You are on page 1of 11

Ecological Economics 160 (2019) 145–155

Contents lists available at ScienceDirect

Ecological Economics
journal homepage: www.elsevier.com/locate/ecolecon

Analysis

The Impact of Social Awareness and Lifestyles on Household Carbon T


Emissions in China
Jun Lia, Dayong Zhanga,c, Bin Sub,

a
Research Institute of Economics and Management, Southwestern University of Finance and Economics, China
b
Energy Studies Institute, National University of Singapore, Singapore
c
Department of Economics, University of Birmingham, UK

ARTICLE INFO ABSTRACT

Keywords: Household carbon emissions account for a significant proportion of the national total emissions. Using a unique
Carbon emissions nation-wide survey data, this paper investigates the impact of different subjective measures of social awareness
China on household carbon emissions in China. Subjective variables are found to have significant impacts on household
Consumer lifestyle carbon emissions. Feeling secure, compliance with rules, and happiness have negative and significant effects on
Household survey
household carbon emissions, whereas the impact of interest in social issues is positive, which indicates the
Social awareness
existence of a gap between awareness and behavior. The impacts of consumer lifestyles on the relationship
between awareness and household emissions are also explored and are found to play a significant role in the
relationship between awareness and emissions. People with a frugal lifestyle not only associate with lower
emissions but also reinforce the role of social awareness. Furthermore, substantial differences are found across
rural/urban areas and regional and income groups in China. The government should further promote a green
lifestyle and give households more incentives to engage in green consumption behavior.

1. Introduction environmental responsibility. In 2009, the Chinese government decided


to reduce the country's carbon intensity (carbon emissions per unit of
Average global temperature was 1.3 °C higher in 2016 than in 1880. the gross domestic product or GDP) by 40–45% in 2020 compared to
Climate experts predict that global temperature will go up 1.1–6.4 °C in the level in 2005 and increase the share of non-fossil fuel-based energy
this century, and sea levels will rise 16.5–53.8 cm (IPCC, 2007). The in primary energy consumption by 15% in 2020 (Jennings et al., 2011).
urgency to control global warming (i.e., limiting the temperature in- In addition, China pledged to peak its carbon emissions by 2030. This is
crease to 2 °C) and mitigate greenhouse gas (GHG) emissions resulted in an ambitious goal, which requires a combination of mitigation policies.
the Paris Agreement signed in the late 2015. Carbon dioxide (CO2) is Moreover, China recently announced the intention to reduce its energy
the main GHG that leads to global warming and climate change (World intensity by 15% and carbon intensity by 18% compared to the level in
Bank, 2007). Emissions by households in particular contribute sig- 2015 over the course of its thirteenth five-year plan (2016–2020) (State
nificantly to the total value. In developed economies, such as the United Council, 2016).
States and the United Kingdom, the share of household emissions can China's household carbon emissions in 2000–2010 account for
exceed 70% of the total (Baiocchi et al., 2010; Bin and Dowlatabadi, 30%–40% of the national total, a percentage that continues to increase
2005). In developing countries, the share is smaller, but improvements (Liu et al., 2011; Wang and Yang, 2014). Household carbon emissions
in living standards have increased consumer demand and thereby can result from the direct use of fossil fuels and electricity, or indirectly
caused rapid expansion in energy requirements, which creates a new from consuming final goods and services. Carbon emissions generated
source of growth in carbon emissions (Adaman et al., 2011; Das and by production are essentially driven by final consumer demand. In this
Paul, 2014; Su and Ang, 2017). sense, production and emissions associate with it are linked to con-
After China's economic reforms and opening up in 1978, the coun- sumer behavior by individuals/households. It is therefore important to
try's CO2 emissions maintained an average annual growth rate of 5.8%, understand the demand side of this issue. In other words, policy makers
and in 2008 China surpassed the United States to become the largest need to understand the factors that drive emissions by exploring Chi-
CO2 emitter. China faces a challenge in balancing economic growth and nese households' consumption behavior.


Corresponding author at: Energy Studies Institute, National University of Singapore, Singapore
E-mail addresses: subin@nus.edu.sg, subin.nus@gmail.com (B. Su).

https://doi.org/10.1016/j.ecolecon.2019.02.020
Received 31 July 2018; Received in revised form 7 January 2019; Accepted 21 February 2019
0921-8009/ © 2019 Elsevier B.V. All rights reserved.
J. Li, et al. Ecological Economics 160 (2019) 145–155

Recent studies have shown increasing interest in this issue and in- (2007) is one of the earliest studies focused on household carbon
vestigated carbon emissions in China at the household level. These emissions based on macroeconomic data in China. They use the con-
studies also pay attention to the factors that may influence emissions sumption lifestyle approach (CLA) to calculate direct and indirect
including income, household size, and household structure. Because of household emissions and investigate the impact of consumer attitudes
the availability of representative micro-level survey data, most of the in China's rural and urban households. Following Wei et al. (2007), the
existing studies are based on censor data (Wei et al., 2007; Liu et al., evolution of Chinese household carbon emissions and the impacts of
2011) or small-scale survey data (Qu et al., 2013; Yang et al., 2016). household consumption on energy use and CO2 emissions have been
Only a few works, such as Golley and Meng (2012) and Xu et al. (2016), explored (e.g. Feng et al., 2011; Liu et al., 2011). Popular methods used
investigate carbon emissions in China using larger-scale survey data but to investigate the factors that influence Chinese household carbon
only for urban areas. emissions include: input-output structural decomposition analysis (Zhu
Behavioral economics suggests that subjective factors, such as be- et al., 2012; Zhang et al., 2017), mixed recursive dynamic general
liefs and preferences, can affect individual decision-making (Kahneman equilibrium model (Dai et al., 2012), gray relational analysis (Feng
and Tversky, 1979). Based on this theory, some recent studies have et al., 2011), and hypothetical extraction method (Zhang et al., 2018).
explored the relationship between household emissions and individual Household level studies in China have appeared recently, for ex-
subjective factors, such as subjective well-being (Wilson et al., 2013) ample, Qu et al. (2013) used a small-scale survey (only 125 rural
and environmental beliefs (Bai and Liu, 2013). Because of the looming households) in Gansu, Qinghai, and Ningxia Provinces and find that it is
threat of climate change and environmental problems in China, re- difficult to reduce household carbon emissions in northwestern arid-
inforced by an increasing amount of public outreach, people now have alpine regions in China. Based on a survey in Beijing (826 respondents),
more information and knowledge than before about carbon emissions. Yang et al. (2016) showed that ecofriendly communities have sig-
The question is however, whether this social awareness will actually nificantly lower residential carbon emissions in Beijing. A few large-
lead to reductions in carbon emissions. In addition, subjective factors sample survey datasets have been used to investigate Chinese house-
are difficult to quantify, and thus require a combination of measures in hold carbon use, such as the 2011 Survey of Consumer Finance com-
a more representative large-scale survey. prising 5761 urban households in 24 cities (Xu et al., 2016) and the
Based on these arguments, this paper contributes to the literature by 2005 Urban Household Income and Expenditure Survey (Golley and
investigating the impact of subjective factors on household carbon Meng, 2012).
emissions and further explores the impact of consumer lifestyle on the Some studies have found that awareness of social and environ-
relationship between awareness and household carbon emissions in mental problems might inspire people to change their lifestyle
China. We use a unique nationwide household level survey data to (Gadenne et al., 2011; Borgstede et al., 2013; Brounen et al., 2013) and
perform empirical analysis. Our specific contributions include the fol- then affect consumption behavior (Carter, 2011; Kaida and Kaida,
lowing: first, this is the first study to consider the impacts of various 2016). Papers in psychology posit that awareness could affect in-
subjective factors—such as a sense of personal safety, compliance with dividual behavior (Ajzen and Fishbein, 1980). Environmental aware-
traffic regulations, happiness, and attention to social issues—on ness has high explanatory power in green purchasing behavior
household emissions in China; second, we consider the differences in (Schlegelmilch et al., 1996), thus, it can affect green purchasing be-
rural and urban areas, across regions, and income levels based on na- havior among consumers (Peattie, 2001). However, awareness might
tionwide large-scale survey data; third, we explore the effect of con- not easily translate into direct pro-environmental behavior (Wang et al.,
sumer lifestyles on the relationship between subjective factors and 2010).
household carbon emissions. In the literature, only a few studies have investigated the relation-
The remainder of this paper is organized as follows: Section 2 re- ship between awareness and household emissions from a micro-per-
views the relevant literature on household carbon emissions and its spective. For example, Wilson et al. (2013) used a sample of 1920 Ca-
determinants. Section 3 introduces the data, the method for calculating nadian respondents and find that carbon emissions and subjective well-
household carbon emissions, subjective factors, and other variables and being are not closely related. The result is similar to that found by
then illustrates the model used in this paper. Section 4 reports and Andersson et al. (2014) in a Swedish survey of 1000 respondents, and
discusses the main results obtained, explaining the sources of household Sekulova and van den Bergh (2013) who used a sample of 840 Spanish
carbon emissions and the impact of subjective factors and attitudes respondents. Bai and Liu (2013) used a sample of 354 respondents to
toward consumption. Section 5 concludes with further discussion on explore the link between low-carbon awareness and behavior among
policy implications. residents in Tianjin, China. They find that the level of behavior is higher
than the level of awareness, indicating a gap between low-carbon
2. Literature Review awareness and behavior. Their findings are similar to the results re-
ported in other studies—that the level of awareness exceeds behavior
Carbon emissions in household level have been a hot topic in recent (Van Raaij and Verhallen, 1983; Owens and Driffill, 2008).
years. Many studies have shown the importance of household emissions Based on these discussions, it is clear that existing studies based on
and further investigated its sources and driving factors. Some of them micro-survey data either use a small-scale dataset or focus only on
focus on the relationship between household energy consumption and urban households, and limited works use a national large-scale micro-
direct carbon emissions (e.g., Biesiot and Noorman, 1999). Others level dataset to investigate household carbon emissions issues in China.
consider consumption patterns that have significant impacts on both In addition, existing studies on the relationship between subjective
direct and indirect household carbon emissions (Weber and Perrels, factors and household emissions only explore the relationship between
2000; Bin and Dowlatabadi, 2005; Duarte et al., 2010; Su et al., 2017). subjective well-being, environmental beliefs or attitudes, with house-
Factors shown to affect household carbon emissions include hold emissions based on a small-scale micro-level dataset. More im-
household income (Druckman and Jackson, 2009; Büchs and Schnepf, portantly, none of them consider the effect of consumer lifestyle on the
2013; Su et al., 2017), the size of households (Weber and Matthews, relationship between subjective factors and household carbon emis-
2008; Shirley et al., 2012; Underwood and Zahran, 2015), energy sions.
supplies (Kerkhof et al., 2009a), education (Büchs and Schnepf, 2013; This paper aims to fill in the gap by studying the impact of sub-
Andersson et al., 2014), and other individual, household, and en- jective factors on household carbon emissions in China using a unique
vironmental characteristics (Brand et al., 2013; Holian and Kahn, nationwide household level dataset. We focus on the emissions reduc-
2015). tion effect of different subjective measures of Chinese households and
Relevant studies on China have only started since 2007. Wei et al. further explore the impact of consumer lifestyles on the relationship

146
J. Li, et al. Ecological Economics 160 (2019) 145–155

between awareness and household carbon emissions. Moreover, using personal sense of safety or security (security), whether they comply with
the nationally representative data, we are able to comment on the traffic regulations (obey), whether they are interested in social issues
rural/urban differences and also cross regional effects in the afore- (focus), and whether they feel happy (happy). These variables are all
mentioned relationships. relevant but represent different aspects of people's awareness. The
general arguments/hypotheses about these factors are: if a person pays
3. Data Processing and Methodology more attention to social problems (focus), he or she is more aware of the
importance of reducing emissions, and thus is more likely to take en-
3.1. Data Sources vironmental friendly actions. Consumer environmental knowledge can
directly affect behaviors (Frick et al., 2004). Environmental knowledge
The survey data used in this study is the China Household Finance is an important factor that constrains sustainable consumption (Press
Survey (CHFS), conducted by the Survey and Research Center for China and Arnould, 2009). A person who is happier (happy) and has a positive
Household Finance at the Southwest University of Finance and feeling of safety or security (security) is more likely to give back to
Economics, China. The dataset provides a high-quality nationwide society or more willing to protect the local environment, and thus is
comprehensive survey of household income, expenditure, assets, debt, more likely to adopt a greener lifestyle. Carter (2011) found positive
insurance, employment, subjective attitudes, and other demographic emotions could effectively stimulate environmental responsibility be-
characteristics. It employs a stratified three-stage probability propor- haviors. People with positive emotions are more likely to be altruistic
tion to size random sampling design that covers both rural and urban and more willing to return to society. A person who complies with
residents. We use the first round of the survey, conducted in 2011 with social norms/rules (obey) can be seen as a good citizen and more likely
a sample of over 8000 households, covering 320 villages/communities to do what is good for society. The altruism is an important psycholo-
from 80 counties in 25 provinces (excluding Tibet, Xinjiang, Inner gical factor to affect pro-environment behaviors (Straughan and
Mongolia, Hainan, Ningxia, Fujian, Hong Kong, Macao, and Taiwan) in Roberts, 1999) and people with this attitude tend to respond to the
China. government's call for emissions reduction.
We compare consumer spending in each category of expenditure
(see below) based on CHFS survey data to National Bureau of Statistics 3.2.2. Measurement of a Consumer Lifestyle
(NBS) consumption data from the China Statistical Yearbook 2011 in Economic theory generally assumes risk aversion in the consumer
Table 1. Here the CHFS results are adjusted by sampling weight. The decision-making process. People prefer to have a smooth level of con-
comparison shows some discrepancies between CHFS data and NBS sumption. Studies in economics and psychology indicate that people
data, such as residential energy consumption, clothing, and health-care) value their habitual level of consumption, and there are great difficul-
consumption. With respect to residential consumption, CHFS data also ties in making them deviate from it. Thus, consumer preferences do not
include property management fees, maintenance fees, and heating fees, vary dramatically (Morduch, 1995), and they might form an environ-
resulting in a higher value than that from the NBS. Health-care mental barrier. Some studies have found environmental barriers that
spending divergences exist because medical expenses are not included weaken the effect of beliefs (Sidiras and Koukios, 2004; Lane and
in the CHFS data. Clothing materials and garment processing costs Potter, 2007). Following this logic, we expect that consumer pre-
excluded from the CHFS lead to lower values than that in the NBS data. ferences or lifestyles matter in the relationship between awareness and
behavior.
3.2. Measurement of Subjective Factors and Consumer Attitudes Two measures of consumer attitudes are used in our empirical
study: the share of eating out in total food expenditures (eatout) and the
3.2.1. Measurement of Subjective Factors share of education and training expenditure in total spending on edu-
Awareness of social and environmental problems has the potential cation, cultural activities, and recreation (edurate). The first measure is
to change people's lifestyles, alter consumer behavior, and reduce more straightforward in showing whether a person follows a frugal
emissions to a certain extent. The question is which subjective factors lifestyle or not. Given that the second measure shows whether someone
matter and how much they can help to reduce emissions. These factors, invests more in education (future) than recreation (present), and thus it
however, are not directly observable from the questionnaire. indicates a person's time preference. We expect to find a stronger
Following Gadenne et al. (2011), who reviewed the literature on awareness-behavior link in this type of households.
psychology and sociology and investigate the role of beliefs, social re-
sponsibility, and attitudes on energy-saving behaviors, we explore re- 3.3. Calculations of Household CO2 Emissions
levant questions in the CHFS survey and come up with four measures of
social awareness. The most important task is to measure households' carbon emis-
In the CHFS survey, people were asked their opinions about sions. Wei et al. (2007) divided household emissions into direct and

Table 1
Summary of per capita expenditure data in CHFS/NBS database and total emissions coefficient by consumption category in China.
Categories CHFS data NBS Data Total coefficient

Amount Share (%) Amount Share (%) Tf (t C/104 RMB)


(RMB) (RMB)

Indirect total
Food 4474.10 39.16 4804.71 37.04 0.218
Clothing 813.84 7.12 1444.34 11.13 0.261
Household facilities and articles 671.81 5.88 908.01 7.00 0.484
Transportation and communication 1892.24 16.56 1983.70 15.29 0.327
Education, cultural activities, and recreation 1849.46 16.19 1627.64 12.55 0.427
Health care 106.10 0.09 871.77 6.72 0.256
Direct
Residential use 1616.20 14.15 1332.34 10.27 1.298
Total (Indirect + Direct) 11,423.75 100.00 12,972.51 100.00

147
J. Li, et al. Ecological Economics 160 (2019) 145–155

indirect emissions. Direct emissions are from residential consumption of 2014). Following these literature, we also include these variables to
electricity, fuel, and other utilities, whereas indirect emissions are control for the effect of individual attributes. In addition, we consider
embodied emissions from (C1) food, (C2) clothing, (C3) household fa- the impact of membership in the Chinese Communist Party (party) and
cilities and articles, (C4) transportation and communications, (C5) the industry in which the head of the household works (industry).
education, cultural activities, and recreation, (C6) health care, and
miscellaneous commodities and services. Indirect household emissions 3.5. Analytical Framework
can be calculated using the CLA method (Bin and Dowlatabadi, 2005;
Wei et al., 2007). Based on the variables discussed in Sections 3.2–3.4, we construct
Following Wei et al. (2007), we calculate the CO2 emissions of a an equation of household per capita CO2 emissions as follows:
household by multiplying the monetary expenditure in each ex-
penditure category by the respective emissions conversion coefficient. ln Carbon = + 1 Subjective + 2 X+ (4)
Household carbon emissions are calculated as follows:
where ln Carbon is the logarithm of per capita emissions; Subjective
Ci, j = Tfi yexp, i, j 44/12 (1) includes security, obey, focus, and happy; X is a vector of control vari-
ables that includes household income, family structure, the number of
Cj = Ci, j cars, marriage, employment, gender, education, age, policy, industry,
i (2) size, rural residence, and dwelling size. The definitions and summary
statistics of these factors are listed in Table 2.
where Tfi is the carbon intensity of ith consumption category (tC/104
RMB), yexp, i, j is the spending in the ith category by the jth household
(104 RMB), Ci, j is carbon emissions associate with the respective 4. Empirical Results and Analysis
spending (t CO2), and Cj is total carbon emissions by the jth household.
According to the CLA method in Wei et al. (2007), we calculate total 4.1. Sources of Household Emissions
carbon intensity coefficients for each consumption category in 2010
based on the China Input-Output Table and sectoral carbon emissions It is worth to clarify that we consider household carbon emissions in
data from the China Emission Accounts and Datasets. With single-re- all the categories except miscellaneous commodities and services. This
gion I-O table, it is difficult to distinguish emission intensities between is consistent with the past studies. Given that expenditure on this ca-
imported and domestic products. Following Su and Ang (2013, 2015), tegory is generally very small in number, excluding them should have
we use the non-competitive imports assumption to account for the no major impact on the empirical results.1
embodied emissions from domestic production only. The same as- Average household CO2 emissions in every consumption category
sumption is also used in the study by Wei et al. (2007) and Liu et al. are illustrated in Fig. 1. The samples are divided into five groups based
(2011). The calculation method can be written as follows: on household disposable income from the lowest to the highest, and
denoted as HH-Q1, HH-Q2, HH-Q3, HH-Q4, and HH-Q5. We find that
Tf = f (I A) 1
(3) CO2 emissions by Chinese households average about 5.240 ton carbon
where f is the vector of direct emissions intensity in each production equivalent (TCE; about 1.837 tons per capita) and are caused mainly by
sector, A is the matrix of domestic production coefficients, and residential use (40.1%), followed by the consumption of food (18.3%),
(I − A)−1 is the domestic Leontief inverse matrix. The results are shown transportation and communications (17.5%), and education, cultural
in Table 1. activities, and recreation (13.8%). The share of emissions in other
consumption categories is relatively low, 10.3%. Indirect CO2 emissions
3.4. Explanatory Variables account for 60.1% of the household total. In addition, both direct and
indirect CO2 emissions increase with income (for more information,
Variables used in past studies are also included to investigate the please refer to Appendix Tables A1-A2).
factors that influence household emissions. They can be generally di- Comparing with the results from other studies, the per capita
vided into two categories: household attributes and individual attri- emissions from our calculation is similar to Maraseni et al. (2016), who
butes. reported per capita emission ranges from 0.8 (rural area in southern
Household attributes are important factors that affect consumer China) to 2.13 ton (urban area in northern China) in 2010. Our average
behavior. For example, household income (income), wealth, and related total emissions is however, less than that in Qu et al. (2013), who re-
expenditures are closely and positively related to household emissions ported an average total household carbon emission of 5.99 ton in
(Duarte et al., 2010; Kerkhof et al., 2009a, 2009b; Shirley et al., 2012; northwestern arid-alpine regions, China. The difference is simply due to
Weber and Matthews, 2008). The number of household members the choice of sample. While Qu et al. (2013) used samples from a sub-
(tolive) has an important impact on total household emissions (Zhu region in China, the CHFS data has obvious advantage to be nationally
et al., 2012; Brand et al., 2013; Qu et al., 2013; Das and Paul, 2014; representative.
Yang et al., 2016; Su et al., 2017). It is found to have negative re- Household emissions also vary across rural and urban households
lationship with per capita emissions (Xu et al., 2016). Household (Appendix Table A3). Urban households have more emissions than
structure (structure) and the number of members of a household who are rural households in all categories, and in HH-Q1 to HH-Q4, the gap
employed (employees) also have significant impact on emissions (Xu narrowed down with increasing income. The richest families (HH-Q5)
et al., 2016). Larger houses (dwelling) and those with more cars have much higher carbon emissions than other households, and it is
(carnum) will produce more emissions, as more fuel is needed for more significant in urban areas because of greater inequality in
heating and driving. Rural or urban residence (rural) is also important household income. The share of indirect CO2 emissions is higher in
(Liu et al., 2011; Brand et al., 2013). These variables are selected to urban households than rural households except for HH-Q2, and the gap
control for the effect of household attributes. widened with increases in income in groups HH-Q1 to HH-Q4 but
Individual attributes can affect consumption decisions and house- slightly narrowed in HH-Q5.
hold carbon emissions. Some studies have shown that education (edu-
cation) can impose a positive effect on emissions (Büchs and Schnepf, 1
According to Wei et al. (2007), the contribution of miscellaneous to total
2013), others show that the impact is negative (e.g. Brand et al., 2013; emission is only 0.98%. Recent studies, like Xu et al. (2016) and Xu and Han
Xu et al., 2016). Marital status (married), gender (gender), and age (age) (2017), also indicate that it is difficult to measure miscellaneous part in a
can also affect household emissions (Andersson et al., 2014; Chancel, survey and thus not included in their studies.

148
J. Li, et al. Ecological Economics 160 (2019) 145–155

Table 2
Descriptions and basic statistics of variables used in the study.
Variables Meaning Obs. Mean Std. dev. Min Max

Dependent variables
Carbon Carbon emissions in a household (ton) 6840 5.240 0.846 0.182 43.051
Per carbon Per capita emissions in a household (ton) 6840 1.837 4.500 0.059 15.024

Independent variables
Security Feeling of personal safety or security 6840 0.532 0.499 0 1
(yes = 1, no = 0)
Focus Focus on social issues (yes = 1, no = 0) 6820 0.491 0.500 0 1
Obey Compliance with traffic regulations 6840 0.815 0.389 0 1
(yes = 1, no = 0)
Happy Feel happy (yes = 1, no = 0) 6840 0.634 0.482 0 1
Eatout Eating out as a share of total food expenditures 6721 0.134 0.225 0 1
Edurate Training expenditure as a share of total spending on 6840 0.343 0.431 0 1
education, cultural activities, and recreation

Control variables
Household characteristics
Income Imputed income in a household 6840 4.474 5.929 0 43.41
(RMB 10,000)
Tolive Number of members of a household 6840 3.270 1.411 1 7
Structure Proportion of young adults (15–60) in the family 6840 0.652 0.332 0 1
Employees Number of members of a household who are working 6840 1.928 1.256 0 5
Dwelling Dwelling size (in sq. m.) 6840 104.900 70.780 12 400
Carnum Number of cars 6840 0.152 0.391 0 2
Rural Habitual residence (rural = 1, otherwise 0) 6840 0.418 0.493 0 1
Temperature Average temperature in county where household lives 6840 14.651 4.528 3.2 23.3
(Celsius)
Distance Distance to the county center (min) 6840 38.576 35.584 0 180

Personal characteristics
Education Education level of the head of household 6840 1.418 0.847 0 4
Married Marital status of the head of household (married = 1, 6840 0.892 0.310 0 1
unmarried = 0)
Gender Gender of the head of household (male = 1, female = 0) 6840 0.249 0.433 0 1
Policy CPC member (Yes = 1, No = 0) 6840 0.161 0.368 0 1
Age Age of the head of household 6840 51.46 13.50 23 84
Industry industry in which the head of household works 6840 3.597 0.776 1 4

Fig. 1. Household carbon emissions by consumption category in urban and rural China (Q1 to Q5 refer to the lowest income group to the highest income group).

To explore the regional differences in household emissions, our this issue, we use the treatment effect model based on maximum-like-
samples are further divided into the eastern, central, and western re- lihood estimation (Cong and Drukker, 2001). The dependent variable is
gions. Households in the eastern region have the highest emissions, and per capita household CO2 emissions, and different subjective factors
households in the western region have the lowest emissions. The share (models 1 to 4) are used to investigate the impact on emissions.
of indirect emissions is lowest in the central region. More detailed in- We use the average temperature in the county in which a household
formation can be found in Appendix Table A4. is located and the distance between the family's house to county center
as instrumental variables (IVs) for the subjective factors. The rational of
4.2. Baseline Results making this choice is the following: climate factors (e.g. temperature)
can affect individual psychology such as cognitive abilities, and also
The model given by Eq. (4) has clear endogeneity problem, and thus physiological behavior such as eating habits (see for example, Parker,
the ordinary least square (OLS) model is not appropriate. To cope with 1995). Maddison and Rehdanz (2011) also suggested that temperature

149
J. Li, et al. Ecological Economics 160 (2019) 145–155

Table 3 awareness-behavior gap documented in the literature (e.g. Bai and Liu,
Regression results of treatment effect models with different subjective factors 2013).
for household per capita emissions in China. Households with more family members produce less per capita
Model 1 Model 2 Model 3 Model 4 carbon emissions due to the sharing of consumer goods by family
members, in other words, the economy of scale. The result here is con-
Income 0.073⁎⁎⁎ 0.074⁎⁎⁎ 0.070⁎⁎⁎ 0.071⁎⁎⁎ sistent with existing literatures, such as Qu et al. (2013) and Xu and Han
(8.485) (7.782) (7.487) (9.237)
(2017). Similar to Xu et al. (2016), more members with jobs in a
Tolive −0.137⁎⁎⁎ −0.138⁎⁎⁎ −0.126⁎⁎⁎ −0.125⁎⁎⁎
(−5.514) (−5.453) (−4.954) (−4.986) household associates with lower per capita emissions due to less home
Structure 0.371⁎⁎⁎ 0.346⁎⁎⁎ 0.313⁎⁎⁎ 0.345⁎⁎⁎ energy consumption. Golley and Meng (2012) suggested that the age of
(3.510) (3.062) (3.027) (3.539) household head is positively associated with emissions in urban China,
Carnum 1.276⁎⁎⁎ 1.319⁎⁎⁎ 1.240⁎⁎⁎ 1.208⁎⁎⁎
while we find otherwise. Elder people tend to be more frugal and con-
(7.323) (7.845) (7.501) (6.909)
Education 0.304⁎⁎⁎ 0.306⁎⁎⁎ 0.312⁎⁎⁎ 0.317⁎⁎⁎
sume less, especially in rural area. Indeed, we find that rural households
(3.665) (3.538) (3.693) (3.613) have substantially lower emissions in China. It is very much different
Employees −0.157⁎⁎⁎ −0.160⁎⁎⁎ −0.156⁎⁎⁎ −0.160⁎⁎⁎ from that in developed countries, for example, Büchs and Schnepf (2013)
(−7.201) (−6.644) (−6.714) (−7.315) concluded that households in UK rural area have higher emissions.
Married −0.351⁎⁎⁎ −0.372⁎⁎⁎ −0.339⁎⁎⁎ −0.414⁎⁎⁎
Significant negative effects on household carbon emissions are
(−4.824) (−4.512) (−4.564) (−4.442)
Age −0.006⁎⁎ −0.007⁎⁎⁎ −0.008⁎⁎⁎ −0.007⁎⁎⁎ found in models 1, 3, and 4. The factors in these models involve positive
(−2.413) (−2.926) (−2.975) (−2.955) emotions, which can affect an individual's prosocial behavior and pro-
Policy −0.013 −0.018 −0.006 −0.011 environmental behavior (Carter, 2011; Kaida and Kaida, 2016). This
(−0.145) (−0.208) (−0.080) (−0.138)
means that feeling safe, feeling happy, or being socially aware can af-
Industry 0.036 0.027 0.018 0.031
(1.069) (0.714) (0.464) (0.997)
fect a person's environmental behavior, causing a decrease in energy
Gender 0.021 0.030 0.008 0.006 consumption, adjustment in the structure of consumption, and move
(0.347) (0.447) (0.128) (0.093) toward a greener lifestyle in daily practice. The findings of negative
Rural −0.237⁎⁎⁎ −0.262⁎⁎⁎ −0.287⁎⁎⁎ −0.300⁎⁎⁎ effects from happiness are interesting as it differs from the existing
(−2.732) (−2.757) (−2.919) (−2.643)
results on developed economies (e.g. Andersson et al., 2014). Brown
ln_dwelling 0.120⁎⁎ 0.137⁎⁎⁎ 0.116⁎⁎ 0.101⁎
(2.389) (2.626) (2.452) (1.740) and Kasser (2005) showed that people with higher subjective well-
Awareness# −2.244⁎⁎⁎ 1.762⁎⁎⁎ −2.686⁎⁎⁎ −2.596⁎⁎⁎ beings tend to behave more ecologically responsible. Our results simply
(−8.492) (4.955) (−11.869) (−12.818) show that being a developing country, China faces much different si-
Cons 2.238⁎⁎⁎ 0.217 3.422⁎⁎⁎ 2.945⁎⁎⁎ tuation from those developed countries.
(8.487) (0.705) (11.344) (9.430)
It is a bit challenging and yet interesting to see that significant
positive effects on household carbon emissions are found in model 2.
First stage Security Focus Obey Subjective Awareness increases per capita emissions instead of reducing emissions
as what we have expected in the first place. It is however, not entirely
Distance 0.002⁎⁎⁎ −0.001 −0.000 −0.000 irrational. Referring to the literature, there is a well-documented issue
(2.872) (−1.428) (−0.573) (−0.888)
Temperature −0.020⁎⁎⁎ 0.016 −0.013⁎⁎⁎ −0.022⁎⁎⁎
on “awareness/attitude-behavior gap”, meaning environmental
(−4.172) (1.309) (−2.751) (−4.432) awareness does not necessarily associate with pro-environmental be-
_cons 0.321⁎⁎⁎ −0.218 1.020⁎⁎⁎ 0.686⁎⁎⁎ havior (Kollmuss and Agyeman, 2002). This phenomenon is not only
(4.293) (−1.382) (14.040) (8.063) relevant to China (e.g. Bai and Liu, 2013), but also to developed
N 6840 6820 6840 6840
countries (e.g. Blake, 1999). Our empirical findings here provide ad-
Lambda 1.463 −1.068 1.601 1.738
Wald test Chi2(1) 59.44⁎⁎⁎ 21.12⁎⁎⁎ 152.95⁎⁎⁎ 424.64⁎⁎⁎ ditional evidence of the existence of such gap. More importantly, the
existing of awareness-behavior gap in China sends policy makers cri-
t Statistics in parentheses. ***, ** and * denote significant at 1%, 5% and
tical message that encouraging environment protection or green con-
10% levels.
# sumption behavior is more complicated than simply letting people
The awareness variable is security (model 1), focus (model 2), obey
(model 3), and happy (model 4). ln_dwelling is the log of housing area
know the issue. A combination of economic and social measures is
(dwelling). needed to make an actual change (Kollmuss and Agyeman, 2002).

4.3. Urban/Rural Differences


is an important factor affecting subjective attitudes (e.g. happiness).
Distance reflects households' ability to access public goods and services, China has clear rural/urban divisions and their differences have
which also indirectly links with their living environment and thus could widened further in recent years. It is therefore worth to investigate the
affect their emotions (e.g. Gao et al., 2014). These are generally exo- impact of subjective factors on carbon emissions by households across
genous variables, which have no direct linkages to household emis- rural/urban areas.
sions. Data on average temperatures in a county are taken from the We interact the rural dummy with all four subjective factors to
China Integrated Meteorological Information Service System (CIMISS). highlight the effect of rural/urban differences in the awareness-emis-
Empirical results are reported in Table 3. In general, the effects of sions relationship. The regression results are shown in Table 4, which
control variables are consistent with our expectations—for example, shows that the impacts of models 3 and 4 have bigger negative effects
structure, carnum, education, income, and dwelling are important factors on household carbon emissions in rural households than urban house-
that cause increases in household carbon emissions. The factors tolive, holds. The impacts of models 1 and 2 on household emissions in rural
employees, married, age, and rural have a negative impact on household areas are similar to those on urban households, which means that the
carbon emissions. abatement effects of some subjective factors are greater in rural
Higher household income level stands for stronger purchasing households than urban households.
power. In other words, richer families will consume more and cause
more emissions. People with higher education level do not necessarily 4.4. Cross-regional Differences
to be more environmental friendly in China. Even if they are more likely
to be informed and aware of environmental issues, their consumption Being a large country, China has clear cross-regional differences in
behaviors do not change accordingly. This is consistent with the terms of economic development, income, and other characteristics. The

150
J. Li, et al. Ecological Economics 160 (2019) 145–155

Table 4 Table 5
Regression results of treatment effect models with different subjective factors Regression results of treatment effect models with different subjective factors
and rural/urban area for household per capita emissions in China. for household per capita CO2 emissions in Eastern, Central, and Western regions
in China.
Model 1 Model 2 Model 3 Model 4
Model 1 Model 2 Model 3 Model 4
Income 0.073⁎⁎⁎ 0.074⁎⁎⁎ 0.070⁎⁎⁎ 0.071⁎⁎⁎
(16.510) (16.530) (16.625) (16.533) Income 0.069 ⁎⁎⁎
0.070 ⁎⁎⁎
0.066
⁎⁎⁎
0.067⁎⁎⁎
Tolive −0.137⁎⁎⁎ −0.138⁎⁎⁎ −0.125⁎⁎⁎ −0.125⁎⁎⁎ (15.587) (15.521) (15.502) (15.541)
(−6.973) (−6.907) (−6.557) (−6.564) Tolive −0.137⁎⁎⁎ −0.136⁎⁎⁎ −0.125⁎⁎⁎ −0.123⁎⁎⁎
Structure 0.371⁎⁎⁎ 0.346⁎⁎⁎ 0.313⁎⁎⁎ 0.340⁎⁎⁎ (−7.009) (−6.878) (−6.588) (−6.504)
(4.337) (3.976) (3.756) (4.085) Structure 0.370⁎⁎⁎ 0.345⁎⁎⁎ 0.308⁎⁎⁎ 0.341⁎⁎⁎
Carnum 1.276⁎⁎⁎ 1.319⁎⁎⁎ 1.239⁎⁎⁎ 1.209⁎⁎⁎ (4.352) (3.991) (3.709) (4.119)
(19.997) (20.419) (20.352) (19.565) Carnum 1.221⁎⁎⁎ 1.259⁎⁎⁎ 1.187⁎⁎⁎ 1.161⁎⁎⁎
Education 0.303⁎⁎⁎ 0.305⁎⁎⁎ 0.309⁎⁎⁎ 0.318⁎⁎⁎ (19.125) (19.454) (19.482) (18.782)
(8.918) (8.829) (9.360) (9.648) Education 0.299⁎⁎⁎ 0.299⁎⁎⁎ 0.306⁎⁎⁎ 0.309⁎⁎⁎
Employees −0.157⁎⁎⁎ −0.160⁎⁎⁎ −0.158⁎⁎⁎ −0.160⁎⁎⁎ (8.848) (8.697) (9.308) (9.436)
(−6.282) (−6.304) (−6.493) (−6.616) Employees −0.141⁎⁎⁎ −0.144⁎⁎⁎ −0.142⁎⁎⁎ −0.145⁎⁎⁎
Married −0.352⁎⁎⁎ −0.371⁎⁎⁎ −0.337⁎⁎⁎ −0.409⁎⁎⁎ (−5.635) (−5.665) (−5.827) (−5.982)
(−4.576) (−4.753) (−4.552) (−5.517) Married −0.349⁎⁎⁎ −0.370⁎⁎⁎ −0.325⁎⁎⁎ −0.409⁎⁎⁎
Age −0.006⁎⁎⁎ −0.007⁎⁎⁎ −0.008⁎⁎⁎ −0.007⁎⁎⁎ (−4.568) (−4.769) (−4.408) (−5.550)
(−2.856) (−3.273) (−3.662) (−3.541) Age −0.007⁎⁎⁎ −0.008⁎⁎⁎ −0.009⁎⁎⁎ −0.008⁎⁎⁎
Policy −0.013 −0.019 −0.006 −0.009 (−3.327) (−3.658) (−4.145) (−3.910)
(−0.192) (−0.274) (−0.089) (−0.134) Policy 0.009 −0.000 0.006 0.011
Industry 0.035 0.026 0.017 0.031 (0.136) (−0.003) (0.084) (0.173)
(1.143) (0.846) (0.565) (1.044) Industry 0.049 0.039 0.031 0.040
Gender 0.022 0.030 0.010 0.006 (1.597) (1.239) (1.043) (0.180)
(0.399) (0.546) (0.188) (0.105) Rural −0.189⁎⁎⁎ −0.208⁎⁎⁎ −0.235⁎⁎⁎ −0.243⁎⁎⁎
ln_dwelling 0.120⁎⁎⁎ 0.137⁎⁎⁎ 0.118⁎⁎⁎ 0.101⁎⁎⁎ (−3.367) (−3.686) (−4.314) (−4.461)
(3.193) (3.587) (3.227) (2.770) Gender 0.020 0.028 0.008 0.010
Rural −0.201⁎⁎⁎ −0.241⁎⁎⁎ −0.454⁎⁎⁎ −0.419⁎⁎⁎ (0.366) (0.504) (0.143) (0.180)
(−2.695) (−3.355) (−4.539) (−5.382) ln_dwelling 0.143⁎⁎⁎ 0.159⁎⁎⁎ 0.141⁎⁎⁎ 0.124⁎⁎⁎
Awareness# −2.210⁎⁎⁎ 1.786⁎⁎⁎ −2.783⁎⁎⁎ −2.687⁎⁎⁎ (3.816) (4.172) (3.850) (3.390)
(−22.166) (12.279) (−31.820) (−33.351) Central −0.468⁎⁎⁎ −0.377⁎⁎⁎ −0.512⁎⁎⁎ −0.598⁎⁎⁎
Awareness ∗ rural −0.067 −0.042 0.211⁎⁎ 0.194⁎⁎ (−6.450) (−5.246) (−4.865) (−7.485)
(−0.726) (−0.456) (1.981) (2.125) West −0.545⁎⁎⁎ −0.279⁎⁎⁎ −0.559⁎⁎⁎ −0.500⁎⁎⁎
Cons 2.225⁎⁎⁎ 0.207 3.483⁎⁎⁎ 3.001⁎⁎⁎ (−5.623) (−3.000) (−4.635) (−5.058)
(8.856) (0.805) (14.083) (12.370) Awareness# −2.375⁎⁎⁎ 1.753⁎⁎⁎ −2.813⁎⁎⁎ −2.792⁎⁎⁎
(−22.496) (11.059) (−28.684) (−31.667)
Awareness ∗ central 0.153 −0.072 0.155 0.341⁎⁎⁎
First stage Security Focus Obey Happy (1.547) (−0.717) (1.317) (3.446)
Awareness ∗ west 0.392⁎⁎⁎ −0.131 0.282⁎⁎ 0.270⁎⁎
Distance 0.002⁎⁎⁎ −0.001⁎⁎ −0.000 −0.000 (3.050) (−1.007) (2.027) (2.148)
(5.064) (−2.304) (−1.151) (−1.057) Cons 2.388⁎⁎⁎ 0.315 3.586⁎⁎⁎ 3.152⁎⁎⁎
Temperature −0.020⁎⁎⁎ 0.016⁎⁎⁎ −0.014⁎⁎⁎ −0.022⁎⁎⁎ (9.500) (1.217) (14.427) (12.970)
(−7.341) (5.214) (−4.374) (−8.234)
_cons 0.321⁎⁎⁎ −0.218⁎⁎⁎ 1.024⁎⁎⁎ 0.686⁎⁎⁎
(7.192) (−4.571) (20.044) (15.684) First stage Security Focus Obey Happy
N 6840 6820 6840 6840
Lambda 1.459 −1.072 1.607 1.746 Distance 0.002 ⁎⁎⁎
−0.001 ⁎⁎
−0.000 −0.000
LR test chi2(1) 136.51⁎⁎⁎ 26.35⁎⁎⁎ 371.28⁎⁎⁎ 442.91⁎⁎⁎ (4.958) (−2.251) (−1.190) (−1.085)
ATE 0.096 0.056 0.098 0.229⁎ Temperature −0.018⁎⁎⁎ 0.013⁎⁎⁎ −0.010⁎⁎⁎ −0.019⁎⁎⁎
(−6.283) (4.393) (−3.294) (−6.940)
t Statistics in parentheses. ***, ** and * denote significant at 1%, 5% and 10%
Cons 0.281⁎⁎⁎ −0.184⁎⁎⁎ 0.977⁎⁎⁎ 0.637⁎⁎⁎
levels. (6.255) (−3.831) (19.016) (14.476)
#
The awareness variable is security (model 1), focus (model 2), obey (model N 6840 6820 6840 6840
3), and happy (model 4). ln_dwelling is the log of housing area (dwelling). ATE Lambda 1.464 −1.037 1.597 1.744
is average treatment effect. LR test chi2(1) 130.91⁎⁎⁎ 18.16⁎⁎⁎ 362.13⁎⁎⁎ 436.84⁎⁎⁎
ATE 0.089 0.048 0.067 0.213⁎

t statistics in parentheses. ***, ** and * denote significant at 1%, 5% and 10%


eastern coastal regions are more developed and have consumption levels.
#
patterns that clearly differ from those in the western regions, so cross- The awareness variable is security (model 1), focus (model 2), obey (model
regional differences in the awareness-emissions relationship in China 3), and happy (model 4). ln_dwelling is the log of housing area (dwelling). ATE is
average treatment effect.
are to be expected.
Table 5 reports results with the inclusion of regional dummies.
Using the eastern regions as reference, we can see clear evidence that increasingly larger urban population over time. Realizing the urban/
the effects of awareness of carbon emissions are significantly different rural differences is critically important to achieve effective carbon mi-
across regions, though the effects vary by factor. For example, feeling tigation. In addition, cross-regional migrations should also be con-
personally secure and compliance with regulations lead to reductions in sidered as China has a large number of migrants from western and
carbon emissions, but the effect is statistically lower in western regions central provinces to eastern areas. Migration could impose strong im-
than in the eastern and central regions. The effects of happiness are also pacts on the mitigation policies.
lower in the central and western regions, however, no significant dif-
ference is seen in model 2. 4.5. Role of a Consumer Lifestyle
The results from models with rural/urban differences and regional
dummies have important policy relevance. It is firstly worth to note that This section investigates whether the consumer lifestyle or con-
China has been carrying on the urbanization process, which leads to an sumer preferences can affect the relationship between subjective factors

151
J. Li, et al. Ecological Economics 160 (2019) 145–155

Table 6 Table 7
Regression results of treatment effect model with different subjective factors Regression results of treatment effect model with different subjective factors
and consumption attitude (eatout) for household per capita emissions in China. and consumption attitude (edurate) for household per capita emissions in China.
Model 1 Model 2 Model 3 Model 4 Model 1 Model 2 Model 3 Model 4

Income 0.066⁎⁎⁎ 0.067⁎⁎⁎ 0.062⁎⁎⁎ 0.065⁎⁎⁎ Income 0.073⁎⁎⁎ 0.074⁎⁎⁎ 0.070⁎⁎⁎ 0.071⁎⁎⁎
(14.813) (14.724) (14.543) (14.895) (16.497) (16.535) (16.578) (16.511)
Tolive −0.124⁎⁎⁎ −0.124⁎⁎⁎ −0.112⁎⁎⁎ −0.114⁎⁎⁎ Tolive −0.135⁎⁎⁎ −0.136⁎⁎⁎ −0.122⁎⁎⁎ −0.123⁎⁎⁎
(−6.315) (−6.212) (−5.905) (−6.028) (−6.537) (−6.494) (−6.107) (−6.179)
Structure 0.313⁎⁎⁎ 0.288⁎⁎⁎ 0.254⁎⁎⁎ 0.285⁎⁎⁎ Structure 0.372⁎⁎⁎ 0.346⁎⁎⁎ 0.316⁎⁎⁎ 0.348⁎⁎⁎
(3.680) (3.327) (3.065) (3.430) (4.345) (3.980) (3.780) (4.173)
Carnum 1.176⁎⁎⁎ 1.216⁎⁎⁎ 1.150⁎⁎⁎ 1.120⁎⁎⁎ Carnum 1.276⁎⁎⁎ 1.318⁎⁎⁎ 1.239⁎⁎⁎ 1.209⁎⁎⁎
(18.481) (18.845) (18.940) (18.143) (19.995) (20.403) (20.348) (19.552)
Education 0.269⁎⁎⁎ 0.267⁎⁎⁎ 0.266⁎⁎⁎ 0.279⁎⁎⁎ Education 0.304⁎⁎⁎ 0.305⁎⁎⁎ 0.312⁎⁎⁎ 0.317⁎⁎⁎
(7.939) (7.749) (8.107) (8.508) (8.935) (8.836) (9.459) (9.634)
Employees −0.164⁎⁎⁎ −0.168⁎⁎⁎ −0.163⁎⁎⁎ −0.163⁎⁎⁎ Employees −0.158⁎⁎⁎ −0.161⁎⁎⁎ −0.159⁎⁎⁎ −0.160⁎⁎⁎
(−6.585) (−6.629) (−6.741) (−6.733) (−6.277) (−6.285) (−6.464) (−6.561)
Married −0.246⁎⁎⁎ −0.258⁎⁎⁎ −0.205⁎⁎⁎ −0.287⁎⁎⁎ Married −0.350⁎⁎⁎ −0.370⁎⁎⁎ −0.336⁎⁎⁎ −0.415⁎⁎⁎
(−3.200) (−3.305) (−2.757) (−3.840) (−4.552) (−4.737) (−4.531) (−5.582)
Age −0.002 −0.003 −0.004⁎ −0.004⁎ Age −0.006⁎⁎⁎ −0.007⁎⁎⁎ −0.008⁎⁎⁎ −0.007⁎⁎⁎
(−1.103) (−1.471) (−1.833) (−1.790) (−2.881) (−3.263) (−3.819) (−3.496)
Policy −0.015 −0.020 −0.008 −0.011 Policy −0.013 −0.018 −0.005 −0.012
(−0.231) (−0.293) (−0.125) (−0.171) (−0.185) (−0.256) (−0.080) (−0.179)
Industry 0.037 0.028 0.017 0.032 Industry 0.035 0.026 0.017 0.031
(1.211) (0.908) (0.560) (1.094) (1.143) (0.817) (0.575) (1.029)
Gender 0.062 0.076 0.062 0.056 Gender 0.022 0.030 0.010 0.007
(1.131) (1.367) (1.168) (1.053) (0.404) (0.546) (0.191) (0.123)
Rural −0.215⁎⁎⁎ −0.243⁎⁎⁎ −0.269⁎⁎⁎ −0.290⁎⁎⁎ Rural −0.236⁎⁎⁎ −0.261⁎⁎⁎ −0.287⁎⁎⁎ −0.301⁎⁎⁎
(−3.874) (−4.350) (−5.000) (−5.363) (−4.238) (−4.660) (−5.304) (−5.553)
ln_dwelling 0.120⁎⁎⁎ 0.138⁎⁎⁎ 0.122⁎⁎⁎ 0.108⁎⁎⁎ ln_dwelling 0.121⁎⁎⁎ 0.138⁎⁎⁎ 0.118⁎⁎⁎ 0.101⁎⁎⁎
(3.198) (3.618) (3.347) (2.947) (3.213) (3.593) (3.210) (2.762)
Eatout 1.439⁎⁎⁎ 0.958⁎⁎⁎ 2.244⁎⁎⁎ 1.609⁎⁎⁎ Edurate −0.028 −0.079 −0.095 0.039
(9.067) (6.446) (10.922) (9.591) (−0.358) (−1.034) (−0.879) (0.470)
Awareness# −2.216⁎⁎⁎ 1.792⁎⁎⁎ −2.495⁎⁎⁎ −2.503⁎⁎⁎ Awareness# −2.249⁎⁎⁎ 1.712⁎⁎⁎ −2.713⁎⁎⁎ −2.560⁎⁎⁎
(−25.826) (14.753) (−33.068) (−35.922) (−23.388) (12.013) (−31.995) (−32.772)
Awareness ∗ eatout −0.426⁎⁎ 0.525⁎⁎ −1.333⁎⁎⁎ −0.739⁎⁎⁎ Awareness ∗ edurate 0.012 0.126 0.074 −0.095
(−2.090) (2.558) (−5.824) (−3.615) (0.117) (1.200) (0.616) (−0.932)
Cons 1.859⁎⁎⁎ −0.182 2.859⁎⁎⁎ 2.497⁎⁎⁎ Cons 2.248⁎⁎⁎ 0.252 3.452⁎⁎⁎ 2.929⁎⁎⁎
(7.366) (−0.708) (11.545) (10.223) (8.912) (0.973) (13.929) (12.047)

First stage Security Focus Obey Happy First stage Security Focus Obey Happy

Distance 0.002 ⁎⁎⁎


−0.001 ⁎⁎
−0.000 −0.000 Distance 0.002 ⁎⁎⁎
−0.001 ⁎⁎
−0.000 −0.000
(5.136) (−2.285) (−1.095) (−0.826) (5.067) (−2.298) (−0.994) (−1.023)
Temperature −0.022⁎⁎⁎ 0.017⁎⁎⁎ −0.015⁎⁎⁎ −0.023⁎⁎⁎ Temperature −0.020⁎⁎⁎ 0.016⁎⁎⁎ −0.013⁎⁎⁎ −0.022⁎⁎⁎
(−7.892) (5.616) (−4.890) (−8.609) (−7.341) (5.198) (−4.303) (−8.221)
Cons 0.344⁎⁎⁎ −0.232⁎⁎⁎ 1.042⁎⁎⁎ 0.702⁎⁎⁎ Cons 0.321⁎⁎⁎ −0.218⁎⁎⁎ 1.020⁎⁎⁎ 0.686⁎⁎⁎
(7.718) (−4.884) (20.326) (15.949) (7.188) (−4.558) (19.995) (15.683)
N 6721 6703 6721 6721 N 6840 6820 6840 6840
lambda 1.484 −1.133 1.607 1.727 Lambda 1.463 −1.065 1.602 1.738
LR test chi2(1) 159.37⁎⁎⁎ 38.04⁎⁎⁎ 379.78⁎⁎⁎ 440.56⁎⁎⁎ LR test chi2(1) 139.89⁎⁎⁎ 25.73⁎⁎⁎ 367.90⁎⁎⁎ 437.90⁎⁎⁎
ATE 0.102 0.052 0.118 0.167 ATE 0.097 0.056 0.098 0.230⁎

t Statistics in parentheses. ***, ** and * denote significant at 1%, 5% and 10% t Statistics in parentheses. ***, ** and * denote significant at 1%, 5% and 10%
levels. levels.
# #
The awareness variable is security (model 1), focus (model 2), obey (model The awareness variable is security (model 1), focus (model 2), obey (model
3), and happy (model 4). ln_dwelling is the log of housing area (dwelling). ATE is 3), and happy (model 4). ln_dwelling is the log of housing area (dwelling). ATE is
average treatment effect. average treatment effect.

even subjectively aware of the necessity of reducing emissions.2


and household emissions. Based on the CHFS questionnaire, we use two People with a higher edurate place greater value on the future. The
variables for this purpose that represent different aspects of pre- results on this factor are shown in Table 7. However, we find that the
ferences. When the first variable (eatout) has a low value, it means that lifestyle represented by edurate has no impact on the relationship be-
the respondent mostly chooses to eat at home and therefore represents a tween subjective factors and household carbon emissions.
frugal lifestyle, whereas when the second variable (edurate) has a high
value, it shows that the respondent household values its future and thus
represents a forward-looking lifestyle. 5. Conclusions and Policy Implications
The results on eatout are reported in Table 6. We find that people's
choice to spend more on eating out can reduce the negative impact of In this paper, we investigate the relationship between subjective
models 1, 3, and 4 on household carbon emissions but strengthens the factors and household carbon emissions and the role of a consumer
positive influence of model 2. This means that having a prodigal life-
style can reduce the abatement effects of subjective factors. This is 2
The results are generally consistent when we divide our sample into five
because consumer behavior is difficult to change in the short term, and quantiles according to income. We thank an anonymous referee for this in-
this type of household (with prodigal lifestyle) are less likely to react sightful suggestion and the results are available upon request.

152
J. Li, et al. Ecological Economics 160 (2019) 145–155

lifestyle/preference on these relationships based on a nationwide CHFS habits and customs can change over the long term, and altering
dataset. Four subjective factor variables (security, obey, focus, and household behavior to achieve a cleaner lifestyle gradually can be
happy) and two measures of consumer lifestyles (eatout and edutate) are beneficial in reducing household emissions. Consumers should be en-
used in the empirical analysis. Using the detailed survey dataset from couraged to use low-carbon products, including clean energy, low-en-
CHFS, we conduct regression analyses at the national level and then ergy-consumption appliances to replace high-carbon products, through
consider rural/urban and regional differences in our model. Treatment a policy with clear economic incentives, for example, giving subsidies
effect model are used to deal with endogeneity problem. to greener consumption or taxing high carbon related consumption.
Results show that the difference in household emissions between Implementation of a carbon-labeling system is also a feasible method to
rural and urban households narrowed with increasing income, but wi- make consumers aware of the carbon content of consumer goods and
dened for the richest quantile. Households in the eastern region have influence their purchase decisions.
the highest emissions, but the share of indirect emissions is lowest in Third, considering the differences in household emissions among
the central region. The abatement effect from subjective factors is larger different income groups—in particular, the richest families have much
in rural households than urban households and is the largest in the higher household emissions in each consumption category—it is ne-
western region, followed by the central region. Further, we found that cessary to take different approaches with respect to reducing emissions
having a prodigal consumption style can reduce the abatement effects for families at different income levels. For example, the allocation of
of subjective factors. emissions quotas at the household level could take into account dif-
The findings have some important implications. First, this paper ferences in household emissions levels across income groups and re-
verifies that subjective factors have important impacts on residential quire wealthier households to pay more, whereas the low income
environmental behavior and household carbon emissions in China. households have much less space to react.
Tangible benefits are important, so the government should take the lead In addition, urban and eastern households produce more emissions
improving social welfare and make people more aware of the en- and are less willing to reduce emissions because they have higher level
vironmental issues. The government and environmental organizations of consumption habit. With income being higher, households in these
should work together to educate the public and disseminate knowledge regions are shifting more toward less frugal lifestyles. While more ef-
about environmental protection and emissions reduction, especially for forts should be given to increase their environmental awareness, it is
those who are willing to respond to the government's encouragement. also important to promote green lifestyles. For example, improving
Second, we demonstrate that consumer lifestyle can affect the transportation infrastructure and encourage people to use public
abatement impacts of subjective factors, which give rise to the needs for transportation and adopting a greener travel habit.
differentiated treatment to different type of consumers. Household
lifestyles and consumer attitudes do not vary dramatically in the short Acknowledgement
term because people prefer a stable path of consumption; they have a
relatively stable marginal propensity to consume, and therefore current We thank financial supports from the National Natural Science
consumption is closely related to current income. However, people's Foundation of China (no. 71573214) and the 111 Project (No. B16040).

Appendix A

Table A1
Household expenditure and emissions by category in China.

Category Expenditure (RMB 10,000) Share (%) Carbon emissions Share (%)
(t-CO2)

Indirect total 2.989 88.87 3.148 60.08


Food 1.445 42.96 0.952 18.17
Clothing 0.263 7.82 0.210 4.01
Household facilities and articles 0.185 5.50 0.328 6.26
Transportation and communications 0.609 18.11 0.912 17.40
Education, cultural activities, and recreation 0.460 13.68 0.721 13.76
Health care 0.027 0.803 0.025 0.48
Direct 0.374 11.12 2.092 39.92
Residential use 0.374 11.12 2.092 39.92
Total (Indirect + Direct) 3.363 100.00 5.240 100.00

Table A2
Household emissions by category and income group in China (t-CO2).

Category HH-Q1 HH-Q2 HH-Q3 HH-Q4 HH-Q5

Indirect total 1.951 1.763 2.131 3.067 6.606


Food 0.584 0.651 0.807 0.998 1.877
Clothing 0.127 0.098 0.141 0.193 0.473
Household facilities and articles 0.183 0.180 0.222 0.345 0.683
Transportation and communications 0.487 0.466 0.478 0.907 2.141
Education, cultural activities, and recreation 0.559 0.357 0.464 0.602 1.571
Health care 0.011 0.011 0.019 0.022 0.061
Direct 1.210 1.114 1.905 2.521 3.614
Residential use 1.210 1.114 1.905 2.521 3.614
Total (Indirect + Direct) 3.161 2.877 4.036 5.588 10.22

Note: HH-Q1 to HH-Q5 refer to the lowest income group to the highest income group.

153
J. Li, et al. Ecological Economics 160 (2019) 145–155

Table A3
Household emissions by category in rural/urban areas in China (t-CO2).

Category Urban Rural Difference

Indirect total 4.233 1.867 2.366


Food 1.205 0.655 0.550
Clothing 0.299 0.104 0.195
Household facilities and articles 0.445 0.189 0.256
Transportation and communications 1.222 0.545 0.677
Education, cultural activities, and recreation 1.025 0.364 0.661
Health care 0.037 0.010 0.027
Direct of which 2.498 1.617 0.881
Residential use 2.498 1.617 0.881
Total (Indirect + Direct) 6.731 3.484 3.247

Note: Difference = Urban – Rural.

Table A4
Household emissions by category in Eastern, Central, and Western regions in China (t-CO2).

Category Eastern Central Western National

Indirect total 4.232 1.890 2.033 3.148


Food 1.203 0.661 0.695 0.952
Clothing 0.285 0.136 0.093 0.210
Household facilities and articles 0.453 0.184 0.201 0.328
Transportation and communications 1.269 0.456 0.657 0.912
Education, cultural activities, and recreation 0.987 0.439 0.374 0.721
Health care 0.035 0.014 0.013 0.025
Direct 2.716 1.515 1.064 2.092
Residential use 2.716 1.515 1.064 2.092
Total (Indirect + Direct) 6.948 3.405 3.097 5.240

References Druckman, A., Jackson, T., 2009. The carbon footprint of UK households 1990–2004: a
socio-economically disaggregated, quasi-multi-regional input-output model. Ecol.
Econ. 68 (7), 2066–2077.
Adaman, F., Karalı, N., Kumbaroğlu, G., Or, İ., Özkaynak, B., Zenginobuz, Ü., 2011. What Duarte, R., Mainar, A., Sánchez-Chóliz, J., 2010. The impact of household consumption
determines urban households' willingness to pay for CO2 emission reductions in patterns on emissions in Spain. Energy Econ. 32 (1), 176–185.
Turkey: a contingent valuation survey. Energy Policy 39 (2), 689–698. Feng, Z.H., Zou, L.L., Wei, Y.M., 2011. The impact of household consumption on energy
Ajzen, I., Fishbein, M., 1980. Understanding Attitudes and Predicting Social Behavior. use and CO2 emissions in China. Energy 36 (1), 656–670.
Prentice-Hall, Englewood Cliffs, NJ. Frick, J., Kaiser, F.G., Wilson, M., 2004. Environmental knowledge and conservation
Andersson, D., Nässén, J., Larsson, J., Holmberg, J., 2014. Greenhouse gas emissions and behavior: exploring prevalence and structure in a representative sample. Personal.
subjective well-being: an analysis of Swedish households. Ecol. Econ. 102, 75–82. Individ. Differ. 37 (8), 1597–1613.
Bai, Y., Liu, Y., 2013. An exploration of residents' low-carbon awareness and behavior in Gadenne, D., Sharma, B., Kerr, D., Smith, T., 2011. The influence of consumers' en-
Tianjin, China. Energy Policy 61, 1261–1270. vironmental beliefs and attitudes on energy saving behaviours. Energy Policy 39 (12),
Baiocchi, G., Minx, J., Hubacek, K., 2010. The impact of social factors and consumer 7684–7694.
behavior on carbon dioxide emissions in the United Kingdom. J. Ind. Ecol. 14 (1), Gao, S., Meng, X., Zhang, L., 2014. Fiscal decentralization and life satisfaction: evidence
50–72. from urban China. Soc. Indic. Res. 119 (3), 1177–1194.
Biesiot, W., Noorman, K.J., 1999. Energy requirements of household consumption. Ecol. Golley, J., Meng, X., 2012. Income inequality and carbon dioxide emissions: the case of
Econ. 28 (3), 367–383. Chinese urban households. Energy Econ. 34 (6), 1864–1872.
Bin, S., Dowlatabadi, H., 2005. Consumption attitudes approach to US energy use and the Holian, M.J., Kahn, M.E., 2015. Household carbon emissions from driving and center city
related CO2 emissions. Energy Policy 33 (2), 197–208. quality of life. Ecol. Econ. 116, 362–368.
Blake, J., 1999. Overcoming the ‘value-action gap‘ in environmental policy: tensions IPCC, 2007. Climate Change Synthesis Report 2007. (Intergovernmental Panel on
between national policy and local experience. Local Environ. 4 (3), 257–278. Climate Change).
Borgstede, C.V., Andersson, M., Johnsson, F., 2013. Public attitudes to climate change Jennings, M., Munuera, L., & Tong, D., 2011. An assessment of China's 2020 carbon in-
and carbon mitigation: implications for energy-associated behaviours. Energy Policy tensity target. Grantham Institute for Climate Change Report GR1, London.
57 (6), 182–193. Kahneman, D., Tversky, A., 1979. Prospect theory: an analysis of decision under risk.
Brand, C., Goodman, A., Rutter, H., Song, Y., Ogilvie, D., 2013. Associations of individual, Econometrica 47, 263–291.
household and environmental characteristics with carbon dioxide emissions from Kaida, N., Kaida, K., 2016. Pro-environmental behavior correlates with present and future
motorised passenger travel. Appl. Energy 104, 158–169. subjective well-being. Environment Development & Sustainability 18 (1), 1–17.
Brounen, D., Kok, N., Quigley, J.M., 2013. Energy literacy, awareness, and conservation Kerkhof, A.C., Benders, R.M.J., Moll, H.C., 2009a. Determinants of variation in household
behavior of residential households. Energy Econ. 38 (2), 42–50. CO2, emissions between and within countries. Energy Policy 37 (4), 1509–1517.
Brown, K.W., Kasser, T., 2005. Are psychological and ecological well-being compatible? Kerkhof, A.C., Nonhebel, S., Moll, H.C., 2009b. Relating the environmental impact of
The role of values, mindfulness, and lifestyle. Soc. Indic. Res. 74 (2), 349–368. consumption to household expenditures: an input-output analysis. Ecol. Econ. 68 (4),
Büchs, M., Schnepf, S.V., 2013. Who emits most? Associations between socio-economic 1160–1170.
factors and UK households' home energy, transport, indirect and total CO2 emissions. Kollmuss, A., Agyeman, J., 2002. Mind the gap: why do people act environmentally and
Ecol. Econ. 90, 114–123. what are the barriers to pro-environmental behavior? Environ. Educ. Res. 8 (3),
Carter, D.M., 2011. Recognizing the role of positive emotions in fostering en- 239–260.
vironmentally responsible behaviors. Ecopsychology 3 (1), 65–69. Lane, B., Potter, S., 2007. The adoption of cleaner vehicles in the UK: exploring the
Chancel, L., 2014. Are younger generations higher carbon emitters than their elders? consumer attitude-action gap. J. Clean. Prod. 15 (11−12), 1085–1092.
Inequalities, generations and CO2 emissions in France and in the USA. Ecol. Econ. Liu, L., Wu, G., Wang, J., Wei, Y., 2011. China's carbon emissions from urban and rural
100, 195–207. households during 1992–2007. J. Clean. Prod. 19, 1754–1762.
Cong, R., Drukker, D.M., 2001. Treatment effects model. Stata Tech. Bull. 10 (55). Maddison, D., Rehdanz, K., 2011. The impact of climate on life satisfaction. Ecol. Econ. 70
Dai, H., Masui, T., Matsuoka, Y., Fujimori, S., 2012. The impacts of China's household (12), 2437–2445.
consumption expenditure patterns on energy demand and carbon emissions towards Maraseni, T.N., Qu, J., Yue, B., Zeng, J., Maroulis, J., 2016. Dynamism of household
2050. Energy Policy 50, 736–750. carbon emissions (HCEs) from rural and urban regions of northern and southern
Das, A., Paul, S.K., 2014. CO2 emissions from household consumption in India between China. Environ. Sci. Pollut. Res. 23 (20), 1–14.
1993–94 and 2006–07: a decomposition analysis. Energy Econ. 41 (1), 90–105. Morduch, J., 1995. Income smoothing and consumption smoothing. J. Econ. Perspect. 9

154
J. Li, et al. Ecological Economics 160 (2019) 145–155

(3), 103–114. Underwood, A., Zahran, S., 2015. The carbon implications of declining household scale
Owens, S., Driffill, L., 2008. How to change attitudes and behaviours in the context of economies. Ecol. Econ. 116, 182–190.
energy. Energy Policy 36, 4412–4418. Van Raaij, W.F., Verhallen, T.M.M., 1983. A behavioral model of residential energy use. J.
Parker, P.M., 1995. Climatic Effects on Individual, Social, and Economic Behavior: A Econ. Psychol. 3, 39–63.
Physioeconomic Review of Research across Disciplines. Greenwood Press, Wang, Z., Yang, L., 2014. Indirect carbon emissions in household consumption: evidence
Westport, CT. from the urban and rural area in China. J. Clean. Prod. 78, 94–103.
Peattie, K., 2001. Golden goose or wild goose? The hunt for the green consumer. Bus. Wang, Z., Zhang, B., Yin, J., Zhang, Y., 2010. Determinants and policy implications for
Strateg. Environ. 10 (4), 187–199. household electricity-saving behaviour: evidence from Beijing. China. Energy Policy
Press, Melea, Arnould, Eric J., 2009. Constraints on sustainable energy consumption: 39 (6), 3550–3557.
market system and public policy challenges and opportunities. J. Public Policy Mark. Weber, C.L., Matthews, H.S., 2008. Quantifying the global and distributional aspects of
28 (1), 102–113. American household carbon footprint. Ecol. Econ. 66 (2–3), 379–391.
Qu, J., Zeng, J., Li, Y., Wang, Q., Maraseni, T., Zhang, L., et al., 2013. Household carbon Weber, C., Perrels, A., 2000. Modelling lifestyle effects on energy demand and related
dioxide emissions from peasants and herdsmen in northwestern arid-alpine regions. emissions. Energy Policy 28 (8), 549–566.
China. Energy Policy 57 (6), 133–140. Wei, Y.M., Liu, L.C., Fan, Y., Wu, G., 2007. The impact of lifestyle on energy use and CO2
Schlegelmilch, B.B., Bohlen, G.M., Diamantopoulos, A., 1996. The link between green emission: an empirical analysis of China's residents. Energy Policy 35 (1), 247–257.
purchasing decisions and measures of environmental consciousness. Eur. J. Mark. 30 Wilson, J., Tyedmers, P., Spinney, J.E., 2013. An exploration of the relationship between
(5), 35–55. socioeconomic and well-being variables and household greenhouse gas emissions. J.
Sekulova, F., van den Bergh, J.C.J.M., 2013. Climate change, income and happiness: an Ind. Ecol. 17 (6), 880–891.
empirical study for Barcelona. Glob. Environ. Chang. 23 (6), 1467–1475. World Bank, 2007. Growth and CO2 Emissions: How Do Different Countries Fare?
Shirley, R., Jones, C., Kammen, D., 2012. A household carbon footprint calculator for Environment Department, Washington, DC.
islands: case study of the United States Virgin Islands. Ecol. Econ. 80 (2), 8–14. Xu, X., Han, L., 2017. Diverse effects of consumer credit on household carbon emissions at
Sidiras, D.K., Koukios, E.G., 2004. Solar systems diffusion in local markets. Energy Policy quantiles: evidence from urban China. Sustainability 9 (9), 1563.
32 (18), 2007–2018. Xu, X., Han, L., Lv, X., 2016. Household carbon inequality in urban China, its sources and
State Council, 2016. Work plan for greenhouse gas emission control during the 13th five- determinants. Ecol. Econ. 128, 77–86.
year plan period. State Council of China, released on 27 October 2016. Yang, Z., Fan, Y., Zheng, S., 2016. Determinants of household carbon emissions: pathway
Straughan, R.D., Roberts, J.A., 1999. Environmental segmentation alternatives: a look at toward eco-community in Beijing. Habitat International 57, 175–186.
green consumer behavior in the new millennium. J. Consum. Mark. 16 (6), 558–575. Zhang, Y.J., Bian, X.J., Tan, W., Song, J., 2017. The indirect energy consumption and CO2
Su, B., Ang, B.W., 2013. Input-output analysis of CO2 emissions embodied in trade: emission caused by household consumption in China: an analysis based on the in-
competitive versus non-competitive imports. Energy Policy 56, 83–87. put–output method. J. Clean. Prod. 163, 69–83.
Su, B., Ang, B.W., 2015. Multiplicative decomposition of aggregate carbon intensity Zhang, Y.J., Bian, X.J., Tan, W., 2018. The linkages of sectoral carbon dioxide emission
change using input–output analysis. Appl. Energy 154, 13–20. caused by household consumption in China: evidence from the hypothetical extrac-
Su, B., Ang, B.W., 2017. Multiplicative structural decomposition analysis of aggregate tion method. Empir. Econ. 54 (4), 1743–1775.
embodied energy and emission intensities. Energy Econ. 65, 137–147. Zhu, Q., Peng, X., Wu, K., 2012. Calculation and decomposition of indirect carbon
Su, B., Ang, B.W., Li, Y.Z., 2017. Input-output and structural decomposition analysis of emissions from residential consumption in China based on the input-output model.
Singapore's carbon emissions. Energy Policy 105, 484–492. Energy Policy 48 (3), 618–626.

155

You might also like