-category of the aggregate for which structural change is to be studied. At the sub-category level the relationship Vi = x1,i x2,i…..xn,i holds. The general index decomposition analysis (IDA) identity is given by The aggregate changes from V 0 =∑ 0 In multiplicative decomposition, we decompose the ratio: In additive decomposition, we decompose the difference: where subscript tot represents the total or overall change and the T and 0 attributed to timespan of T and 0.
In the logarithmic mean Divisia index (LMDI) approach, the general formulae for the effect of the kth factor on the right-hand side of Equations (5) and (6) are respectively: Where L(a,b)=(a-b)/ (ln a-ln b), where both a and b positive numbers and a=b as defined in Ang, 2004 [29]. Therefore, the change of CO2 emissions from four combustion industries of Turkish economy between a base year o and a target year t, denoted by ∆Ctot under the LMDI can be decomposed into the five effects as follows; (i) the changes in the economic activity effect (denoted by ∆Cact ); (ii) the changes in the structure effect (denoted by ∆Cstr ); (iii) the changes in the sectoral energy intensity effect (denoted by ∆Cint); (iv) the changes in the sectoral energy-mix effect (denoted by ∆Cmix); and the changes in the emissions coefficient effect (denoted by ∆Cemf ) in additive form, as shown in Equation (7): ∆C tot = − 0 = ∆C act + ∆C str + ∆C int + ∆C mix + ∆C emf (7) The LMDI formulae can be expressed as:

Analysis of Data
We   .5% for residential use). The use of primary energy sources for electricity production over 1998-2018 is depicted in Fig. 2(b). The average annual growth rate of gas for electricity corresponded to 12.4%. Solid fuels grew by 4.1% per year, and liquid fuels by 3.6% per year over the same period [31].

The economic activity of Turkey
During 1990e2007, real GDP grew at an annual average rate of 4.4%. However, the Turkish economy was hit by three years of contraction. In 1994, structural problems, including high budget deficits, skyrocketing interest and inflation rates plunged the economy into a recession when GDP declined by 5% in real terms. The following turnaround did not last long as the economy was affected in 1998 by the Russian financial crisis and domestic political turmoil.
Conditions worsened in 1999 when general elections were held and the country was hit by a major earthquake -real GDP declined by 4.7% the same year. An economic crisis peaked in 2001, resulting in a sharp contraction of 7.5% as depicted in Fig. 3.  The development of sectoral output in physical terms (joule, ton, ton-km) is depicted in Fig. 5.  As can be seen from the GHG emissions by transport mode are given in Table 3

Analysis of energy-related GHG emissions
In this section, initially the shares of sectors that are covered in the analysis are reviewed. A noticeable development in the study period is the decline in the share of agriculture, forestry and fisheries sector and the increase in the share of transport sector. The share of agriculture, forestry and fisheries has declined from 22% in 1998 to 16% in 2013 while the share of transport sector has increased from 16% to 21% during the same period. Share of manufacturing industries and construction on the other hand was 44% in 1990 and increased to 52% in 2013.
Shares of residential and public electricity and heat production sectors are steady around 8% and 4%, respectively, during the study period.
Manufacturing industry sectors that are covered in this study are the ones for which emissions data are available. Most of these are dirty industries that have relatively higher energy intensities. Overall, total share of these dirty industries in industrial production constitutes 32% as the period average. Among these sectors, food processing, beverages and tobacco has the highest share in manufacturing industries and construction sector's output, which is around 14% as the period average. This sector is followed by iron and steel, petroleum refining and pulp paper and print with average shares, 5.2%, 3% and 2.9%, respectively. Finally, chemicals and non-ferrous metals sectors have respective shares of 1.5% and 1% as the period averages.
An overview of the energy intensities of the sectors helps to evaluate the decomposition results.
The energy intensities of the sectors show that for GDP sectors the highest energy intensity is realized in the public electricity and heat production sector, followed by the residential sector and the lowest energy intensity is observed in the agriculture, forestry and fisheries sector.  The structure effect on industry is the most effective one followed by services and agriculture ( Fig. 8). Industry is positive in 12 years (1981-1986, 1989, 1991, 1993 and 1995-1997) and services are positive in 13 years (1981, 1983-1985, 1987, 1989-1990, 1992-1993, 1996-1997, 1999-2000), whereas agriculture is positive only in three years (1990,1994,1998). The most consistent structural changes in industry occur positively between 1981 and 1986 and negatively between 1998 and 2000. From 1987 to 1997, it fluctuates between positives and negatives, reaching a maximum at 1995. This indicates that, although insignificant in aggregate decomposition, the most significant structural changes in industry occurred during the beginning of the studied period. This is roughly applicable also for services. However, the most significant structural changes in agriculture occurred during 1989-1990. Before these years, they are all negative, but after these years, they are fluctuating between positives and negatives.

Intensity effect
In contrast to the structural effect, the sectoral decomposition of the intensity effect shows thatservices has the biggest positive and negative effects followed by industry and agriculture ( Fig. 9). Services are all negative except for the years 1982, 1988, 1994-1995 and 1999, but agriculture is all positive except for the years 1988,1990,1992,1998 and 2000. On the other hand, industry is positive in 12 years (1982, 1985, 1987, 1989-1991, 1994, 1996-2000), resulting in an increasing pattern in intensity values. In the years 1982, 1988-1989, 1994 and 1999, when the intensity effect is positive in the aggregate decomposition (see Fig. 6), all sectors are positive except for the year 1988.  The aggregate and sectoral decomposition analyses can, therefore, be summarized as follows: (1) the major contributing effect of the primary energy consumption in the Turkish economy from 1980 to 2000 is the production effect, and the structure and intensity effects are insignificant; (2) the Turkish economy can be separated into three relatively stable periods between 1983-1987, 1990-1993 and 1995-1997, each one characterizing positive production and structural effects but negative intensity effects; (3) the most prominent changes in the sectoral energy use pattern in the Turkish economy appears to occur during 1982 and 1988-1989 and between these periods, while the 1994 and 1998-2000 breaks are directly related to the economic crises. Fig. 11. CO2 emissions

Results and discussion
In this study, the changes in GHG emissions of four combustion sectors in Turkey were analyzed by the LMDI method that is the complete decomposition method to identify and analyze the factors for reach GHG emissions reduction goals and determined to achieve a low- As the installation of capacity-efficient production equipment is added, it will be crucial to increase industry efficiency and reduce fuel requirements and greenhouse gas emissions.
According to the results obtained in this study, the proposed policies mentioned above would have active effects to reduce emissions in Turkey.
It should be noted that the energy intensity of the Turkish manufacturing and commercial industries decreased respectively by 39.3 and 11.5 Mt GHG. Unfortunately, there is no progress on the energy intensity of the transport sector and it causes 10 Mt CO2 emissions increasing.
The results of the decomposition analyses confirm the economic programs of the governments, which are traditionally separated into three periods as 1980-1983, 1983-1987 and 1988 onwards. The Stabilization Program, also called the ''24 January Decisions'', which was announced during the beginning of the 1980-1983 period, included a series of economic policy changes aimed at stabilization, liberalization and integration into the worlds. During the 1983-1987 period, the strong inward orientation of Turkeys trade and industrialization changed into an outward orientation coupled with a good integration with the international markets [33]. The reason why this period is also known as ''the golden years'' is related to the outstanding success in export performance, well above the rate of growth of world trade and exports by the newly industrializing countries.
This study shows that the sectoral energy use in Turkey from 1980 to 2000 has undergone significant changes owing to the transformation from an agricultural to an industrial economy enhanced by rapid urbanization between 1983 and 1987. However, the country is still in its early stage of development and energy demand should be increasing faster than national income until the energy intensity of the country reaches a peak.
The major driving force to improve the energy-economy relationship of the Turkish economy appears to be governmental policies. Single party governments usually have more power to implement development policies than coalition governments. These policies should include changes in the structure of final energy demand, increases in the efficiency of materials and energy use and the substitution of materials and fuels that are more efficient.
The energy efficiency in Turkish industry is traditionally low compared to other similar countries [38] but can be increased by some policy implications [14]. Some studies clearly showed that Turkey can reduce emissions considerably without slowing economic growth dramatically [34].
Finally, decomposition analyses of total primary energy consumption into production, structure and intensity effects of the Turkish economy from 1980 to 2000 offer useful tools for a better understanding of overall and sectoral changes. However, further decomposition into secondary and tertiary sectors is definitely needed for detailed investigations.