1

Determinants of Household Carbon Emission: Pathway toward Eco-Community

in Beijing

Zan Yang, Ying Fan, Siqi Zheng

Hang Lung Center for Real Estate and Department of Construction Management,

Tsinghua University, Beijing, China

Abstract: Based on the household survey of “household energy consumption and

living conditions in Beijing”, in this paper, we estimate the determinants of household

carbon emissions and its relation to a community’s building, neighborhood and

location attributes. We find that a community with energy-saving building attributes

and with greater amenities in neighborhood significantly reduces the level of

residential household carbon emissions. Meanwhile, a community that is far from

public facilities tends to increase the probability of car purchase and results in higher

transportation carbon emissions. Given the implications these findings have on

eco-communities and urban sustainability, we further discuss the importance of the

role of a community when conducting ecological studies and its role for designing

strategies for urban ecological infrastructure (UEI) planning toward eco-friendly

economy and society.

Key words: carbon emission; residential community; eco-community; urban

ecological infrastructure

2

1. Introduction

The urbanization process refers to more than population growth and the physical

expansion of cities; it involves changes in people’s economic and social activities.

Environmental changes in cities thus depend on both population size and energy

consumption, and this is environmentally significant since such change is associated

with a wide range of human social and economic activities as well as the

transformation of materials and energy (Stern et al., 1997). Human energy

consumption has radically revamped Earth’s carbon cycle. At the global level,

residential energy use has become second only to industrial energy use with 70EJ

(22% of global energy use) consumed annually in the mid-1990s (IEA, 2000); it is

expected that energy consumption demand will continue to increase over the next 20

years (IEA, 2010). Integrating human consumption into ecosystem studies is critical

to our understanding of ecosystem drives and patterns, as well as in developing

policies for sustainable urban development.

Household energy consumption is determined by each household’s characteristics and

lifestyle, as well as by the characteristics of the living community that defines and

constrains residents’ social activities. A residential community provides more than

just a physical space to inhabit; it directly affects the resources needed to support

household activities through its spatial characteristics and its conjunction with public

services and thus the extent of environmental pressure. From an ecological

perspective, a residential community functions as an infrastructure that links nature,

public service, and society; further, it socially grounds the nature of human

environmental behavior. It is a mechanism through which the drivers of household

carbon emissions can be captured. More importantly, it paves one possible path

toward urban ecological infrastructure (UEI) by processing eco-community that

integrates eco-logical associations among households and communities as well as the

environment and related services (Li and Yang, 2015). Recently, an understanding of

the ecological community has begun to develop, however, few studies have connected

community attributes to household energy consumption and understand its role in the

ecological infrastructure. In this study, we will analyze Chinese household carbon

emissions and their relation to community attributes in Beijing.

3

China has witnessed an unprecedented urbanization, growing from 19.39% of total

population living in urban areas in 1980 to 54.77% in 2014 (MacroChina Database). It

is projected that by 2030, 1 billion Chinese will live in cities (Zhu et al. 2011). This

has been accompanied by increased housing demand and car ownership. According to

the National Bureau of Statistics of China, the average living area per person living in

urban areas increased from 24.5 square meters in 2002 to 32.9 square meters in 2014.

Over this same period, the number of private cars increased by more than 23.5 times.

Changes in household living environment and lifestyle are becoming more important

components of energy consumption and carbon emissions in China. As of 2009, China

has become the world’s biggest producer of greenhouse gases, representing total

emissions of 7527 million tons and 21% of the global emissions of carbon dioxide.

China’s per capita carbon emissions also exceeded the global average (World Bank,

2011). Among them, energy consumption in the residential sector amounted to 396.66

million tons standard coal, which represents 11% of total energy consumed in China

in 2014 (National Bureau of Statistics of China, 2014). China has committed to

reducing its carbon dioxide emissions per GDP by 40 – 45 percent before 2020, from

base levels recorded in 19901. The major burden of such carbon mitigation polices

will fall on energy consumers, including both firms and households. Understanding

household energy consumption patterns and how the patterns relate to their attached

community is important to capturing determinants that shape household activities.

This is essential to defining strategies for urban ecological planning aimed at

eco-friendly economics and society.

Based on the survey of “household energy consumption and living conditions in

Beijing” conducted in 2009, we estimate both residential carbon emissions and

transportation carbon emissions. We empirically explore household social/economic

characteristics and preferences of energy consumption and their effect on carbon

emissions, integrating the role of the community in terms of its building,

neighborhood and location attributes. We find that energy-saving building attributes

and greater amenities in neighborhoods have a significant positive role on limiting

household residential carbon emissions. Meanwhile, better access to public facilities

1 On the eve of the 15th Conference of Parties (COP15) of United States Framework Convention on Climate

Change (UNFCCC)

4

tends to decrease the probability of car purchase and thus carbon emissions from

transportation. Given the implications of these empirical results, this paper offers

insight into both social and economic policy aimed at eco-communities.

Our study contributes to previous studies on eco-communities from several

perspectives. Firstly, existing studies focus on explaining the concepts and indicators

of eco-communities, as well as technologies or strategies needed to develop

eco-communities (Han at al., 2008; Zhang et al., 2013). Few studies have connected

household performance with community characteristics or investigated the

mechanism through which community characteristics impact on household

performance. In our paper we indicate that eco-community not only provides

residential service, but also impacts total household energy consumption, and the way

of their energy consumption because these specific neighborhood and location

attributes enable us to bind this information to a specific community. To our

knowledge, this is the first paper that connects household carbon emissions from

housing and neighborhood attributes to capture the characteristics of the

eco-community. Being able to draw these connections has important policy

implications for urban planning. Second, though the role of housing has been

emphasized in environmental studies, most studies mainly conduct their analysis from

the perspective of physical attributes of the housing in the existing emissions studies

(Reid and Houston, 2013; Chen and Zhu, 2013). In this paper, we combine the

physical and spatial features of housing in a more comprehensive way. Given the

rapid increases in urbanization and the housing market in China, examining housing

and neighborhoods in energy consumption is a critical perspective for capturing what

drives carbon emission and future emission trends. The special case in China also

provides a new way to understand the eco-community from the perspective of urban

and market transformation. Third, this paper supplements the scarce studies on

household energy consumption in China, which suffer from limited data. Based on the

detailed information of household consumption and corresponding living conditions,

this paper offers a way to comprehensively explore household consumption behavior

and thus provides new evidences on existing environmental studies. Specifically,

using micro information, we also connect the provision of public good around

housing (including green land and public infrastructures) with household energy

consumption behaviors, and calculate the positive externality of public goods.

5

In Section 2, related literatures are reviewed and the conceptual framework of the

study is plotted. Section 3 introduces the household survey and section 4 provides

empirical results. Conclusions are in section 5.

2. Related Literatures and Research Design

Energy consumption is a human-environment transaction, which reflects the social

and economic activity of households (Dalton et al, 2007). A large amount of

inter-disciplinary academic research is required to understand the role of economic

and social factors in shaping household consumption and its effect on carbon

emissions. Related literatures mainly focus on the role of household characteristics in

household carbon emissions and less on the role of community (Dalton et al, 2007;

Yang, et al., 2010; Braubach and Fairburn, 2010; Kohlhuber et al., 2006; Chen and

Zhu, 2013).

The study of environmental effects from a household perspective is complicated by

heterogeneous household performance and interactions with social and political

factors. Socio-economic and demographic characteristics of households (Dalton et al,

2007), household demand for appliances (Druckman and Jackson, 2009; Chen and

Zhu, 2013), household attitudes and beliefs concerning energy use (Dietz et al, 2009),

and household lifestyle (Bin and Dowlatabadi, 2005) have all been described as

interactively influencing carbon emissions (Stern, 2000; Holden and Norland, 2005;

Van Raaij and Verhallen, 1999; Seryak and Kissock, 2003; Clapham, 2005). In China,

the linkage between household emissions and household lifestyle has also been

investigated (such as, Yang, et al., 2016; Li et al., 2013; Wang, 2015; and Du et al.,

2015).

Housing has long been central to the issue toward low-carbon consumption

particularly from a technological perspective (Reid and Huston, 2012; Choy et al.,

2013; Zhu et al., 2013). Existing researches find that residential carbon emissions

6

vary within different houses due to heterogeneous building structures (Genjo et al.,

2005), size and types (Alfredsson, 2004), as well as building materials, which include

brick, wood, and concrete (Hammond and Jones, 2008; Börjesson and Gustavsson,

1998).

The role of residential location on environment effects has been given attention due to

its connection to commute behavior (Kaza, 2010; Chang et al., 2010; Qin and Han,

2013; Ma et al., 2014), and because of its relationship to environmental justice

(Kohlhuber et al., 2006; Braubach and Fairburn, 2010). Its role is also highlighted as

it relates to urban spatial development (Ewing and Rong, 2008), which implies that

planning parameters including building density and land use mix are important factors

that impact household carbon emissions (Qin and Han, 2013). In addition,

endogenous decisions of household choice and transportation have been investigated

(Pinjari et al., 2009; Waddell, 2001) and the importance of residential location choice

on household energy demand has been studied (Yu et al, 2011).

In these previous studies, correlations between housing and carbon emissions have

focused more on physical entities, including spatial characteristics of housing.

Household consumption is thus “adhered” to the building attributes for fixed location.

This has restricted ecological efforts that have been more focused on technology

innovation and so have met with limited success (Reid and Houston, 2010). However,

if we understand housing as an important component in residential community, the

new insight concerning community attributes might direct future ecological study.

From this perspective, housing, neighborhood and location are naturally connected.

More importantly, community can be regarded as the infrastructure that encourages

people begin to look at designing communities rather than simply fitting into existing

communities (Flora and Flora, 1993). According to this framework, communities that

are connected both within and outside the community will engage with and shape

household energy consumption, and this improvement will benefit each household in

a community and expand outward into the whole society. This framework provides a

way to acknowledge the “inextricably social nature of technological change” (Shove,

2010), which also provides a straightforward path way toward urban ecological

infrastructure (UEI) by achieving eco-community. Unfortunately, neither the role of

community in ecological infrastructure, nor the linkage between community and its

7

environmental neighborhood attributes have yet been given much attention.

In this paper, household carbon emissions are investigated from the perspective of

both residential carbon emissions and transportation carbon emissions, which

dominate household total carbon emissions. We study household

demographic/economic characteristics and preferences of energy consumption

behaviors and in particular investigate how emissions are shaped by the residential

community according to heterogeneous building, neighborhood and location

attributes.

The framework of the paper is plotted in Figure 1 for use in future empirical study.

Household carbon emissions include both residential and transportation carbon

emissions. For determinates, household demographic/economic characteristics and

preferences, as well as community attributes in terms of building, neighborhood and

location attributes, will be investigated to understand household energy consumption

behaviors, and thus carbon emissions.

Figure 1. Research framework of the paper

Note: The solid arrow denotes a direct impact, while a dashed line denotes an indirect impact to

the ecological infrastructure.

8

3. Background and Data

As of 2009, China has become the world’s biggest greenhouse gas producer. With

8.50 Gt CO2 from burning fossil fuels and producing cement in 2012, China’s total

carbon emissions already equal the emissions from the U.S. and the E.U. combined,

and are responsible for 25% of global carbon emissions (World Bank, 2013). In per

capita terms, China’s carbon emissions per capita already exceed the global average

level, reaching 4.89 t CO2/per person (World Bank, 2007) in 2006. As demonstrated in

Figure 2, both carbon emissions in total and per capita level have shown rapid growth

since the beginning of the 21st century, which has been referred to as a “century for

cities”. From 2000 to 2011, the annual growth rate of total carbon emissions in China

reached 9.26%, which is much higher than that of other developing countries,

including India (5.20%) and Brazil (2.69%), and significantly exceeds the world

average (3.08%).

According to China’s Carbon Emissions Report in 2012, China’s carbon emissions

result mainly from fossil fuel combustion (90%) and cement production (10%). 90%

of China’s energy consumption is primarily derived from fossil fuel combustion: 68%

from coal consumption, 13% from oil and 7% from gas. More importantly, the

high-speed urbanization process has been considered as another major driver of

China’s carbon emissions growth. 85% of China’s direct carbon emissions are from

urban areas, highlighting the significant ecological role played by cities (Dhakal,

2009). As the capital of China, Beijing has also experienced dramatic growth in

carbon emissions over the last 10 years. From 2000 to 2012, the average growth rate

of carbon emissions in Beijing reached 4.25%. In 2012, carbon emissions in Beijing

exceeded 199 million tons, representing 2% of total carbon emissions in China.

9

Figure 2 Total Carbon Emission in China and Beijing

(a) (b)

Data source: World Bank (World Development Indicators) and MacroChina Database. Figure (a)

exhibits the total carbon emissions in China and other counties (million tons). Figure (b) exhibits

the carbon emissions and its growth rate in Beijing (million tons).

3.1 Beijing Household Survey Design

For the study, a self-constructed survey is used in Beijing. The survey was carried out

by “The Institute of Real Estate Studies” at Tsinghua University in 2009. It was

designed to obtain information on household socioeconomic characteristics, energy

consumption and living conditions. In addition, neighborhood and spatial

characteristics for these residential communities are also included in the survey.

All twelve districts in Beijing are covered in the survey, namely the inner city area

(Dongcheng, Xicheng, Chongwen, Xuanwu), the inner suburban area (Chaoyang,

Fengtai, Haidian, Shijingshan) and the outer suburban area (Changping, Daxing,

Shunyi, Tongzhou). We employ a two-stage sampling method. In the first stage, the

total number of communities is assigned to one of these twelve districts in proportion

to local population sizes. We also consider three major community types in Beijing,

including privatized public housing (public housing built before the housing reform

and later sold to employees at deeply subsidized prices, fanggaifang), commodity

10

housing (developed by real estate developers after the 1990s), and

government-subsidized affordable housing (built after the housing reform). According

to the National Bureau of Statistics of China, the relative shares of these three types of

housing stock are approximately 4:5:1. We let the distribution of the surveyed

communities resemble this ratio. (Figure 3 presents the spatial distribution of these 38

communities in Beijing).

In the second stage, within each community, housing units are randomly selected, and

the household head is interviewed. We collected 900 questionnaires. 74 of them were

incomplete and were dropped. We are left with 826 household observations.

Figure 3. Spatial distribution of 38 residential communities surveyed in Beijing

The survey was carried out by face-to-face interviews and the questionnaire responses

were checked carefully to maximize accuracy. Questions regarding the household as a

whole where asked of only one person, preferably the head. We can assess the quality

of the data by comparing our survey results to that based on official micro statistics,

such as the urban household survey (UHS) conducted by the National Bureau of

Statistics of China from 2002 to 2009. For example, the total household disposable

income in our survey (117.30 thousand RMB) is quite comparable to UHS (109.29

thousand RMB in 2009), while the average housing area in our survey (90.79 square

11

meters) is also close to UHS (85.38 square meters in 2009).

The questionnaire consists of four components: The first is about consumption for

daily life, including water, electricity, coal gas, Liquefied Petroleum Gas (LPG),

natural gas, and heating mode including central heating or decentralized heating. In

order to investigate household preferences for daily life, we also include questions on

temperature settings, fueling patterns and car ownership of households. The second

part contains information on everyday transportation of households, including

commuting, dining, shopping and driving children to school. We further asked about

transportation choices such as public and private vehicles. We also collected the

information of household living conditions, including building attributes (size,

structure and floor of the unit), neighborhood attributes (especially ecological

characteristics of the community, namely green rate, surrounding environmental status

in the 1.5 km buffer) and location attributes (distance to the nearest subway, school,

convenience store) of the dwelling. In the last part, the social and economic situation

of the household was ascertained: age of head of household, income, Hukou2,

education and employment status; family size and disposable income. In addition,

household’s self-grading information on the community is investigated, including the

satisfaction level for community surrounded environment and community location.

Though the survey used in this paper is 2009, it has relatively strong

representativeness. This is the only year that the survey that supports the current study

is available. But more importantly, during 2009 – 2016, although the whole economy

and housing market have been dynamic in Beijing, the related variables that influence

our study in fact tend to be stable. Since 2009, housing and facility construction have

been more concentrated in the sub – area of Beijing (Fifth ring or even further), which

is less covered in our sample and less influences our conclusion. In addition, we also

plot several certain key variables (such as housing structure, area, housing

2 Hukou refers to a household-registration record that officially identifies an individual as a resident of a given

area. It is one of China’s most important defining characteristics, as it defines individuals’ social-economic status

and access to welfare benefits.

12

neighborhood, household carbon emissions) to show how they are likely to be stable

after 2009. This suggests that the correlation between emissions and housing and its

neighborhood are reliable and the results of the study can be valuable in

understanding eco – community in today’s Beijing.

3.2 Household Energy Consumption and Carbon Dioxide Emissions

In measuring carbon emissions, we calculate five main sources of carbon emissions

from household energy consumption: household electricity, home heating and

residential cooking energy use, which are associated with residential energy use; as

well as gasoline consumption from private car gasoline consumption and taxi use,

which are associated with private transportation energy use. At present, private

transportation, especially the use of private cars in Beijing, is undergoing rapid

development, boosted by rapid urban sprawl and constantly increasing household

income. Considering the complexity of the mechanism and the lack of a unified

method to measure public emissions, we do not calculate carbon emissions caused by

public transportation, which will be left for further investigation.

More detailed methods used to calculate the total carbon emissions caused by

household energy consumption are described in the appendix. As shown in Table 1

below, we can see that residential carbon emissions and carbon emissions from

private transportation separately account for 70.2 percent and 29.8 percent of the total

carbon emissions of household. While the residential emissions mainly consist of

emissions from household electricity and home heating, which constitute 46.3 percent

and 36.0 percent respectively, emissions from residential cooking is 17.7 percent of

the total amount, representing a relatively low proportion. Regarding gasoline

consumption for transportation, emissions from private car use is the main component

(88.5 percent of the total transportation emission), and the remaining 11.5 percent is

from taxi use.

13

Table 1 Descriptive statistics for household carbon emissions

Terms Obs. Average

proportion of

total carbon

dioxide

emissions

Descriptive statistics of annual carbon dioxide from

household living (Ton/household)

Mean Std. Dev Max Min

Residential carbon emissions

Electricity 826 32.48% 2.403 1.812 17.195 0.057

Heating 643 12.44% 2.401 1.042 10.434 0.169

Cooking 795 25.26% 0.956 1.105 8.516 0.011

Transportation carbon emissions

Private car 393 26.40% 4.104 3.655 29.489 0.084

Taxi 417 3.42% 0.502 0.922 11.655 0.019

Household living carbon emissions

Total 826 7.399 4.717 40.947 0.437

Note: Due to the heterogeneous energy consumption patterns of different communities (for

example, some communities use decentralized heating rather than central heating), as well as the

missing information in some items, the observation numbers in the sub-samples are not the same.

Figure 4 statistically compares the difference in carbon emissions for communities

with different building attributes. Community A represents a typical eco-community;

while Community B represents a typical non-eco-community. It is found that

heterogeneity in heating mode, size, floor, orientation and structure results in different

levels of carbon emissions. For example, a community characterized by central

heating mode, larger size, lower floor, inferior orientation and non-brick structure tend

to have higher carbon emissions.

14

Figure 4 Comparison of household residential carbon dioxide emissions

Data source: Beijing Household Survey. The figure demonstrates the average residential carbon

emissions of different types of communities. Community A denotes eco-community, which is

featured with decentralized heating mode, small (less than mean value of the whole sample) size,

higher-floor (less exceeding than mean value of the whole sample), good orientation (mainly

facing south), and brisk structure. Community B denotes non-eco-community, which is featured

with central heating mode, large (exceeding mean value of the whole sample) size, low-floor

(lower than exceeding mean value of whole sample), inferior orientation, and non-brisk structure.

3.3 Distribution of Household Carbon Emissions: Role of Community

In order to verify the distribution of household carbon emissions, we calculated the

indicators of total, residential and transportation carbon emissions at both household

and community level. Referencing income inequality indicators, we plot the Carbon

Lorenz Curve of the whole sample based on our survey (Figure 5). Consistent with

Groot (2010), we find relatively high levels of inequality in household carbon

emissions based on our survey. In particular, the inequality curve of transportation

carbon emissions is more convex than the curve of residential carbon emissions,

which highlight the role of location of the community.

From the structural perspective, distribution of carbon emissions reveals similar facts

(Figure 6). For example, the communities concentrated by older buildinglocated in the

inner city area exhibit a higher share of central heating relative to total household

15

energy consumption. The communities in the central area and outer city area

demonstrate a higher share of transportation carbon emissions due to the use of

private cars.

Figure 5 Lorenz curve of household carbon emissions

Data source: Beijing Household Survey.

The distribution of carbon emissions indicates that community plays a significant role

in household carbon emission. Thus, we further design the empirical regressions to

investigate the role three different attributes in household residential and

transportation carbon emissions.

Figure 6 Inner structure of residential and transportation carbon emissions at community

level

(a) (b)

16

Note: In figure (a), the size of the circle denotes the comparison of average residential carbon

emissions at community level. The darkest area of the pie denotes the share of electricity carbon

emissions from total household residential carbon emissions. The lightest area of pie denotes the

share of cooking carbon emissions from total household residential carbon emissions. The

remaining area of the pie denotes the share of fueling carbon emissions from total household

residential carbon emissions. In figure (b), the size of the circle denotes the comparison of average

transportation carbon emissions at community level. The dark area of the pie denotes the share of

private car carbon emissions from total household transportation carbon emissions. The light area

denotes the share of taxi carbon emissions from total household transportation carbon emissions.

4. Regression Analysis

In this section, we firstly run a basic regression to investigate the impact of household

social/economic characteristics on carbon emissions. Then, we further explore the role

of each community’s building, neighborhood and location attributes on household

carbon emissions. In the end, we provide a robust check with alternative proxies for

community neighborhood and location effect, as well as sample-bias correction.

Variables employed in the empirical analysis are exhibited in Table 2. We find

significant differences in household characteristics by household’s car ownership

status and Hukou (see Table 3). Richer and younger households and those with larger

household size and local Hukou are more likely to own private cars. Households with

local Hukou are statistically richer and older.

Table 2 Variables descriptive statistics and definitions

Variable Definition Unit Mean Std. Dev.

Household carbon emissions from living

totalc household's annual carbon emissions Ton 7.399 4.717

residentialc household's residential carbon emissions Ton 5.193 2.499

transportc household's carbon emissions from

transportation

Ton 2.998 3.580

Household demographic/economic characteristics

income household annual disposable income 1000RMB 117.298 130.079

Hukou binary: 1=have local Hukou, 0=otherwise - 0.864 0.343

age the age of head Year 46.494 13.801

hhsize family size 1 3.127 1.105

Household preference

t_surround household's air-conditioner setting

temperature in summer

℃ 30.712 2.770

17

clean binary: 1=use coal gas, 0=otherwise - 0.094 0.293

carownership binary: 1=own private car, 0=otherwise - 0.558 0.658

Community building attributes

central binary: 1=the community has central

heating, 0=otherwise

- 0.778 0.416

size size of housing m2 90.788 38.793

hfloor the floor of the unit in the building 5.188 4.612

brick binary: 1=the building structure is brick,

0=otherwise

- 0.512 0.500

orientation binary: 1=the building faces south,

0=otherwise

- 0.745 0.436

Community neighborhood attributes

greenrate the green rate of the community 1 0.305 0.128

environment binary: 1=have park in 1.5km buffer,

0=otherwise

- 0.627 0.484

Community location attributes

subway distance to the nearest subway line m 1034.052 1114.779

edu distance to the nearest junior/middle school m 916.849 592.068

cvs distance to the nearest convenient store m 1044.485 797.013

tiananmen distance to Tiananmen m 13717.210 6898.698

Self-satisfaction grading

selfenvironment satisfaction level of the convenience of

community surrounded environment

1 3.644 0.953

selfconvennient satisfaction level of the convenience of

community location

1 3.658 0.940

Table 3 Differences in household characteristics (by carownership and Hukou)

Variables

without

cars

owning

cars Mean

Different

non-local

Hukou

local

Hukou Mean

Different Mean Mean Mean Mean

income 79.949 146.869 -66.920*** 104.196 119.353 -15.157

hukou 0.836 0.886 -0.050**

age 48.251 45.305 2.946*** 33.938 48.464 -14.526***

hhsize 2.912 3.291 -0.379*** 2.937 3.157 -0.220*

Note: *** denotes p <0.01, ** denotes p<0.05, *denotes p<0.1. The null hypothesis of t statistics is that

there is no difference in mean value between the given two groups. Significant Mean Diff indicates

that there exists a statistically meaningful difference in mean values between two sub-groups.

4.1 Role of Household Characteristics and Preference in Carbon Emissions

In this section, associations between household carbon dioxide emissions (including

residential and transportation emissions) and household demographic/economic

18

characteristics (including income, age, Hukou, and household size) and preferences

are investigated. The result is exhibited in Table 4.

Column (1) only includes household income, household size and age of household

head in the test. It indicates that the elasticity of household income to carbon

emissions is significant with a value of 0.32. Column (2) indicates the role of

household preference in household carbon emissions. We can find that t_surround is

significantly negative correlated with carbon emissions, which suggests that

household consumption preferences for air-conditioner temperature significantly

impact emissions. In addition, the positive coefficient of d_coalgas implies that

households that prefer green energy, such as coal gas, produce less carbon emissions.

Coal gas has a lower Default Carbon Content than LPG and natural gas.3 The positive

coefficient of carownership indicates that carbon emissions are strongly positively

associated with the possibility of owning a car.

In columns (3) and (4), we further divide the total emissions into residential carbon

emissions and transportation carbon emissions. This indicates both family size and

age of head have strong positive association with residential emissions.4 Regarding

household preference, we find that household consumption preferences measured by

preferred temperature in summer, green consumption attitude and car ownership

status, are all significant.

Table 4 Role of household in carbon dioxide emission

(1) (2) (3) (4)

lntotalc lntotalc lnresidentialc lntransportc

Household demographic/economic characteristics

lnincome 0.323*** 0.118*** 0.122*** 0.285***

(0.0272) (0.0253) (0.0237) (0.0603)

Hukou 0.167*** 0.0665 0.121** -0.0730

(0.0599) (0.0507) (0.0524) (0.118)

3 Default Carbon Content means that the carbon emissions per Joule and the above three cooking energy’s Default

Carbon Content are separately 12.1TC/TJ, 17.2 TC/TJ and 15.3 TC/TJ. 4 We also involve the square of income into the regression. The square is not significant.

19

age 0.0000750 0.00277** 0.00538*** -0.00618*

(0.00151) (0.00130) (0.00133) (0.00316)

hhsize 0.0750*** 0.0371** 0.0660*** -0.00419

(0.0176) (0.0149) (0.0153) (0.0367)

Household preference

t_surround -0.0151*** -0.0109*

(0.00577) (0.00595)

d_coalgas -0.180*** -0.160***

(0.0558) (0.0577)

carownership 0.463***

1.523***

(0.0277)

(0.0638)

Other control variables

education controlled controlled controlled controlled

occupation controlled controlled controlled controlled

standard errors clustered by

community

controlled controlled controlled controlled

_cons -2.234*** 0.415 -0.0541 -3.757***

(0.318) (0.345) (0.344) (0.695)

N 825 792 792 608

R-sq 0.205 0.425 0.119 0.586

Note: 1. The dependent variable in columns (1) and (2) is the logarithm of household total carbon

dioxide emissions. The dependent variable in column (3) is the logarithm of household

residential carbon dioxide emissions. The dependent variable in column (4) is the

logarithm of household carbon dioxide emissions from private transportation.

2. Robust standard errors are reported in parentheses: *** denotes p <0.01, ** denotes

p<0.05, and *denotes p<0.1.

3. Prefix l denotes taking the logarithm of the variable.

4.2 Role of Community Attributes in Household Carbon Emission

Based on the benchmark result in Table 5, we further investigate the role of

community in household carbon emissions.

Table 5 Role of community in household carbon emissions

(1) (2) (3) (4)

lntotalc lnresidentialc lnresidentialc lntransportc

Community building attributes

central 0.252*** 0.346*** 0.344***

(0.0374) (0.0354) (0.0371)

lnsize 0.528*** 0.607*** 0.601***

(0.0373) (0.0367) (0.0364)

hfloor -0.00547 -0.0144*** -0.0133***

(0.00356) (0.00347) (0.00350)

20

orientation -0.0738** -0.0809** -0.0767**

(0.0351) (0.0354) (0.0352)

brick -0.0195 -0.0695** -0.0613**

(0.0312) (0.0305) (0.0308)

Community neighborhood attributes

greenrate 0.00649

-0.0344**

(0.0166)

(0.0160)

environment -0.0338***

-0.0251**

(0.0125)

(0.0126)

Community location attributes

lnsubway -0.00128

0.0229

(0.00782)

(0.0195)

lnedu 0.0413**

0.159***

(0.0209)

(0.0561)

lncvs 0.0579***

0.0708*

(0.0146)

(0.0373)

Household demographic/economic characteristics

controlled controlled controlled controlled

Household preference

controlled controlled controlled controlled

Other control variables

education controlled controlled controlled controlled

occupation controlled controlled controlled controlled

standard errors clustered by

community

controlled controlled controlled controlled

_cons -1.298*** -1.073*** -1.344*** -5.862***

(0.387) (0.314) (0.336) (0.877)

N 759 763 761 792

R-sq 0.592 0.392 0.408 0.184

Note: 1. The dependent variable in (1) is the logarithm of household total carbon emissions. The

dependent variable in (2), (3) and (4) is the logarithm of residential carbon emissions.

The dependent variable in (5) is the logarithm of transportation carbon emissions.

2. Robust standard errors are reported in parentheses: *** denotes p <0.01, ** denotes

p<0.05, and *denotes p<0.1.

3. Prefix ln denotes taking the logarithm of the variable.

Column (1) exhibits the compound impact of both household and community factors

on household total carbon emissions. Basically, after controlling the household’s basic

demographic/economic characteristics and preference, very few community attributes

are significant. Considering the heterogeneity of carbon emissions in residential and

transportation factors, we further decompose total emissions into two exclusive parts.

21

Columns (2) and (3) show the role of building and neighborhood attributes of the

community in residential carbon emissions respectively. In column (2), we find that

house size significantly impacts residential emissions and the elasticity of house size

to residential emissions is 0.61. We also find a negative role of hfloor and orientation

on residential carbon emissions, which denotes that units located on the higher floors

and facing the south side have better daylight conditions, which might significantly

reduce energy consumption for lighting and heating. More interestingly, we find that a

community with better thermal insulation structures, such as brick-concrete structures,

can effectively improve energy efficiency and reduce household energy consumption

and emissions. We also find an energy saving role for the decentralized heating model.

Although central heating has relatively low energy consumption per unit size, it

constantly heats for 24 hours each day, and actually results in higher energy

consumption than does decentralized heating.

We further study the relationship between neighborhood characteristics of community

and residential emissions, which is exhibited in column (3). It is found that, all else

equal, emissions from household living are significantly decreased in a community

with a higher green rate, and with a park within its 1.5 km neighborhood. This could

be directly related to the positive externality of greening (Boyle and Kiel, 2001).

Given that a large proportion of household energy consumption is used to maintain

indoor temperature through heating and cooling systems, the thermal properties of

green land, such as heat flux reduction and solar reflectivity, are highlighted in energy

consumption. Some experiments indicate that enclosure greening could lead to 31% to

44% energy savings compared with non-insulated materials (Castleton et al., 2013;

Liu and Minor, 2005)5, which suggests that greening can optimize heat exchange

mechanisms by keeping indoor temperature and thus reducing energy consumption

and carbon emissions. Meanwhile, a high-amenity environment can raise public

environmental concern, which has significantly positive effects on shaping

5 Enclosure greening could absorb the radiant heat of the sun in the summer and reduce wind speeds in the winter,

eliminating the temperature difference, which helps enhance a building’s thermal resistance.

22

low-carbon behaviors and behavioral intentions (Minton and Rose, 1997), and thus

may reduce energy consumption and carbon emissions.

The last column explores the location effect of community on household

transportation carbon emissions. After controlling car ownership status, we still find a

significant positive role for lnedu and lncvs, which indicates the distance between

residence and junior/middle school, as well as the distance between residence and

convenience store have positive impacts on transportation carbon emissions. Driving

children to school and purchasing daily necessities are suggested as important energy

consumption activities in addition to regular commuting. This suggests a connection

between urban infrastructure, services and household energy consumption.

4.3 Robust Test

Considering some latent variables might be unobserved in investigating the role of

community in household carbon emissions, we also use self-report information

regarding community environment and location condition for a robust test. Compared

with specific indicators included in Table 5, self-reported information is regarded as

containing more comprehensive information, which could compensate for the

endogenous problem caused by the latent variable. selfenvironment in Table 6 denotes

the satisfaction level regarding the neighborhood environment, as reported by the

individual living in this community, while selfconvenient refers to the satisfaction

level with the convenience of the community location. We discover that the higher the

level of satisfaction with neighborhood attributes, the lower the household’s

residential carbon dioxide emissions. Similarly, the higher the level of satisfaction

with community location attributes, the lower the carbon emissions from

transportation. It is interesting to notice that after controlling other factors, we find a

higher absolute elasticity of the “satisfaction level with the convenience of the

community location” than the “satisfaction level regarding the neighborhood

environment”, which implies that higher satisfaction regarding location is more

23

efficient in reducing household energy consumption than satisfaction regarding

neighborhood. This also suggests the important role of spatial planning of residential

project in the city.

Table 6 Role of community in household carbon emissions (self-reported information)

(1) (2) (3)

lntotalc lnresidentialc lntransportc

Household demographic/economic characteristics

lnincome 0.00907 -0.0193 0.302***

(0.0240) (0.0224) (0.0605)

hukou -0.000105 0.00216* -0.00602*

(0.00119) (0.00119) (0.00315)

age 0.0101 0.0390*** -0.0124

(0.0138) (0.0138) (0.0368)

hhsize 0.00907 -0.0193 0.302***

(0.0240) (0.0224) (0.0605)

Household preference

t_surround -0.0122** -0.00763

(0.00517) (0.00517)

d_coalgas -0.130** -0.0998**

(0.0505) (0.0506)

carownership 0.410***

1.526***

(0.0252)

(0.0638)

Community building attributes

central 0.262*** 0.343***

(0.0359) (0.0357)

lnsize 0.553*** 0.602***

(0.0373) (0.0369)

hfloor -0.0101*** -0.0146***

(0.00347) (0.00348)

brick -0.0248 -0.0622**

(0.0312) (0.0307)

orientation -0.0623* -0.0779**

(0.0353) (0.0354)

Community neighborhood attributes (self-reported)

selfenvironment -0.0302** -0.0234*

(0.0128) (0.0126)

Community location attributes (self-reported)

selfconvenient -0.0142

-0.103***

(0.0157)

(0.0400)

Other control variables

education controlled controlled controlled

24

occupation controlled controlled controlled

standard errors clustered by community controlled controlled controlled

_cons -0.572* -0.985*** -3.566***

(0.337) (0.318) (0.704)

N 759 759 604

R-sq 0.580 0.395 0.592

Note: 1. The dependent variable in column (1) is the logarithm of household carbon dioxide

emissions. The dependent variable in column (2) is the logarithm of household

residential carbon dioxide emissions. The dependent variable in column (3) is the

logarithm of household carbon dioxide emissions from private transportation.

2. Robust standard errors are reported in parentheses: *** denotes p <0.01, ** denotes

p<0.05, and *denotes p<0.1.

3. Prefix ln denotes taking the logarithm of the variable.

It should be noted that in Table 4, Table 5 and Table 6, the significantly positive role

of car ownership might be questioned due to a structural difference between

households with private cars and households without private car. Thus, in this section,

we further compare the heterogeneity between these two sub-samples. In order to

avoid selection bias, we use a two-stage Heckman model (Heckman, 1979) to study

the carbon emissions from private cars (the carbon emissions from a taxi is only 3.4

percent of the total carbon emissions from household living) and its main

determinants. Using the Heckman procedure, we are able to control selectivity and

simultaneity bias in the test. We incorporate an Inverse Mills Ratio (lambda) into the

equation, , where demonstrates the normal density function.

Inverse Mill’s ratio (lambda) is estimated from probit model with full sample. The

result is shown in Table 7.

From the first step regression, we find the influential factors on the probability a

household owns a private car. We use the distance from residential communities to

Beijing’s main center (Tiananmen) to control the degree of spatial match for each

household residence and workplace.6 From the perspective of household social or

6 This process is relatively rough. Hence, we will calculate the distance from household’s residential community

to their workplace for each sample through improving GIS spatial database system and use these data to study the

25

economic characteristics, we find that residents with higher income, younger age and

larger family size tend to have higher probability of purchasing cars. In addition, the

spatial location in which households live also has significant impact on the probability

of purchasing a car.

From the second step regression, we focus on the different level of transportation

carbon emissions in sub-groups. The regression result shows that for households who

already own cars, gasoline consumption is not significantly impacted by household

income. However, for those who do not own private cars, the household income

matters in vehicle choice, because higher income groups might take a taxi rather than

taking the subway or the bus.

Table 7 Comparisons between household with private car and household without private car

(1) (2) (3)

carownership lntransportc Lntransportc

Household demographic/economic

characteristics

lnincome 0.499*** -0.549 2.389**

(0.0679) (0.461) (0.926)

hukou 0.322** -0.853** 1.062

(0.145) (0.344) (0.669)

age -0.00978*** 0.0102 -0.0406**

(0.00370) (0.0103) (0.0201)

hhsize 0.146*** -0.237 0.337

(0.0418) (0.144) (0.282)

lntiananmen 0.190*

(0.0783)

Community location attributes

lnsubway 0.00848 0.0101

(0.0193) (0.0394)

lnedu 0.260*** 0.172

(0.0536) (0.110)

lncvs -0.00720 0.146**

(0.0409) (0.0741)

Other control variables

lambda -2.320 3.615

impact of spatial match degree on carbon dioxide emissions from private car.

26

(1.463) (2.771)

education controlled controlled controlled

occupation controlled controlled controlled

standard errors clustered by community controlled controlled controlled

_cons -6.414*** 8.853 -33.67**

(1.110) (7.014) (14.00)

N 825 329 279

Log likelihood -505.651

LR chi2 99.97

Prob > chi2 0.000

R-sq 0.131 0.345

Note: 1. The dependent variable in column (1) is the dummy variable of household car ownership.

The dependent variable in column (2) and (3) is the logarithm of household carbon

dioxide emissions from private transportation.

2. Robust standard errors are reported in parentheses: *** denotes p <0.01, ** denotes

p<0.05, and *denotes p<0.1.

3. Prefix ln denotes taking the logarithm of the variable.

5. Conclusion

Based on the household survey of “household energy consumption and living

conditions in Beijing”, conducted by the Institute of Real Estate Studies at Tsinghua

University in 2009, we estimate the carbon emissions from household living,

including residential carbon emissions and carbon emissions from private

transportation of 826 urban households living in 38 residential communities. We

empirically explore the impacts of building attributes, neighborhood attributes and

location attributes on household carbon emissions. We find that these three

dimensions are integrated together to shape household energy consumption behaviors.

Specifically, for building attributes, within-building positions (in terms of orientation

and height) with better daylight conditions, thermal insulation structure and

decentralized heating system all contribute to energy savings and lower carbon

emissions. For neighborhood attributes, green space leads to lower emissions directly

and indirectly. For location attributes, better accessibilities to desirable urban

amenities such as schools, parks and convenient stores will decrease a household’s car

ownership probability and thus reduce transportation carbon emissions.

27

Our major findings are consistent with existing studies in terms of the correlations

between residential carbon emissions and building/location attributes. We also make

several new contributions to the existing literature. We provide the robust and

bias-corrected new evidences of the energy consumption elasticity in China, and

highlight the role of neighborhood attribute (amenity) as well. Meanwhile, the role of

neighborhood and location attribute of community in household energy consumption

behavior is tested from both subjective and objective measures. In addition, we also

find some additional interesting results. Firstly, compared with residential carbon

emissions, carbon emission from private transportation is more unevenly distributed

at community level, which implies a relatively important role of community’s location

in shaping household carbon emissions. Secondly, the neighborhood greening could

help to optimize local heat exchange and promote public environmental concerns,

thus reduce the residential carbon emission. Existing studies on this issue is mainly

qualitative analysis and focus on the field of greening in enclosure structures. Thirdly,

after controlling other factors, we find that higher satisfaction regarding location is

more efficient in reducing household energy consumption than satisfaction regarding

environment.

Our empirical findings have clear implications for both urban planners and real estate

developers. Urban planners should pursue a better spatial match between where

people are and the locations of urban amenities. This is very important because large

Chinese cities are undergoing fast population suburbanization, but the provision of

local public services is clearly lagging behind and most high-quality local public

services are still concentrated in the inner city. If people have to travel long distances

to access those amenities, they are more likely to drive cars, which will cause traffic

congestion and also lead to rising energy consumption and carbon emissions. From

the real estate developers’ perspective, all else equal, if homebuyers know they can

reduce energy expenditures in such “green” communities, they are willing to pay a

higher purchase price. Such a green premium will generate the incentive for

profit-driven developers to pursue energy-efficient designs for both buildings and

communities.

28

Appendix: Measurement of Carbon Emissions

Household carbon emissions are divided into residential carbon emissions and private

transportation carbon emissions. The measurement process is as follows.

We consider the three main sources of residential carbon emissions: household

electricity, home heating and residential cooking; the translation from energy

consumption to carbon emissions is relatively simple. For each household, we begin

by estimating every single component and then aggregate them into the total

residential carbon emissions. In the case of each component, we use conversion

factors to translate energy used into carbon emissions. To determine the carbon

emissions impact of different kinds of energy, we use net calorific value and default

carbon content for each component.

Carbon emissions from private transportation mainly consist of gasoline consumption

for private car and taxi. For each household travelling by private transportation, we

separately estimate the carbon emissions from gasoline consumption for private car

and taxi, and then aggregate them into the total carbon emissions from private

transportation. But private car and taxi involve a relatively complex translation from

the information from our survey to carbon emissions, as can be seen in the following

equations.

where

29

Gasoline consumption per km equals 0.1L/km. Taxi price in Beijing is composed of

two parts: one is 10 RMB yuan for the first three kilometers and the other is 2 RMB

yuan per kilometer beyond the first three kilometers. denotes the rate of a

taxi when it is driving without passenger, which reflects the fact that the taxi

consumes gasoline when there is no passenger in it. The data source and calculation

methodology of household carbon emissions from both residence and private

transportation are shown in Table A.

Table A The calculation matrix of household carbon emissions

Terms Item Data and Methodology

Residential carbon emissions

electricity formula Household electricity consumption× carbon dioxide emission factor

per kWh.

source “Household electricity consumption” (our micro data)

“carbon dioxide emission factor of the North Grid” (Office of National

Coordination Committee on Climate Change, a department within the

National Development and Reform Commission)

central

heating

formula Household dwelling size× carbon dioxide emission factor of Central

heating per m2

source “House size”(our micro data)

“carbon dioxide emission factor of Central heating” (Research Centre

for Building Energy-Saving at Tsinghua University)

cooking formula Household energy consumption (coal, coal gas and LPG) × carbon

dioxide emission factor of the specific energy per unit

source “Household energy consumption of coal LPG and natural gas” (our

micro data)

“carbon dioxide emission factor of the specific energy” (IPCC2006)

transportation carbon emissions

private

car

formula Household expenditure on gasoline/ gasoline price× carbon dioxide

emission factor of gasoline

source “Household expenditure on gasoline of private car” (our micro data)

“gasoline price” (National Development and Reform Commission)

“carbon dioxide emission factor of gasoline” (IPCC2006)

taxi formula Household expenditure on gasoline of taxi/ unit price of taxi ÷

gasoline consumption per 100KMs of taxi ÷(1—vacant)×carbon

dioxide emission factor of gasoline

30

source “Household expenditure on gasoline of taxi” (our micro data)

“unit price of taxi” (National Development and Reform Commission)

“gasoline consumption per km of taxi and its vacant” (National

average data)

“gasoline price” (Department of National Development and Reform)

31

References

1. Alfredsson, E. C. (2004). “Green” Consumption—No Solution for Climate

Change. Energy, 29(4), 513-524.

2. Bin, S., & Dowlatabadi, H. (2005). Consumer Lifestyle Approach To US Energy

Use and the Related CO 2 Emissions. Energy Policy, 33(2), 197-208.

3. Börjesson, P., & Gustavsson, L. (1998). Greenhouse Gas Emission from Building

Construction in a Life-Cycle Perspective--Wood or Concrete Buildings? (No.

CONF-980815--). Lund University.

4. Boyle, M., & Kiel, K. (2001). A Survey of House Price Hedonic Studies of the

Impact of Environmental Externalities. Journal of Real Estate Literature, 9(2),

117-144.

5. Braubach, M., & Fairburn, J. (2010). Social Inequities in Environmental Risks

Associated With Housing and Residential Location—A Review of Evidence. The

European Journal of Public Health, 20(1), 36-42.

6. Can, A. (1992). Specification and Estimation of Hedonic Housing Price

Models. Regional Science and Urban Economics, 22(3), 453-474.

7. Castleton, H. F., Stovin, V., Beck, S. B. M., & Davison, J. B. (2010). Green Roofs;

Building Energy Savings and the Potential for Retrofit. Energy and

buildings, 42(10), 1582-1591.

8. Chang, H., Parandvash, G. H., & Shandas, V. (2010). Spatial Variations of

Single-Family Residential Water Consumption in Portland, Oregon. Urban

Geography, 31(7), 953-972.

9. Chen, F., & Zhu, D. (2013). Theoretical Research on Low-Carbon City and

Empirical Study of Shanghai. Habitat International, 37, 33-42.

10. Choy, L. H., Ho, W. K., & Mak, S. W. (2013). Toward a low carbon Hong Kong:

A proposal from the institutional perspective. Habitat International, 37, 124-129.

11. Clapham, D. (2005). Identity and Lifestyle in the Residential Environment. In

conference Proceeding, Paper Voorgedragen Op De Conferentie ‘Doing,

Thinking, Feeling Home: The Mental Geography Of Residential Environments’,

32

Delft (Vol. 14).

12. Dalton, M., Jiang, L., Pachauri, S., & O’Neill, B. C. (2007, March). Demographic

Change and Future Carbon Emissions in China and India. In Unpublished Update

of the 16 March 2007 Draft Presented At The Annual Meeting Of The Population

Association Of America, New York, NY (Pp. 29-31).

13. Dhakal, S. (2009). Urban Energy Use and Carbon Emissions from Cities in China

and Policy Implications. Energy Policy, 37(11), 4208-4219.

14. Dietz, T., Gardner, G. T., Gilligan, J., Stern, P. C., & Vandenbergh, M. P. (2009).

Household Actions Can Provide A Behavioral Wedge To Rapidly Reduce US

Carbon Emissions. Proceedings Of The National Academy Of Sciences,106(44),

18452-18456.

15. Druckman, A., & Jackson, T. (2009). The Carbon Footprint of UK Households

1990–2004: A Socio-Economically Disaggregated, Quasi-Multi-Regional

Input–Output Model. Ecological Economics, 68(7), 2066-2077.

16. Du, Y. W., Huang, T. Z., & Kang, G. D., (2015). A Study of Urban Household

Carbon Emission Characteristic and Its Impact Factors from the Micro

Perspective: from the Survey Data of the Urban Household Activities in Jiangsu

Province. Population & Economics. (2), 30-39.

17. Ewing, R., & Rong, F. (2008). The Impact Of Urban Form On US Residential

Energy Use. Housing Policy Debate, 19(1), 1-30.

18. Flora, C. B., & Flora, J. L. (1993). Entrepreneurial Social Infrastructure: A

Necessary Ingredient. The Annals of the American Academy of Political and

Social Science, 529(1), 48-58.

19. Gastwirth, J. L. (1972). The Estimation of the Lorenz Curve and Gini Index. The

Review of Economics and Statistics, 306-316.

20. Genjo, K., Tanabe, S. I., Matsumoto, S. I., Hasegawa, K. I., & Yoshino, H. (2005).

Relationship between Possession of Electric Appliances and Electricity for

Lighting and Others in Japanese Households. Energy and Buildings, 37(3),

259-272.

21. Groot, L. (2010). Carbon Lorenz Curves. Resource and Energy Economics, 32(1),

33

45-64.

22. Hammond, G. P., & Jones, C. I. (2008). Embodied Energy and Carbon in

Construction Materials. Proceedings of the Institution of Civil

Engineers-Energy, 161(2), 87-98.

23. Han, Y., Dai, L. M., Zhao, X. F., Yu, D. P., & Wu, S. N. (2008). Construction And

Application of an Assessment Index System for Evaluating the Eco-Community’s

Sustainability. Journal of Forestry Research, 19(2), 154-158.

24. Holden, E., & Norland, I. T. (2005). Three Challenges for the Compact City as a

Sustainable Urban Form: Household Consumption of Energy and Transport in

Eight Residential Areas in the Greater Oslo Region. Urban Studies, 42(12),

2145-2166.

25. Kaza, N. (2010). Understanding the Spectrum of Residential Energy

Consumption: a Quantile Regression Approach. Energy Policy, 38(11),

6574-6585.

26. Kohlhuber, M., Mielck, A., Weiland, S. K., & Bolte, G. (2006). Social Inequality

in Perceived Environmental Exposures in Relation to Housing Conditions in

Germany. Environmental Research, 101(2), 246-255.

27. Li, Z., Li, P., Guo, J., & Zeng, X. F. (2013). Impact Factors Estimation and

Research on the Differences across Cities of Residential CO2 Emissions in

Chinese Major Cities. China Population, Resource and Environment, 23(10),

87-94.

28. Li, J., & Yang, T. (2015). China's Eco-City Construction. Springer, New York.

29. Liu, K., & Minor, J. (2005). Performance Evaluation of an Extensive Green Roof.

Presentation at Green Rooftops for Sustainable Communities, Washington DC,

1-11.

30. Ma, J., Heppenstall, A., Harland, K., & Mitchell, G. (2014). Synthesizing Carbon

Emission for Mega-Cities: A Static Spatial Microsimulation of Transport CO2

from Urban Travel in Beijing. Computers, Environment and Urban Systems, 45,

78-88.

31. Minton, A. P., & Rose, R. L. (1997). The Effects of Environmental Concern on

34

Environmentally Friendly Consumer Behavior: An Exploratory Study. Journal of

Business Research, 40(1), 37-48.

32. Pinjari, A. R., Bhat, C. R., & Hensher, D. A. (2009). Residential Self-Selection

Effects in an Activity Time-Use Behavior Model. Transportation Research Part B:

Methodological, 43(7), 729-748.

33. Qin, B., & Han, S. S. (2013). Planning Parameters and Household Carbon

Emission: Evidence from High-and Low-Carbon Neighborhoods in Beijing.

Habitat International, 37, 52-60.

34. Reid, L. A., & Houston, D. (2013). Low Carbon Housing: a ‘Green’ Wolf in

Sheep's Clothing? Housing Studies, 28(1), 1-9.

35. Seryak, J., & Kissock, K. (2003). Occupancy and Behavioral Affects on

Residential Energy Use. In Proceedings of the Solar Conference (Pp. 717-722).

American Solar Energy Society.

36. Stern, P. (2000). Toward a Coherent Theory of Environmentally Significant

Behavior. Journal of Social Issues, 56(3), 407-424.

37. Stern, S. A., Kawakami, H., Houde, A. Y., & Zhou, G. (1997). U.S. Patent No.

5,591,250. Washington, DC: U.S. Patent and Trademark Office.

38. Van Raaij, W. F., & Verhallen, T. M. (1983). A Behavioral Model of Residential

Energy Use. Journal of Economic Psychology, 3(1), 39-63.

39. Waddell, P. (2001). Towards a Behavioral Integration of Land Use and

Transportation Modeling.

40. Wang, Q. C. (2015). Effects of Household Size on Energy Consumption and

Carbon Emissions. Resources Science, 37(2): 0299-0307.

41. Yang, Z., Wu, S. P., & Cheung, H. Y., (2016). From Income and Housing Wealth

Inequalities to Emission Inequality: Carbon Emission of Household in China.

Journal of Housing and Built Environment, 2016. 1-22.

42. Yu, B., Y., Zhang, J., & Fujiwara, A. (2012). Analysis of the Residential Location

Choice and Household Energy Consumption Behavior by Incorporating Multiple

Self-Selection Effects. Energy Policy, 46, 319-334.

35

43. Zhang, X., Shen, G. Q., Feng, J., & Wu, Y. (2013). Delivering a Low-Carbon

Community in China: Technology vs. Strategy? Habitat International, 37,

130-137.

44. Zhu Z, Shen Y Q, Huang M. (2011). Empirical Study on Low- Carbon

Consumption and Factors of Carbon Emission-Based On Hangzhou. Survey

Resource Development and Market, 27(09), 831-834.

45. Zhu, J., Chew, D. A., Lv, S., & Wu, W. (2013). Optimization Method for Building

Envelope Design to Minimize Carbon Emissions of Building Operational Energy

Consumption using Orthogonal Experimental Design (OED). Habitat

International, 37, 148-154.