Energy eﬃciency in Indonesia’s manufacturing industry: a perspective from Log Mean Divisia Index decomposition analysis

This research discusses energy intensity in Indonesia's manufacturing sector from 1980 to 2015. The manufacturing sector is the second-largest energy-consumer in Indonesia (after the transportation sector) and one of the largest contributors to Indonesia’s output. Thus, it is important to know the energy usage performance of this sector. This study discusses the factors affecting changes in energy consumption in various subsectors of Indonesia's industry and investigates the energy intensity across manufacturing subsectors. This paper analyses the specific characteristics of energy intensity in the manufacturing sector in Indonesia from 1980 to 2015. This has not been investigated a great deal in the past, particularly when employing the Log Mean Divisia Index II method. The overall energy intensity of Indonesia’s manufacturing sectors has seen a strong and continuous decline, with a reduction of 65% over the 35 yr, reinforced by some limited changes in industry structure towards lower intensity. Over the entire period, this reduction was dominated by increases in energy efficiency within industries, as indicated by a 62% fall in the within-industry intensive index. By contrast, the effect of moving to a less intensive industry structure was much less important (a 9% fall in the structural index). The greatest rise in energy efficiency within the industry happened before the financial crisis (from 1980 to 97). the shock of the financial crisis saw an unexpected reaction when value-added fell by 13% but energy use remained largely unchanged, implying a rise in energy intensity. From 2000 to 2015 the earlier trends resumed, but at a more subdued pace, where over this period aggregate intensity fell by 23%.


Introduction
Energy demand and greenhouse gas emissions in Indonesia have increased rapidly in recent years [1]. The energy consumption in the manufacturing sector accounts for around 40% of total final energy consumption in Indonesia in 2016, which is also the second-highest energy-consuming sector after the transportation sector. Based on International Energy Agency (IEA) [2], energy consumption growth in Indonesia has not coincided with a reduction of energy intensity, this is mainly due to efficiency improvements from new investment in the industrial sector and structural shifts in the economy.
The manufacturing sector is one of the most vital sectors in Indonesia as it contributes significantly to national energy consumption and output (Gross Domestic Product/GDP).
According to the Indonesian Statistic Bureau (ISB), since the 1990s the manufacturing sector has become the largest contributor to Indonesia's energy consumption and national output. The manufacturing sector has accounted for around 27% of Indonesia's GDP in 2016 [3]. In the last 5 yr, the value-added from this sector has grown averagely at a rate of 5% per year Along with the manufacturing output growth, the energy consumption also rose in the previous 5 yr from 3 to 4% growth annually. Interestingly, the rate of growth in energy consumption is slightly lower than the rate of growth in value-added. Based on this estimate, it can be said that the energy efficiency of the manufacturing sector has improved over time. Hence, looking at this trend, it is essential to examine and identify which sectors have the greatest potential to improve energy intensity in the manufacturing sector.
This study aims to examine the driving forces affecting changes in aggregate energy intensity in various sectors of Indonesia's manufacturing industry and investigate the energy intensity and value-added performance across manufacturing sectors. This study analyses the characteristics of energy intensity in the manufacturing sector in Indonesia from 1980 to 2015. This area has not had a great deal of research in the past of Indonesia's literature particularly by employing the Log Mean Divisia Index-II (LMDI-II) method. The analysis of industrial energy consumption has been conducted for several countries in the international context, but it is found insufficient in the Indonesian case. This study aims at filling this gap by focussing on identifying the factors affecting changes in energy intensity resulting from decomposition.
By conducting the decomposition method across Indonesia's manufacturing sector, this will provide further insight into the energy intensity performances. This study provides valuable evidence and information which adds to the knowledge/research into the study of energy use in Indonesia. The findings will help subsequent researchers and policymakers to understand what forces drive energy intensity, so this can inform policy aimed at reducing energy consumption in Indonesia in the manufacturing sector.

Literature review
Many studies have been conducted to investigate the relationship between energy consumption, economic growth and environmental aspects in Indonesia [4][5][6]. However, there are very few studies employing the decomposition method in Indonesia that measure the factors affecting energy intensity [7,8] and environmental analysis [9]. Previous research generally focused on the manufacturing sector's energy consumption, while this study extends the investigation by adding subsectoral data to provide further insight into the energy system.
The first attempt to investigate the factors affecting energy consumption in Indonesian manufacturing was conducted by Sitompul [9] by using three-digit level data disaggregation (9 sectors and 30 subsectors) from 1980 to 2000, and applied decomposition analyses developed by Sun [10] to separate the energy intensity into technical effects and fuel mix effect. The result shows that the major contributor to the energy intensity changes in Indonesia is a technical effect.
Hartono et al. [7] decomposed the changes in energy intensity in the manufacturing sector into activity and efficiency effects from 2002 to 2006. In the analysis, they partitioned the industry into nine subsectors and divided the industry into two types: medium enterprises (firms that have less than 100 employees) and large enterprises (firms that have more than 100 employees). They found that the level of energy intensity in each industry level varies across subsectors and the changes in energy intensity in the national level are determined by large enterprises. He also found that the level of energy intensity is influenced by capital intensity, wages and capital share.
Vivadinar et al. [8] measured the factors affecting energy intensity and consumption in the manufacturing sector, focusing on high energy consumption industries: steel, pulp and paper, cement and glass, chemical and non-metallic industries. They applied the decomposition method employing annual industry data from 2001 to 2007. They found that the changes in the energy demand and intensity were a result of the technology factor, whereas the role of production output was relatively small. However, the structural effect significantly affected the energy intensity in the glass and pulp industries.
Ramstetter and Narjoko [11] investigated whether multinational corporations were more energy-efficient than state-owned enterprises in Indonesian industries, focussing on medium to large manufacturers over the period 1996 to 2006. By using a translog production function model, they found that the relationship between energy intensities and ownership amongst Indonesia's manufacturing sectors was relatively weak. Another recent study by Setyawan [12] discovered that the aggregate energy intensity in the Association of South East Asian Nations countries was decreasing. By benchmarking the economy-wide energy intensity performances of Singapore, Malaysia, Vietnam, the Philippines, Thailand and Indonesia, he found that all these countries showed a shift in industry value added to more energy-intensive industries.

Material and methods
This study employs the LMDI-II in multiplicative form, following the model developed by Ang [13,14] This study will refer to decomposition analysis described below, where Indonesia's energy intensity includes two factors: the intensity effect of each subsector in manufacturing (Dint) and the structural effect of manufacturing (Dstr). The decomposition method is computed as follows: Where denotes total energy intensity change in year t, relative to the reference year; denotes changes in aggregate energy intensity due to changes in each subsector energy intensity; denotes changes in aggregate energy intensity due to changes in the structure of the economy; denotes the ratio of output of subsector i to the aggregate output; wi,t is a weight function with factor i in time t; wi,o is a weight function with factor i in time 0; L is the logarithmic average of two positive numbers.
This study divides the driving forces affecting energy intensity into two factors: (1) whether an increase in energy consumption in Indonesia is related to the shift to more or less energy-intensive industries (between sector changes/structural effect); and (2) whether it is a result of the improvement or deterioration of energy efficiency (changes within sector energy intensity/intensity effect). Additionally, it will also determine which sectors have the most improved energy consumption, output and energy intensity. By employing energy intensity as an indicator of energy efficiency, this study will investigate whether there was an improvement in energy efficiency in Indonesia's manufacturing sector during the period 1980 to 2015.

Description of data sources
The primary database used in this study is the Medium and Large Manufacturing Firms • Industries refer to three-digit categories. In this study, the data were disaggregated into the two-and three-digit level.
For reasons of compatibility and consistency, this study follows the ISIC Revision 2 As mentioned earlier, this study utilized a long data history analysis, therefore for compatibility and consistency reasons, this study follows the ISIC Revision 2 for the data analysis. The ISIC-2 Indonesia's economic activities used in this study is shown in Table 1.
The classification of ISIC revision 2 is as follows: The manufacturing sector in this study is limited to the medium and large the scale manufacturing sector, which accounted for nearly 90% of aggregate manufacturing valueadded. This study is limited to the manufacturing sector as this sector has comprehensive and detailed data (up to the subsector level disaggregation) that makes it easier to elaborate on the effects of industrial restructuring. Data disaggregation is important in exploring the detail necessary to study the effect of changes in the structure and intensity within subsectors. It is also essential to identify energy-intensive industries among the manufacturing subsector.

Procedures in constructing the dataset
The ISB manufacturing dataset has been considered by many researchers as one of the best datasets that provides a long period dataset in manufacturing sector statistics [15][16][17].
Notwithstanding this, the ISB dataset has several drawbacks which need some adjustments to produce a valid and consistent data. Reliable empirical research results can only be obtained from a valid and consistent dataset [18].
Therefore, to construct a consistent dataset, several adjustments have been adopted in this study, as below: • Phase 1: Set definitions for each variable As the ISB changed the definition of each variable during the period of observation, this study verifies and compares each variable in the dataset (for the particular year of the study period) to ensure the consistency and validity of the variables. If there is an inconsistency in the variables and definitions, then this study redefines the inconsistent variables to get a consistent definition over the selected period of study. nil for other years then adjustments are made by correcting the value of 0% to 100%.
• Phase 3: Computing total output/value-added The ISIC code provided by the ISB has five digits. The data need to be aggregated into twodigit and three-digit industrial codes to make it comparable across the study period. Output in this study is measured as value-added of Indonesian Rupiah's (in million Rupiah at a constant 1980 price).
• Phase 4: Summation of input (energy consumption) Energy consumption in this study is the end-use energy consumption by the manufacturing sector. All data are in total fuel consumption, which is the sum of fuel used for manufacturing processes and power generation. The dataset provides different types of fuel used, which include ten types of energy primarily utilized in the manufacturing sector, including gasoline (in L), kerosene (in L), automotive diesel oil (in L), industrial diesel oil (in L), fuel oil (in L), liquid petroleum gas (in kilograms), coal (in km), coke (in km) and electricity (by kWh). These fuel types values were standardised and converted into a standard energy unit: TJ.

Result and discussion
This section describes the energy-economic characteristics of the manufacturing sector's energy consumption, the manufacturing sector's value added and the results of energy intensity decomposition analysis of the manufacturing sectors.

Energy consumption by type of the manufacturing sector
The manufacturing sector has a substantial share of energy use in Indonesia at approximately 40% of aggregate final energy consumption in 2015, and its consumption has increased by around 4% per year since the 2000s. Figure 1 shows the proportion of coal used in manufacturing increased significantly from around 2.5% in 1980 and to 35% in 2010, although there was a decrease after this period to 21% in 2015. This decrease in coal consumption was followed by an increase in gas and electricity use during the same period from 29 and 8% in 2010 to 36 and 12% in 2015, respectively. Oil was the prominent fuel source in the manufacturing since the 1980s, however, its share decreased from 40% in 1980 to 31% in 2015.
One potential reason behind the changes in the manufacturing energy mix is Indonesia's limited natural resources that lead to it restructuring its national energy mix policy. By revising Energy Plan [19], Indonesia's government has introduced several changes to its energy policy planning. The new regulation focuses on re-balancing energy mix to focus on indigenous energy supplies, which includes in reducing oil use, reducing the consumption of coal and renewable energy, and optimising the production and the use of gas. This regulation provides the platform to achieve an energy mix transformation by 2025 to include 30% coal, 25% natural gas, 23% renewable resources and 22% oil.

Energy consumption and value-added share (ISIC-2 digit)
The ISB data at the two-digit level ( Table 2) Table 3). The food industry is one of the most prominent contributors to aggregate valueadded in the manufacturing sector. This is potentially due to the government's policy aiming for this sector to achieve self-sufficiency. Fulfilling a sufficient food supply is an essential factor for economic development, hence the government's action in prioritizing the food industry by enacting certain protecting government regulation. Developing resource-based industries like food and beverages is a significant target for Indonesia, like many other developing countries in Asia [21].
After the Food sector, the second-largest value-added share in 2015 came from the share of the textile industry increased from 1980 to 1990 from 13.5% to 16.1%, due to the growing export at the beginning of the 1980s and the increasing demand of Indonesia's domestic market [22]. Moreover, Pangestu and Sato [23] listed several driving forces for the increased export opportunities, including the comparatively low labour cost, undervalued real exchange rate, under-utilized export quotas and various government incentives, for example, interest rate subsidies for credit exports and export subsidies. However, after it peaked at around 18% in 2000, the textile output share declined to around 11% in 2015. Dhanani [24] and Patunru and Rahardja [25] (14).
Aggregate energy intensity in the manufacturing sector has improved (decreased) in all key industry subsectors, even though the trends are not similar (see Table 4). Over the 35 yr of

Result of decomposition analysis on energy intensity (ISIC-2 digit)
The decomposition results (see Fig. 2) show that the main driving force reducing the aggregate energy intensity in the manufacturing sector is the intensity effect, while the structure effect only plays a small role in the changes of the aggregate energy intensity in Indonesia's manufacturing sector from 1980 to 2015.
From 1980 to 1990 and 1990 to 1997, the aggregate energy intensity in the manufacturing sector decreased by 40 and 34%, respectively. Most of the decrease in energy intensity in these periods were attributable to changes within industry energy intensity/intensity effect for around 35 and 36%, respectively. The intensity effect had the highest effect that caused the aggregate energy intensity in the manufacturing sector to decline (see Fig. 2).
Additionally, from 1980 to 1990 the role of Dstr contributed to decreasing the aggregate energy intensity by 8%, but from 1990 to 1997, the structural effect increased the aggregate energy intensity by 4% compared to the base period in 1990.
Throughout the financial crisis period from 1997 to 2000, the aggregate energy intensity increased to 13% compared to the base year of 1997. This increase occurred due to significant changes within industry energy intensity/intensity effect that increased the aggregate energy intensity by 13%. During this period, energy consumption was found to not change much, while the total value-added fell significantly. During this period the role of the structural effect was found to be negligible to the overall changes in the aggregate energy intensity. The increase of aggregate energy intensity in this period indicated the energy intensity became less efficient, which is a result of the decrease in overall value-added in the manufacturing sector. Second, this period is called as "the period during the financial crisis." During the crisis period of 1997 to 2000, changes within industry increased the aggregate energy intensity in the manufacturing sector. The increasing of energy intensity in this period is most likely due to falling of overall output in the key manufacturing sector. Tables 2 and 3 show that the annual growth of value-added is slower than the annual growth of energy consumption, which results in deterioration in aggregate energy intensity in the manufacturing sector.
A rapid industrialisation process took place in the manufacturing sector which had experienced quite significant structural changes throughout the study. Manufacturing sectors output grew significantly from 1980 to 1997. However, during the economic crisis from 1997 to 2000, the growth rate of all subsectors slowed down. Hill [27] concluded that the key factor of industrial transformation in Indonesia between the period of the 1970s to 1990s was due to its fast diversification. Additionally, Rodrik [28] also claims that the major dimension to economic development is product diversification. Ultimately, the structural shifts in Indonesia's manufacturing sector indicate the change from light industries with labour intensive to heavy industries with more capital and technology-intensive [29].
The historical trend of decomposition results reveals that the intensity effect reduction in the manufacturing sector had changed through different periods. This trend of decomposition demonstrates the consequences of the financial crisis on the changes of aggregate energy intensity which could limit the growth of certain manufacturing sectors from structural and intensity changes.

Conclusions
The main objective of this study was to investigate the energy efficiency performances that regulates all of the energy users greater than 6 kt of oil equivalent [30]. Additionally, the Indonesian government had also reformed its energy subsidy which significantly increased the energy prices specifically for the manufacturing sector. IEA [2] highlights the tendency of higher energy prices to foster industry efficiency. Indeed, all of these policy measures had encouraged manufacturers to enhance their competitiveness and minimize their energy intensity, including investing in new and efficient technologies to enhance its energy productivity.
The analysis also revealed that the greatest energy intensity sector is the Non-metal sector (particularly the cement and lime sector). This sector consumed more energy per valueadded and involves a large share of Indonesia's manufacturing energy use compared to other sectors. Furthermore, the decomposition results reveal that the intensity effect reduction in the manufacturing sector changes through different periods. The trend helps to understand the consequences of economic incidents which could limit the growth of some industries from structural change and energy use. However, besides these driving forces, some other factors might also have influenced the overall trends and it is hard to offer a single explanation.