Issuu on Google+

Energy Research & Social Science 2 (2014) 1–11

Contents lists available at ScienceDirect

Energy Research & Social Science journal homepage: www.elsevier.com/locate/erss

Original research article

Exploring the sensitivity of residential energy consumption in China: Implications from a micro-demographic analysis Chonghui Fu a , Wenjun Wang b,∗ , Jian Tang c a School of Humanities and Management, GuangDong Medical College, XinCheng Road 1#, DongGuan SongShan Lake High-Tech Zone„ Dongguan, GuangDong 523808, PR China b GuangZhou Institute of Energy Conversion, Chinese Academy of Science, Neng Yuan Road 2#, Tian He District, GuangZhou 510000, GuangDong, PR China c ShenZhen Godsdata Institute of Applied Statistics, Xin Zhou Road 4009#, Fu Tian District, ShenZhen 518000, GuangDong, PR China

a r t i c l e

i n f o

Article history: Received 23 November 2013 Received in revised form 11 April 2014 Accepted 11 April 2014 Keywords: Demographic sensitivity Residential energy consumption Scale effect

a b s t r a c t Energy consumption in the residential sector is one of the main parts of the total consumption in China, and demographic factors are the fundamental parameters affecting total energy use. Using residential energy consumption (REC) data from household surveys, demographic data from population censuses and macroenergy statistics, and the research assesses a theoretical model of the demographic sensitivity of REC in urban China. The method of population component is adopted to explore the demographic sensitivity on REC. Our research reveals different micro-demographic processes have different effects on REC, even when macro-demographic levels are identical or similar. Natural population change, urbanisation and aging are sensitive to REC. However, the population age structure is not sensitive to REC except for the 60+ age group. The scale effect plays a pivotal role in correlations between REC and demographic changes; decreasing per capita REC correlates with increasing family size. Because of the multiple sensitivities of population to REC, population size cannot be the exclusive demographic indicator with which to judge changes in REC. The effects of demographic structural factors surpass those of demographic quantitative factors. Finally, the findings of demographic sensitivity are used to simulate the scenarios of REC in 2015 under different assumed micro-demographic processes. © 2014 Elsevier Ltd. All rights reserved.

1. Introduction This study examines the demographic sensitivity of residential energy consumption (REC) and evaluates how and to what extent demographic factors cause significant changes in REC in urban China from the perspectives of micro-demographic processes. We argue that different micro-demographic processes may cause various scenarios of REC even when macro-demographic levels are identical or similar. We expect that the findings of this study will improve forecasting (and adapting to) changes in REC through the consideration of detailed demographic insights. What we talk about in the paper focuses on demographic processes which would change or influence individual and household interactions with urban residential energy system in China. REC in this paper refers to the energy directly consumed for human life in household, including cooking, heating, electricity, transportation

and traffic. The sensitivities of urbanisation, age-structure change and natural population change on REC are detected through the method of population component. However, the energy consumption structure in residential sector is excluded from the analytic frame, in which the amount of REC is included only. Global warming is one of the most important threats to the sustainable development of human beings. The primary cause of global warming is likely the emission of greenhouse gases as a result of human activities; this assertion is backed statistically at a confidence level greater than 90% [1]. According to data collected by the World Bank,1 China surpassed the U.S.A. in 2006 to become the most prolific carbon dioxide (CO2 )-emitting country in the world. A combination of energy conservation and emissions reduction has become the national energy strategy of China; however, each component of this strategy poses complex challenges to policymakers across domains, from the economic to the social. Although the level

∗ Corresponding author. Tel.: +86 1371460315. E-mail addresses: 461541231@qq.com (C. Fu), 13714603315@139.com (W. Wang), 1907094809@qq.com (J. Tang).

1 http://data.worldbank.org/indicator/EN.ATM.CO2(e).KT/countries/1W? display=default,2012-2-3.

http://dx.doi.org/10.1016/j.erss.2014.04.010 2214-6296/© 2014 Elsevier Ltd. All rights reserved.


2

C. Fu et al. / Energy Research & Social Science 2 (2014) 1–11

of REC in China accounts for only approximately 10% of total energy consumption [2], the huge population size and rapid demographic shift implies considerable growth in REC in the future as the population urbanises and lifestyles shift, as is occurring already. Energy consumption is affected by a wider range of factors, such as levels of income, lifestyle, technology and other factors, but population is a fundamental parameter affecting total energy use. Whether micro-demographic changes alone (i.e., accompanied by little or no macro-demographic change) can result in different levels of REC is an important scientific issue in the domain of energy consumption research. Therefore, the composition of the population and micro-demographic changes that may induce changes in population structure should be studied scientifically to ensure that the understanding of REC is based on appropriate demographic principles in addition to considering population size in such an analysis. The present study is organised as follows. Section 2 reviews the extant literature on the relationship between population and energy consumption and/or CO2 emissions. Section 3 provides definitions for demographic sensitivity and the scale effect of REC and proposes a theoretical model that bridges the gap between micro-demographic processes and the macro-scale results of REC. Section 4 presents an empirical analysis that includes a description of the data, scale effects, demographic sensitivity and scenario analysis. Section 5 discusses the findings, offers suggestions for future research and concludes the study.

2. Literature review Numerous studies have suggested that population plays a pivotal role in energy consumption, carbon emissions and/or climate change, although the specific mechanism through which population acts remains unclear. Almost all studies of energy consumption, CO2 emissions and climate change utilise demographic factors as important parameters. Therefore, this section focuses on studies investigating population and REC; all analytical methodologies examining the relationship between population and energy consumption and/or CO2 emissions will be taken into account. Energy consumption in the residential sector is one of the main parts of the total consumption in China [3]. REC is the second largest energy use category (10%) in China and urban residents account for 63% of the REC. Although the increasing of productive energy consumption would slow down, or even the amount tend to be stable, the total energy consumption will tend to grow significantly thanking for the fast increase of residential energy consumption [4]. Scale factors including increased urban population and income levels have played a key role in the rapid growth of REC [5]. In order to promote energy conservation in the residential sector, and to predict the CO2 emission, it is important to examine the residential energy consumption so that policy makers and energy experts in different countries can learn from each other in the policy-making of residential energy standards [3]. The basic theoretical framework of the relevant studies is derived from the Impact of Population, Affluence and Technology (I-PAT) model [6] and its extended model [7], in which methods of factor decomposition or regression are applied generally. Because the I-PAT model has important advantages in terms of data availability and application convenience, many statistical studies have been conducted to explore the net effects of population on energy consumption and carbon emissions. However, this type of model also has obvious disadvantages, including limitations in the compatibility of demographic factors and the comparability of findings, among other issues. Therefore, Dietz et al. [8–10] proposed the Stochastic Impacts by Regression on Population, Affluence,

and Technology (STIRPAT) model to test elasticity with respect to coefficients for population, affluence and technology; STIRPAT may account for more demographic indicators and is more sophisticated than I-PAT. With the development of insights into correlations between population and energy consumption and/or CO2 emissions, the research community has recognised that population size might not account for the complex correlations completely because a wide range of demographic factors may be associated with energy consumption and/or CO2 emissions. For example, the Intergovernmental Panel on Climate Change examined 40 demographic scenarios in its fourth report [11], but it is believed that the effects of population on climate change were underestimated because the demographic component of the IPCC report included only changes in population size, whereas important factors of population structure were excluded [12]. Recent studies integrate indicators of population structure into models to reflect the complex effects of demographic structural factors on energy consumption [13–17]. However, although there are many such studies, their theoretical frameworks do not break through the domain of the STIRPAT model; indeed, these studies also belong to the linear model based on macro-demographic indicators. For example, there are methodological limitations to addressing age structures in 1-year groups, family patterns and urban-rural distributions simultaneously. In recent years, efforts have been made to analyse linkages between population and energy consumption, in which systematic models commonly used by climate change researchers are employed. Models of this type may contain more detailed parameters of demographic dynamics, technological change and economic development than linear models. In addition to population size, population composition (e.g., age, urban–rural residence, household structure, etc.) can be analysed in detail. Dalton et al. [18] adopted a population, environment and technology (PET) model to analyse CO2 emissions in China and India, in which the urbanrural distribution, family size and age structure were considered. Including the age of the householder, family size and urban-rural distribution in the demographic component, O’Neill et al. [19] explored the effects of global demographic trends on future carbon emissions. However, the systematic model addresses most demographic composition factors indirectly, which seems to be unsuited for a study of REC. For instance, in systematic models, the labour supply is taken as an intermediate variable to transfer the impact of population aging on energy consumption or carbon emissions. Generally, in addition to population size, age, urban–rural residence and household structure—all of which are factors that influence energy consumption directly or indirectly—demographic compositional factors are also closely related to energy consumption [20]. Because households are the basic units of energy consumption (particularly REC), the scale effect of energy consumption is the theoretical basis for how households affect energy consumption [21]. Because of the loss of economies of scale, the per capita energy consumption of smaller households is significantly higher than that of larger households [20], which has been shown empirically [22,23]. Similarly, it is necessary to consider the impact of urbanisation if there are obvious differences in energy consumption or patterns of carbon emissions between rural and urban populations [24,25]. Conclusions from the analyses of the effects of urbanisation on energy consumption and/or carbon emissions are mixed. In developed countries, the level of CO2 emissions in urban areas is lower than in rural areas, but this pattern is reversed for developing countries [26,27]. In addition, populations worldwide are aging, and the impact of such older populations on energy demand should be considered. Because the income level and consumption behaviours of the elderly vary regionally and over


C. Fu et al. / Energy Research & Social Science 2 (2014) 1–11

time, the effects of population aging on energy consumption or CO2 emissions may vary among different populations [28]. Therefore, the effects of changes in age structure on energy consumption may be uncertain and vary across regions of the globe. Some studies in the past decades have significantly promoted and added to the understanding of the mechanisms and policy implications of interactions between population and energy consumption and/or CO2 emissions. However, there are several issues that still must be addressed. Studies about energy consumption and/or CO2 emissions based on macro-demographic indicators remain the subject of debate and might not completely reflect the character of the micro-demographic processes that could induce changes in demographic composition. Although demographic analyses have incorporated insights for understanding and projecting population and household dynamics [29], there are methodological and practical difficulties that must be overcome if the comprehensive variables of demographic composition are to be integrated into traditional models. It would be difficult to interpret the results because of the volume of demographic parameters, which might limit the theoretical value and usage of these types of studies. Thus, introducing the micro-demographic processes underlying macro-population levels into the operational models of energy consumption is an important scientific issue, which may provide new thoughts to overcome the shortages of the methods based on macro-demographic indicators.

3. Development of the theoretical model Accumulating scientific evidence reveals that the contribution of population to REC is underestimated if population size is included as the sole demographic variable in models because changes in population composition are linked to REC. Demographic studies [30–32] have discovered that future demographic dynamics depend on changes in population size and in demographic composition that are associated with micro-demographic processes. It is reasonable to think that different micro-demographic processes operating under similar macro-demographic levels may result in various changes in REC. The demographic sensitivity of REC is defined in this study as the extent to which the system of REC is affected by demographic changes during a period, which includes changes in the amount of REC and transitions of patterns of REC. The intensity of demographic sensitivity corresponds to the magnitude of changes in the amount of REC. In addition to population growth, demographic factors also involve changes in population age structure, urban-rural distribution and the micro-demographic processes that induce those changes. We develop the definition of the scale effect of REC similar to economies of scale in economics [33]. To maintain the normal daily life of a household, certain amounts and types of energy consumption are necessary. In our study, the energy consumed in households is classified into fixed-energy consumption and variable-energy consumption. The former includes energy whose amount of consumption does not change or changes slightly with the number of persons in a household, such as the energy used for lighting and heating. For a certain household, its residential pattern is determined mainly by their socio-economic status in China, which will influence its energy consumption used for lighting and heating no matter how many persons the household has in the case of urban family size 2.83 in 2010. The latter includes energy whose amount of consumption increases or decreases significantly with the number of persons in a household, such as the energy used for transportation and cooking. Therefore, the more persons in a household, the less per capita fixed energy and residential energy

micro-demographic changes

natural change

urbanisation

3

scale effect

macro results

family pattern

pattern of REC

aging

amount of REC

Fig. 1. The theoretical framework of REC and demographic sensitivity.

is consumed because of economies of scale (i.e. the scale effect of REC). If the scale effect of REC could be verified in a certain population, the theoretical hypotheses of demographic sensitivity of REC would be deduced as follows. Deduction one. If urbanisation affects the family pattern significantly, population urbanisation is sensitive to REC under a scale effect. Deduction two: If aging affects family patterns significantly, population aging is sensitive to REC under a scale effect. Deduction three: If population naturally changing affects family patterns significantly, natural population change is sensitive to REC under a scale effect. From the theoretical deductions discussed above, we construct a theoretical framework for the current study as shown in Fig. 1. The formation of REC may be described theoretically as follows. Based on population size, the family patterns of a certain population are formed under micro-demographic processes, and persons with different demographic features are distributed among different types of households (i.e., households with 1, 2, 3, 4 or 5+ persons). Combined with certain patterns of REC, the amount of REC is determined under the corresponding demographic sensitivity and scale effects of REC. Obviously, the scale effect is formed under specific economic and technical conditions, and simulating the scale effect reflects the effects of economic and technical factors on REC partly. Moreover, the scale effect would play the role of control variable such that the effects of economic and technical factors on energy consumption would be minimised when the net effects of demographic sensitivity on REC were calculated. It is demographically evident that certain macro-demographic levels, e.g., the urbanisation rate, the natural growth rate and the proportion of aged persons, might result from different micro-demographic processes. For instance, the change in the urbanisation rate might be induced by population migration or natural population changes associated with different microdemographic process, but these two types of demographic processes will result in completely different household patterns because floating families2 are often smaller in size than local families in China [34,35] and the family size of childbearing households is often large. Because different micro-demographic changes under identical macro-demographic levels may induce various changes in REC in response to the scale effect, it is necessary to consider all of the possible micro-demographic processes when studying REC. In

2 In China, migrants without local household registration (hukou) are defined as floating population.


4

C. Fu et al. / Energy Research & Social Science 2 (2014) 1–11

Table 1 Comparison of REC between 2005 and 2010 (unit: Chinese Yuan). 2010a

2005 Mean

Standard deviation

Mean

Standard deviation

Electricity Fuels Heating Transportation fuels Traffic costsb Total (cost of REC)

316.78 165.25 133.56 181.15 252.13 1048.88

689.06 330.37 354.33 703.52 770.73 1879.35

632.54 331.68 265.80 326.05 442.18 1998.25

1270.38 609.95 815.50 1285.86 1038.13 3306.28

Weighted cases

2138.4 thousand persons

a b

2070.4 thousand persons

The values in 2010 are measured in constant 2005 Chinese Yuan. Traffic costs mean cost of public transportation in a year.

this study, the above-discussed rationale will be tested by analysing demographic sensitivity. 4. Empirical research To explore the theoretical ideas discussed above, we employ actual data from population censuses and household surveys to test the scale effect of REC and then calculate demographic sensitivity such that changes in REC during a given period (2005–2010) might be attributable to changes in population dynamics. The present study takes the REC in China’s urban area as an example. 4.1. Data description This study matches energy and population data from the China Urban Household Survey (CUHS),3 Chinese energy statistics and national population censuses. Energy data used in this study are from the 2005 and 2010 CUHS, an annual survey conducted by the State Statistical Bureau of China. The urban survey started in 1955 and covered a national sample of households. Each household was required to keep a record of income and expenditures and received a small payment in return. The survey ceased during the Cultural Revolution in 1966–1976, and resumed in 1980. The questionnaires and sample sizes have expanded in recent years. In order to avoid the problem of sample aging, provision was made for the annual replacement of a proportion of the households, such that the entire sample was replaced every 4 years. The questionnaire included sufficient detail to provide a general picture of household energy use. We use the weighted sampling data from CUHS from 10,000 households4 The following five types of energy consumption, measured in Chinese Yuan, are examined: electricity (cost of electricity used by household in a year); fuels (cost of fuels used for cooking in a year); heating (cost of energy for heating in a year); transportation fuels (cost of fuels used for private vehicles); traffic costs (cost of public transportation in a year). Table 1 provides a summary and average amount of different types of REC in 2005 and 2010. In the composition of REC, the consumption of electricity is highest, and the consumption of heating fuel is the lowest in 2005 and 2010. In REC related to transportation, traffic cost is the primary component, but energy consumed by private vehicles is increasing. In sum, the total REC approximately doubles from 2005 to 2010. It is noteworthy that, for purposes of simplifying calculation processes and avoiding unnecessary bias, we have not transformed one unit of Chinese Yuan into a Ton of Standard Oil Equivalent (toe) because

3 The specification of the survey can be found in the China yearbook of household survey 2011(pp. 379-390). 4 The number of total households surveyed in 2005 and 2010 was 15,000 and 64,500, respectively.

energy prices in different cities and towns are different slightly. To measure relevant demographic variables, we use data from the 2005 1% National Population Sample Survey (NPSS) and 2010 National Population Census (NPC) in China. The urban population of China5 increased by 104 million persons, and the proportion of urban population to overall population was up by 6.69% from 2005 to 2010. Almost 50% of the total population lived in urban areas in 2010. The development of urbanisation resulted from the migrating population primarily, and the migrating population was 261 million persons in 2010, after experiencing 77.39% growth since 2005. Accompanying the change in population age structure, the proportion of the population aged 65 and over to the overall population was 8.87% in 2010, which increased by 1.19% from 2005 to 2010. Moreover, the average urban family size decreased by 0.13 persons, from 2.97 persons in 2005 to 2.84 persons in 2010. 4.2. Scale effect According to the definition of the scale effect of REC, whether the scale effect of REC exists in the urban population in China can be identified in two steps. The first step is to test whether the differences in the average REC among types of households are significant. Thus, the above-discussed five types of REC will be classified into fixed- or variable-energy consumption. The second step is to fit a curve for the scale effect, which represents the benchmark of per capita REC and the velocity with which per capita REC decreases as family size rises. Step 1. To examine the differences of average REC among types of household, we estimate a series of one-way analysis of variance (ANOVA) models. The general model is set up to involve any two types of household. Households are classified into five types according to family size: 1-person, 2-person, 3-person, 4-person and 5 persons and above. Thus, there are ten pairs of types of household (C52 ) for each type of REC. Along with determining whether the amount of average REC depends on the type of household, we set up a priori contrast to determine whether the amount of average REC differs for different types of household. With respect to any one type of REC, if the average consumption rate of each type of household is mostly equal (i.e., the p-values of ANOVAs are greater than 0.05), this type of REC might be identified as fixed-energy consumption. Otherwise, it would be identified as variable-energy consumption. The results show that ANOVAs of electricity, fuels and heating are mostly not significant (p < 0.05), and few ANOVAs are significant both in 2005 and 2010 (see Table 2a–2c). By contrast, ANOVAs

5 To match the data from the CUHS, only the household population is included here.


C. Fu et al. / Energy Research & Social Science 2 (2014) 1–11

5

Table 2 ANOVA of relationship between REC and family size. (a)

1

2

3

4

5+

(b)

1

1 2 3 4 5+

   

  

 



(c)



 

  

   

1 2 3 4 5+

   

  

 



1

2

3

4



 

  

5+

(d)

1

2

3

4

5+

   

1 2 3 4 5+

    

 

  

  

 

   



2

3

4

5+



 

  

   

1 2 3 4 5+

   

  

 



(e)

1

2

3

4

5+

(f)

1

1 2 3 4 5+

    

 

  

   

1 2 3 4 5+

   

  

 



2

3

4

5+



 

  

   

  

 



Note: (1) The upper triangles of the tables are the results calculated from the 2005 data. The lower triangles of the tables are the results calculated from data in year of 2010. (2)  denotes p > 0.05;  denotes p < 0.05. (3) ANOVA of (a) electricity and types of household, (b) fuels and types of household, (c) heating and types of household, (d) transportation fuels and types of household, (e) traffic costs and types of household and (f) REC and types of household.

of transportation fuels and traffic costs are mostly significant, and few ANOVAs are not significant both in 2005 and 2010 (see Tables 2d and 2e). Therefore, statistically, electricity, fuels and heating can be classified as fixed energy consumption, and transportation fuels and traffic costs belong to variable energy consumption. The amount of REC is measured as the sum of the amounts of these five types of energy consumption. As shown in Table 2f, all the ANOVAs of REC are significant, which indicates that there is a significant correlation between family size and average REC statistically. Step 2. To further characterise the scale effect of REC, we also examine how and to what extent per capita REC changes with family size rising to describe the scale effect accurately. Through a curve-fitting analysis, the per capita RECs of types of household are fitted with linear, logistic and exponential curves in 2005 and 2010. The best fit of these three curves is that of the exponential curve. Therefore, we adopt the exponential curve to represent the scale effect using the following formula: Per capita REC = exp(a + b∗familysize)

(1)

where a2010 = 7.82, b2010 = −0.26; a2005 = 7.27, b2005 = −0.24; values of F-statistic test for 2005 and 2010 are 19.14 (p-value =0.022) and 95.23 (p-value =0.002) respectively. As Fig. 2 shows, formula (1) is a monotone decreasing function with increasing family size. That is, per capita REC decreases as family size increases. Thus, the significant differences of average REC among types of household and the feature of monotone decreasing per capita REC are verified, which is consistent with the theoretical assumptions of the scale effect. Thus, we can conclude that the scale effect of REC exists both in 2005 and 2010 but that the parameters of the scale effect are different in 2005 and 2010. 4.3. Demographic sensitivity According to the definition of demographic sensitivity, those demographic processes that might result in changes in family size may also affect REC significantly because of the scale effect, such as urbanisation, age structure change, natural population change and other factors.

Chinese Yuan 2500 measured value(2010) fitted value(2010) measured value(2005) fitted value(2005)

2000

1500

1000

500

0

1

2

3

4

5 persons

Fig. 2. The curve of the scale effect of the relation between per capita REC and family size. Note: REC in 2010 is measured in constant 2005 Chinese Yuan

4.3.1. Sensitivity of age structure If age-specific per capita REC is different among ages, then changes in the population age structure will induce changes in REC. The extent of the change in the population age structure determines the magnitude of the change in REC. In this section, ages are divided into 5-year groups. Fig. 3 provides a chart whose xcoordinate is age-group and y-coordinate is age-specific per capita REC. The shape of the chart suggests that per capita REC is not sensitive to age structure before 60 years of age, whereas the age-specific per capita REC of people aged 60 and over decreases sharply. Thus, we speculate that REC may be sensitive to population aging. The CUHS only counted the REC of the entire family for the year; thus, we must transform the amount of REC per household into a per individual amount to calculate the age-specific per capita REC. We assume that residential energy is equally consumed among persons in a household, such that the amount of per capita REC in a household might be treated as the individual amount of REC in our study. Because households are the basic units of energy consumption, differences in the amount of REC among persons in a household may not be distinguished accurately. Although the method of transformation in this study is not perfect and may


6

C. Fu et al. / Energy Research & Social Science 2 (2014) 1–11

Table 3 Statistical results of the relation between REC and age structure. 2010

R2 Constant a

2005

Model 1People of all ages

Model 2People aged 0–59

Model 3People of all ages

Model 4People aged 0–59

0.5027 1208.61*** −26.86***

0.0002 1064.93*** 0.37

0.5085 570.67*** 13.76**

0.0372 662.70*** −3.25

Note: * denotes p < 0.05, ** denotes p < 0.01, *** denotes p < 0.001, with two-tailed tests, respectively. Table 4 Statistical tests of differences of level of per capita REC between people aged 0–59 and 60 and over. 2010

0–59 60+

2005

Samples (weighted)

Mean (Chinese Yuan)

183,422 30,413

1069.61 923.84

T test (Sig.)

Samples (weighted)

Mean (Chinese Yuan)

T test (Sig.)

0.00000

181,001 26,048

668.75 636.45

0.00001

Table 5 Comparison of REC and family size between people aged 0–59 and 60 and over. Family size

1 2 3 4 5+

Per capita REC in 2010 (Chinese Yuan)

Household composition of people aged 0–59 and 60 and above (%)

0–59 years

60+ years

2005

3068.94 1400.47 1045.67 811.25 786.28

1224.05 998.67 820.76 813.52 851.67

0–59 years

60+ years

0–59 years

2.62 12.89 29.79 25.38 29.31

9.02 31.93 15.31 14.89 28.84

3.46 17.40 30.76 22.48 25.90

induce bias, it is acceptable on the condition that no further data supports can be acquired. To test the observed phenomenon in statistics, we construct four regression models whose dependent variable is the per capita REC and whose independent variable is age. As shown in Table 3, the analytical results of models 3 and 4 are not significant; thus, for people aged 0–59 years, changes in age structure may not significantly result in changes in REC. In view of these analytic results, we conclude statistically that the population age structure is not sensitive to REC except for the 60 and over age group; however, population aging may be sensitive to REC. In the next section, the extent to which the micro-demographic process of population aging affects REC will be investigated. Chinese Yuan 1400 1200

2005 2010

1000 800 600 400 200

60+ years 9.65 34.51 16.22 14.78 24.84

4.3.2. Sensitivity of population aging To identify the sensitivity of population aging, we must verify whether the family size of the elderly is different from other family sizes. If so, population aging would affect REC under a scale effect. However, it is difficult to judge the changing direction of family size of the elderly before detailed calculations are conducted. There are factors resulting in the elderly’s family size decreasing (such as empty nests), in addition to factors resulting in the elderly’s family size increasing (such as co-residence with adult children for daily life care). In addition, the changing lifestyle and consumption patterns of the elderly would also affect their REC. In particular, population aging may reduce transportation demand [36] because the elderly may be more likely to be withdrawing from mainstream social life but may increase energy use for the heating and cooling of floor space [37]. From the results of the T-test in Table 4, it can be observed that the per capita REC of the elderly significantly differs from that of people aged 0–59 in both 2005 and 2010. Moreover, as shown in Table 5, the per capita REC of the elderly in different family sizes are obviously lower than for people aged 0–59. Although the family size of the elderly is smaller than that of people aged 0–59 (and shrank from 2005 to 20106 ), the level of REC of the elderly is far lower than that of people aged 0–59. Therefore, because the positive impact of the scale effect on REC resulting from the elderly’s family size shrinking is offset by the far lower level of REC of the elderly, the sensitivity of population aging is negative.

4 59 10 -1 4 15 -1 9 20 -2 4 25 -2 9 30 -3 4 35 -3 9 40 -4 4 45 -4 9 50 -5 4 55 -5 9 60 -6 4 65 -6 9 70 74 75 -7 9 80 +

0

0-

per capita REC

2010

Fig. 3. Relation of per capita REC and age structure.

age group

6 The proportion of the elderly with family sizes of 1, 2 and 3 persons increased from 2005 to 2010, which resulted in a reduction of the family size of the elderly in 2010.


C. Fu et al. / Energy Research & Social Science 2 (2014) 1–11

Thus, we may consider that population aging is sensitive to REC, and the increase of population aging level will cause a decrease of REC. To measure the intensity of the sensitivity of population aging, we can adopt a back-stepping method. Thus, if the population aging level in 2010 remains stable from 2005 and the other demographic parameters used in the model remain at their actual levels in 2010, the proportion of the assumed increment of REC from 2005 to 2010 to the actual increment is the intensity of the sensitivity of population aging. The process for calculating the sensitivity of population aging is as follows. A EN2010 =

5 

[popA2010 (x, 0∼59) × en2010 (x, 0∼59)

+ popA2010 (x, 60+) × en2010 (x, 60+)]

(2)

A where EN2010 denotes the amount of REC in 2010 and the population aging level in 2010 is assumed to equal that of 2005, popA2010 (x, 0∼59) denotes persons aged 0–59 with family size x and the population aging level in 2010 is assumed to equal that of 2005, popA2010 (x, 60+) denotes persons aged 60 and over with family size x if the population aging level in 2010 is assumed to equal that of 2005, en2010 (x, 0∼59) denotes the average REC of people aged 0–59 with family size x in 2010 and en2010 (x, 60+) denotes the average REC of people aged 60 and above with family size x in 2010. u popA2010 (x, 0∼59) = P2010 × [1 − (R A − R A)] ×

u popA2010 (x, 60+) = P2010 × (R A − R A) ×





pop2010 (x, 0∼59) P2010 (0∼59)

pop2010 (x, 60+) P2010 (60+)



(3)



(4)

u where P2010 denotes the number of urban dwellers in 2010, R A denotes the proportion of urban dwellers aged 60 and over to the total urban dwellers in 2010, R A denotes the difference of R A between 2005 and 2010, pop2010 (x, 0∼59) and pop2010 (x, 60+) are the numbers of urban dwellers aged 0–59 and 60+ with family size x in 2010, respectively, and P2010 (0∼59) and P2010 (60+) are the numbers of urban dwellers aged 0–59 and 60+ in 2010, respectively.

ENt =

5 

popt (x) × ent (x),

(5)

x=1

where ENt denotes the actual amount of REC in t year; ent (x) denotes the actual per capita REC of people with family size x in t year; popt (x) denotes the number of urban dwellers with family size x in t year; the subscript t represents the year of observation (i.e., 2005 and 2010, respectively); and x represents family size, i.e., 1, 2, 3, 4 and 5+ persons. Then, the sensitivity of population aging is determined as follows.

Sensitivity of population aging =

A (EN2010 − EN2005 ) − (EN2010 − EN2005 )

(EN2010 − EN2005 )

4.3.3. Sensitivity of urbanisation To measure the extent to which urbanisation affected REC from 2005 to 2010, we must know the impact of urbanisation on the number of urban dwellers and their family size. Accordingly, assume that the urbanisation level in 2010 is equal to that in 2005, and the other demographic parameters used in the assumed model remain the actual levels in 2010. Thus, the difference in the increment of REC between actual and assumed demographics is the impact of urbanisation on the change in REC from 2005 to 2010. The percentage change of the increment of REC will be taken as the intensity of sensitivity of urbanisation. The formulas are as follows. u EN2010 =

x=1

(6)

where EN2010 is the actual amount of REC in 2010 and EN2005 is the actual amount of REC in 2005. Using appropriate data from NPSS and NPC, the population indicators in the formulas can be calculated. We can obtain the intensity of sensitivity of population aging by combining these with raw data from CUHS, which is −1.89%. The population aging results in 1.89% of the decrease of the increment of REC from 2005 to 2010. This situation is reversed for developed countries, such as the U.S.A [38], in which elderly persons use more residential energy than younger persons.

7

5 

popu2010 (x) × en2010 (x),

(7)

x=1 u represents the amount of REC in 2010 if the urbanwhere EN2010 isation level is assumed to be unchanged from 2005 to 2010 and popu2010 (x) represents the number of urban dwellers with family size x in 2010 if the urbanisation level is assumed to be unchanged from 2005 to 2010. Thus, the assumed amount of REC in 2010 will be mainly determined by demographic parameters because, in formula (7), en2010 (x) is derived from data of household survey and does not vary under different urbanisation levels.



u popu2010 (x) = (P2010 ∗ R2005 )∗

pop2010 (x) u P2010



,

(8)

u where P2010 represents the overall population figures in 2010, R2005 represents the proportion of urban dwellers to overall population in 2005 (i.e., the urbanisation level in 2005), pop2010 (x) represents u the number of urban dwellers with family size x in 2010 and P2010 represents the number of urban dwellers in 2010.

Sensitivity of urbanisation =

u (EN2010 − EN2005 ) − (EN2010 − EN2005 )

EN2010 − EN2005

(9)

Using corresponding data of population and REC, we can calculate the intensity of the sensitivity of urbanisation, which is 12.89%. That is, urbanisation accounted for 12.89% of the increment of REC from 2005 to 2010. The sensitivity of urbanisation results from dual demographic driving forces, namely, the increasing number of urban dwellers and the decreasing urban family size [28], and both of these demographic changes promote the growth of REC. 4.3.4. Sensitivity of natural population change Natural population change indicates the demographic changes caused by births and deaths. In this section, we are concerned with the change in the number of people and in family patterns caused by natural change. Because the natural change in urban population was positive from 2005 to 2010, we only discuss the number of people added caused by natural population changes during that period and its effects on family patterns. Similarly, natural population change is assumed to be zero from 2005 to 2010, and the other demographic conditions used in the assumed model remain the actual levels in 2010. Thus, the difference in the increment of REC between the actual and assumed demographic situations is the impact of natural population change on the change in REC from 2005 to 2010. The percentage change of increment of REC will be taken as the intensity of sensitivity of natural change. In the following three situations, the natural population change will be zero: (a) the number of births from 2005 to 2010 remains the actual level, and the number of deaths increases to equal that of births; (b) the number of deaths from 2005 to 2010 remains the actual level, and the number of births decreases to equal to that of deaths; and (c) the numbers of births and deaths from 2005 to 2010 both change but equally. The first situation goes against the social


8

C. Fu et al. / Energy Research & Social Science 2 (2014) 1–11

Table 6 Calculating formulas of pop2010 (y) in the case of zero natural change. Family size

Measured persons

1 2 3 4 5 6+

Pop(1) Pop(2) Pop(3) Pop(4) Pop(5) Pop(6+)

Persons required to be subtracted

5 Dx (y)

=

5 Px (y) × 5 × 5 mx (y) , (1 + (5 − 5 ax ) × 5 mx (y))

(10)

where 5 Dx (y) represents deaths aged x–x + 5 with family size y from 2005 to 2010, 5 Px (y) represents the number of people aged x − x + 5 with family size y in 2005, 5 mx (y) represents the age-specific mortality rate of people aged x − x + 5 with family size y in 2005 and 5 ax is 2.5, which indicates an even distribution of deaths. D(y) =

80+ 

5 Dx (y),

(11)

x=0

where D(y) represents the number of deaths with family size y from 2005 to 2010. D=

5+ 

D(y)

Pop(1) Pop(2) + 2N(3) Pop(3) − 3N(3) + 3N(4) Pop(4) − 4N(4) + 4N(5) Pop(5) − 5N(5) + Pop(6+) − N(6+)

N(3) N(4) N(5) N(6+)

norms, and the third situation has numerous possible permutations of the numbers of deaths and births, which reduce the chances for understanding and interpreting this scenario. The present study takes the second situation as the assumed precondition.

(12)

Persons after subtraction (pop2010 (y))

n where EN2010 represents the amount of REC if the natural change is assumed to be zero from 2005 to 2010. As shown in Table 6, pop2010 (y) represents the number of people with family size y after N(y)s are subtracted.

Sensitivity of natural population change =

n − EN2005 ) (EN2010 − EN2005 ) − (EN2010

EN2010 − EN2005

(18)

Using appropriate data of population and REC, we can calculate the result of formula (18), which is 7.37%, i.e., the natural population change accounts for 7.37% of the increment of REC from 2005 to 2010. Generally, with respect to a naturally increasing population, natural change would increase the population size and enlarge the family size of households with childbearing events. These two demographic processes have conflicting effects on REC.

4.4. Scenario analysis

y=1

Through the above empirical study, we discover that the scale effect exists in the REC of the urban population in China and that population aging, urbanisation and natural population change are sensitive to REC. Thus, these analytical results and methodology may be applied to the scenario analysis of REC in the future.

where D is the number of total deaths from 2005 to 2010. B(y) =

6 

5/2[W2005 (15 + 5x, y) + W2005 (10 + 5x, y)

x=0

× L(15 + 5x)/L(10 + 5x)]f (15 + 5x)

(13)

where B(y) represents the number of births by women with family size y, W2005 (15 + 5x, y) represents the number of women in the age interval (15 + 5x,19 + 5x) with family size y in 2005, L(15 + 5x) represents the person-years lived of women in the age interval (15 + 5x, 19 + 5x) from 2005 to 2010 and f(15 + 5x) represents the age-specific fertility rate of women in the age interval (15 + 5x,19 + 5x). B=

5+ 

B(y)

(14)

y=1

where B represents the number of total births from 2005 to 2010. N = B − D

(15)

To obtain the assumed household composition in the case of zero natural change, N should be subtracted from all types of household in an appropriate ratio, and the change in family size caused by N should be taken into account. N(y) = N ×

B(y) B

(16)

where N(y) represents the number of people required to be subtracted from the number of people whose family size is y. The values 6+ of y are 3, 4, 5 and 6+. B, is B(y). 3 n EN2010 =

5+  x=1

pop2010 (x) × en2010 (x)

(17)

4.4.1. Scenario setup Because of limited space, this section only analyse the effects of demographic sensitivity on REC from 2010 to 2015, and the purpose of this study is to show the probable effects of demographic changes on REC. However, this is not a prediction of REC in 2015 because REC is also affected by many other factors, such as levels of income, lifestyle, technology and other factors, which are beyond the scope of the current study. In this section, these other factors (except for demographic indicators) will be controlled through the scale effect. Section 4.2 showed that the scale effect of REC is exponential (see Eq. (1)), that a determines the benchmark level of REC and that b denotes the intensity of the scale effect (i.e., the velocity of the decrease in REC as family size increases). This study sets up the following two scenarios for a scale effect: (a) in 2015, a2015 is the extrapolated value of a from 2005 to 2010, and b2015 in 2015 is the same as that in 2010, which is denoted as SE1; (b) a2015 and b2015 in 2015 are the extrapolated value of a and b from 2005 to 2010, respectively, which is denoted as SE2. The arrangement of the scenario constructs of the scale effect is to reflect the increasing demand on REC caused by the increase in the standard of living. Family size has been continuously decreasing in recent years. The shrinking family size was caused by a continued decline of fertility, an increase of population migration and the independent living arrangements of young couples after marriage. It is assumed that this trend will continue in the near future. Thus, family size in


C. Fu et al. / Energy Research & Social Science 2 (2014) 1–11

9

Table 7 Constructs of scenario parameters. Scenarios 1 2 3 4 5 6 7 8 9 10 11 12

Parameters of scale effect

Family size

Level of urbanisation

Level of population aging

Natural change

SE1

FS-M FS-H FS-L FS-M FS-H FS-L

Urb Urb Urb Urb Urb Urb

Aging 1 Aging 1 Aging 1 Aging 2 Aging 2 Aging 2

NC 1 NC 1 NC 1 NC 2 NC 2 NC 2

SE2

FS-M FS-H FS-L FS-M FS-H FS-L

Urb Urb Urb Urb Urb Urb

Aging Aging Aging Aging Aging Aging

NC 1 NC 1 NC 1 NC 2 NC 2 NC 2

1 1 1 2 2 2

Table 8 The results of scenario analysis of REC in 2015. Scenario construct

REC (10,000 toe) (1)

Sensitivity (%)

Increment (10,000 toe)

Natural change (2)

Urbanisation (3)

Aging (4)

Natural change (1)*(2)

Urbanisation (1)*(3)

Aging (1)*(4)

2005–2010 1 2 3 4 5 6 7 8 9 10 11 12

3803.99 4359.76 4359.76 4359.76 4359.76 4359.76 4359.76 4359.76 4359.76 4359.76 4359.76 4359.76 4359.76

7.37 8.19 6.72 9.32 9.23 7.78 10.26 9.01 7.32 11.18 10.71 8.39 12.11

12.89 13.28 12.37 14.01 11.12 10.74 11.67 15.54 13.36 16.67 12.57 11.71 12.84

−1.89 −2.01 −1.73 −2.56 −1.76 −1.59 −2.14 −2.27 −1.89 −2.82 −1.94 −1.73 −2.57

280.36 357.06 292.98 406.33 402.41 339.19 447.31 392.77 319.35 487.60 466.79 365.99 527.83

490.34 578.98 539.30 610.81 484.81 468.24 508.78 677.40 582.45 726.85 547.83 510.38 559.66

−71.90 −87.63 −75.43 −111.61 −76.73 −69.32 −93.30 −99.02 −82.22 −122.77 −84.41 −75.56 −111.96

2015 is set at 2.72 persons through extrapolation7 and denoted as FS-M. To demonstrate the impact of demographic sensitivity through the scale effect on REC, high and low family sizes are constructed, and their values are set at 2.94 and 2.5 persons, respectively; these are denoted as FS-H and FS-L, respectively. Natural population change is affected primarily by fertility and the mortality rate. The mortality rate will not change much in the short term, and the fertility rate is mainly affected by family planning policies in the short term. The scenarios of natural change are set as follows. (a) The mortality and fertility rate is constant from 2010 to 2015. (b) The mortality rate is constant, and family planning policy is changed to a 2-child policy in 2015. These scenarios are denoted as NC 1 and NC 2, respectively. The adjustment of the family planning policy in China remains the subject of debate. Although it will occur in the near future, it is unlikely that the effects of policy adjustment on fertility will be completely in place before 2015 because of the short period of implementation. However, the purpose of constructing this fertility scenario is to simulate and elucidate the impact of micro-level changes in fertility behaviours on REC. From NPC in 2010, the total fertility rate (TFR) in 2010 was 1.18 persons, which obviously underestimated the fertility level of China in 2010 [39]. Through a comprehensive comparison, we employ the constructs of relevant studies [26,28] to estimate the TFR from 2010 to 2015 in the case of a two-child policy. Thus, the scenario of TFR is constructed to be that the TFR of the urban population will increase evenly from 1.2 persons in 2010 to 1.8 persons in 2015.

7

Urban family size in 2005 is 2.97, and that in 2010 is 2.85.

Urbanisation is the clearest tendency in China’s future and is enforced by national development policy.8 Thus, the scenario for the urbanisation level is set to 54.68%, which assumes that the proportion of urban population to overall population will grow by 1% per year from 2010 to 2015. The setup of urbanisation is denoted as Urb. In addition to changes in the fertility rate, population migration and floating in China is the other demographic factor affecting population aging. Because floating people have obvious economic characteristics, their age structure is concentrated in the main working-age group, which would contribute to reduce the extent of population aging. Thus, scenario 1 of population aging is constructed to be the demographic consequence of population aging under the scenario of Urb and NC 1; scenario 2 of population aging is constructed to be the demographic consequence of population aging under the scenario of Urb and NC 2. These constructs are denoted as Aging 1 and Aging 2, respectively. Combining the constructs discussed above, 12 scenarios are listed in Table 7. 4.4.2. Discussion of scenario analysis According to the formulas of demographic sensitivity and scenario setups, the demographic sensitivity of REC and the results of scenario analyses in 2015 may be estimated based on appropriate data resources. The summary statistics for the scenario analyses of changes in REC in urban China in 2015 are shown in Table 8. For the purpose of clearly displaying the correlations between demographic

8

http://finance.qq.com/zt2012/zyjjgz/,2013-01-01.


10

C. Fu et al. / Energy Research & Social Science 2 (2014) 1–11

sensitivity and changes in REC, we also transform the form of the results percentage score into the standard measure unit of energy (i.e., ton of standard oil equivalent, toe). Therefore, it is necessary to estimate the absolute amount of REC in 2015 in advance. The present study adopts an auto-regression model to extrapolate the estimated amount of REC with unit toe in 2015 using the time series data of REC from the China Energy Statistical Yearbook [2] from 2005 to 2010 year by year. The absolute amount of REC in 2015 in this section is used as a constant; thus, we are confident that its accuracy will not significantly mislead explanations of the results. Several conclusions emerge from the analysis in our scenario.

a. Although the macro-demographic levels are identical or similar, different micro-demographic processes may also cause varied scenarios of REC. For example, the macro levels of urbanisation, population aging and natural change in scenarios 1–3 and 7–9 are constructed to be equal, but the greatest difference (between scenarios 9 and 2) of the increment of REC in 2015 reaches 3.35 million toe (or 7.68% as a percentage calculation), whereas the difference is lowest (between scenarios 3 and 2) at 1.49 million toe (or 3.41% as a percentage calculation). It is noteworthy that the above-described differences include the effects of different parameters of scale effects that partly represent the effects of non-demographic factors on REC; however, the difference in the increment of REC between scenarios 9 and 8, 2.72 million toe or 6.24% as a percentage calculation would be the net difference under the identical scale effect and macro-demographic levels. The calculation processes in Section 4.3 show that those macro-demographic levels might be the consequences of a range of micro-demographic processes and are affected by the micro-demographic conditions in 2010 (such as sex–age structure and age-specific household composition) and by micro-demographic processes from 2010 to 2015, such as individual behaviours of migration, fertility and mortality. Therefore, the micro-demographic processes under certain macro-demographic levels have significant potential to affect REC, which indicates that it makes political and theoretical sense to re-consider the demographic component in energy consumption studies. b. The effects of demographic sensitivity on REC will be continuously increasing, but population is not the most important factor with respect to REC. Under the impact of the scale effect, the increment of REC in 2015 caused by demographic sensitivity is between 7.38 million toe (scenario 5) and 10.92 million toe (scenario 9), which is between 0.39 million toe to 3.93 million toe more than in 2010. Comparing the maximum increment of REC to consumption of total energy by sectors, the increment of REC surpasses the amount of energy consumption in 16 subsectors of industry, including the manufacture of foods and the production and distribution of gas. Nevertheless, the increments of REC caused by demographic sensitivity make up 16.93–25.04% of the total increment of REC from 2010 to 2015. Approximately three-fourths of the increment of REC that may be attributed to non-demographic factors such as technological progress and quality of life increase remains unexplained by demographic sensitivity. c. Among the sensitive demographic factors, urbanisation is the strongest, which may be the chief cause for the increase of REC in urban areas in China. The sensitivity of urbanisation is negatively correlated with family size. With decreasing family size, the intensity of the sensitivity of urbanisation would increase, vice versa. The range of sensitivity of urbanisation is from 10.74% (scenario 5) to 16.67% (scenario 9), values that exactly correspond

to the highest (2.94 persons) and lowest (2.5 persons) constructs of family size. d. The sensitivity of natural population change is influenced by the fertility level, and its intensity of sensitivity ranges from 6.72% to 12.11%, which is lower than that of urbanisation. Whereas the fertility level in scenario 5 is higher than in scenario 1, the intensity of sensitivity of natural change in scenario 5 is lower than in scenario 1. This phenomenon may be interpreted to mean that the increase of REC caused by population growth (i.e., the quantitative factor) will be partly offset by the changes in population structure (i.e., the structural factor). e. Accompanying the shrinking family sizes and decreasing consumption of the elderly, the sensitivity of population aging is negative, which will offset part of the increment of REC. Although population size is considerably sensitive to REC, the idea of saving REC through population regulation and control does not seem realistic. Neither increasing the population aging level to save energy nor decreasing the urbanisation level is consistent with the nature and purposes of social development. Because of the multiple sensitivities of population to REC, population size cannot be the exclusive demographic indicator with which to judge changes in REC, and the argument [40] that controlling population size would help save energy should be tested more carefully. f. The scale effect plays a key role in correlations between REC and demographic changes. The effects of changes in demographic factors are connected to changes in REC through the scale effect. As shown in our results, the demographic sensitivity of scenarios 7–12 is more intensive than those of scenarios 1–6 because of different constructs of the scale effect. We speculated that, with respect to demographic factors, there would be essential differences between production energy consumption and REC if the scale effect of REC cannot be proven to exist in a system of production energy consumption. Thus, if a study investigating the social total energy consumption, which includes REC and production energy consumption, uses the same population module, there may be unavoidable bias due to their different mechanisms of population affecting energy consumption. Until more is known about the nature of the scale effects in the two sectors, it would be unwise to draw far-reaching conclusions from one demographic module without a substantive justification of that choice because the scale effect of energy consumption would be defferent between REC and productive consumption. 5. Conclusion Based on the theoretical model constructed here, this study uses data from household surveys, population censuses and macro-scale energy statistics to verify the theoretical rationality of demographic sensitivity. As a general methodology, the paper will also be meaningful to other analysts. Overall, our research findings have several implications. (a) The sample size of micro-scale data of energy consumption is not large as a result of limitations of time and economic cost, and the demographic variables do not always meet the basic requirements of micro-demographic analysis. It is important to match data about REC with those of population censuses and employ micro-demographic analysis because abundant information about population structure cannot be represented completely through macro-demographic indicators alone. (b) It is noteworthy that there are several limitations of the current study that invite further attention and possible research. First, this study focuses on urban populations and their behaviours with respect to REC; rural populations are not involved because


C. Fu et al. / Energy Research & Social Science 2 (2014) 1–11

of the lack of raw data from a rural household survey. However, urbanisation affects the size and composition of urban populations and rural populations, which indicates that the complex demographic scenarios in this study are simplified. Whereas we believe this approach and theoretical model provides a useful complement to relevant case-study research, it also indicates that this framework cannot be applied to national REC. Another limitation of our study is that the demographic parameters in the section on scenario analysis are relatively simple. If population and household predictions are applied and the time span for prediction is lengthened, more meaningful results may be obtained. In addition, there are important differences in the behaviour of REC between migrating and floating families, and these populations are subsumed by data on resident families because migrating or floating information cannot be identified from household survey data. Indeed, this limitation indicates that we must look more deeply at these household data. Finally, the theoretical relevance of studying demographic sensitivity is to provide concise and accurate demographic information for other relevant studies, but much of the data do not involve the applicable issues of demographic sensitivity of REC. Acknowledgements

[12] [13] [14] [15] [16]

[17]

[18] [19]

[20] [21] [22] [23] [24]

This research is funded by Humanities and Social Science Foundation of Ministry of Education of China (Grant No. 13YJAZH022), and doctoral starting projects of GuangDong Medical College (Grant No. B2012075). Great debts are owed to many editors here and four anonymous referees. References

[25] [26]

[27] [28] [29]

[1] Parry M, Canziani O, Palutikof J, Linden P, van der Hanson C, editors. Climate change. Impacts, Adaptation and Vulnerability. Contribution of Working Group II to the Fourth Assessment Report of the IPCC. Intergovernmental Panel on Climate Change, vol. 3. Cambridge: Cambridge University Press; 2007. p. 12–43. [2] Department of energy statistics, National Bureau of Statistics, P.R.C. China energy statistical yearbook, vol. 1. China Statistics Press; 2011. p. 62–3 [Chinese edition]. [3] Zhang Q. Residential energy consumption in China and its comparison with Japan: Canada, and USA. Energy Build 2004;36(12):1217–25. [4] Jie F, Pingxin L. Analysis on the future energy consumption and preliminary discussion on carbon emission of China from the perspective urbanisation. Adv Earth Sci 2011;26(1):57–65. [5] Zhao X, Li N, Ma C. Residential energy consumption in urban China: a decomposition analysis. Energy Policy 2012;41:644–53. [6] Ehrlich PR, Holdren JP. Impact of population growth. Science 1971;171(3977):1212–7. [7] Kaya, Y. Impacts of carbon dioxide emission control on GNP growth: interpretation of proposed scenarios. Paper presented to the IPCC Energy and Industry Subgroup, Response Strategies Working Group, Paris; 1990. [8] Dietz T, Rosa EA. Rethinking the environmental impacts of population, affluence and technology. Human Ecol Rev 1994;1:277–300. [9] Dietz T, Rosa EA. Effects of population and affluence on CO2 emissions. Proc Natl Acad Sci U S A 1997;94(1):175–9. [10] Dietz T, Rosa EA, York R. Driving the human ecological footprint. Front Ecol Environ 2007;5(1):13–8. [11] Nakicenovic N, Alcamo J, Davis G, de Vries B, Fenhann J, Gaffin S, et al. Special report on emissions scenarios: a special report of Working Group III of the

[30]

[31]

[32]

[33] [34] [35] [36] [37] [38] [39] [40]

11

Intergovernmental Panel on Climate Change No. PNNL-SA-39650. Richland, WA, US/US: Pacific Northwest National Laboratory/Environmental Molecular Sciences Laboratory; 2000. Hoepf Y, Malea KM, Karen H. Projecting Population, Projecting Climate Change: Population in IPCC Scenarios; 2009. PAI Working Paper. Wei Y, Liu C, Fan Y. Report of China (2008): Carbon dioxide emission study, vol. 6. Science Press; 2008. p. 101–19 [Chinese edition]. Chen J, Peng X, Zhu Q. Impacts of household pattern on carbon emission. Chin J Popul Sci 2009;3:68–79 [Chinese edition]. Peng X, Zhu Q. Impacts of population change and consumption pattern on carbon emission in China. Popul Res 2010;1:48–56 [Chinese edition]. Liddle B, Lung S. Age-structure, urbanisation, and climate change in developed countries: revisiting STIRPAT for disaggregated population and consumptionrelated environmental impacts. Popul Environ 2010;31(5):317–43. Pachauri S. An analysis of cross-sectional variations in total household energy requirements in India using micro survey data. Energy Policy 2004;32(15):1723–35. Dalton M, O’Neill B, Prskawetz A, Jiang L, Pitkin J. Population aging and future carbon emissions in the United States. Energy Econ 2008;30(2):642–75. O’Neill BC, Dalton M, Fuchs R, Jiang L, Pachauri S, Zigova K. Global demographic trends and future carbon emissions. Proc Natl Acad Sci U S A 2010;107(41):17521–6. Jiang L, Hardee K. How do recent population trends matter to climate change? Popul Res Policy Rev 2011;30(2):287–312. Jiang L. The impacts of demographic change on climate change. Popul Res 2010;1:59–67 [Chinese edition]. Liu J, Daily GC, Ehrlich PR, Luck GW. Effects of household dynamics on resource consumption and biodiversity. Nature 2003;421(6922):530–3. MacKellar FL, Lutz W, Prinz C, Goujon A. Population, households, and CO2 emissions. Popul Dev Rev 1995:849–65. Jiang L, O’Neill BC. The energy transition in rural China. Int J Global Energy Issues 2004;21(1):2–26. Pachauri S, Jiang L. The household energy transition in India and China. Energy Policy 2008;36(11):4022–35. Guzmán JM, Martine G, McGranahan G, Schensul D, Tacoli C, editors. Population change and climate change. New York: United Nations Population Fund and International Institute for Environment and Development; 2009. Sovacool BK, Brown MA. Twelve metropolitan carbon footprints: a preliminary comparative global assessment. Energy Policy 2010;38(9):4856–69. Chen J. The impact of population on carbon emissions. Northwest Popul 2011;2:23–33 [Chinese edition]. Jiang L, O’Neill BC. Impacts of demographic trends on US household size and structure. Popul Dev Rev 2007;33(3):567–91. Zeng Y, Land KC, Wang Z, Danan G. US family household momentum and dynamics – extension of ProFamy method and application. Popul Res Policy Rev 2006;25(1):1–41. Yi Z, Wang Z, Leiwen J, Gu D. Projection of family households and elderly living arrangement in the context of rapid population aging in china – a demographic window of opportunity until 2030 and serious challenges thereafter. GENUS: Int J Demogr 2008;64(1–2):9–36. Zeng Y, Wang Z. Projection and policy analysis on population and households aging in 21 century in eastern, middle and western regions of China. Popul Econ 2010;2:1–10 [Chinese edition]. Mankiw NG. Principles of microeconomics. South-Western Pub.; 2011. Liang Z, Ma Z. China’s floating population: new evidence from the 2000 census. Popul Dev Rev 2004;30(3):467–88. Wang F, Zou X. Inside China’s cities: institutional barriers and opportunities for urban migrants. Am Econ Rev 1999;89(2):276–80. Prskawetz A, Leiwen J, O’Neill BC. Demographic composition and rejections of car use in Austria. Vienna Yearbook of Population Research; 2004. p. 175–201. O’Neill BC, Chen BS. Demographic determinants of household energy use in the United States. Popul Dev Rev 2002;28:53–88. Tonn B, Eisenberg J. The aging US population and residential energy demand. Energy Policy 2007;35(1):743–5. Fu C, Zhang L, Li Y. Characteristics of the changes of population fertility in China based on the sixth census. Stat Res 2013;30(1.):68–75 [Chinese edition]. Zhao B. Population program and responding to climate change. Popul Res 2010;1:43–7 [Chinese edition].


Exploring the sensitivity of residential energy consumption in china