Urbanization and lifestyle trends in China

1
Urbanization and lifestyle trends in China: Implications for the carbon
emissions problem
Nicholas Apergis, Northumbria University, U.K., [email protected]
Jun Li, Curtin University, Australia, [email protected]
1
Urbanization and lifestyle trends in China: Implications for the carbon
emissions problem
ABSTRACT
Demographic changes have significant impacts on a country’s long term growth trajectory
through the savings, consumption and labour market channels. Population changes, including
ageing, migration and urbanization may affect significantly the growth prospects for fastgrowing developing countries, like China. Rural population migrates to cities and will need
more energy services and will produce more CO 2 emissions since urban lifestyle turns out to
be generally more energy-intensive. Population and household structures also keep changing
across the majority of Chinese cities. Migration and urbanization together are likely to drive
China’s energy consumption, CO2 emissions and air pollution upwards if the current trend is
to continue. It is, thus, necessary for China to draw useful lessons from experiences in
advanced economies in terms of reconciling population development and environment issues.
This study aims to analyse, for the first time, the challenge of environmental sustainability,
resulting jointly from population and lifestyle changes in China over the period 1978-2012.
3
The empirical analysis, through cointegration and instrumental variables methodologies,
generates empirical findings documenting that urbanisation, population changes and
consumption behavioural changes contribute significantly to increased carbon emissions. The
results seem to be highly significant for policy makers who need to adopt new policies to
mitigate the environmental problem China has to cope with over the next years.
Keywords: Carbon dioxides; Urbanization; Life-style; China
JEL Classifications: Q43; R23; D11
1. Introduction
Over the past years, in the development process of the majority of developed
countries, it has been documented that industry accounts for the largest proportion of carbon
emissions. However, recently, a number of studies highlight that the contribution of
residential energy consumption to carbon emissions has exceeded that of industrial sectors.
Therefore, new research has focused on exploring the effect of population growth, the
associated residential consumption and the adopted new lifestyles of the population on carbon
emissions (Bin and Dowlatabadi, 2005; Druckman and Jackson, 2009). Moreover, the study
of contemporary urbanization and its impact on the environment is increasingly central to a
wide variety of scientific disciplines. Increasing attention is due to the rapid growth of cities
in many parts
of the world, the growing awareness of the importance of society-wide social, biophysical,
and infrastructural changes that accompany urbanization, and how these dramatic shifts
influence trends across a range of environmental issues at multiple geospatial and temporal
scales (Douglas, 2013). Of particular concern is the role of cities and the urbanization process
in climate change and specifically the carbon cycle (Seto and Reenberg, 2014). There are
ongoing debates about either whether urbanization is fundamentally detrimental to
environmental quality (Srinivas, 2000; Brown, 2001) or, with appropriate governance,
incentives, and cultural capacities, urbanization can be a potential path toward low carbon
societies (Owen, 2009; Glaeser, 2011)].
3
The literature has recently started to closely monitoring urbanization levels, given that
these levels are highly relevant to residential consumption as well as to consumption
structures. Urbanization generally affects carbon emissions through three mechanisms: i) the
use of energy in production is concentrated primarily in cities; as a result, residential
consumption increases in line with urbanization. The total outcome leads to higher energy
demand, resulting in carbon emission increases, ii) the requirements for infrastructure and
dwelling houses grow, leading to higher levels of the demand for building materials, which
are important sources of carbon emissions, and iii) urbanization involves the conversion of
grasslands and woodlands, with such land-use changes increasing carbon emissions.
The goal of this paper is to analyse the challenge of environmental sustainability
resulting from population changes in the Chinese economy spanning the period 1978-2012.
In particular, in this study we investigate, for the first time in the literature, the joint effect
from both lifestyle characteristics and urbanization levels in China on carbon emissions. By
doing so, we are expecting to get more complete and accurate information reflecting the
impact of those drivers on carbon emissions. To this empirical end, we make use of both
cointegration and Instrumental Variables (IV) methodologies. The empirical findings are
expected to provide substantial guidelines for policy makers on how to mitigate carbon
emissions, by explicitly intervening and targeting urbanization and lifestyle, without,
however, reducing the households’ standard of living.
China’s demography has been evolving rapidly. In Europe, population ageing started
early but evolved slowly, it took 50-100 years to double the population aged 65 and over, or
from 7% to 14% of the total population, whereas it took only 25 years to complete this
process in China. Ageing population has long term implications for lifestyle changes and
consumption behavioural shifts of the society. Today, urbanization in China is just over 50%.
Rural population migrated to cities will need more financial support from the social security
system, while it will produce more CO2 emissions, provided that urban lifestyle turns out to
be more energy-intensive. Migration, urbanization and family miniaturizing together are
likely to drive China’s energy consumption, CO 2 emissions and air pollution upwards if no
control is put in place. More specifically, since the 1990s, the total living expenditures in both
urban and rural areas have doubled. Across the Chinese territory there have been similar
trends of changes in consumption patterns with a pronounced reduction of food and clothing
and a gradual growth of shares for education and recreation, medical services, and
communication and transportation, while the gap in the standard of living between urban and
rural areas continued to widen. However, migration to urban areas contributed to a number of
5
serious problems. In particular, the urban housing shortage has been a dominant factor in
contributing to poverty levels. The Housing Reform Policy launched in 1980 to solve the
problems of urban housing shortages and poor housing conditions (Hubacek et al., 2007)
promoted commercialization of the housing sector and private ownership, allowing people to
buy their own apartments. At the same time, the government, state owned enterprises,
domestic private companies and overseas developers invested significant funds into the urban
housing development (Hubacek et al., 2007). In addition, the latest consumer items, such as
air conditioners, personal computers, mobile phones and automobiles, which were previously
only the sign of the wealthy part of the population, increased significantly as well (Feng et
al., 2009). As a result, urbanization levels, new lifestyles, as well as the consumption of those
new consumer items dramatically increased, therefore, giving rise to substantial problems
related to rising pollution levels and, thus, they will signify that China will follow western
trajectories with regards to CO2 emissions.
In China, air pollution is a serious issue for the entire country. Particles smaller than
2.5 micron (PM2.5), may pass through the lungs to affect other organs, leading to health
hazards such as heart disease, altered lung function and lung cancer. The concentration of
PM2.5 in China is the highest in the world and air pollution is more pronounced in cities than
in rural areas. From 2000 to 2010, the incidence of lung cancer in Beijing increased by
56.35%. China’s air pollution and CO2 emissions both contribute to an important part in
global environmental changes. It is, thus, necessary for China to draw useful lessons from
experiences in advanced economies in terms of reconciling population growth and
environmental quality.
Overall, population dynamics and lifestyle changes reflected upon consumption
patterns are expected to substantially influence China's energy use and generated carbon
emissions. The further investigation of these issues is expected to facilitate improvements in
decision making for low carbon development.
The paper is structured as follows: Section 2 reviews the literature, while Section 3
introduces methodological issues and data description. Section 4 presents the empirical
analysis and, finally, Section 5 concludes the paper.
5
2. Literature review
Most environmental degradation can be traced to the behaviour of consumers either
directly, through activities like the disposal of garbage or the use of cars, or indirectly through
the production activities undertaken to satisfy them (Daly, 1996; Duchin, 1998). Cultural and
behavioural changes also affect energy use and carbon emissions. It is thus important to
understand lifestyle changes in order to support urban planning policy making. Lifestyle
factors may include societal values, norms, disciplines, cultural climate and others.
Urbanisation in developing countries is likely to reshape household lifestyles in many cases,
thus causing a shift in consumption behaviour (e.g. more intensive use of motorised vehicles
or electric appliances and electronic devices).
In general, the literature that investigates the relationship between population
characteristics and carbon emissions has followed two major strands: One strand investigates
the causalities and mechanisms of interaction between population characteristics and carbon
emissions, and through the employment of statistical and econometric methodologies
evaluates the impact of those population characteristics on carbon emissions. Lutzenhiser
(1991) discusses different physical, economic, psychological and sociological models and
points out that lifestyle research needs to take account the cultural perspective on household
energy use. Knapp and Mookerjee (1996) discuss the nature of the relationship between
global population growth and CO2 emissions through Granger causality tests. Their results
document the absence of a long-term equilibrium relationship, and only the presence of a
short-term dynamic relationship between CO2 emissions and population growth. Shui and
Dowlatabadi (2003) indicate that a sectoral approach (categorization based on industrial,
transportation, commercial, and residential sectors) is limited in its capacity to reveal the total
impacts of consumer activities on energy use and its related environmental impacts proposed
by the Consumer Lifestyle Approach (CLA) to explore the relationship between consumer
activities and environmental impacts in the U.S. Their results highlight that more than 80% of
the energy used and the CO2 emitted in the U.S. are a consequence of consumer demands and
the economic activities to support these demands, while direct influences due to consumer
activities (i.e., home energy use and personal travel) are 4% of the U.S. GDP, but they
account for 28% and 41% of the U.S. energy use and CO2 emissions, respectively. Moreover,
indirect influences (i.e., housing operations, transportation operations, and food) involve
more than twice the direct energy use and CO2 emissions. Therefore, the characterization of
both direct and indirect energy use and emissions is critical to the design of more effective
7
energy and CO2 emission policies. Satterthwaite (2009) investigates the CO2 emission levels
in various nations for an extended time period (e.g., 1950 to 2005). His empirical findings
indicate low association between rapid population growth and high emission increases, while
Jiang and Hardee (2009) illustrate that consumption and production patterns across various
population groups differ. However, a major drawback in these studies is that population size
is the only demographic variable considered, while lifestyle trajectories and the role of
urbanization are ignored. Hence, paying more attention to the variables of urbanization and
lifestyle trends is an important necessity in investigating the impact of population on carbon
emissions. The second mechanism is the effect of population growth-related emissions on
deforestation. Birdsall (1992) concludes that reductions in population growth matter, but are
not the key factor in levelling off carbon emissions.
A different strand of the literature gives emphasis on the role of urbanization for
carbon emissions. Satterthwaite (2009) considers the implications of population growth and
urbanization for climate change. His empirical findings document that the dominant factor for
higher greenhouse gas emissions is the increasing number of urban consumers along with
their consumption levels. Poumanyvong and Kaneko (2010) empirically investigate the
effects of urbanization on energy use and CO2 emissions. Their findings show that the impact
of urbanization on carbon emissions is positive across all income groups, but that this effect
is more pronounced in the middle-income group. The literature has also identified that
urbanization affects the carbon cycle indirectly and directly by facilitating releases and
absorptions of carbon dioxide (Hutyra et al., 2014). As cities grow, the concentration of
population and concomitant changes in social, political, behavioural, and economic activities
accelerates carbon releases through land use changes and increased use and consumption of
energy and energy-related materials. Alternatively, urban vegetation can absorb carbon from
the atmosphere and cities can temporarily store carbon in building materials and other
infrastructure (Pataki et al., 2006). Although direct carbon sequestration by urban plants and
soils is negligible as compared with urban carbon emissions, local cooling effects that reduce
energy use can be also substantial (Pataki et al., 2011).
In the strand of literature that focuses on the Chinese economy, there exists a
significant number of studies investigating energy consumption and CO2 emissions (FisherVanden et al., 2004; Hubacek et al., 2007; Peters et al., 2007; Auffhammer and Carson, 2008;
Guan et al., 2008; Lin et al., 2008). Pachauri and Jiang (2008) compare the household energy
transitions in China and India by analysing aggregate statistics and nationally representative
household surveys. The authors demonstrate that urban households in both countries consume
7
a disproportionately large share of commercial energy and are much further along in the
transition to modern energy. Feng et al. (2009) trace lifestyle changes in China. Their
descriptive statistics analysis shows that household consumption on a regional scale follows
similar trajectories, driven by changes in income and the increasing availability of goods and
services. They also find that technological improvements have not been capable of fully
compensating for higher emissions levels due to population growth and increasing wealth. Li
and Li (2010) by making use of data from 30 Chinese provinces, explore how the population
structure and technology levels impact carbon emissions. Their findings display that the
population driver affects positively carbon emissions, while Wang (2010) finds both positive
(i.e., GDP per capita, population size, family income) and negative (i.e., energy intensity,
transportation average length of transportation routes, residential energy intensity) effects on
carbon emissions. Du et al. (2012) investigate the driving forces, emission trends and
reduction potentials of China's CO2 emissions. Their empirical findings document that
economic development, technology progress and industry structure are all the most important
factors affecting China's carbon emissions, while the impact of energy consumption structure,
trade openness and urbanization level is negligible. Their scenario simulations show that CO2
emissions in China will increase continuously up to 2020, but the reduction potential is
significant. Finally, Zhu and Peng (2012) examine the impact of population size, population
structure, and consumption levels on carbon emissions. By making use of a ridge regression
model and expanding the stochastic impact on population, affluence, and technology, they
highlight empirical findings revealing that changes in both consumption levels and
population structure are the dominant drivers for carbon emissions. It is shrinking household
sizes that significantly contribute to higher levels of residential consumption, resulting in
higher carbon emissions.
3. Data and methodology
This study makes use of annual data spanning the period 1978-2012, obtained from
various sources, including United Nations (UN) Population Division, World Bank’s World
Development Indicator database, National Bureau of Statistics of China (NBS), International
Energy Agency, U.S. Energy Information Administration (EIA), and UN Food and
Agriculture Organisation. The Appendix summarises the sources of data, while Table 1
provides a number of descriptive statistics in relevance to all the variables used in the
empirical analysis.
9
In terms of the methodological approaches used, we first specify a log-linear model to
estimate the determinants of per capita CO2 emissions in China. In this version we do not
include disaggregated lifestyle consumption, only aggregate consumption.
ln(PCO2)t = a + b1 ln(PCY)t + b2 ln(PCY)2t + b3 ln(URB)t + b4 ln(TRAD)t + b5 ln(CONS)t + ut
(Model 1)
where, PCO2 denotes per capita carbon missions, PCY is per capita income, URB is the
urbanization measure, TRAD defines trade openness, CONS denotes total population
consumption and, finally, u is the error term.
In addition, we test the impact of both urbanization and lifestyle consumption through
a model that explicitly introduces the two major components that dictate lifestyle
consumption, i.e. housing (HOUS) and motor vehicle (MOTOR) expenses. The new model
yields:
ln(PCO2)t = a + c1 ln(PCY)t + c2 ln(PCY)2t + c3 ln(URB)t + c4 ln(TRAD)t + c5 ln(MOTOR)t
+ c8 ln(HOUS)t + ut
(Model 2)
Next, given that the economic system is commonly characterised by inertia, i.e. the effect of a
random exogenous shock in previous periods may be carried over through a time path,
although the behaviour of the economy is mainly influenced by explanatory factors during
the current period (Hill et al., 2008). This characteristic is also relevant in our modelling
framework. Carbon emissions at time t are inherently correlated with the emissions levels
during a certain time period k precedent time t due to the strong inertia of their technical and
urban infrastructures. For instance, it is impossible or extremely costly to replace all the
power plants and freight infrastructure or remove all the private vehicles from expressways
overnight. Upgrading road transport network turns out to be extremely capital and time
intensive process.
Thus, we proceeded to estimate the model by modifying explicitly
introducing carbon dioxide lags. This allows us to draw further conclusions about the shortand long-term impacts of urbanization and lifestyle consumption on carbon emissions. The
new version of the model yields:
9
ln(PCO2)t = a + b1 ln(PCY)t + b2 ln(PCY)2t + b3 ln(URB)t + b4 ln(TRAD)t + b5 ln(CONS)t +
b6 ln(PCO2)t-1 + b7 ln(URB)t-1 + b8 ln(CONS)t-1 + ut
(Model 3)
and
ln(PCO2)t = a + c1 ln(PCY)t + c2 ln(PCY)2t + c3 ln(URB)t + c4 ln(TRAD)t + c5 ln(MOTOR)t
+ c6 ln(HOUS)t + c7 ln(PCO2)t-1 + c8 ln(URB)t-1 + c9 ln(MOTOR)t-1
+ c11 ln(HOUS)t-1 + ut
(Model 4)
Finally, we will examine how a number of demographic changes contributes to the evolution
of carbon intensity. To this end, we add life expectancy (LIFE), fertility (FERTILITY) as well
as three age variables representing the population up to 14 years old (POP14), the population
between 15 and 64 years old (POP1564) and finally, the population above 64 years old
(POP>64). This form of the model yields:
ln(PCO2)t = a + b1 ln(PCY)t + b2 ln(PCY)2t + b3 ln(URB)t + b4 ln(TRAD)t + b5 ln(FERT)t
+ b6 ln(CONS)t + b7 ln(LIFE)t + b8 ln(POP14)t + b9 ln(POP1564)t
+ b10 ln(POP>64)t + ut
(Model 5)
and
ln(PCO2)t = a + c1 ln(PCY)t + c2 ln(PCY)2t + c3 ln(URB)t + c4 ln(TRAD)t + c5 ln(FERT)t +
c6 ln(LIFE)t + c7 ln(MOTOR)t + c8 ln(HOUS)t + c9 ln(POP14)t
+ c10 ln(POP1564)t + c11 ln(POP>64)t + ut
(Model 6)
11
4. Empirical analysis
4.1. Unit Root Testing
To test the stationarity properties of the variables involved in the modelling process, we make
use of two modified Dickey–Fuller tests with good power. These are the DF-WS test,
proposed by Park and Fuller (1995), making use of the WSLS estimator, which is more
efficient than the OLS estimator in estimating autoregressive parameters, and the DF-GLS
test, proposed by Elliot et al. (1996), which analyses the sequence of Neyman–Pearson tests
of the null hypothesis of the presence of a unit root. The results are reported in Table 2. They
indicate that all the variables are integrated of order one.
[Insert Table 2 about here]
4.2. Cointegration Testing and Estimates
Next, Johansen and Juselius (1990) cointegration tests are performed in the cases of
Models 1, 2, 5, and 6. They reveal evidence in favour of cointegration between carbon
emissions and the above mentioned control variables across all versions of the modelling
approach. The cointegration results are reported in Table 3. Both the eigenvalue test statistic
and the trace test statistic indicate that there is a single long-run relationship across the
variables under study as well as across the various versions of the model examined.
[Insert Table 3 about here]
Once the presence of a cointegrating relationship is established, the analysis is carried out by
obtaining the estimations of the cointegration vector. In particular, normalizing the
cointegration vector on carbon emissions, we obtain the estimates reported in Table 4.
[Insert Table 4 about here]
GDP growth is illustrated to exert a strong impact on CO2 emissions, reflecting both the
income effect and the specificity of industry dominance in China’s economic growth model.
The negative sign of squared GDP growth is in accordance with the diminishing marginal
income effect, indicating the presence of an inverted-U shaped Kuznets effect associated with
income and suggesting that China is in a position to mitigate the negative spillovers from
carbon emissions and air pollution as the country gets more affluent. Our results are in
accordance to those reached by de Leon Barido and Marshall (2014) who provide supportive
11
evidence for the positive association between income/level of development and carbon
emissions across many countries.
Urbanisation is also shown to contribute to carbon emission intensity per capita in a
positive manner. This is mostly due to the presence of a higher demand for goods and
services supplied as well as lifestyle changes. The positive effect of urbanization on carbon
emissions indicates that China is not characterized by strong environmental policies,
therefore, underscoring the importance of pollution problems. Our results receive statistical
report by those in Elliott and Clement (2014) who provide empirical evidence about the
strength of urbanization in leading to higher levels of carbon emissions. These findings
strongly indicate that urbanization requires substantial and ongoing inputs related to energy
and other sectors, which seems to occur through highly inefficient and environmentally
destructive means. The three major characteristics associated with urbanization, i.e.
population concentration, landscape transformation and systematic interaction, all seem to be
substantially conducive to the positive impact on carbon emissions.
In terms of lifestyle consumption, the effect on carbon emissions is also positive.
Higher levels of such consumption by Chinese households indicate that these households tend
to consume more electronics and electrical appliances as a result of becoming more affluent,
indicating a spillover effect on carbon emissions growth. Our findings are in accordance with
those provided by Wei et al. (2007) who suggest that a very high percentage of higher energy
production and their spillover effects on carbon emissions are a consequence of people’s
lifestyles, and the economic activities that support these demands. Both urbanization and
changing lifestyles lead to more energy intensive consumption, shaped by the familiar
western societies and values brought via the mass media.
The analysis also shows that population variables are significantly important as
influential drivers for carbon emissions. In addition, both extended life expectancy and an
ageing population has negative impacts on per capita emissions, indicating that older people
generally have lower consumption of energy intensive goods and services than people in
younger generations, highlighting again the importance of the lifestyle driver for carbon
emission projections. With respect to the population variables, our findings seem to be
receiving statistical support by Shi (2003) and Cole and Neumayer (2004) who find a firm
positive relationship between population changes and carbon dioxide emissions, mostly due
to higher demand for energy consumption.
Extended life expectancy leads to lower carbon intensity, whereas fertility declines
over the last five decades seem to contribute to increasing carbon intensity. These two
13
findings have significant policy implications for China’s demographic and environmental
policy. The literature has provided mixed results. For instance, O’Neill and Chen (2002)
provide evidence for the U.S., showing that the increase in the life expectancy increase
residential energy use, while it decreases transportation energy use, with substantial
consequences for carbon emissions. By contrast, Fan et al. (2006) generate conclusions
similar to ours, given that their study documents a negative impact on CO2 emissions for
countries with high income and a positive impact for countries at other income levels.
The demographic shift has complex effects on carbon emissions intensity changes.
The positive sign of the fertility driver suggests that the demographic changes over the next
decades will continue to increase carbon intensity across the Chinese cities, as the general
trend would be higher levels of life expectancy and a declining child-bearing rate, as a result
of economic development. In effect, both fertility and lifetime variables embody the
underlying lifestyle and consumption patterns changes driving carbon emissions upward. The
findings on the positive association between fertility and carbon emissions find empirical
support by Stephenson et al. (2010) who document that although the contribution of highfertility countries to global carbon emissions is low, it is increasing with the economic
development it needs to reduce their poverty levels. Rapid population growth endangers the
capacity of these communities to adapt to climate changes. Significant mass migration,
mostly to urban areas, occurs as a response to climate changes and should be regarded as a
legitimate response to the effects of such climate changes.
4.3. Robustness Checks: Instrumental Variables (IV) Estimates
Finally, to avoid potential endogeneity problems across our primary variables under
study, i.e. carbon dioxide, urbanization and lifestyle consumption, in estimating models 3 and
4, we make use of the instrumental variable least squares (IVLS) methodological approach.
In the presence of endogeneity, OLS estimators are inconsistent and biased. Nevertheless, the
choice of a valid instrument is always a complicated task. The results appear in Table 5 and
are based on White’s standard errors to avoid potential problems of heteroskedasticity. They
also remain robust across both models. All variables show the expected sign, while they are
statistically significant at the 1% level. The empirical findings are similar to those obtained in
Table 4.
There also exists a lagged income effect. In particular, it is also interesting to note that
the long-run income elasticity of carbon emission is higher than its short-run counterpart,
suggesting that China would be less capable of controlling the carbon dioxide problem as the
13
country becomes more affluent; in other words, carbon intensity would increase at a faster
pace in the longer run, unless a number of serious policy responses are undertaken soon that
will eventually mitigate the pollution problem.
[Insert Table 5 about here]
5. Conclusions and policy implications
This paper examined the impacts of population structure, urbanization and
consumption life-style changes on carbon emissions in China from 1978 to 2012. Through
the methodologies of cointegration and instrumental variables (IV) the empirical analysis
managed to rectify the positive influence of urbanization, consumption life-style, population
structure, fertility and life expectancy on carbon emissions.
Policymakers should find a way to control emissions without sacrificing standards of
living. The empirical analysis revealed that the population structure played a highly
significant role in affecting carbon emissions. Therefore, policies that respond to changes in
urbanization level and age structure should be also seriously considered, given that China will
continue to undergo urbanization in the next several decades, and the pressure of emission
increase to grow will be stronger. Policy makers in China have reoriented their policies
towards urbanization from ‘control the scale of large cities, rationally develop medium cities,
and actively develop small towns’ to ‘develop large, medium, and small cities in a
synchronized manner and form city groups with large radiation effects’. This conversion
reflected the awareness of the vital function of concentrating the use of resources and
environmental treatment in large-scale or medium-scale cities. Policymakers should more
strongly emphasize the assembly and scale effects of cities to reduce emissions and initiate
low carbon development.
The working age population profile also indicates that the trend of population aging
will be the major characteristic of China's age structure in the future. Given this backdrop, the
empirical findings imply that changes in the age structure will present an alleviationdominant impact on carbon emissions in the future. In terms of consumption life-styles and
experiencing a shift from household budgets dominated by expenditures on food and clothes
to increasing shares of expenditures for services, housing and luxury items, such as
automobiles, driven by changes in income and the increasing availability of goods and
services, China will keep following ascending trajectories regarding CO2 emissions. Based
on those findings, there is a lot of room for improvement on the consumer side as well as an
awareness of the interaction of infrastructure and consumption. Future expenditures and
15
consumer behaviour depend to a large extent on the infrastructure that is built today.
Consumers' choices are bound by the availability of alternatives (i.e., public transportation).
Thus, the trajectory of future emissions needs to be addressed by the infrastructure choices of
today (Hubacek et al., 2007). In terms of the consumption items, there are only a few
alternatives in sight as Chinese consumers try to emulate Western lifestyles and thus,
inadvertently, western levels of emissions.
Overall, policies regarding urbanization and derived residential consumption will
extensively influence carbon emission trends. Policymakers should immediately initiate
overall planning for population development and residential consumption by establishing the
necessary assessment frameworks and guidance systems that promote low carbon
development in China. For instance, policies directed to enhance efforts to target home
energy reduction should focus on building attributes. Such efforts should focus on changing
behaviours by taking an approach that makes efficiency more convenient, increases
motivation, and provides more actionable and pertinent information (Carrico et al., 2011).
Therefore, an effective reduction in energy use and emissions will come from building
efficiency, energy conservation actions based on human behaviours, and conservation
education.
Discussions around the processes of urbanization should incorporate the political
discourses surrounding fossil fuels, land use, and lifestyle choices when accounting for
pollutants. Recent research has examined the politics and power dynamics engaged in
resistance to low-carbon transitions (Geels, 2014). A directed effort to embed studies of
energy consumption in the political and institutional aspects of urbanization is necessary. A
greater deployment of policy analysis tools, methods, and theoretical approaches would
substantially contribute to an improved understanding of the way policy interventions try to
accomplish in a mission to influence the trajectory of urbanization as it relates to both energy
use and carbon emissions.
Appendix
Data sources
Data series
Population (total, by age group)
Life expectancy
Fertility
GDP ( PPP, MER)
Source
UN population division
WDI
UN, WDI
WDI
15
Urbanisation
Per capita GDP
Trade openness
Export/import
Energy consumption
Energy productivity (GDP/energy)
Electricity production and consumption
Energy price (oil)
CO2 emissions
PM10
Meat consumption (mutton, pork, beef)
number of motor vehicle per capita
WDI
WDI
WDI
WDI
WDI, EIA, IEA
WDI
WDI, EIA, IEA
EIA, BP
IEA, WDI
WDI
FAO , research publications
NBS, research publications
References
Auffhammer, M., Carson, R.T., 2008. Forecasting the path of China's CO2 emissions using provincelevel information. Journal of Environmental Economics and Management 55, 229-247.
de Leon Barido, D.P., Marshall, J.D., 2014. Relationship between urbanization and CO2 emissions
depends on income level and policy. Environmental Science and Technology 48, 3632-3639.
Bin, S., Dowlatabadi, H., 2005. Consumer lifestyle approach to US energy use and the related CO2
emissions. Energy Policy 33, 197-208.
Birdsall, N., 1992. Another look at population and global warming. Working Paper, No. WPS 120,
Washington, DC: The World Bank.
Brown, L.R., 2001. Eco-Economy, Building an Economy for the Earth. W.W. Norton & Co., New
York.
Carrico, A.R., Vandenbergh, M.P., Stern, P.C., Gardner, G.T., Dietz, T., Gilligan, J.M., 2011. Energy
and climate change: Key lessons for implementing the behavioural wedge. Journal of Energy and
Environmental Law, 61-67.
Cole, M.A., Neumayer, E., 2004. Examining the impact of demographic factors on air pollution.
Population and Development Review 26, 5-21.
Douglas, I., 2013. Cities: An Environmental History, I. B. Tauris & Co. Ltd, London.
17
Druckman, A., Jackson, T., 2009. The carbon footprint of UK households 1990–2004: A socioeconomically disaggregated, quasi-multi-regional input–output model. Ecological Economics 68,
2066-2077.
Du, L., Wei, C., Cai, S., 2012. Economic development and carbon dioxide emissions in China:
Provincial panel data analysis. China Economic Review 23, 371-384.
Elliot, G., Rothenmberg, T.J., Stock, J.H., 1996. Efficient tests for an autoregressive unit root.
Econometrica 64, 813-836.
Elliott, J.R., Clement, M.T., 2014. Urbanisation and carbon emissions: A nationwide study of local
countervailing effects in the United States. Social Science Quarterly 95, 795-816.
Fan, Y., Liu, L.C., Wu, G., Wei, Y.M., 2006. Analysing impact factors of CO2 emissions using the
STIRPAT model. Environmental Impact Assessment Review 26, 377-395.
Feng, K., Hubacek, K., Guan, D., 2009. Lifestyles, technology and CO2 emissions in China: A
regional comparative analysis. Ecological Economics 69, 145-154.
Fisher-Vanden, K., Jefferson, G.H., Liu, H., Tao, Q., 2004. What is driving China's decline in energy
intensity? Resource and Energy Economics 26, 77-97.
Geels, F.W., 2014. Regime resistance against low-carbon transitions: Introducing politics and power
into the multi-level perspective. Theory of Cultural Society 31, 21-40.
Glaeser, E., 2011. Triumph of the City: How Our Greatest Invention Makes us Richer, Smarter,
Greener, Healthier, and Happier, Penguin Press, New York.
Guan, D., Hubacek, K., Weber, C.L., Peters, G.P., Reiner, D.M., 2008. The drivers of Chinese CO2
emissions from 1980 to 2030. Global Environmental Change 18, 626-634.
Hubacek, K., Guan, D., Barua, A., 2007. Changing lifestyles and consumption patterns in developing
countries: A scenario analysis for China and India. Futures 39, 1084-1096.
Hutyra, L.R., Duren, R., Gurney, K.R., Grimm, N., Kort, E., Larson, E., Shrestha, G., 2014.
Urbanization and the carbon cycle: Current capabilities and research outlook from the natural sciences
perspective. Earth’s Future, doi:10.1002/2014EF000255.
17
Hubacek, K., Guan, D., Barua, A., 2007. Changing lifestyles and consumption patterns in developing
countries: A scenario analysis for China and India. Futures 39, 1084-1096.
Jiang, L., Hardee, K., 2009. How do recent population trends matter to climate change? PAI work
paper. http://www.populationaction.org/Publications/Working_Papers/April_2009.
Johansen, S., Juselius, K., 1990. Maximum likelihood estimation and inference on cointegration—
With applications to the demand for money. Oxford Bulletin of Economics and Statistics 52, 169-210.
Knapp T., Mookerjee, R., 1996. Population growth and global CO2 emissions. Energy Policy 24, 3137.
Li, G., Li, Z., 2010. The impacts of population, economy and technology on carbon dioxide
emissions: A study based on the dynamic panel model. Population Studies 3, 32-39.
Lin, J., Zhou, N., Levine, M., Fridley, D., 2008. Taking out 1 billion tons of CO2: The magic of
China's 11th Five-Year Plan? Energy Policy 36, 954-970.
O’Neill, B.C., Chen, B.S., 2002. Demographic determinants of household energy use in the United
States. Population and Development Review 28, 53-88.
Owen, D., 2009. Green Metropolis: Why Living Smaller, Living Closer, and Driving Less Are the
Keys to Sustainability, Riverhead Books: New York.
Pachauri, S., Jiang, L., 2008. The household energy transition in India and China. Energy Policy 36,
4022-4035.
Pataki, D.E., Carreiro, M.M., Cherrier, J., Grulke, N.E., Jennings, V., Pincetl, S., Pouyat, R.V.,
Whitlow, T.H., Zipperer, W.C., 2011. Coupling biogeochemical cycles in urban environments:
Ecosystem services, green solutions, and misconceptions. Frontiers in Ecology 9, 27-36.
Park, H.J., Fuller, W.A., 1995. Alternative estimators and unit root tests for the autoregressive process.
Journal of Time Series Analysis 16, 415-429.
19
Pataki, D.E., Alig, R.J., Fung, A.S., Golubiewski, N.E., Kennedy, C.A., McPherson, E.G., Nowak,
D.J., Pouyat, R.V., Romero-Lankao, P., 2006. Urban ecosystems and the North American carbon
cycle. Global Change Biology 12, 2092-2102.
Peters, G., Weber, C.L., Guan, D., Hubacek, K., 2007. China's growing CO2 emissions – A race
between lifestyle changes and efficiency gains. Environmental Science and Technology 41, 59395944.
Poumanyvong, P., Kaneko, S., 2010. Does urbanization lead to less energy use and lower CO2
emissions? A cross-country analysis. Ecological Economics 70, 434-444.
Satterthwaite, D., 2009. The implications of population growth and urbanization for climate change.
Environmental Urbanization 21, 545-567.
Seto, K.C., Reenberg, A., 2014. Rethinking Global Land Use in an Urban Era, MIT Press, Cambridge,
Mass.
Shi, A., 2003. The impact of population pressure on global carbon dioxide emissions, 1975-1996:
Evidence from pooled cross-country data. Ecological Economics 44, 29-42.
Sims, C.A., 1980. Macroeconomics and reality. Econometrica 48, 1-48.
Srinivas, H., 2000. Focusing on the real environmental culprits—Urban areas: UNU’s City
Inspirations Initiative. Global Environmental Change 10, 233-236.
Stephenson, J., Newman, K., Mayhew, S., 2010. Population dynamics and climate change: What are
the links? Journal of Public Health 32, 150-156.
Wang, Q., 2010. Research on the relationship of population and carbon emissions based on the
nonlinear hypothesis. Population Studies 1, 3-12.
Wei, Y.M., Lan, C.L., Ying, F., Gang, W., 2007. The impact of lifestyle on energy use and CO2
mission: An empirical analysis of China’s residents. Energy Policy 35, 247-257.
Zhu, Q., Peng, X., 2012. The impacts of population change on carbon emissions in China during
1978–2008. Environmental Impact Assessment Review 36, 1-8.
19
Table 1
Descriptive statistics
___________________________________________________________________________
Variable
Mean
Median
Minimum
Maximum
Std. Dev.
_______________________________________________________________________________
Fertility
3.07681
2.69600
0.00000
6.16100
1.69880
Per capita GDP 6.05826
5.95215
4.44869
8.11612
1.10851
Per capita CO2 7.53933
7.59504
6.35291
8.80373
0.68783
Per capita motor
vehicles
3.57647
3.57521
1.29048
6.58893
1.60998
Urbanization
3.24792
3.17562
2.80817
3.94692
0.37380
Consumption
530.186
460.141
143.8334
1229.966
311.319
Life expectancy 65.2392
68.4617
0.00000
75.04200
12.1210
Squared per
capita GDP
37.9076
35.4303
19.79090
65.87140
13.8380
Housing
32030.4
9750
1103
159313
43402.3
Trade openness 33.5788
31.6746
5.31418
70.56707
18.8394
Popul. up to 14 31.2878
30.1907
17.98437
41.13266
7.75763
Popul.15-64
63.0201
64.1573
55.24876
73.50738
6.27837
Population>64 5.69216
5.65193
3.618581
8.679630
1.49548
__________________________________________________________________________
21
Table 2
Unit root tests
___________________________________________________________________________
Variable
DF-WS test
DF-GLS test
___________________________________________________________________________
co2
-1.20(3)
-1.18(3)
Δco2
-5.46(2)*
-5.73(2)*
pcy
-1.34(3)
-1.30(3)
Δpcy
-5.27(1)*
-5.58(2)*
2
pcy
-1.32(3)
-1.26(4)
Δpcy2
-5.30(2)*
-5.81(3)*
urb
-1.18(4)
-1.14(3)
Δurb
-5.83(2)*
-6.16(2)*
trad
-1.15(3)
-1.09(3)
Δtrad
-5.76(2)*
-5.96(1)*
fert
-1.26(3)
-1.21(3)
Δfert
-5.28(2)*
-5.62(1)*
life
-1.29(4)
-1.17(3)
Δlife
-6.11(2)*
-6.58(1)*
motor
-1.35(3)
-1.24(3)
Δmotor
-6.24(1)*
-6.46(2)*
hous
-1.37(3)
-1.28(3)
Δhous
-5.69(2)*
-5.87(1)*
pop14
-1.24(3)
-1.15(3)
Δpop14
-5.11(1)*
-5.47(2)*
pop1564
-1.29(4)
-1.20(3)
Δpop1564
-5.48(3)*
-5.72(1)*
pop>64
-1.18(3)
-1.12(3)
Δpop>64
-5.39(2)*
-5.85(1)*
___________________________________________________________________________
Notes: Small letters indicate variables in logarithms. Δ indicates first differences. Numbers in parentheses denote
the optimal number of lags used in the augmentation of the test regression and were obtained through the Akaike
criterion. * indicates the rejection of the unit root null hypothesis at 5%.
21
Table 3
Cointegration tests
___________________________________________________________________________
r
n-r
ml
95%
Tr
95%
___________________________________________________________________________
Model 1: Lags = 4
r=0
r=1
48.914
40.219
95.773
86.544
r≤1
r=2
33.209
34.397
72.329
75.328
r≤2
r=3
23.549
28.167
49.806
53.347
r≤3
r=4
18.741
21.894
30.163
35.068
r≤4
r=5
10.844
15.752
14.395
20.168
r≤5
r=6
2.356
9.094
2.356
9.094
Model 2: Lag = 3
r=0
r=1
r≤1
r=2
r≤2
r=3
r≤3
r=4
r≤4
r=5
r≤5
r=6
r≤6
r=7
57.893
38.429
31.936
21.883
16.126
10.249
2.014
51.258
40.219
34.397
28.167
21.894
15.752
9.094
109.852
82.315
70.308
45.627
29.318
12.537
2.014
102.326
86.544
75.328
53.347
35.068
20.168
9.094
Model 5: Lag = 3
r=0
r=1
r≤1
r=2
r≤2
r=3
r≤3
r=4
r≤4
r=5
r≤5
r=6
r≤6
r=7
r≤7
r=8
r≤8
r=9
r≤9
r=10
r≤10
r=11
110.935
81.637
65.328
52.624
40.925
31.863
24.658
20.326
15.637
8.924
1.452
95.962
84.007
72.549
60.016
51.258
40.219
34.397
28.167
21.894
15.752
9.094
226.547
160.836
124.702
105.863
90.725
69.527
52.831
36.725
19.836
9.711
1.452
197.435
172.814
146.739
129.817
102.326
86.544
75.328
53.347
35.068
20.168
9.094
23
Model 6: Lag = 2
r=0
r=1
126.594
118.753
248.994
231.709
r≤1
r=2
87.515
95.962
157.782
197.435
r≤2
r=3
80.005
84.007
139.613
172.814
r≤3
r=4
59.814
72.549
115.204
146.739
r≤4
r=5
51.426
60.016
101.316
129.817
r≤5
r=6
38.515
51.258
87.515
102.326
r≤6
r=7
28.371
40.219
64.714
86.544
r≤7
r=8
22.810
34.397
49.005
75.328
r≤8
r=9
17.626
28.167
31.139
53.347
r≤9
r=10
13.711
21.894
15.638
35.068
r≤10
r=11
8.427
15.752
4.119
20.168
r≤11
r=12
0.246
9.094
0.246
9.094
___________________________________________________________________________
Notes: r is the number of cointegrating vectors, n-r is the number of common trends, ml = maximum eigenvalue
statistic, Tr = Trace statistic. The number of lags was determined through Likelihood Ratio tests, developed by
Sims (1980).
23
Table 4
Long-run estimates – Models 1, 2, 5, and 6
___________________________________________________________________________
Variables
Model 1
Model 2
Model 5
Model 6
__________________________________________________________________________
Intercept
0.964
0.873
1.006
0.993
(1.09)
(0.94)
(1.32)
(1.08)
pcy
0.438*
0.416*
0.386*
0.392*
(6.58)
(6.24)
(6.14)
(5.84)
pcy2
-0.061*
-0.048*
-0.042**
-0.039**
(-4.13)
(-3.86)
(-2.64)
(-2.75)
urb
0.228*
0.215*
0.204*
0.211*
(6.19)
(6.42)
(5.91)
(6.09)
trad
0.185*
0.172*
0.153*
0.162*
(5.61)
(5.28)
(5.09)
(5.46)
cons
0.307*
0.258*
(6.53)
(5.63)
motor
0.195*
0.173*
(5.41)
(5.29)
hous
0.158*
0.129*
(5.26)
(5.14)
fert
0.116*
0.093*
(5.18)
(4.95)
life
0.084*
0.069*
(5.28)
(5.02)
pop14
0.249*
0.226*
(6.15)
(5.94)
pop1564
0.191*
0.174*
(6.03)
(5.82)
pop>64
0.072**
0.046**
(2.61)
(2.37)
Diagnostics
R2
LM
0.62
[0.32]
0.57
[0.28]
0.64
[0.38]
0.69
[0.41]
25
RESET
[0.24]
[0.22]
[0.31]
[0.35]
___________________________________________________________________________
Notes: Numbers in parentheses denote t-statistics. LM and RESET are tests for serial correlation and model
functional misspecification, respectively. Figures in brackets denote p-values. *, ** denote significance at the
1% and 5% statistical levels, respectively.
Table 5
IV Estimates – Models 3 and 4
___________________________________________________________________________
Variables
Model 3
Model 4
__________________________________________________________________________
Intercept
0.758
0.762
(1.16)
(1.03)
Δpcy
0.393*
0.377*
(6.15)
(5.86)
2
Δpcy
-0.053*
-0.042*
(-4.28)
(-3.59)
Δurb
0.214*
0.192*
(6.35)
(6.28)
Δtrad
0.167*
0.151*
(5.83)
(5.58)
Δcons
0.348*
(6.61)
Δmotor
0.226*
(5.80)
Δhous
0.184*
(5.71)
Δco2(-1)
0.583*
0.514*
(6.17)
(5.92)
Δurb(-1)
0.086*
0.064*
(5.29)
(5.11)
Δcons(-1)
0.118*
0.103*
(6.03)
(5.82)
Δmotor(-1)
0.095*
(5.68)
Δhouse(-1)
0.047*
(5.41)
Diagnostics
R2
Sargan
0.66
[0.28]
0.61
[0.24]
25
LM
[0.25]
[0.26]
RESET
[0.21]
[0.18]
___________________________________________________________________________
Notes: The instruments used for Model 3 are: a constant, three lags of income, 4 lags of urbanization and 4 lags
of consumption, while those for Model 4 are: a constant, 4 lags of income, 4 lags of urbanization, 3 lags of
motor vehicle expenses and 3 lags of housing expenses. Sargan tests the null validity of instruments used. The
remaining notes remain similar to those in Table 4.