DECOMPOSITION OF ENERGY CONSUMPTION AND DECOUPLING ANALYSIS IN THE INDONESIAN INDUSTRY: AN ANALYSIS OF GREEN INDUSTRY

This study aims to identify ways that efficiently reduce the energy consumption of the industrial sector. We use the logarithmic mean divisia index method to measure the impact of the various driving forces of energy consumption during the period 2010-2014. We then apply the index decoupling to analyze the correlation between energy consumption and industrial growth. The findings indicate that industrial growth is a major driver of energy consumption, while reductions in energy intensity and industrial structure play an important role in limiting energy consumption. In addition, energy consumption and energy intensity follow different patterns in each sub-sector; we therefore conclude that the application of different sub-sector policies is preferred over global policies. Globally, decoupling has not been identified during the period 2010-2014, however, decoupling occurs for more detailed year periods.


INTRODUCTION
Energy consumption in Indonesia continues to increase, in the period 2000 and 2015 increased by 44% (MEMR, 2016). The increase in energy consumption is predominantly derived from industrial and transportation sectors, with a total share of over 75% (MEMR, 2016). Industry sector contribution is around 35-40%. The cause of this increase is a sustained expansion of economic activity by increasing gross domestic product (GDP). Fossil fuels of oil, gas and coal still dominate and represent 95% of Indonesia's primary energy consumption total (MEMR, 2016). The high energy consumption of fossil fuels in this energy mix in the industrial and transport sectors causes high CO 2 emissions in the sector (Ministry of Industry, 2012). Even in the industrial sector, fossil fuel energy consumption increased by more than 100%, accompanied by an increase in CO 2 emissions between 2010 and 2014 (Statistics Indonesia, 2016). This drastic increase in fossil fuel energy consumption will be very worrying in terms of energy availability and environmental sustainability, especially climate change. Industrialization will result in such proportions will continue to increase significantly in the future. On the other hand, Indonesia's energy resources reserves are worried and feared that they can not meet domestic demand for energy over the long term (MEMR, 2018).
Energy efficiency becomes a key element in achieving safe energy supplies as the country faces an energy crisis, rising energy costs and rising concerns about climate change (Duran et al., 2015). Energy efficiency can be considered as an additional energy source resulting from a reduction in energy consumption while maintaining a given level of output. This reduction in energy consumption is achieved by improving technology or process efficiency. In the energy economy literature, changes in energy intensity are often used as a proxy of changes in energy efficiency (Choi and Ang, 2010) while energy intensity is defined as the amount of energy consumed per unit of output. Increased energy efficiency enables industries to reduce energy intensity to achieve a certain level of output at lower costs and less pollution (Reddy and Ray, 2011;Duran et al., 2015). The International Energy Agency (IEA) has stated that improving the energy efficiency of the industrial sector is vital in maintaining a sustainable energy supply. To achieve this, the Indonesian government promotes decoupling between energy consumption growth and economic growth (MEMR, 2016). The industrial sector is expected to achieve the highest levels of energy saving and CO 2 mitigation among all other priority areas (buildings, equipment, lighting and transportation) if appropriate measures are implemented (IEA, 2011).
Decoupling is an important concept to connect between economic variables, energy variables and environmental impacts (Enevoldsen et al., 2007). Initially, the term decoupling was used to describe the relationship between economic activity and environmental degradation (Freitas and Kaneko, 2011). The growth of economic activity and the growth of environmental pressure (e.g., CO 2 emissions) over a period of time has become a feature of decoupling (Freitas and Kaneko, 2011). Specifically, the definition of decoupling refers to the relative growth rate of an environmental pressure and an economic variable that is easily linked. According to the Organization for Economic Cooperation and Development (OECD), decoupling is the relationship between "economic goods" and "environmental crime" and began presenting decoupling as an indicator of sustainable development in 2002.
There are several different definitions of decoupling and so are the measurement methods. At present there is no consensus about the relative strengths and weaknesses of each (Zhong et al., 2010). OECD (2002) developed the concept of measurement to distinguish between absolute decoupling and relative decoupling. In the case of absolute decoupling, decoupling occurs when the growth variable of economic activity increases, the environment variable shows a stable trend or the trend decreases over time. However, relative decoupling occurs when the growth rate of environmental variables shows a positive trend but at a lower level than the growth of economic activity. Climent and Pardo (2007) then use this concept to examine the correlation of GDP and energy consumption in Spain. Diakoulaki and Mandaraka (2007) adopted it to evaluate progress in reducing CO 2 emissions from industrial growth in 14 European Union countries. Freitas and Kaneko (2011) used it to examine CO 2 emissions and the growth of economic activity in Brazil during [2004][2005][2006][2007][2008][2009]. According to Freitas and Kaneko (2011), this decoupling analysis method (specifically index) is more effective when combined with other analytical methods such as decomposition analysis methods.
This study aims to help identify the focus of energy efficiency policy in the future to achieve a reduction in cost-effective energy consumption and CO 2 emissions mitigation in the Indonesian industrial sector. A more specific objective of this study is to analyze the factors that explain changes in energy consumption in various Indonesian industrial sub-sectors. This information can help policy makers to determine which sub-sectors and which firms should prioritize to focus policy, reduce energy consumption, minimize costs and mitigate CO 2 emissions. The international energy agency recommends the use of the index decomposition analysis (IDA) to isolate the impact of changes in energy intensity from other factors that affect energy consumption, given that energy intensity can be affected by government through energy policies (Duran et al., 2015). This information can help gauge the real impact of energy policies and focus these policies on strategic sectors, based on their energy reduction potential (IEA, 2014).
Index decomposition analysis especially logarithmic mean Divisia index method (Ang, 2005) was used in this study to identify and analyze the relative impact of various factors on energy consumption, especially changes in energy intensity in various Indonesian industrial sub-sectors. The benefits of the logarithmic mean divisia index (LMDI) method are simple to use, complete decomposition possible without residual, can accept additive decomposition and multiplication, and can be applied for short time series (Su and Ang, 2012). Decoupling concept analysis (decoupling) is also used to identify industrial sub-sectors that have efficiency in using energy.
The main contributions of this study to the existing literature are: (1) Expansion of analysis for inclusion of characteristics of the company such as sub-sector and energy intensity in energy policy design; (2) applying the concept of decoupling index in analyzing energy efficiency, that is correlation of energy consumption and economic activity. This information helps policy makers in better subsectoral and cost-effective policy design. Previous studies using the IDA only consider different sub-sectors based on the classification of industrial activity (Ma and Stern, 2008;Xu and Ang, 2014). To our knowledge, have considered the decoupling index in energy efficiency analysis.
The organization of this paper is as follows. Section 1 introduction. Section 2 methodolgy, presenting the IDA framework, LMDI approach and decoupling analysis. Section 3 describes the data used. Section 4 shows the LMDI decomposition and decoupling analysis of Indonesian industrial energy consumption based on industry sub-sector. Finally, Section 5 concludes the paper and presents policy recommendations.

Energy Decomposition
We used the LMDI method in this study. LMDI is the most widely adopted and preferred method in the context of index decomposition analysis, IDA (Shao et al., 2011) because of its ease of formulation. Some of the practical advantages of the LMDI method are: Easy to use and applicable for short time series (Su and Ang, 2012;Ang and Liu, 2001), perfectly decomposed, no residuals (Tunç et al., 2009), zero values in the set of datasets can be replaced by a small positive constant value , consistent in aggregation, this means sub-group results can be added to aggregate results (Ang and Liu, 2001;, strong theoretical basis, adaptability, which is clear between additive and multiplicative versions, the results do not depend on the base year, and use a rather simple formula with direct interpretation of the results (Ang, 2004). Ang (2005) provides a practical guide to using this method.
IDA estimates the effects of three different effects on changes in energy consumption. These effects are: 1. Activity effects. This effect considers changes in the scale of economic activity or in the overall economic output level, assuming that this increase in output levels involves an increase in energy consumption. 2. Economic structure effects. This effect considers changes in the mix of economic activity. These changes have an impact on energy consumption because each activity has a different energy intensity. 3. Intensity effects. This effect considers the impact on energy consumption from changes in energy intensity. This is considered a proxy for a good energy efficiency change.
The decomposition presented in equation 1 describes the total energy consumption. Sub-section i becomes sub-category of aggregate energy consumption whose changes in economic structure are studied: Where E is the total energy consumption in the industry, Q (=∑ i Q i ) is the total industry activity output level, S i (= Q i /Q) is the industrial economic activity share and I i (=E i /Q i ) is the energy intensity of the sub-category i.
For IDA additive, the change in energy consumption (E) between two periods (0, T) as follows: Where, The subscripts act, str, and int each show the effect of activity, the effect of the economic structure and the intensity effect. The interpretation of the effects of the economic structure depends on the categorization applied and considers the impact on energy consumption of the changes in the division of different components in each of the sub-categories mentioned above.

Index Decoupling Formulation
We apply decoupling index method to analyze decoupling of energy consumption and growth of industrial economic activity of Indonesia. Based on the definitions given by Diakoulaki and Mandaraka (2007), decoupling is a measure of measures to reduce energy intensity (lower energy intensity), switch to fuels (cleaner energy consumption), and shift toward industries with more energy use slightly (improvement of industrial economic structure) by referring to any action that directly or indirectly uces the reduction of CO 2 emissions associated with energy consumption. Thus, assuming the same trend pattern and the correlation between energy consumption and high CO 2 emissions, the index decoupling formulation for CO 2 emissions would be similar to the decoupling formulation for energy consumption. Therefore, the total absolute effort (ΔF T ) during the period from 0 to T can be explained as the difference between changes in total energy consumption and changes in energy consumption due to the influence of industrial economic activity or as the amount of energy consumption of 2 (two) effect factors in Eq.
(2). Therefore, we use ΔF T to represent the total effect of decreasing energy consumption as follows: If the amount of change in energy consumption of the two negative effect factors, then ΔF T will be negative which means resulting in a decrease in energy consumption. To assess the extent to which this effort is effective in terms of separating energy consumption from the growth of industrial economic activity, the D T decoupling index during the period from base year 0 to year T is defined as follows: Since Equation (7) refers to the influence of positive industrial economic activity, the decoupling index can be determined for each equation effect, and the value obtained can help to identify the relative contribution of each of the effect factors to the overall decoupling process. Thus, the decoupling index of the whole process will be equal to the number of partial decoupling ices (Eq. 9).
Where D T is the total industry decoupling index; D str T � and D int T � is the decoupling index of each influence of the economic structure of the industrial economy, the energy intensity of the industry. If D T ≥1, showing the absolute decoupling effect, we can say that the amount of effect factor that energy efficiency is greater than the effect factor of industrial economic activity. Meanwhile, if 0<D T <1, it indicates a relative decoupling effect and we can conclude that the effect of energy efficiency seems to be weaker than the effect factor of industrial economic activity. Finally, if D T <0, it shows no decoupling and we conclude that the number of energy efficiency effect factors is not strong enough to significantly reduce energy consumption. If the value of D str T � and D int T > 0 , we can say that the efficiency factors of energy consumption such as economic structure of industrial economy and industrial energy intensity are substantive enough to contribute to decoupling. Conversely, if less than 0 (<0), these factors do not contribute to decoupling (Wang et al., 2013;Zhang and Da, 2015).

DATA
This study uses data from industrial Annual Statistics Indonesia 2010-2014 period. This data provides information about all manufacturing firms with 20 or more workers employed for at least 6 months and includes more than 20,000 firms each year. This survey contains information on energy consumption, outputs and other characteristics at the enterprise level, such as industry sub-sector and technology intenstiy. This survey provides a unique identifier for each company, which does not change over the period 2010-2014.

Results of Decomposition Analyisis
This section presents the results of additive decomposition analysis of changes in energy consumption in Indonesian industry. We report aggregate decomposition results in the study period, which includes all firms. The results of the report are presented using graphical instruments. Figure  Over the past few years, the possibility of decoupling growth in economic activity and growth in energy consumption has not occurred. Our results show that energy consumption has continued to increase significantly over the last few years in the industrial sector, which can be indexplained by the high increase in output. This means that energy efficiency improvements have not occurred yet. Thus, it is possible that energy consumption will increase steadily as economic activity increases if energy efficiency does not occur.

Decomposition Analysis by Sub-sector
In this section, we present descriptive results according to the industry standard classification (sub-sector). This sub-sector classification considers firms with the same business line (ISIC, 2009). Based on ISIC (2009), the company is classified into 24 industry sub-sectors. For explanation, we divide this sub-sector into three (3) groups based on the average intensity of energy usage; (1) sub-sectors with high intensity energy consumption (HEI), (2) sub-sectors with medium intensity energy consumption (MEI), and (3) sub-sectors with low intensity energy consumption (LEI).  (15); and on LEI (Figure 4), in the sub-sector of machinery and equipment (28), beverage (11), repair and installation services of machinery (33). In the HEI, the effects of economic structure show significant changes in sub-sector output share, thus reducing energy consumption in rubber and plastic sub-sectors (22), paper (17), basic metals (24). The effects of economic structure also occur in all MEI sub-sectors indexcept tobacco processing (12). In LEI, the effects of economic structures occur in the sub-sector of pharmaceuticals and medicine products (21), printing and reproduction of recording media (18), repair and installation services of machinery (33), other processing (32), products from coal and petroleum refineries (19).

Results of Decoupling Analysis
The result of decoupling analysis between energy consumption and economic activity of Indonesian industry along with the influence of each economic structure of industrial economy and industrial energy intensity on decoupling effect index or decoupling effect in 2010-2014 period in aggregate is presented in Table 2. The total decoupling index was negative (<0) for most of the study period, indicating that there was no decoupling effect between energy consumption and industrial economic activity. The decline in energy consumption derived from the inhibiting effect factor is less strong than the effect factor caused by the growth of industrial economic activity. That is, as industrial economic activity grows, energy consumption also increases. The inhibiting factors, that is, the economic structure of the industrial economy and the energy intensity of the industry have no role in reducing energy consumption and instead contribute to the increase of energy consumption along with the growth of industrial economic activity. In particular, in the period 2010-2011, 2011-2012 and 2013-2014, the increase in energy consumption is 5.5 million tce, 1.75 million tce and 3.68 million tce respectively with an average of 2.85 tce in the entire study period 2010-2014.
However, it should be noted that the decoupling index in the period 2012-2013 is 0.81 (0<D T <1); it shows that there is a relative decoupling effect in this period, which means the growth of industrial economic activity acfirms the reduction of energy consumption. However, the reduction in energy consumption derived from inhibiting effect factors such as industrial energy intensity and industrial economic structure appears to be weaker than the effects of industrial economic activity. To that end, the Indonesian government is committed to reducing energy consumption by issuing Presidential Regulation no. 22 of 2017 on the National General Energy Plan. The government's energy efficiency target is to achieve a 17% reduction in energy consumption by 2025 (efficiency) projected from 2015 (BaU). To meet this target, the new action plan requires that 10-30% of the expected reductions occur in the industry (MEMR, 2016). Implementation of this commitment is very important because in the nindext period, 2013-2014, the decoupling index again becomes negative.   The effect of changes in industrial energy intensity on decoupling (D int ) is negative during 2010-2014, which implies that decoupling between energy consumption and economic activity or growth in manufacturing industry economic activity has not occurred.
Overall D int accounted for 67.22% of total decoupling index (D tot ) during 2010-2014. In particular, the contribution of industrial energy intensity fluctuates in the decoupling process, emerging significantly in the period 2011-2012 and 2012-2013 with relative decoupling effects over other periods. In 2011-2012, there has not been decoupling because it is not supported by industrial economic structure while in 2012-2013, decoupling is relatively due to the decrease of industrial energy intensity with index 0.52 and supported by industrial structure structure with index 0,3 (Table 2). This means that energy intensity is the most important energy inhibiting factor and the largest contributor to the decoupling process. Finally, there is a relative decoupling effect between energy consumption and growth in industrial activity of Indonesia in the period 2012-2013, although in other periods it does not occur (Table 2). This shows that the growth of industrial economic activity that affects decoupling over time. Therefore, it is important to utilize the target factors that contribute to the decoupling of energy consumption and the growth of industrial sector economic activity. So the Indonesian government must take effective measures to reduce energy intensity, change industry (industrial economic structure), and support the transition of economic change to low carbon energy sources.

Results of Decoupling Analysis by Sub-sector
The result of decoupling analysis of energy consumption and industrial economic activity of Indonesia and the influence of decoupling of industrial economic structure and industrial energy intensity on decoupling index or decoupling effect in detail of industrial sub-sector in 2010-2014 period is shown in Table 3.
The textile sub-sector (13), basic metals (24), garments (14), and pharmaceuticals (21) show a relative decoupling effect between energy consumption and industrial economic activity, which means growth in industrial economic activity accompanied by decreased consumption energy, although the decline in energy consumption derived from the inhibiting effect factor is still weaker than the effect factor caused by the growth of industrial economic activity. While metal goods, not machineries (25), leather and footwear (15) and repair and installation services (33) show the absolute decoupling effect between energy consumption and industrial economic activity which means decreased energy consumption derived from more inhibiting effect factors stronger than the effect factor caused by the growth of industrial economic activity. Industrial economic activity grows, energy consumption decreases. The inhibiting factors, that is, the economic structure of the industrial economy and the energy intensity of the industry have contributed in reducing energy consumption even though each sub-sector is different. The contribution of these two inhibiting factors is in the sub-sectors of basic metals (24), metal goods, not machinery (25), garments (14), leather and footwear (15), repair and installation services (33) in sub sector tindextile (13) the main contribution is energy intensity and in sub sector pharmaceuticals (21) the main contribution is industrial economic structure.

CONCLUSIONS AND IMPLICATIONS
The aggregate results show that changes in total energy consumption fluctuate but tend to increase in the period 2010-2014. This fluctuation is caused by the main drivers of the dynamics of energy consumption, namely changes in industrial economic activity (output) and industrial energy intensity. This result can be explained by the increase in the level of economic activity (output) is high and unstable and even low increase in energy intensity (efficiency of energy) or it can be said efficiency has not happened. This is an important fact because it shows that decoupling between industrial economic activity (output) and energy consumption does not occur. Thus, the possibility of energy consumption will continue to increase when the economic activity of the industry increases. Therefore, to achieve and realize the occurrence of decoupling, energy efficiency policy must be implemented and realized in more detail and carefully.
More detailed results are based on sub-sectors. The role played by potential energy intensity factors varies across industry sub-sectors. Therefore, energy efficiency policies that take into account the specific behavior of firms in each sub-sector need to be addressed. To date, the Indonesian energy agenda considers only a single, homogeneous energy efficiency policy for the industry as a whole. Our results show that focusing policies on the most energyintensive and least efficient firms, the non-metallic minerals subsector (23) will be cheaper and more efficient. Cogeneration is one of the policies considered by the government; however, each subsector requires completely different cogeneration technologies.
The IDA (LMDI) method provides a detailed overview of industrial energy consumption in Indonesia. Different corporate characteristics can help policy makers to focus energy efficiency policies in the future only on certain sub-sectors, the most intensive and inefficient firms. Different policies should reduce implementation and targeted costs. However, different policies can face several obstacles. The application of policies to particular groups can be seen as an injustice to some firms. The policymaker must balance the benefits with the target. Mandatory energy efficiency targets should be applied to specific sub-sectors. Corporate identification is another challenge of different policies. However, as sustainable development has become an important global topic, the Indonesian government should not only focus on economic efficiency but also improve energy conservation and environmental quality. Based on the results obtained in this study, the strategic measures for sustainable development should aim to (1) Reduce the intensity of energy consumption, especially in the energy-intensive industrial sector, (2) promote shifting industrial to industrial structures with less energy intensive, (3) promoting low-carbon energy sources in energy mixed structures, and (4) encouraging the import of energy-dense products.