Residential energy consumption in urban China: A decomposition analysis

Residential energy consumption (REC) is the second largest energy use category (10%) in China and urban residents account for 63% of the REC. Understanding the underlying drivers of variations of urban REC thus helps to identify challenges and opportunities and provide advices for future policy measures. This paper applies the LMDI method to a decomposition of China's urban REC during the period of 1998–2007 at disaggregated product/activity level using data collected from a wide range of sources. Our results have shown an extensive structure change towards a more energy-intensive household consumption structure as well as an intensive structure change towards high-quality and cleaner energy such as electricity, oil, and natural gas, which reflects a changing lifestyle and consumption mode in pursuit of a higher level of comfort, convenience and environmental protection. We have also found that China's price reforms in the energy sector have contributed to a reduction of REC while scale factors including increased urban population and income levels have played a key role in the rapid growth of REC. We suggest that further deregulation in energy prices and regulatory as well as voluntary energy efficiency and conservation policies in the residential sector should be promoted.

Residential energy consumption (REC) is the second largest energy use category (10%) in China although it lags behind the industry by far (Fig. 1, Fig. 2).1 Urban residents account for most of the REC (63%).2 With urban population expected to grow by 20 million per year and residential building area increasing by 2 billion square meters every year through 2020, residential energy consumption (REC) is likely to continue its rapid growth (Zhou et al., 2009). Other factors may also contribute to a rapid growth of REC, including elevated income levels, changing consumer preferences, as well as penetration of electric appliances and private transportation vehicles. On the other hand, continued energy price reforms, energy efficiency policies and energy conservation awareness may help to restrain further REC growth.

1. Download full-size image

Fig. 1Data source: China Energy Statistical Yearbook (2008).

1. Download full-size image

Fig. 2Data source: CSYs (1999–2008), SCE—standard coal equivalent.

Understanding the underlying drivers of variations of China's REC thus helps to identify challenges and opportunities and sheds some light on China's future energy policy. Residential building energy consumption contributes significantly to total REC. Some studies (Ouyang et al., 2009, Chen et al., 2008, Yoshino et al., 2006) explored the evaluation or saving of energy consumption in residential building sector. Meanwhile, the transitions of fuel types, electricity-using appliances, private vehicle ownership, and house heating also exert important influences on changes in REC (Sathaye and Tyler, 1991, Glicksman et al., 2001, Taylor et al., 2001, Brockett et al., 2002, Riley, 2002, Deng, 2007, Ni, 2008, Zhou et al., 2009). The other factor affecting REC that has attracted much attention is the energy efficiency. Some studies have illustrated the energy efficiency indicators in residential sector (Haas, 1997), the impact of refrigerator efficiency standards on the environment, manufacturers and consumers (Tao and Yu, 2011), and the efficiency improvement of residential building energy consumption (Yu et al., 2009, Li, 2009, Zhong et al., 2009, Morrissey and Horne, 2011, Zhao et al., 2009). Our study differs in that we provide a comprehensive analysis of drivers behind China's urban REC at disaggregated product/activity level over the period of 1998–2007. In this study, we look at the major energy-using products in the following four categories: (1) private transportation; (2) electrical appliances; (3) central heating; and (4) activities using coal, natural gas, coal gas and liquefied petroleum gas (LPG).

The rest of the paper proceeds as follows. Section 2 reviews previous studies on REC. Section 3 discusses our selection of index decomposition method and applies the selected method to a REC model. Section 4 discusses the data. Section 5 presents our findings. Section 6 concludes the paper with some policy suggestions.

2. Literature review

There is an extensive literature on China's industrial energy consumption and total energy consumption. Besides the sheer importance of the industrial sector, the quality of the statistical data also contributes to the large number of studies in this area. Consistent and detailed data on residential energy consumption, however, is less prevalent. The existing studies are primarily case studies that rely on surveys. Much of the existing studies focus one of the following four themes: improvement of residential building energy efficiency, change of REC structure, factors affecting the ownership and use of private vehicles, and impacts of household characteristics on energy saving and emission based on direct and indirect energy consumption.

2.1. Improvement of residential building energy efficiency

Studies have explored housing energy efficiency from different perspective. First, promoting retrofit is a key factor for the improvement of residential building energy efficiency. Li (2009)explored some renovation measures. Zhao et al. (2009) put forward the technical scheme of heat metering and energy efficiency retrofit of existing residential buildings in northern heating areas of China. Second, implementing incentive policy is also important. Zhong et al. (2009) showed that the performance-based incentives were more important. Third, residential building envelope should be efficient. Yu et al. (2009) illustrated that envelope shading and exterior wall thermal insulation were the best strategies to decrease the air conditioner electric consumption, and there was a large potential to decrease the energy consumption of the existing strip residential building by reconstructing envelope. Fourth, energy efficiency standards are also a crucial factor. Glicksman et al. (2001) and Taylor et al. (2001) have pointed out low energy efficiency standards are a major reason for escalating energy demand for house heating; increased total floor area and growing income have both contributed to the growing energy consumption for residential house heating.

2.2. Change of REC structure

Sathaye and Tyler (1991) looked at the transitions of REC from traditional biomass to commercial fuel types. Asia's households have witnessed dramatic increases in the use of commercial fuels in recent years. Most strikingly, electricity has come to penetrate most households. The results of the household surveys conducted in China, India, the Philippines, Thailand, and Hong Kong suggest that changing patterns of activity and livelihood in these households underlie this growth of commercial fuel consumption. Zhang (2004) illustrated also that the direct consumption of coal in China's residential sector had been decreasing while the consumption of electricity and gases (natural gas, coal gas and LPG) had been growing since 1990s. Meanwhile the annual energy consumption per household in China on the secondary energy basis tended to increase, while that on the primary energy basis kept on decreasing.

Meanwhile, the electricity consuming products have showed new trends. Brockett et al. (2002) conducted a survey on electricity consuming products in households in five Chinese cities. They found that the wealthier cities have a higher rate of electric appliances ownership illustrating an income or expenditure effect, while land-rich but poorer areas had larger households, affecting total lighting consumption. The small size of households combined with average low income of a developing country had constrained adoption of appliances such as dishwashers and clothes dryers which were both common in developed countries. The survey highlighted the commonalities of Chinese food structure by the widespread ownership of rice cookers and the predominance of direct gas flame for cooking. It also identified the trend away from coal towards electricity and gas fuel had taken place in all cities driven by the structural change of energy-using activity.

2.3. Factors affecting the ownership and use of private vehicles

Riley (2002), Deng (2007) and Ni (2008) have studied the impact of demographic and economic changes as well as policy differences on ownership and use of private vehicles. Riley (2002) illustrated that rapid economic development, and the relaxation of policies designed to restrict private vehicle ownership seemed to be the primary factors influencing motor vehicle growth in China. Deng (2007) pointed out that income effect was strong at both national level and within regions. Meanwhile, changes and fees imposed on private car owners by authorities at different levels might have a strong influence on car ownership level. Ni (2008) made a survey in Shanghai, China, and argued that in terms of the purchase and use behavior, variables such as gender, perceptions of different aspects of the utility of different travel means, as well as personal or household income were significant.

2.4. Impacts of household characteristics on energy saving and emission based on direct and indirect energy consumption

Golley et al. (2008) conduct a cross-sectional study on China's urban direct and indirect energy consumption based on a household survey. Their focus is to identify impacts of household characteristics on energy consumption and associated emissions and they find that while richer households do indeed emit more per capita, poorer households tend to be more emission intensive—that is, generating higher emissions per Yuan spent.

Wei et al. (2007) and Golley et al. (2008) also consider indirect energy consumption defined in terms of the energy inputs needed in the production of goods and services consumed ultimately by households. Based on the application of a Consumer Lifestyle Approach, Wei et al. (2007) quantify the direct and indirect energy use and related carbon emissions during the period of 1999–2002 and find that residents' lifestyle can have an important and significant impact on energy use and related carbon emissions.

In sum, the above studies mainly have explored specific characteristics of China's REC. Our study differs in that we provide a comprehensive analysis of drivers of China's REC growth by looking the major energy-using products in the four categories: (1) private transportation; (2) electrical appliances; (3) central heating; and (4) activities using coal, natural gas, coal gas and liquefied petroleum gas (LPG).

At the same time, our study focuses on the direct energy consumption at disaggregated product/activity level. We provide a time-series analysis of China's urban REC during the period 1998–2007 based on bottom-up modeling, thus sheds new light on how urban REC structure has changed over time and what driving forces underlies the change. Similarly, Zhou et al. (2009) project China's REC up to 2020 based on a single-year decomposition of REC on electrical appliances and house heating in 2000, and assumptions about future development in population, urbanization, residential living area, household size, appliance penetration, as well as end-use energy efficiency. We provide a time-series analysis of urban REC based on 17 energy-using products/activity.

3. Method

Two broad categories of decomposition techniques, input–output techniques – structural decomposition analysis (SDA) and disaggregation techniques – index decomposition analysis (IDA), have been widely used to identify the magnitude of some predetermined driving factors of changes in observed energy indicators. Both approaches have several variants. Hoekstra and Van den Bergh (2003) investigate the different types of indexes that are used in IDA and SDA. The SDA approach is based on input–output coefficients and final demands from input–output tables while the IDA framework uses aggregate input and output data that are typically at a higher level of aggregation than input–output tables. This basic difference also determines the advantages and disadvantages of the two methods. One advantage of SDA is that the input–output model includes indirect demand effects – demand for inputs from supplying sectors that can be attributed to the downstream sector's demand – so that SDA can differentiate between direct and indirect energy demands. Chai et al. (2009)show that impact of indirect energy demand can be substantial. Rose (1999), Hoekstra and van den Bergh (2003) and Hoekstra (2005) provide overview of the Environmental SDA literature.

IDA is only capable of capturing impacts of direct energy demands; however, we consider the IDA approach sufficient for a study of residential energy consumption in which case the direct energy use is the area that consumers have the most power to act. The IDA approach has gained some favorable momentum in recent years partly due to readily available data and flexibility of application to disaggregation at different levels, which eases empirical applications and comparisons. Empirical studies using IDA have been conducted at different disaggregation levels and covered a wide range of countries and country groups in North America, Europe and Asian regions. Ang and Zhang (2000) and Liu and Ang (2007) present excellent accounts of empirical IDA applications. Among the variants of IDA approaches including Laspeyres index decomposition approaches and Divisia index decomposition approaches, several scholars (Ang and Zhang, 2000, Ang and Liu, 2001, Ang, 2004) have argued that the logarithmic mean Divisia index (LMDI) approach is a preferred method with the advantages of path independency, ability to handle zero values and consistency in aggregation. Details about the LMDI approach can be found in Ang and Liu (2001).3

The index decomposition technique has been widely used in the areas of energy economics and environmental economics. Variants of the IDA approach have been applied to a large number of analyses of industrial energy consumption, total energy consumption and energy-related emissions, etc. (Ang and Zhang, 2000, Liu and Ang, 2007), but it is rarely applied to REC which is largely due to data scarcity. Achão and Schaeffer (2009) applied the LMDI method in its additive form4 to residential electricity consumption in Brazil by disaggregating residential consumers by consumption classes and regions of the country and decomposing electricity consumption by four electrical appliances: television, refrigerator, freezer, and washing machine. In this study, we also use the LMDI approach in its additive form to explore the driving factors behind the rapid growth of China's urban REC, however, with a more comprehensive dataset, we are able to study 17 energy-using products in four broad categories which constitute a complete residential energy consumption structure.

The total annual urban REC is factored into the following expression:

(1)$E=\sum _j\sum _i\frac{E_{ij}}{Y_{ij}}\frac{Y_{ij}}{Y}\frac{Y}{L}\frac{L}{P}P=\sum _j\sum _{i}PRS1S2EPPO$

where E is the total annual urban REC in China, Eij is the urban REC due to ith energy-using product in jth category, Yij is the energy expenditure on Eij (in constant prices), Y is the total energy expenditure on urban REC, L is the total urban living expenditure, P is the total urban population, j is the four categories: private transportation; electrical appliances; central heating; and other energy-using activities; i is the 17 energy-using products: private vehicle, motorcycle, ten electrical appliances, central heating, activities using coal, natural gas, coal gas, and LPG, respectively.

For ease of presentation, we introduce five intermediate terms –PR, S1, S2, EP and PO – to represent the five terms in the first half of Eq. (1), respectively.

Applying LMDI in its additive form, the change in total urban REC between any two years (tand t−1) –ΔEtot– is decomposed as follows:

(2)$\mathrm{\Delta }{E}_{tot}=E_t-E_{t-1}=\mathrm{\Delta }{E}_{PR}+\mathrm{\Delta }{E}_{S1}+\mathrm{\Delta }{E}_{S2}+\mathrm{\Delta }{E}_{EP}+\mathrm{\Delta }{E}_{PO}$

where Et, Et−1 is REC in Year t and t−1 respectively,

(3)$\mathrm{\Delta }{E}_{PR}=\sum _j\sum _i{w}_{ijt}\ln \left(\frac{P{R}_{ij,t}}{P{R}_{ij,t-1}}\right)$

$\mathrm{\Delta }{E}_{S1}=\sum _j\sum _i{w}_{ijt}\ln \left(\frac{S{1}_{ij,t}}{S{1}_{ij,t-1}}\right)$$\mathrm{\Delta }{E}_{S2}=\sum _j\sum _i{w}_{ijt}\ln \left(\frac{S{2}_t}{S{2}_{t-1}}\right)$$\mathrm{\Delta }{E}_{EP}=\sum _j\sum _i{w}_{ijt}\ln \left(\frac{E{P}_t}{E{P}_{t-1}}\right)$$\mathrm{\Delta }{E}_{PO}=\sum _j\sum _i{w}_{ijt}\ln \left(\frac{P{O}_t}{P{O}_{t-1}}\right)$

where ΔEPR, ΔES1, ΔES2, ΔEEP, and ΔEPO are called price effect, structure effect 1 (intensive structure effect), structure effect 2 (extensive structure effect), expenditure effect, and population effect, which represent the REC change due to energy price changes, change of energy expenditure share on household product i in total REC expenditure, change of REC expenditure share in total living expenditure, change of per capita living expenditures and change of population size, respectively.

And wij is the logarithmic weighting scheme, specified in the following:

(8)$w_{ijt}=L(E_{ij,t},E_{ij,t-1})=\frac{(E_{ij,t}-E_{ij,t-1})}{\ln (E_{ij,t}/E_{ij,t-1})}$

Hence, Eq. (2) can be expressed as follows:

(9)$\mathrm{\Delta }{E}_{tot}=E_t-E_{t-1}=\sum _j\sum _i\frac{E_{ij,t}-E_{ij,t-1}}{\ln E_{ij,t}-\ln E_{ij,t-1}}\ln \left(\frac{P{R}_{i,t}}{P{R}_{i,t-1}}\right)+\sum _j\sum _i\frac{E_{ij,t}-E_{ij,t-1}}{\ln E_{ij,t}-\ln E_{ij,t-1}}\ln \left(\frac{S{1}_{i,t}}{S{1}_{i,t-1}}\right)+\sum _j\sum _i\frac{E_{ij,t}-E_{ij,t-1}}{\ln E_{ij,t}-\ln E_{ij,t-1}}\ln \left(\frac{S{2}_t}{S{2}_{t-1}}\right)+\sum _j\sum _i\frac{E_{ij,t}-E_{ij,t-1}}{\ln E_{ij,t}-\ln E_{ij,t-1}}\ln \left(\frac{E{P}_t}{E{P}_{t-1}}\right)+\sum _j\sum _i\frac{E_{ij,t}-E_{ij,t-1}}{\ln E_{ij,t}-\ln E_{ij,t-1}}\ln \left(\frac{P{O}_t}{P{O}_{t-1}}\right)$

4. Data—urban REC and expenditures

Research on China's REC has suffered from the lack of consistent and detailed data. Previous studies have collected data through surveys (Sathaye and Tyler, 1991, Brockett et al., 2002, Ni, 2008). Our study takes a bottom-up approach to modeling China's urban REC using data from a wide variety of sources. In this study, we only examine urban residential sector for the following reasons: (1) data on quantities and prices of various commercial energy consumption are relatively complete for the urban residential sector but incomplete for the rural areas; (2) non-commercial energy such as biomass still accounts for a substantial proportion of rural residential energy consumption; however, statistics is rather incomplete and subject to substantial measurement error. For the REC in urban areas we consider oil consumption for private transportation, electricity consumption for electrical appliances, directly purchased heat for central heating, and, directly purchased coal and gases for other energy-using activities (mainly cooking and individual heating). This section describes modeling of energy consumption and expenditures for each of these elements. Data sources and some modeling results are summarized in Appendix A Data description and sources, Appendix B Energy consumption (Mtce) and expenditure (RMB 100 million). Oil, electricity, coal, natural gas, coal gas and LPG are all converted to million tons of standard coal equivalents (Mtce) where aggregation is needed and conversion factors are provided in Appendix C. All expenditures are in 1998 constant prices.

4.1. Private transportation

Energy consumption and expenditure on private transportation are modeled as follows:

(10)$E_{T_{}i}=N_{T_{}i}⁎e_{T_{}i}⁎M_{T_{}i}$

$Y_{Ti}=p_{Ti}⁎E_{Ti}$

where ETi, NTi, eTi, MTi, YTi and pTi represent annual energy consumption of urban private transportation i (i=private vehicles, motorcycles), annual number of urban private vehicles and motorcycles, fuel efficiency of private vehicles and motorcycles (L/100 km), annual average mileage of private vehicles and motorcycles, annual expenditure on REC of urban private transportation i and annual average retail gasoline price, respectively.

4.2. Electrical appliances

Urban energy consumption and expenditure on electrical appliances are modeled as follows:

(12)$E_{Ai}=N_{Ai}⁎C_{Ai}⁎T_{Ai}$

$Y_{Ai}=p_e⁎E_{Ai}$

where EAi, NAi, CAi, TAi, YAi and pe represent annual energy consumption of urban residential electrical appliance i (i=air-conditioner, washer, refrigerator, electric cooking, color TV, monochrome TV, home computer, water heater, range hood, microwave oven, electric fan), annual number of urban electrical appliance i, annual average capacity of electrical appliance i, annual average using time of appliance i, urban expenditure on energy consumption of electrical appliance i and annual average retail electricity price, respectively.

4.3. Central heating

Urban energy consumption and expenditure on central heating are modeled as follows:

(14)$E_c=A_c⁎e_c$

$Y_c=A_c⁎p_c$

where Ec, Ac, ec, Yc and pc represent urban energy consumed for central heating, total central heating area in urban areas, annual energy consumption per unit central heating area, expenditure on urban central heating and urban central heating price (per unit area), respectively.

4.4. Other energy-using activities

Although central heating has penetrated rapidly in urban areas in recent years, individual heating systems are still widely used where central heating is unavailable or inadequate especially in southern provinces where central heating is not officially provided. On the other hand, cooking has not been completely electrified. Coal products and gases are still widely used for heating and cooking purposes. Energy-using activities not included in 4.1 Private transportation, 4.2 Electrical appliances, 4.3 Central heating are modeled by fuels. The expenditure on the ith fuel – Yoi – is modeled as follows:

(16)$Y_{oi}=p_{oi}⁎E_{oi}$

where poi and Eoi are the price of fuel i (i=coal, natural gas, coal gas, and LPG) and urban residential consumption of fuel i, respectively.

4.5. Lighting

Data on urban residential lighting energy consumption is very limited. We are thus forced to leave out lighting in our primary analysis. To examine the extent to which our analysis is robust to the omission of lighting, we conduct a sensitivity analysis in the next section using available urban lighting data during 2002–2004. The expenditure Yl on urban residential lighting energy consumption El is calculated by

(17)$Y_l=E_l⁎p_e$

where pe is annual average retail electricity price.

5. Results and discussion

5.1. Results

5.1.1. Energy price effect promotes REC decrease

Table 1 and Fig. 3 present the complete decomposition results. As expected, the price effect (ΔEPR) is mainly negative. The Chinese government has launched a series of reforms in the energy sector in an effort to improve the economic efficiency. One essential element is the price deregulation which has resulted in an overall increase in energy prices (Ma and He, 2008, Zhao et al., 2010). Our results illustrate that rising energy prices have had a dampening impact on the urban REC. The total urban REC would have increased much more rapidly without recent deregulation reforms in energy prices.

If you are interested in this research article, you can download the full length article through the download page using the link below, both PDF file and Word file are provided.

https://www.sciencedirect.com/search/advanced?tak=LMDI&show=25&sortBy=relevance