Exploring the mitigation potential for carbon dioxide emissions in Indonesia’s manufacturing industry: an analysis of firm characteristics

Abstract This study investigates ways to effectively reduce carbon dioxide (CO2) emissions in Indonesia’s manufacturing industry, by firm characteristics. It is important to determine the firm characters that have the greatest potential to decrease CO2 emissions. The Logarithmic Mean Divisia Index (LMDI) method is used to decompose CO2 emissions into the key factors influencing changes in CO2 emissions, such as economic activity, industrial structure, energy intensity, energy structure, and emissions coefficient during the 2010–2018 period. The findings indicate that changes in CO2 emissions in industrial sub-sectors vary. High technology firms had the lowest average emissions compared to firms with other technology. Large-sized firms had the lowest emissions than small and medium firms. Foreign private firms had lower emissions than national private firms did. Firms in the Java–Bali location had, on average, highest emissions than those outside Java–Bali. Exporting firms had lower average emissions intensity compared to non-exporting firms. This study’s novelty is an analysis of the effect of components that affect changes in CO2 emissions in firm groups based on their characteristics so that policymakers can focus on the potential reduction in CO2 emissions in certain groups of firms, namely firms that use the most energy intensively, is inefficient, and uses low-quality energy. Comparative analysis using firm characteristics reveals that energy-intensive firms’ economic growth determines changes in CO2 emissions in Indonesia’s manufacturing industry.


Introduction
The manufacturing industry plays an important role in the economy and is a major contributor to final energy consumption [1,2] and consequently environmental impact such as carbon dioxide (CO 2 ) emissions [3,4]. In Indonesia, the manufacturing sector makes the largest contribution to total gross domestic product (GDP) as compared to other sectors [5]. The performance of the Indonesian manufacturing industry experienced significant development during the 2010-2018 period, growing 28.04% with an average annual growth rate of 6.38% [5]. In line with the manufacturing sector's growth, the consumption of final fossil fuels also increased. The manufacturing and transportation industry sectors dominated the increase in Indonesia's final energy consumption for this period, with a total contribution of more than 75%. The manufacturing sector's contribution was between 35% and 40% [2]. The industrialization process will undoubtedly cause this proportion to continue to increase significantly in the future.
Growth of economic activity that is rapid as well as sustainable, which is marked by an increase in GDP, is usually followed by growth in the demand for energy, which tends to be difficult to control [6]. Indonesia's final energy demand is estimated to increase by an average of 5.3% annually during the 2016-2050 periods, assuming constant GDP growth of 6.04% annually [7]. Fossil fuel energynamely, fuel oil (BBM), natural gas and coal -still dominate and represents more than 95% of total primary energy consumption [2]. Fuel demand is currently increasing, with an average annual growth rate of 4.7%; natural gas demand is estimated to increase almost eight times by 2050, with at an average annual growth rate of 6.3%; Coal as a domestic supply of fuel is increased to 93% by 2050 [7].
The dramatic increase in fossil fuel consumption is concerning from the perspective of energy availability and environmental sustainability, especially considering climate change. The availability of natural resource reserves of fossil fuels in Indonesia is decreasing and is expected to soon be depleted [2]. If this occurs, the country will become vulnerable to experiencing an energy crisis or it could result in an energy supply shortage, which will affect the economic crisis [7]. Climate change occurs in consequence of a significant increase in greenhouse gas (GHG) emissions resulting from the accelerated consumption of fossil fuels. The World Resources Institute estimates that about 61.4% of global GHG emissions come from fossil fuel consumption [8]. The biggest contributor to GHG emissions is CO 2 gas emissions, accounting for 60% of all GHGs [4,9,10]. The manufacturing industry is responsible for almost one-third of the fossil fuel consumption worldwide and 36% of global CO 2 emissions [8]. Therefore, it is important to reduce the consumption of fossil fuels to mitigate CO 2 emissions by the manufacturing industry.
Reducing CO 2 emissions is quite important for developing sustainable manufacturing industries worldwide. The Indonesian government must outline several CO 2 emissions mitigation strategies and investigate the component factors affecting changes in CO 2 emissions. Which components contribute to changes in CO 2 emissions? Information and clear exposition of the role that the economy and fossil fuel energy play in changes in CO 2 emissions are required. Therefore, this study attempts to address these issues.
One component factor that is important in influencing the growth of environmental impact is expanded economic activity [11][12][13][14][15]. Growth in economic activity engenders growth in environmental impact in the form of an inverted U-shaped curve. Environmental degradation at first shows a rise and then falls with the increasing economic activity. With an inverted U-shaped curve, the growth of industrial economic activity implies that environmental quality will improve [11]. In the early stages of economic growth, degradation and pollution increase. However, after going through certain levels of income per capita, which varies for different indicators, the trend reverses. Thus, at high-income levels, economic growth leads to environmental improvement [12]. Industrial growth aligned with increased internalization is likely to benefit environmental quality. The growth of economic activity can reduce the degradation of natural resources if producers internalize the effects of feedback on production, although there are theories that increased economic activity will increase pollutant emissions [11]. Based on that opinion, it is important to determine the conditions required to internalize environmental problems because sustainable development depends upon this condition.
There are three component factors that affect the relationship between growth in environmental impact (CO 2 emissions) and growth in economic activity [10,16], namely: (i) changes in technology, (ii) shifts in the composition of energy inputs and (iii) a shift in the composition of outputs. New discoveries and innovations (i.e. technology changes) can improve energy conservation, leading to higher economic growth (i.e. GDP) than input growth rates or environmental impact. Changing from low-quality fuel to higher-quality fuel (i.e. a change in energy consumption) will change the composition of energy inputs (i.e. a change in energy share) and reduce energy intensity, considering that less energy is required to produce a unit of output. Changes in output composition (i.e. a change in output share) can also affect the relationship between energy and income because different industries have different energy intensities. Thus, increasing technology change in the manufacturing sector is important for maintaining sustainable energy availability and reducing environmental impact [16,17]. CO 2 emissions mitigation can be achieved by replacing environmentally friendly energy or shifting the composition of energy inputs [18][19][20]. The industrial sector is expected to achieve the highest level of CO 2 emissions mitigation among all sectors if the appropriate measures are implemented [17]. Therefore, minimizing waste generation (i.e. CO 2 emissions mitigation) must be considered important by the Indonesian manufacturing sector. Effective and well-targeted energy efficiency (i.e. a reduction in energy intensity) and energy diversification (i.e. a shift in the composition of energy input structures) is expected to meet the energy consumption and CO 2 emissions targets for the manufacturing industry as capital for the future of the manufacturing industry [21].
This study investigates ways to effectively reduce changes in CO 2 emissions in the Indonesian manufacturing sector based on firm characteristics. The firm characteristics considered include industry sub-sector, technology intensity, island location, firm size, capital ownership type and export capability. To achieve these objectives, the following is required: (1) estimates of CO 2 emissions and (2) investigations of the effects of components that affect changes in CO 2 emissions in Indonesia's manufacturing industry based on firm characteristics. The novelty of this study is in its analysis of CO 2 decomposition based on specific firm characteristics.

Literature review
In the literature, most of the existing studies on emissions driver analysis were conducted based on the decomposition method. The decomposition method is very popular and widely applied by researchers and practitioners in the field of energy and emissions [22][23][24][25][26][27][28][29][30][31][32][33], which mainly contain two types: index decomposition analysis (IDA) and structural decomposition analysis (SDA) [25]. The fundamental difference between the IDA methods and the SDA method is that IDA uses index numbers [22][23][24][25][26][27][28], only needs to use sectors' aggregate data, while SDA uses an input-output model, which is based on an input-output system [29][30][31][32][33]. In practical implementation, the IDA method is better able to analyses time and country dynamics in detail because of data availability [30]. IDA is essentially an accounting approach; it is used to quantify the impacts of various factors on a change in energy consumption or CO 2 emissions. Because of its low data requirement, flexibility in modelling and ease of result interpretation, IDA has been widely used to assist policymaking [34]. Meanwhile, SDA can be used to conduct a more systematic analysis, decompose models with more influencing factors, and analyses the impacts of various factors on the change in energy consumption and emissions; however, this method has higher requirements for data collection. Wang et al. [35] presented a recent survey on IDA and SDA applied to energy and emissions. They also made a comparison between the two methods.
More recently, a new type of decomposition analysis known as the production-theoretical decomposition analysis (PDA) has emerged. This method is used to study changes in an aggregate indicator within a framework of productive efficiency, which is different from IDA and SDA. Given a set of inputs and outputs, the production theory starts from a theoretical definition of general production technology. A best practice frontier is constructed using distance functions to characterize the production technology [34,36]. An entity's distance from the frontier represents its production inefficiency. The strength of the production theory framework lies in modelling the production frontier according to production economics from which entities' efficiency and technology levels can be assessed [37]. Inherent in this analytical framework, PDA can capture the impacts of efficiency and production technology on aggregate emission change. However, the current decomposition approach based on PDA and weak disability may cause deviation on potential emission reduction [38,39]. Furthermore, decomposition results associated with these methods may be biased and incomplete [36].
IDA techniques there are two, namely the IDA Laspeyres method and the IDA Divisia method [23]. The IDA Divisia method has an advantage over the IDA Laspeyres method because the IDA Divisia can be applied to short time series [31] and can accommodate zero-value data [40,41]. In calculating the index, the IDA Divisia may use two different methods. The first method is the arithmetic mean Divisia index and the second one is the logarithmic mean Divisia index (LMDI). Currently, many studies on CO 2 emissions related to energy use have been performed. The LMDI decomposition method has been widely applied, especially to identify factors affecting CO 2 emissions. The LMDI method has been widely used for its advantages such as comprehensive decomposition generating no residuals and easy to use [42,43]. A summary of several previous studies that have analyzed the manufacturing industry sector used decomposition method especially LMDI is presented in Table 1.
Akbostanci et al. [3] used the LMDI decomposition method to analyses CO 2 emissions of the Turkish manufacturing industry for the 1995-2001 period. The LMDI method is used to decompose the changes in the CO 2 emissions of the manufacturing industry into five components: changes in activity, activity structure, energy intensity, energy mix and emission factors. CO 2 emissions are calculated by using the fuel consumption data. The study covered 57 industries and found changes in total industrial activity and energy intensity to be the primary factors determining the changes in CO 2 emissions during the study period. Moreover, among the fuels used, coal is the main determining factor. Liu et al. [44] analyzed changes in China's industrial CO 2 emissions from final fuel use using the LMDI method in the 1998-2005 period. The changes in industrial CO 2 emission are decomposed into carbon emissions coefficients of heat and electricity, energy intensity, industrial structural shift, industrial activity and final fuel shift. The overwhelming contributors to the change in China's industrial sectors' carbon emissions were industrial activity and energy intensity. Meanwhile, the impact of emission coefficients of heat and electricity, fuel shift and structural shift was relatively small. Wang and Nie [45] found that CO 2 emissions were in the form of a U-shaped curve during the 1995-2007 period and that there was a turning point in 2001. They used the LMDI decomposition method to estimate the contribution of economic growth, energy intensity, energy mix and economic structure on changes in CO 2 emissions. The study results demonstrated that economic growth is the main factor in emissions intensity growth and that energy intensity is the most significant factor in reducing emissions. Similar conclusions were drawn by Hammond and Norman [46], Chang et al. [47], Ren et al. [48], and Yan and Fang [8], who all explored the effects of different factors using the same method.
In another study, Jeong and Kim [49] decomposed Korean industrial manufacturing greenhouse gas (GHG) emissions using the LMDI method, both multiplicatively and additively. The results indicate that the structural and intensity effects played roles in reducing GHG emissions; the structure effect played a more significant role than the intensity effect. Moreover, the energy-mix effect increased GHG emissions, whereas the emission-factor effect decreased GHG emissions. Similar studies on the use of LMDI methods in decomposing CO 2 emissions in industrial manufacturing sectors continue to develop with novelty respectively, such as Zhu et al.  [55] and Wen and Li [56]. A special study in Indonesia has been conducted by Zaekhan et al. [4], who found that the use of various types of fossil fuels and an increase in total output were the main causes of high CO 2 emissions in the manufacturing sector. The study results demonstrated that economic growth is the main factor in emissions intensity growth and that energy intensity is the most significant factor in reducing emissions. In addition, the LMDI method was employed for studies of carbon emissions in other industries or sectors [59][60][61][62] and national or global levels [63][64][65][66].

Data and descriptive statistics
This study uses data from survey statistics of Indonesian large-and medium-sized industries were prepared by Statistics Indonesia (Indonesian Central Bureau of Statistics) in 2010-2018. These data provide information about all manufacturing firms with 20 or more laborers employed for at least six months and include more than 20,000 firms each year. This survey contains information about each firm's real output and energy consumption (based on fuel type). These basic data are used in this study. The output is measured as value-added in trillions of Indonesian rupiah based on 2010 constant prices. Since this study covers the 2010-2018 period, the value of output is based on constant prices in 2010, the most recent base year available. The 2010 figure remains valid because there is no more recent base year. Meanwhile, the 2016 data were not included in this study because the data were not available. Moreover, the statistical survey of large and medium industries in Indonesia was not conducted that year. The types of fuel consistently consumed by Indonesian manufacturing include petrol, diesel oil, kerosene, liquefied petroleum gas (LPG), coal, natural gas, and electricity. Fuel energy consumption data are available in different physical units. To obtain a firm's total fuel consumption, these physical values were standardized to barrel oil equivalents (BOE) using coefficients issued by the Ministry of Energy and Mineral Resources and then converted to tons of coal equivalent (TCE) to adjust to the CO 2 emissions coefficient standards published by the Intergovernmental Panel on Climate Change (IPCC) [67,68]. CO 2 emissions can be estimated from fossil fuel consumption data using the IPCC method. Descriptive statistics of aggregate data per period are summarized in Tables 2 and 3.
In addition to data on energy consumption and output, the Statistics Indonesia data include other firm characteristics such as industrial sub-sector groups, technology intensity, firm size, type of ownership, island location and export capability. Industrial sub-sector groups are based on the Indonesian business field raw classification system, which is adjusted to the International Standard Industrial Classification (ISIC), Revision 4. The firm is classified into 24 sub-sectors of industrial groups. The firm's technology intensity is based on classifications by the Organization for Economic Co-operation and Development (OECD), namely high (H), medium-high (M-H), medium (M) and medium-low technology (M-L). Firm size groups are based on the number of laborers-small firms have less than 100 laborers, medium-sized firms between 100 and 1000 laborers and large firms more than 1000 laborers. Based on the type of ownership group, firms can be classified into domestic investment (PMDN) and foreign investment (PMA) firms. Based on geographical location, the firm is classified as being located the Java-Bali island location, Sumatra, Kalimantan, Sulawesi, or Maluku-Papua. Based on export capability, firms are classified as exporting or non-exporting firms. The descriptive statistics of the data by firm characteristics are shown in detail in Appendix A.

CO 2 emissions estimation
The method used to estimate CO 2 emissions is the IPCC guidelines for the National Greenhouse Gas Inventory. This calculation method is based on fossil fuel type consumption and carbon emissions factors [67,68]. CO 2 emissions can be estimated using Equation (1): where subscript i denotes firm characteristics such as industrial sub-sector, technology intensity, firm size, ownership type, island location and export capability. Subscript j is fuel type, C denotes total CO 2 emissions, Cij is CO 2 emissions of sector i from fuel consumption j, Eij denotes fuel consumption j in sector i, gj denotes carbon emissions factor of fuel consumption j, Oj denotes carbon oxidation fraction of fuel consumption j, 44/12 is the ratio of the relative molecular weight of CO 2 and carbon and fij denotes CO 2 type emissions factor of fuel in sector i after conversion to a standard unit of coal equivalent (TCE). Table 4 presents the f values for each type of fuel.

CO 2 emissions decomposition
Decomposition analysis is one of the most effective and widely applied methods for investigating the mechanisms that influence energy consumption and environmental side effects, especially CO 2 emissions [25,30,45,[69][70][71][72][73]. Decomposition analysis, especially IDA, has become a widely used technique for tracking energy efficiency trends [23,[74][75][76]. This technique examines the impacts of various component factors that contribute to changes in energy consumption and changes in CO 2 emissions. This study uses the IDA technique, specifically, the LMDI method due to the advantages such as decomposes completely without residuals [42,43,77], is consistent in aggregation [42,78], can be applied to short time series and is easy to use [31], strong theoretical basis, adaptability, and interpretability [74]. LMDI is the preferred method due to the ease of its formulation [79][80][81]. LMDI IDA considers five component effects as impacting CO 2 emissions changes in industry [40]: 1. The industrial economic activity effect, represented by the total output. 2. The industrial economic structure effect, represented by the relative share of industrial output in the total output. 3. The energy intensity effect, represented by the ratio of industrial fossil fuel consumption and the industrial output. 4. The energy mix effect, represented by the composition of fossil fuel energy in the total energy consumption. 5. The emissions coefficient effect, represented by the average emissions factor of fossil fuel energy use. The emissions factor is assumed constant for conventional fuels [8].
The CO 2 emissions decomposition is as follows: where i denotes sector, j denotes fuel type, C is total CO 2 emissions, and Cij denotes the CO 2 emissions from fuel consumption j in sector i. Qi denotes output in sector i, Q (¼ RiQi) denotes the total output of industrial economic activity, and Si (¼ Qi/Q) denotes the proportion of economic activity in sector i. Ei denotes the total fuel consumption in sector i and Ii (¼ Ei/Qi) denotes energy intensity in sector i. Eij denotes fuel consumption j in sector i (Ei ¼ RjEij). Mij (¼ Eij/Ei) denotes the energy mix in the amount of fuel consumed in sector I and Kij (¼ Cij/Eij) denotes the coefficient of CO 2 emissions from fuel consumption j in sector i. Specifically, the decomposition of CO 2 emissions additives into their component parts [40] is shown in Equation (3): Subscripts act, str, int, mix and emf of changes in CO 2 emissions on the right-hand side of Equation (3) show the effects associated with industrial economic activity, industrial economic structure, industrial energy intensity, industrial energy mix and emissions coefficients, respectively. The components are written as follows: A summary of the operational definitions of the study variables is presented in Table 5.

Descriptive analysis
This section describes a descriptive analysis of the data according to firm characteristics, such as the industrial sub-sector group, technology intensity, firm size, ownership type, island location and export capability (Appendix A). Firstly, we based on the industrial sub-sector characteristics. From the data set, the sub-sectors with the most number of firms, that is, more than 1000 firms, in the 2010-2018 period were food (10), chemicals (20), wood, bamboo, rattan (16), rubber and plastic (22), non-metallic minerals (23), garments (14), textiles (13), repair and installation machinery (33) and motor vehicles and trailers (29). Firms in this subsector represent more than 72% of the total firms in Indonesia's manufacturing industry. However, not all sub-sectors contributed the highest to energy output and consumption either. Only a few sub-sectors have the highest share of output and consumption, such as food (10), chemicals (20), rubber and plastics (22), motor vehicles and trailers (29) and textiles (13), with a total output of more than 55%. Meanwhile, the sub-sectors with the highest energy consumption are non-metal minerals (23), food (10), textiles (13), chemicals (20) and rubber and plastics (22), with a total share of more than 63%. Considering the shares of output and energy consumption, firms that are energy-efficient (low energy intensity) may be firms in the motor vehicle and trailer sub-sector (29), and firms that are not efficient/wasteful of energy (high energy intensity) are firms in the non-metallic mineral sub-sector (23). Based on average energy intensity data, the five (5) sub-sectors with the highest energy intensity are non-metal minerals (23), repair and machine installation services (33), textiles (13), paper (17), and wood, bamboo, and rattan (16). By contrast, the sub-sectors with the lowest energy intensity are processing tobacco (12), furniture (31), other transportation equipment (30), other processing (32), and computers, electronics and optics (26).
Second, we based on the characteristics of the technology intensity. From the data set, mediumlow technology (M-L) firms represent more than 69% of total firms and high-tech (H) firms only around 2.45%. The rest are medium (M) and medium-high (M-H) firms. Medium-low technology (M-L) firms account for more than 52% of the output share and more than 44% of total energy consumption, whereas high-tech (H) firms only contribute less than 5% of the output share and less than 2% of total energy consumption. Meanwhile, medium (M) and medium-high technology (M-H) firms represent approximately 43.48% of output share and 53.17% share of total energy consumption, respectively. In terms of energy intensity, high technology (H) firms have a small average energy intensity compared to firms with lower technology. Therefore, firms with high technology (H) use less energy, on average, with a large output value. Additionally, firms with high technology (H) may be more efficient than firms with other technology. Medium technology (M) firms have higher average energy intensity than firms with other technologies, which means firms with medium technology (M) intensity are more wasteful or less efficient.
Third, we based on the characteristics of firm size. From the data set, firms with less than 100 laborers represent more than 69% of the total firm. However, this group of firms had the lowest share of output, which only contributed less than 12% of the output share (11.26%), and the share of energy consumption was less than 10%. Meanwhile, medium-sized firms (100-999 laborers) represent about 27% of the total number of firms, but the shares of output and total energy consumption are 46.75% and around 43%, respectively. Large firms (>1000 laborers) account for less than 5% of total firms and account for more than 40% (41.98%) of the output share and more than 45% (47.7%) share total energy consumption. In terms of energy intensity, small firms have a smaller average energy intensity compared to the medium and large firms. Medium and large firms dominate the CO 2 emissions resulting from energy consumption because of having high energy intensity. Small firms (20-99 laborers) contributed a small share of energy consumption (less than 10%), and many firms have low energy intensity. Smaller firms may have less energy because they have a high proportion of manual labor. Meanwhile, the size of the firm is increasing, the energy intensity tends to be higher because it is more capital intensive. Fourth, we based on the characteristics of the type of ownership. Ownership classification refers to foreign investment participation, domestic investment, namely, private or state-owned firms and other investments. Firms that are members of the ownership of PMA and PMDN are firms that invest using government facilities, whereas those that do not use the facilities are non-PMA/PMDN. Accordingly, more than 21% of the firms are owned by the national private sector (PMDN), more than 10% are foreign-owned firms (PMA), and the rest are non-PMA/PMDN. Although national private-owned firms account for around 21%, they are responsible for more than 48% of total energy consumption. Meanwhile, foreignowned firms (PMA) account for more than 25%. In addition, non-PMA/PMDN firms contributed to energy consumption by 25.8%. PMDN firms contributed the highest share of output, whereas non-PMA/PMDN firms contributed the smallest share even though the number of firms was more than 68% of the total firms. Foreign private-owned firms (PMA) have a stable share of output and energy consumption during the study period and have lower average energy intensity than firms with other ownership types.
Fifth, we based on the characteristics of the island location. The classification of island location considers the position of a firm. This position concerns logistics and transportation costs for energy transfer and product marketing. From the data set, firms in the Java-Bali location represent more than 84% of the total firms and account for more than 72% share of output and more than 65% share of total energy consumption. However, the energy intensity of firms in the Java-Bali location is the lowest on average, so it is more efficient than firms located outside Java-Bali. Locations with the next largest number of firms are as follows: Sumatra (10.58%), Sulawesi (2.53%), Kalimantan (1.94%), and Maluku-Papua (0.42%). Firms in the Maluku-Papua location account for less than 1% of total firms and contribute less than 0.5% of total output and less than 1% of total energy consumption. In terms of energy intensity, firms in this location have a large average energy intensity compared to that in other locations (apart from Sulawesi). This shows that firms in Maluku-Papua use more wasteful and less productive energy because of the small share of output.
Sixth, we based on the characteristics of export capability. The exporter classification is a firm that carries out export activities or not. This classification is divided into 3 groups, Yes, No, and Others. Another group is firms that do not fill out the survey in the statistics of medium to large industries. From the data set, exporting firms represent less than 10% of total firms, but account for more than 20% of the output share and more than 30% share total energy consumption. Non-export firms represent more than 43% of total firms but account for about 18.95% share of output and approximately 19.5% share total energy consumption. Although the number of firms that do not export is more than exports, the contribution of output and energy consumption of firms that do not export is smaller than that of exporting firms. This shows that firms that export are more productive than non-export firms. In terms of energy intensity, firms that do not export have an average energy intensity lower than the average energy intensity of export firms. Firms that do not export have less energy because they are less energy-intensive. On the other hand, although export firms share a higher output, a higher share of energy also makes energy intensity higher because they tend to be energy-dense.

Analysis of CO 2 emissions changes in each effect over time
This section presents the analysis of changes in CO 2 emissions in each effect over time in the manufacturing industry aggregate. Table 6 presents the results of the decomposition of changes in CO 2 emissions in Indonesia's manufacturing industry in the 2010-2018 period. These results indicate that in general the main drivers of the dynamics of changes in total CO 2 emissions are changes in industrial economic activity (DCact) and changes in industrial energy intensity (DCint). The effect of changes in industrial structure and changes in energy composition is still very small. In Table 6, it can be seen that changes in total CO 2 emissions  Table 6 show that there is a different effect pattern in each characteristic. This can be explained according to the group of firm characteristics as follows. According to the industrial sub-sector, the increase in CO 2 emissions in the 2010-2011 period was mainly in consequence of an increase in industrial energy intensity and industrial economic activity of 7.95 million tons of CO 2 and 6.06 million tons of CO 2 , respectively. The small increase in CO 2 emissions in the 2011-2012 period can be explained by an increase in CO 2 emissions because industrial economic activity was offset by a decrease in CO 2 emissions in consequence of changes in industrial energy intensity and changes in the structure of industrial energy composition, each of 4.08 million tons of CO 2 , À2.18 million tons of CO 2 and À0.25 million tons of CO 2 . The increase in CO 2 emissions in the Indonesian manufacturing industry observed in the 2012-2013 period was even smaller, in consequence of the continuing decline in industrial energy intensity and a change in the industrial economic structure that reduced CO 2 emissions. In the 2013-2014 period, the increase in CO 2 emissions was caused by each component of the effect. In the 2014-2015 period, the increase in CO 2 emissions was caused by an increase in industrial economic activity, changes in the industrial economic structure, and industrial energy intensity. In this period there was a change in the structure of the industrial energy composition towards environmentally friendly energy (energy with small emission intensity). The reduction in CO 2 emissions in the Indonesian manufacturing industry that was observed in the 2015-2017 period was very significant, in consequence of high energy efficiency or a large decrease in industrial energy intensity and a change in the energy composition structure in the industry, although the increase in industrial economic activity was also high. There was an increase in CO 2 emissions again in the 2017-2018 period, this was in consequence of an increase in industrial economic activity, industrial energy intensity, and the structure of the industrial energy composition even though there was a slight decrease in CO 2 emissions in consequence of changes in the industrial economic structure. Energy efficiency in this period does not occur. In general, the biggest contributors to changes in CO 2 emissions in the 2010-2018 period were an increase in industrial economic activity (61.85%) and a decrease in industrial energy intensity (32.45%). Meanwhile, changes in industrial structure and changes in the structure of industrial energy composition still contributed little (4.72% and 0.97%).
According to technology intensity, the increase in CO 2 emissions in the 2010-2011 period was caused by an increase in industrial energy intensity and industrial economic activity of 8.34 million tons of CO 2 and 6.13 million tons of CO 2 , respectively, but there was a decrease in CO 2 emissions in consequence of structural changes. industry À0.38 million tons of CO 2 . The small increase in CO 2 emissions in the 2011-2012 period can be explained by the increase in CO 2 emissions because of industrial economic activity and an increase in industrial energy intensity, which was offset by a decrease in CO 2  According to the firm size, the effect of each component on changes in CO 2 emissions in each period is consistent with the results on the characteristics of the industrial sub-sector except in the 2013-2014 and 2015-2017 periods. During that period, there was a decrease in CO 2 emissions in consequence of changes in the industrial structure in the firm size characteristic group. In this period there was a change in the industrial structure towards a less energy-intensive industry and a change in the structure of the industrial energy composition towards environmentally friendly energy. Contributors to changes in CO 2 emissions in the 2010-2018 period were an increase in industrial economic activity (41.84 million tones CO 2 ) and changes in industrial structure (0.07 million tones CO 2 ), while a decrease in energy intensity (À18.42 million tones CO 2 ) and a change in the structure of the energy composition (À0.38 million tons of CO 2 ) is a barrier to increasing CO 2 emissions.
According to ownership, the increase in CO 2 emissions in the 2010-2011 period was caused by an increase in all effects, respectively 6.00 million tons of CO 2 (DCact), 0.62 million tons of CO 2 (DCstr), 6.66 million tons of CO 2 (DCint), and 1.24 million tons of CO 2 (DCmix). The small increase in CO 2 emissions in the 2011-2012 period can be explained by the increase in CO 2 emissions in consequence of industrial economic activity and an increase in industrial structure, which was offset by a reduction in CO 2 emissions in consequence of changes in industrial energy intensity and changes in the structure of industrial energy composition. Meanwhile, a small increase in CO 2 emissions in the 2012-2013 period was in consequence of an increase in CO 2 emissions in consequence of industrial economic activity, changes in industrial structure, and changes in energy structure, which were offset by a reduction in CO 2  According to export capability, the increase in CO 2 emissions in the 2010-2011 period was caused by an increase in all effects, respectively 6.12 million tons of CO 2 (DCact), 0.01 million tons of CO 2 (DCstr), 8.02 million tons of CO 2 (DCint), and 0.38 million tons of CO 2 (DCmix). The increase in CO 2 emissions in the 2011-2012 period can be explained by the increase in CO 2 emissions because of industrial economic activity and changes in industrial structure, which were offset by decreases in CO 2 emissions in consequence of changes in energy intensity and changes in the structure of industrial energy composition. Meanwhile, the increase in CO 2 emissions in the 2012-2013 period was caused by an increase in CO 2 emissions in consequence of industrial economic activity and changes in the energy structure, which were offset by a decrease in CO 2 emissions in consequence of changes in industrial structure and changes in industrial energy intensity. There was a reduction in CO 2 emissions in the 2013-2014 period in consequence of changes in industrial structure (À0.02 million tons of CO 2 ), while in the 2014-2015 period the reduction in CO 2 emissions was in consequence of changes in the energy structure. The effect of each component on changes in CO 2 emissions in the 2015-2017 period is consistent with the results on firm size, ownership, and island location characteristics. Contributors to changes in CO 2 emissions in the 2010-2018 period were an increase in industrial economic activity (41.84 million tones CO 2 ) and changes in industrial structure (0.07 million tones CO 2 ), while a decrease in energy intensity (À18.42 million tones CO 2 ) and a change in the structure of the energy composition (À0.38 million tons of CO 2 ) inhibiting the increase in CO 2 emissions.
Mitigation potential of CO 2 emissions by firm characteristics CO 2 emissions analysis industrial Sub-Sector Table 7 presents the results of the decomposition of CO 2 emissions analysis in the industrial sub- sectors. Changes in CO 2 emissions in consequence of industrial economic activity in all sub-sectors are positive, which means industrial economic activity always results in increased CO 2 emissions. Furthermore, changes in CO 2 emissions in consequence of the influence of the industrial structure vary in value in each sub-sector; some are positive and some are negative. In the heavy polluters' industrial sub-sectors, the effect of the positive industrial economic structure is found in sub-sectors (23), (10), and (27), while the negative is found in sub-sectors (13), (20), (22), (17) and (24). In the medium polluters' industrial sub-sectors, the effect of the positive industrial economic structure is found in sub-sectors (29), (14), (16), and (15), while the negative is found in sub-sectors (25), (26), (30), and (12). In the low polluters' industrial sub-sectors, the effect of industrial economic structure is almost positive for all sub-sectors, except for subsector (21) and (32). There was a significant reduction in CO 2 emissions for heavy polluters in industrial sub-sectors (13), (20), (22), (17) and (24); for medium polluters in sub-sectors (25), (26), (30), and (12); and for all low polluters, except in subsector (21) and (32). CO 2 emissions decreased because of changes in energy intensity for heavy polluters in almost for all industrial sub-sectors except for sub-sector (22); for medium polluters, in industrial sub-sectors (14), (16), (26), and (12); and for low polluters, in almost for all industrial sub-sectors except sub-sector (21). These results indicate efforts to mitigate CO 2 emissions from these sub-sectors by decreasing energy intensity (i.e. increasing energy efficiency). There was also a reduction in CO 2 emissions in consequence of changes in the industrial energy structure for heavy polluters in the industrial sub-sector (23), (20), (17), (24), and (27), for medium polluters, in industrial sub-sectors (25), (26), and (30) and for low polluters, there is no reduction in CO 2 emissions in all sub-sectors. Based on these results, the structure of industrial energy composition has not played a major role in CO 2 emissions reduction in all Indonesian industrial sub-sectors compared to other effects, such as industrial economic structure and industrial energy intensity. Thus, the effect of the energy composition structure needs to be addressed and realized by transitioning to a lowcarbon energy fuel source or from low-quality to high-quality fuel. The effect of the emission coefficient on changes in CO 2 emissions was not found in this study. This is because the emission coefficient value did not change in the 2010-2018 period. Table 8 presents the results of the CO 2 emissions decomposition analysis based on technological intensity. In the 2010-2018 period, the largest increase in total CO 2 emissions was found for firms with medium (M) technology, followed by firms with medium-low (M-L) technology, medium-high technology (M-H) and high technology (H). This shows that firms with medium technology (M) are heavy polluters as compared to firms with other technologies. The decomposition results in Table 8 demonstrate that the effects of industrial economic activity still dominate the increase in CO 2 emissions. The increase in CO 2 emissions by firms with medium technology (M) is dominated by changes in CO 2 emissions because of economic activity effects and energy intensity effects. For firms with medium technology (M), the increase in CO 2 emissions in consequence of economic activities is higher than the effect of its energy intensity. Conversely, there is a negative energy intensity effect for firms with medium-low technology (M-L), medium-high technology (M-H), and high technology (H) which means that changes by these firms in energy intensity can reduce CO 2 emissions and can even limit the increase in CO 2 emissions caused by industrial economic activity.

CO 2 emissions analysis by technological intensity
A reduction in CO 2 emissions in consequence of industrial structural effects occurs in firms with high technology (H) and medium technology (M). This shows that there are important changes in the share of output that can reduce total CO 2 emissions. There is a shift in high-energy-intensive firms to low energy-intensive firms. The reduction in CO 2 emissions because of changes in the energy composition structure occurs in almost for all technology, except in medium-high technology (M-H).  This shows that there are important changes in the structure of energy that can reduce total CO 2 emissions, except in medium-high technology (M-H). The energy composition structure effects on reducing CO 2 emissions in firms with medium-high technology (M-H) does not yet exist. Thus, the effect of the energy composition structure needs to be addressed and realized by transitioning to a low-carbon energy fuel source or from low-quality to high-quality fuel in firms with this mediumhigh technology. Table 9 presents the results of the CO 2 emissions decomposition analysis based on firm size. In the 2010-2018 periods, large firms (i.e. number of laborers > 1000) had the smallest increase in total CO 2 emissions, but the energy intensity of this large group of firms is the highest (Appendix A.4). Even though this group only represented < 5% of all Indonesian manufacturing firms (Appendix A.1), these large firms used fossil fuels very intensively and had a tendency to be 'wasteful' because their average share of energy consumption (47.70%) is greater than their average share of output (41.98%). In general, the increase in CO 2 emissions in this large group of firms was dominated by the positive effects of economic activity. Meanwhile, the industrial structure, energy intensity, and energy composition structure have negative effects so that they play a role in reducing the increase in total CO 2 emissions, which means that these effects can balance total CO 2 emissions which are dominated by increases in CO 2 emissions because industrial economic activity. The same effect also occurs in a medium group of firms with 200-499 laborers. Based on these findings, it means that there is potential to mitigate CO 2 emissions in both groups of firms.

CO 2 emissions analysis by firm size
Besides, based on the results presented in Table  9, there are negative structural economic effects for firms with 20-99 (À0.01), 200-499 (À0.98), and >1000 (À2.67) laborers, showing that there is an important change in the shared output of these firms in the decline of CO 2 emissions. Structural effects of the industrial economy occurred, which means that there is a shift from high-energy-intensive firms to low energy-intensive firms in this group. Changes in energy intensity effect on changes in CO 2 emissions across firm size groups are all negative, which means that changes in energy intensity by firms in each of these groups can balance total CO 2 emissions which are dominated by increases in CO 2 emissions because of industrial economic activities. There is a direct relationship between the effects of energy intensity and energy composition structure. In addition to energy efficiency, a large group of firms (laborers > 1000) and a group of medium-sized firms with 200-499 laborers also transitioned towards higherquality fuels, or lower CO 2 emissions. Therefore, the focus of public policy, especially energy policy, should be directed towards medium and large firms. Table 10 presents the results of the CO 2 emissions decomposition analysis based on capital ownership. The increase in total CO 2 emissions in domestically owned firms (PMDN) was higher than that of foreign-owned firms (PMA). The increase in total CO 2 emissions is in consequence of the influence of various components. The increase in CO 2 emissions is dominated by the positive effects of industrial economic activities. The effects of changes in industrial structure and changes in energy intensity on changes in CO 2 emissions in foreign-owned firms (PMA) and group of domestically owned firms (PMDN) are all negative, which means that changes in industrial structure and changes in energy intensity by firms in each of these groups can balance the total CO 2 emission which is dominated by the increase in CO 2 emission in consequence of industrial economic activities. There has been a shift in the industrial economic structure from high-energy-intensive firms to low-energyintensive firms, and PMA and PMDN firms have shown energy efficiency. The contribution of the structural effect of the energy composition is also negative for PMA firms, indicating that there has been a shift towards high-quality fuels (low CO 2  emissions) by these firms. Meanwhile, the contribution of the structural effect of the energy composition is still positive for PMDN firms, which shows that there has been no shift towards highquality fuels in these firms. This shows that each group of firms needs different policies. One policy option is to focus on PMDN firms because PMDN firms make a large contribution to increasing CO 2 emissions and can be an early stage in determining energy and environmental policies for a larger and more heterogeneous group. Table 11 presents the results of the CO 2 emission decomposition analysis of based on island locations. The largest increase in total CO 2 emissions occurred in the Java-Bali location, then in Kalimantan, Sulawesi Sumatra, and Maluku-Papua locations. In general, the effects of industrial economic activity on changes in CO 2 emissions are all positive, the highest in the Java-Bali location followed by Sumatra, Kalimantan, Sulawesi, and Maluku-Papua. These results indicate that changes in industrial economic activity that are getting higher are not necessarily followed by changes in high CO 2 emissions, as can be seen in the case in Sumatra. The component effects that cause changes in CO 2 emissions such as industrial structure effects, energy intensity effects, and energy composition structure effects are all negative, which means they will balance the increase in total CO 2 emissions in consequence of industrial economic activities. In Table 11, there is a decrease in CO 2 emissions in consequence of structural effects (À1.14), energy intensity effects (À6.20), and energy composition structural effects (À0.36). This shows that in Sumatra there has been a shift in firms from high-energy-intensive to low-energyintensive, energy efficiency, and this indicates a shift towards higher-quality fuel or lower fuel emissions. Apart from Sumatra, the effect of energy intensity in other locations is also negative, except in Kalimantan. The increase in CO 2 emissions in the Kalimantan region is in consequence of all effects; changes in economic activity, industrial structure, energy intensity, and energy structure. This shows that firms in the Kalimantan region are not efficient or wasteful in their use of energy and are still using polluting energy. Firms in the Java-Bali region have the lowest average energy intensity compared to other regions, but this low average is not able to reduce total CO 2 emissions in consequence of the high increase in industrial activity. There was a decrease in CO 2 emissions in consequence of a decrease in energy intensity (À12.03) and a change in the energy composition structure (À0.94) in the Java-Bali region, which indicates that there is energy efficiency and a shift towards higher-quality fuels or lower fuel emissions in Java-Bali although the impact is small. Table 12 presents the results of the CO 2 emission decomposition analysis based on export capability. Based on these results, the exporting firm group that caused the increase in CO 2 emissions was higher than the non-exporting firm group in the 2010-2018 period, although the difference was small (5.13 and 4.59). This result is proportional to a large number of firms in each group, where the exporting firm has a higher number of firms (Appendix A.1). The decomposition results also show that the effect of industrial economic activity dominates changes in CO 2 emissions. The negative influence of the industrial economic structure on export and non-export firms indicates that there is a reduction in CO 2 emissions, which can balance the increase in CO 2 emissions in consequence of industrial economic activities. The balance of the increase in total CO 2 emissions by export and nonexport firms is also affected by changes in the effect of negative energy intensity. The magnitude of the negative effect of energy intensity on the exporting firm is greater than that of non-exporting firms, this shows that the exporting firm is more energy-efficient than the non-exporting firm. An indication of this result is that the exporting firm is more competitive than the non-exporting firm. On the other hand, the effect of the energy composition structure on changes in CO 2 emissions for export firms and non-exporters is different, respectively negative and positive. This indicates that the exporting firms have switched to higher-quality fuels or lower CO 2 emissions, although the impact is small, while the nonexporting firms have not. Thus, the focus of public policy should focus on export-oriented firms to reduce CO 2 emissions and encourage non- exporting firms to follow the steps that have been taken by exporting firms.

Conclusions
This study investigated ways to effectively reduce CO 2 emissions by the Indonesian manufacturing industrial sector based on six categories, namely, (1) industrial sub-sector, (2) technological intensity, (3) firm size, (4) capital ownership, (5) island location and (6) export capability. Some important findings are presented as follows.
The main drivers of changes in CO 2 emissions in Indonesia's manufacturing industry were industrial economic activity and industrial energy intensity. Industrial energy intensity was the driving force behind reductions in CO 2 emissions. However, industrial economic activity appears to dominate CO 2 emissions. Therefore, the effect of energy intensity does not automatically imply changes in reducing CO 2 emissions. A decrease in CO 2 emissions in consequence of the effects of energy intensity occurred in the characteristic group of industrial sub-sectors: for heavy polluters in almost for all industrial sub-sectors except for sub-sector rubber and plastic (22); for medium polluters, in industrial sub-sectors (14), (16), (26), and (12); and for low polluters, in almost for all industrial subsectors except sub-sector (21).
Another important finding concerns the differences in industrial and energy-mix structures, which were the decisive factors behind changes in CO 2 emissions although still very small. In addition to energy intensity, changes in CO 2 emissions were also influenced by industrial structure and energy structure effects. A decrease in CO 2 emissions in consequence of the effect of industrial structure occurred in these sub-sectors: for heavy polluters, in sub-sectors (13), (20), (22), (17) and (24); for medium polluters, in sub-sectors (25), (26), (30), and (12); and for low polluters, in sub-sectors (21) and (32). A decrease in CO 2 emissions in consequence of the energy structure occurred in these sub-sectors: for heavy polluters, in sub-sectors (23), (20), (17), (24), and (27); for medium polluters, in sub-sector (25), (26), and (30); and for low polluters, there is no reduction in CO 2 emissions.
Based on firm characteristics: By the characteristic group of technological intensity, the potential reduction in CO 2 emissions in consequence of the effect of energy intensity occurred in firms with high technology intensity (H), medium-high technology intensity (M-H), and medium-low technology intensity (M-L); the effects of the industrial structure occurred in firms with both high (H) and medium (M) technology intensities; the effect of the energy structure occurred in firms with high technology intensity (H), medium technology intensity (M) and medium-low technology intensity (M-L). By the characteristic group of firm size, the potential reduction in CO 2 emissions in consequence of the effect of energy intensity occurred across all sized firm groups; the effect of the industrial structure occurred in firms with 20-99 and 200-499 laborers; the effect of energy structure occurred in firms with 200-499 and >1000 laborers. By the characteristic group of capital ownership, the potential reduction in CO 2 emissions in consequence of the effect of energy intensity occurred across all ownership type firm groups; the effects of the industrial structure occurred in firms owned by PMA and PMDN; the effect of the energy structure in the PMA firm. There is no potential reduction in CO 2 emissions because of the energy structure in PMDN firms. By the characteristic group of island location, the potential reduction in CO 2 emissions in consequence of the effect of energy intensity occurred in all locations except in Kalimantan; the effects of the industrial structure occurred in Sumatra, the effects of energy structure occurred in firms in Java-Bali and Sumatra locations. By the characteristic group of export capability, the potential reduction in CO 2 emissions in consequence of industrial structure and energy intensity occurred in export-oriented and non-exporters; the effect of the energy structure occurred in export-oriented firms.

Policy implications
Reducing CO 2 emissions is important for developing sustainable manufacturing industries, so the Indonesian government must outline several CO 2 emissions mitigation strategies for improving environmental quality. Exploring the identification of various significant firm characteristics will enable policymakers to promote energy conservation and diversification in improving environmental quality by focusing only on a particular group of firm characteristics. The different firm characteristics will help policymakers concentrate energy and environmental policy only on certain groups and only on the most energy-intensive and inefficient firms in the future.
Based on the results obtained in this study, the practical implications that can be recommended are as follows. First, the government should provide incentives or rewards to firms that have the potential to reduce CO 2 emissions for their energy efficiency and energy diversification towards environmentally friendly energy, that are firms in manufacturing industry sub-sectors of electrical equipment (27), and computers, electronics (26); high technology intensity firms; firms with >1000 laborers; those owned by foreign investment firms; in Sumatra region; and export-oriented firms. This incentive or reward is expected to encourage the behavior of other groups of firms to follow in the footsteps of the non-polluting group of firms. Second, the government should provide subsidies and increase the capitalization to firms that tend to become heavy polluters, namely medium technology intensity firms; small firms (firms with 20-99 and 100-199 laborers); domestic investment firms (PMDN); in Kalimantan and Sulawesi regions; and not exported firms so that they are willing to use fossil fuels high quality (low emission). Third, the government should immediately impose carbon taxes or high taxes on non-environmentally friendly fuels so that firms try to behave according to the pattern of energy conversion from nonenvironmentally friendly fuels to environmentally friendly fuels (gas or renewable energy). Promote low-carbon energy sources (gas or renewable energy) in the energy mix structure to firms that tend to be heavy polluters-namely, sub-sectors non-metallic minerals (23), food (10), textile (13), chemicals (20), rubber and plastics (22), paper (17), and basic metals (24); firms with medium-high technology intensity, firms with 20-199 and 500-999 laborers, domestic-owned firms (PMDN), firms located outside the Java-Bali and Sumatera regions and non-exported firms-to encourage the use of high-quality fuel, which will promote emission reductions. Fourth, the government should implement technological changes and revitalize the inefficient machines in firms belonging to sub-sector rubber and plastic (22), motor vehicles and trailers (29), metal goods (25), other transport equipment (30), leather and footwear (15), and pharmaceuticals (21); firms have medium technology intensity and firms are located in Kalimantan region. Fifth, the government should promote a shift in the industrial economic structure to industries with less-intensive energy use, especially in the most energy-intensive industry manufacturing sub-sectors of glass and non-metal mineral products (23), food (10), electrical equipment (27), motor vehicles and trailers (29), garment (14) and wood, bamboo (16); medium-high and medium-low technology intensity firms; firms with 100-199 and 500-999 laborers; in all region except for Sumatera. Finally, the policymakers and stakeholders must have a strong commitment to immediately implement policy implications related to achieving climate change mitigation targets especially CO 2 emissions in the manufacturing industry sector. Management commitment is also needed to improve the firm's energy efficiency performance by transformation using the latest technology that is relatively more energy efficient. The transformation is can save costs and at the same time improve environmental quality. 4