You are on page 1of 20

China Economic Review 36 (2015) 146–165

Contents lists available at ScienceDirect

China Economic Review

Population growth and the environmental Kuznets curve


Sophie Xuefei WANG a,⁎, Yu Benjamin FU b, Zhe George ZHANG b,c
a
China Center for Human Capital and Labor Market Research, Central University of Finance and Economics, Beijing, China
b
Beedie School of Business, Simon Fraser University, Burnaby, B.C., Canada
c
Department of Decision Sciences, College of Business and Economics, Western Washington University, Bellingham, WA, United States

a r t i c l e i n f o a b s t r a c t

Article history: This paper provides a specific application of the environmental Kuznets curve (EKC) in order to
Received 31 January 2015 explain the effect of population growth on the environment. The main purpose is contributing
Received in revised form 31 August 2015 to enhance the connection between theoretical and empirical analysis. We develop an overlap-
Accepted 31 August 2015
ping generations (OLG) model that featured with an inverted U-shaped relationship between pol-
Available online 16 September 2015
lution emission and income and we examine the effect of population growth on this relationship.
Simulations illustrate the model's predictions that positive population growth makes the EKC
Keywords: steeper and have higher peak, but it does not fundamentally change the pollution–income rela-
Environmental Kuznets curve
tionship. The econometric analysis finds evidence supporting our model's predictions using Chi-
Population growth
nese data at the province level.
Overlapping generations model
Fixed effects © 2015 Elsevier Inc. All rights reserved.
Pollution

1. Introduction

The environmental Kuznets curve (EKC) hypothesizes that the relationship between the level of environmental degradation and
economic growth follows an inverted U-shape: as income per capita increases, measured levels of environmental degradation, such as
pollution emissions, increase at first, but then, after some turning point, start decreasing. Previous studies on the EKC often focus on
two questions: Is the EKC hypothesis plausible? And, if it is plausible, where is its turning point? In this paper, we try to deepen the
understanding of an important question for policy: what role does population growth play in changing the environmental quality, by
untangling the complex interaction among environmental quality, economic growth and population growth. More specifically, what
effect does population growth have on the environmental quality and the EKC?
To answer this question, we first provide theoretical foundation for the interaction among environmental quality, economic
growth and population growth. We begin with deducing the EKC using an overlapping generations (OLG) model in which agents
live for two periods and pollution is an unavoidable by-product of production. The properties of both production and abatement tech-
nologies are the joint determinants that shape the pollution-income path. Under our model's assumptions, the pollution-income re-
lationship is inverted U-shaped as described in the EKC hypothesis. Then, we examine the effects of population growth on this
relationship. In the OLG model, agents face a trade-off between economic growth and environmental quality. On the one hand, the
positive population growth puts pressure on consumption and, consequently, production, which generates more pollution. On the
other hand, the altruistic agents are more concerned about the environmental quality if they have more children, as children are

⁎ Corresponding author at: China Center for Human Capital and Labor Market Research, Central University of Finance and Economics, 39 Xueyuan South Road, Haidian
District, Beijing, 100081, China. Tel.: +86 10 62288391, +86 18811426675.
E-mail addresses: sophiewang2004@gmail.com (S.X. Wang), yuf@sfu.ca (Y.B. Fu), gzhang@sfu.ca (Z.G. Zhang).

http://dx.doi.org/10.1016/j.chieco.2015.08.012
1043-951X/© 2015 Elsevier Inc. All rights reserved.
S.X. Wang et al. / China Economic Review 36 (2015) 146–165 147

susceptible to poor environmental quality.1 Our model predicts that the production effect dominates the environmental effect, and
population growth makes the EKC steeper and exerts a burden on the environment. However, the negative effect of population
growth on the environment is limited, unlike the neo-Malthusians' doomsday prediction. Simulations are provided to illustrate the
model's predictions. Next, using panel data from China, we empirically analyze the EKCs for industrial waste gas emissions and indus-
trial solid wastes in China, and examine the effects of population growth on the EKCs. Thus, our study focuses on the link between
industrialization and environment, extending the existing literature which investigates the effect of population growth on the EKC
using deforestation as a measure of environmental degradation (Barbier & Burgess, 2001; Bhattarai & Hammig, 2001; Cropper &
Griffiths, 1994; Koop & Tole, 1999; Scrieciu, 2007).
The rest of this paper is organized as follows. Section 2 reviews the related literature. Section 3 shows the model set-up and the-
oretical analysis. Section 4 presents numerical simulations for the model. Section 5 provides empirical evidence from China, using
two-way fixed effects models. Section 6 concludes the paper.

2. Literature review

2.1. The population and environment debate

Debate on the impact of population on environment can be traced back to Malthus's concern over the relationship between pop-
ulation growth and food production 200 years ago. Malthus (1798) argues that population grows geometrically, while food produc-
tion can only grow arithmetically. A growing population forces the cultivation of land of poorer and poorer quality. This
environmental degradation lowers the marginal product of labor. As a result, population growth will outstrip growth in food supply
and the consequence can be catastrophic. More than a century later, after witnessing rapid technological progress, Boserup (1965,
1976, 1981) argues that the pressure of population growth is the driving force for economic development through technological in-
novation and substitution of scarce factors with more abundant ones. These developments will keep food production apace with pop-
ulation growth, enabling the natural environment to support a larger population at the same level of welfare. Land degradation may
take place but technological advances can limit the damage, especially when markets are well-functioning. Soon the debate was ex-
tended to other aspects of environmental resources and environmental quality, such as, pollution, natural resources, deforestation and
climate change. The neo-Malthusians predict a pessimistic future: without controlling population growth, the growth of population
and output will exceed the earth's carry capacity either in terms of resource availability or in terms of assimilative capacity for
waste.2 With growing population against limited resources, the limits of arable land will be reached, resources such as energy and
minerals will be exhausted, and pollution levels will rise beyond the earth's assimilative capacity. On the contrary, the “cornucopians”
are much more optimistic and predict that population growth provides both the incentive and the means for advancement that con-
tinuously pushes back resource limits and creates wealth.3 Therefore, population growth and environmental quality are not negative-
ly related, as predicted by the Malthusians and neo-Malthusians. The views of the neo-Malthusians and the “cornucopians” are the
two extremes, and the truth is likely to be somewhere in between.

2.2. The EKC theory and evidence

There is an extensive literature on the theory and empirical evidence for the EKC, the former of which was developed in the early
1990s. One plausible explanation is based on the perspective of a natural progression of economic development: the economy starts
from a clean agrarian economy, develops into a polluting industrial economy, and then develops into a clean service-based economy
(Arrow et al., 1995; Munasinghe, 1999). Other theories focus on scale, technique, and composition effects (Grossman & Krueger, 1991;
Kaufmann, Davidsdottir, Pauly, & Garnham, 1998; Panayotou, 1993). If as an economy grows, the scale of all activities increases pro-
portionally, pollution will increase with economic growth. If growth is not proportional but is accompanied by a change in the com-
position of goods produced, and if richer economies produce proportionally fewer pollution-intensive products, because of changing
tastes or patterns of trade, this composition effect can lead to a decrease in pollution as the economy grows. Finally, if richer countries
use less pollution-intensive production techniques, as environmental quality is a normal good, growth can lead to falling pollution. In
the earlier stage of economic development, the scale effect dominates, but it will eventually be outweighed by the composition and
technique effects. The EKC summarizes the interaction of these three processes.
More recent theoretical works on the EKC often build dynamic models based on neoclassical macroeconomic theories, with a few
exceptions.4 Chimeli and Braden (2005), Dinda (2005), López (1994); Selden and Song (1995) and Stokey (1998) extend the basic
dynamic growth model with infinitely lived agents, by including the environment and the disutility of pollution. The EKC can also
be derived from an OLG model as in John and Pecchenino (1994) and Cao, Wang, and Wang (2011). There are two major differences
between their models and the OLG model here. First, John and Pecchenino (1994) and Cao et al. (2011) assume that agents are young
and work in the first period and become old and consume only in the second period, while our model assume agents are children and
make no economic decision in the first period and enter adulthood, work and are altruism toward their children in the second period.

1
WHO suggests that children aged less than five years that live in developing countries are the most vulnerable population in terms of total deaths attributable to
pollution.
2
The neo-Malthusian school is represented by Paul Ehrlich, Dennis Meadows and his fellow Club of Rome members.
3
The “cornucopia” is represented by Herman Kahn and Julian Simon.
4
Andreoni and Levinson (2001) and Lieb (2002) developed static models to explain the EKC.
148 S.X. Wang et al. / China Economic Review 36 (2015) 146–165

Second, in John and Pecchenino (1994) and Cao et al. (2011), each young agent supplies his one unit of labor inelastically and makes
the decision on how to divide his wage between saving and investing in environmental maintenance, while in our model, we assume
each adult agent makes the decisions on how to divide her human capital between production and pollution abatement and how to
allocate her income between her consumption and the consumption of her children.
Empirically, the EKC hypothesis is often tested by estimating a polynomial function of income. Mixed results have been found
across studies and even in the same studies for different environmental indicators. Studies that find evidence for the existence of
the EKC include Baek and Kim (2013); Carson, Jeon, and McCubbin (1997); Cole, Rayner, and Bates (1997); Grossman and Krueger
(1991); Hettige, Lucas, and Wheeler (1992); Hilton and Levinson (1998); Martinez-Zarzoso and Maruotti (2011); Panayotou
(1993, 1997); Selden and Song (1995); Shafik and Bandyopadhyay (1992); Song, Zheng, and Tong (2008); Wagner (2008). Studies
that find no significant relationship, monotonic increasing relationship, or N-shaped relationship between the measured levels of en-
vironmental degradation and income include Caviglia-Harris, Chambers, and Kahn (2009); Grossman and Krueger (1995); Holtz-
Eakin and Selden (1995); Koop and Tole (1999); Roy and van Kooten (2004); Yang, He, and Chen (2015). Studies that find mixed re-
sults include Brajer, Mead, and Xiao (2011); Cropper and Griffiths (1994); Onafowora and Owoye (2014) and Shafik (1994). The turn-
ing points of the EKC vary for different environmental indicators, but for most of the pollution indicators, the estimated income
turning points fall in the threshold of $3000–10,000 measured in 1985 USD (Dinda, 2004).

2.3. Population and the EKC

Following the population and environment debate, some EKC studies accommodate population in their models, and most of them
find negative impact of population on environment, with a few exceptions. Using cross-section data, Panayotou (1993) finds that the
turning point of the EKC for deforestation is delayed by a higher population density, and Panayotou (1997) proves that population
density raises the height of the EKC for sulfur dioxide at every level of income. Using panel data in Malaysia, Vincent (1997) finds
that higher population density is associated with higher total suspended particulates (TSP) concentration and worse water quality.
Using pooled cross-country data, Shi (2003) finds that global population change is positively associated with growth in carbon dioxide
emissions, and the impact of population change on emissions is much more pronounced in developing countries than in developed
countries. Using cross-national panel data, Selden and Song (1994) find a negative relationship between population density and air-
borne emissions (sulfur dioxide, carbon monoxide, and suspended particulate matter). They attribute this negative relationship to less
stringent environmental standards in countries with lower population densities. Shen (2006) finds mixed results for air and water
pollutants in China: population density is negatively related to dust fall, positively related to cadmium in water, and not related to sul-
fur dioxide, arsenic and chemical oxygen demand (COD) in water.
Studies on population growth and the EKC concentrate in the field of deforestation. Barbier and Burgess (2001); Bhattarai and
Hammig (2001); Cropper and Griffiths (1994) and Scrieciu (2007) find that population growth shifts up the EKC for deforestation
and the negative environmental effect is quite large in magnitude, while Koop and Tole (1999) find an insignificant positive effect
of population growth on deforestation.

3. Model

3.1. Assumptions

We consider an OLG model in which agents live for two periods. Generation t is defined as the adults living in period t; thus, gen-
eration t is born in period t -1 in which they are children and become adults in period t. Each adult has η children at the beginning of
period t, and lives with them for one period. Each child is endowed with nothing, and makes no economic decision, while adults are
altruistic toward their children and make all the decisions. The initial population is normalized to 1, N1 =1. Thus, population evolves
according to Nt = ηt-1.
Each adult of generation t has preferences defined over her consumption and her children's welfare weighted by the number of
children. Children's welfare is measured in two dimensions: their consumption and the environmental quality during their child-
hood.5 Thus, the utility function of an adult of generation t is given by

 η α
U ðcat ; cct ; Et Þ ¼ cat ðcct Et Þ ð1Þ

where η is the gross population growth rate, α N 0 measures the degree of altruism toward children, cat is the consumption of the
adult, cct is the total consumption of the children in the household and Et is the environmental quality. For simplicity, we assume
α = 1.6

5
Here, we assume environmental quality enters adult's utility through its effect on her children's health, as children are more susceptible to poor environmental
quality than adults.
6
The model can be easily extended by relaxing this constraint.
S.X. Wang et al. / China Economic Review 36 (2015) 146–165 149

Each adult of generation t is endowed with Ht units of human capital, and she decides how to allocate her human capital between
production and the development and adoption of pollution abatement technology. Human capital accumulates at an exogenous rate,
h. The initial level of human capital is normalized to 1, H1 =1. Thus, the human capital constraint is given by

t1
Hpt þ zt ¼ Ht ≡ ð1 þ hÞ ð2Þ

where Hpt is the level of human capital supplied to production, and zt is the level of human capital supplied to the development and
adoption of pollution abatement technology. Each adult owns a firm in a competitive market, and supplies human capital to her firm
inelastically at wage rate, w. The production function is yt = f(Hpt), with f '' (⋅) b 0 b f ' (⋅). She then divides her incomes between her own
consumption and consumption of her children. Therefore, agent's budget constraint is given by the following equation

cat þ cct ¼ wt Hpt þ πt ð3Þ

where πt is the firm's profit. Pollution is a by-product of production; that is, pollution increases in the level of output. Firms have op-
tions to develop and adopt pollution abatement technologies which requires input in terms of human capital, zt. zt = 0 implies no pol-
lution abatement. Aggregate pollution emission is determined by the function below
 
P t ¼ Nt g ðzt Þf H pt ð4Þ

where Nt acts as an index of the number of firms, zt is the human capital input in pollution abatement in period t, and g(zt) measures
the pollution intensity, with g(zt) N 0, g'(zt) b 0 b g''(zt), g' ' '(zt) ≤ 0, and gðzt Þ∈ ½g ; g ̅, where gð0Þ ¼ g and lim gðzÞ ¼ g ̅ , as a result of
z→∞
abatement technological constraint. Eqs. (2) and (4) imply a trade-off between ̅ pollution abatement and output growth. Pollution
emission per worker is given by pt =g(zt)f(Hpt).
Environmental quality evolves over time according to
   
Et  E ¼ ð1  bÞ Et1  E  max 0; P t  P

ð5Þ

where b ∈ (0, 1), measuring the self-adjustment ability of the nature. Thus, without human activity, environment quality eventually
converges to E⁎, which is a stable equilibrium. This assumption is plausible, because ecology and biology literature shows that the
earth can absorb and purify minor quantities of pollutants, possibly because of the chemical properties of some vegetation.7 Without

loss of generality, we assume E0 = E⁎. Yet, the amount of pollution that can be absorbed by the ecosystem is limited, and the limit is P,
 in period t.
with P b E . Therefore, production deteriorates environmental quality by the amount of max½0; P t  P
Taking the levels of human capital, Ht, existing environmental quality, Et-1, and population growth rate, η as given, a representative
agent solves the following optimization problem

η
max U ðcat ; cct ; Et Þ ¼ cat ðcct Et Þ ð6Þ
cat ;cct ;Et ;H pt ;zt

s:t:cat þ cct ¼ wt H pt þ π t ð7Þ

Hpt þ zt ¼ Ht ð8Þ

   
Et  E ¼ ð1  bÞ Et1  E  max 0; P t  P

ð9Þ

 
t1
Pt ¼ η g ðzt Þf H pt : ð10Þ

A competitive equilibrium for this economy is a sequence of allocations {cat, cct, Et, Hpt, zt, Pt, πt} and prices {wt}, such that, in each
period t,
1) agents maximize Eq. (6) subject to .Eqs. (7)–(10);
2) firms maximize profits;
3) markets clear.

7
There are studies documenting the biological purification of sewage from chemical plants and the ability of ornamental plants to absorb and purify environmental
pollutants (Koren'kov, 1991; Wang et al., 2006). The self-adjustment ability of the earth was also assumed by other economists, such as John and Pecchenini (1994).
150 S.X. Wang et al. / China Economic Review 36 (2015) 146–165

3.2. The EKC

We analyze the model starting with deducing the pollution–income relationship by assuming zero population growth, η = 1. Ap-
plying the equilibrium conditions, the representative agent's problem becomes

max U ðcat ; cct ; Et Þ ¼ cat cct Et ð6′Þ


cat ;cct ;Et ;H pt ;zt

 
s:t:cat þ cct ¼ f H pt ð7′Þ

Hpt þ zt ¼ Ht ð8′Þ

   
Et  E ¼ ð1  bÞ Et1  E  max 0; P t  P

ð9′Þ

 
P t ¼ g ðzt Þf H pt : ð10′Þ

To solve this maximization problem, we divide the time horizon into three stages and analyze each stage individually: At the very
 That is, pollution is at
early stage, as agent has very low level of human capital and thus the output level is so low that gð0Þf ðH t Þ≤ P.
such low level that it has no effect on environmental quality even without abatement. At the second stage, as human capital accumu-
lates to a certain level, firms generate more pollution than P and environmental quality degrades. Now, agents start considering pol-
lution abatement in order to balance economic growth and environmental quality. The third stage begins with a level of human capital
at which agents adopt a pollution abatement technology such that the pollution level decreases back to P.  Thus, the environmental
quality converges to its natural equilibrium during the third stage.

 where gð0Þf ðHÞ


3.2.1. Stage 1: H t ≤ H  ¼ P.
At this stage, the result is trivial. Even if agents do not invest in pollution abatement, the levels of pollution emission are still lower
than the limit that can be absorbed by the ecosystem and thus environmental quality are not affected. Consequently, agents have no
incentive to sacrifice consumption to adopt abatement technology, zt = 0. If environmental quality starts at its natural equilibrium,
E0 =E⁎, it will stay at the level of E⁎ at stage 1. Therefore, the time paths for key variables at stage 1 are as follows: output is increasing,
pollution emission is increasing, environmental quality is constant at the level of E⁎, and investment in pollution abatement is constant
at the level of 0.

3.2.2. Stage 2: H b H t b He where gð0Þf ðHÞ


 ¼ P and gðz ðHeÞÞ f ðHe  z ðHeÞÞ ¼ P for z ðHeÞN0. z ðHeÞ is the solution to agent's optimization prob-
lem at the level of human capital of H . e
Entering stage 2, the level of pollution emission will exceed the level that can be absorbed by the earth without abatement (gð0Þ
 and the environmental quality starts to degrade. Thus, agents start considering investing in pollution abatement to
f ðH t ÞNP),
balance economic growth and environmental quality. Agent's optimization problem is represented in the following Lagrangian
function:
  
cat cct Et þ λt ð1  bÞEt1 þ bE  g ðzt Þf ðH t  zt Þ þ P  Et þ μ t ½ f ðH t  zt Þ  cat cct 

L¼ max
cat ;cct ;Et ;Hpt ;zt ;λt ;μ t

The corresponding first order conditions (FOCs) are

cat : cct Et ¼ μ t ð11Þ

cct : cat Et ¼ μ t ð12Þ

Et : cat cct ¼ λt ð13Þ

 0  0
zt : λt g0 ðzt Þf ðH t  zt Þ  g ðzt Þf ðHt  zt Þ ¼ μ t f ðH t  zt Þ ð14Þ

 
λt : Et  E ¼ ð1  bÞ Et1  E  gðzt Þf ðHt  zt Þ þ P

ð15Þ
S.X. Wang et al. / China Economic Review 36 (2015) 146–165 151

μ t : cat þcct ¼ f ðHt  zt Þ ð16Þ

Substituting Eqs. (11), (12), (13) and (16) into Eq. (14) gives:

2 ðzt Þ
½ f ðHt  zt Þ g 0 ¼ ½2Et  f ðH t  zt Þg ðzt Þf 0 ðH t  zt Þ: ð17Þ

Eqs. (15) and (17) jointly determine the optimal level of investment in pollution abatement z⁎(Ht). Note that it is never optimal to
 Otherwise, agents can lessen the abatement investment, increase human capital input in production,
abate pollution to a level below P.
increase consumptions and thus become better off without affecting environmental quality. Notice also that a necessary condition for
the existence of solutions for Eq. (17) is f(Ht -zt)g(zt)b 2Et which we assume holds.
Eq. (17) will have corner solutions when Et is large and Ht is small, since the decreasing LHS of Eq. (17) is always smaller than its
increasing RHS for all values of zt. At the corner solutions, z⁎ = 0. Therefore, at the beginning of stage 2, as agent's human capital is
lower than some threshold, H, ^ and the level of environmental quality is high, agent's marginal utility of consumption is larger than
her marginal utility of environmental quality without pollution abatement, and therefore, agent will allocate all her human capital
into production and no pollution abatement technology is adopted. And then as the level of human capital reaches H in period T,
the environmental quality deteriorates to such a level that the optimization problem starts to have interior solution where z⁎ N 0.
For the rest of this subsection, we will focus on the properties of the interior solutions to the optimization problem for periods t ≥ T
at stage 2.
^ b Ht b H
Proposition 1. Agents invest more in the pollution abatement technology as their human capital accumulates, for H e. That is,
0 b dzt/dHt b 1, if f' '(⋅) b 0b f' (⋅) and g'(⋅) b 0b g''(⋅).

Proof. Totally differentiate Eqs. (15) and (17) and solve the equation system of the two differentiation equations for dzt/dHt. Under
the assumption that f ''(⋅) b 0 b f ' (⋅) and g'(⋅) b 0 b g ' ' (⋅), we have 0 b dzt/dHt b 1. The detailed proof is shown in the Appendix A. qed.
Proposition 1 suggests that agents prefer to invest more in pollution abatement as their human capital accumulates and the economy
grows, after their levels of human capital reach H. Eq. (19) shows that there are two effects of increasing human capital. First, the direct effect
in the first term is negative. It shows that as human capital accumulates, agents are capable of producing more output and thus more pol-
lution, and consequently, the environmental quality deteriorates. The second effect is an indirect effect: as human capital accumulates after it
^ agents invest more in pollution abatement, as shown in Proposition 1, which exerts a positive effect on the environ-
reaches the level of H,
mental quality. If the direct production effect dominates, the environmental quality deteriorates as the economy grows; otherwise, economic
growth will be associated with improving environmental quality. What is the relationship between the levels of pollution emission and in-
come? This is shown in Proposition 2.

Proposition 2. The relationship between the levels of pollution emission and income is inverted U-shaped at stage 2. The turning point of the
Þ=g0 ðz
EKC is at the output level that the corresponding investment in pollution abatement z satisfies d½gðz Þ þ fðH   0
z Þ=f ðHt  zÞ=dz ¼ 0.
t

Proof: If the EKC hypothesis holds here, environmental degradation measured by pollution emission per worker and income mea-
sured by output per worker should be first positively related, and then after some turning point, negatively related. Thus, we will ex-
amine the sign of dpt/dyt. Since dpt/dyt and dpt/dHt have the same sign, we will examine the sign of dpt/dHt, instead. By differentiating
Eq. (10) with respect to Ht and substituting dzt/dHt from the proof of Proposition 1 in the Appendix A into it, we can prove the exis-
tence of a turning point. The detailed proof is shown in the Appendix B. qed.
A numerical example to deduce the inverted U-shaped EKC is as follows. f(x) = xθ, and g(z) = 1/(zγ + K), with K N 0 and 0 b θ b γb 1.
Then,

f ðxÞ x
0 ¼
f ðxÞ θ

!
f ðH t  zt Þ
d 0
f ðH t  zt Þ 1
¼ N0
dðH t  zt Þ θ

γ
g ðzt Þ z þK
0 ¼  t γ1
g ðzt Þ γzt

" #
g ðzt Þ
d 0

g ðzt Þ γ γ1 1
¼ Kzt 
dzt γ γ
152 S.X. Wang et al. / China Economic Review 36 (2015) 146–165

Therefore, we have


gðz Þ f ðH  zt Þ
d  0 t þ 0 t

g ðzt Þ f ðHt  zt Þ 1γ γ 1 1
¼ Kzt þ 
dzt γ γ θ


gðz Þ f ðH  zt Þ
d2  0 t þ 0 t
g ðzt Þ f ðHt  zt Þ γ1
¼ ð1  γÞKzt b0
dz2t

At the turning point of the EKC, the level of pollution abatement investment, z, satisfies



1γ γ 1 1
Kz þ  ¼ 0
γ γ θ

or

1
θð1  γÞK γ
z ¼ :
γθ

Thus, when zt b z, pollution emission is increasing as the economy grows; when zt N z, pollution emission is falling as the economy
grows. Together, we will observe an inverted U-shaped EKC. For example, if θ =1/3 and γ =1/2, at the turning point of the EKC, z ¼ 1.
The turning point level of income can be calculated by substituting z into the equation system of Eqs. (15) and (17).
The economic interpretation of Proposition 2 is that as human capital accumulates and environmental quality deteriorates, agents
start to adopt abatement technology to achieve a balance between economic growth and environmental quality. Proposition 1 shows
that agents always invest an increasing level of their human capital in adopting abatement technology after their levels of human cap-
^ From Eq. (17), the allocation of human capital is in equilibrium when the marginal benefit of production equals the mar-
ital reach H.
ginal benefit of pollution abatement. Nevertheless, at low levels of pollution abatement, the technology effect is not large enough to
offset the production effect, and therefore the level of pollution emission continues rising. As human capital accumulates, agents will
adopt sufficient pollution abatement which is large enough to offset the negative effect from production growth, the level of pollution
emission starts to fall.

Lemma 1. The time path of the investment in pollution abatement is increasing for periods t≥ T at stage 2. The time path of the pollution
emission is increasing at first and then decreasing at stage 2.

Proof: A continuous-time model is more appropriate to formally prove Lemma 1. However, since the proof is quite straightforward,
we will skip the formal proof using maths, and instead explain it in words. Proposition 1 states that the level of investment in pollution
abatement increases in the level of human capital for periods t ≥T at stage 2. And from the proof of Proposition 2, the level of pollution
emission first increases in the level of human capital, and then it decreases as human capital accumulates. Thus, as time passes at stage
2, human capital accumulates; at the same time, the level of investment in pollution abatement increases for periods t ≥T, and the level
of pollution emission increases at first, and then decreases. qed.
The time paths for key variables at stage 2 are as follows: output is increasing, pollution emission is first increasing and then de-
creasing, environmental quality is first decreasing and then increasing, and investment in pollution abatement is first constant at the
level of 0 and then increasing.

e where gðz ðH
3.2.3. Stage 3: H t ≥H eÞÞf ðH
e  z ðH
eÞÞ ¼ P.

From Proposition 2, pollution emission falls after the income level passes some threshold. Because dPt/dHt b 0 and PðH eÞ ¼ P,
 at
stage 3, PðH ≥H eÞ≤ P.
 However, abating pollution below the level of P is never optimal, and therefore, pollution emission will be kept
 In order to keep the level of pollution emission constant, pollution abatement needs to be increasing to catch up
at the level of P.
with the output growth. In addition, at the level of P ¼ P, pollution emission has no effect on environmental quality, and Et converges
back to E⁎ at stage 3. The time paths for key variables at stage 3 are as follows: output is increasing, pollution emission is constant at the
 environmental quality is increasing and converging back to E⁎, and investment in pollution abatement is increasing.
level of P,

3.3. Positive population growth

When the population growth rate is positive, η N 1, the representative agent's problem is defined by Eqs. (6), (7), (8), (9) and (10).
Transform the objective function (Eq. (6)) by taking its natural logarithm. This positive increasing transformation will not change the
S.X. Wang et al. / China Economic Review 36 (2015) 146–165 153

solution to our optimization problem, and the corresponding Lagrangian function is:
n    
ln cat þ ηln cct þ ηlnEt þ λt ð1  bÞEt1 þ bE  max 0; P t  P  Et þ μ t ½ f ðHt  zt Þ  cat cct 

L¼ max
cat ;cct ;Et ;Hpt ;zt ;λt ;μ t ;δt
h  io
t1
þδt P t  η g ðzt Þf H pt :

 the solution is trivial, z⁎t = 0, as in stage 1 discussed in subsection 3.2. Thus, we will focus on the case of P t NP.
For P t ≤ P,  In this case,
the Lagrangian function is simplified as
n h i o
ln cat þ ηln cct þ ηlnEt þ λt ð1  bÞEt1 þ bE  η g ðzt Þf ðH t  zt Þ þ P  Et þ μ t ½ f ðH t  zt Þ  cat cct : :
 t1
L¼ max
cat ;cct ;Et ;Hpt ;zt ;λt ;μ t

Higher rate of population growth increases the weights on the children's consumption and environmental quality in the agent's
utility function, and thus, it has two effects on her optimal choices and the environment. First, agent must produce more to feed
her increasing number of children. This increase in output leads to a higher level of pollution emission and worse environmental
quality. Second, as the agent puts more weight on the environmental quality, she may invest more in pollution abatement to avoid
harming the environment. The corresponding first order conditions (FOCs) of the Lagrangian function are

1
cat : ¼ μt ð18Þ
cat

η
cct : ¼ μt ð19Þ
cct

η
Et : ¼ λt ð20Þ
Et

t−1  0 0  0
zt : −η λt g ðzt Þf ðH t −zt Þ−g ðzt Þf ðHt −zt Þ ¼ μ t f ðH t −zt Þ ð21Þ

  
λt : Et −E ¼ ð1−bÞ Et−1 −E −η g ðzt Þ f ðH t −zt Þ þ P ̅
t−1
ð22Þ

μ t : cat þcct ¼ f ðHt  zt Þ ð23Þ

Substituting Eqs. (18), (19), (20), and (23) into Eq. (21) gives
n o
t 0 2 0 0
η −g ðzt Þ½ f ðHt −zt Þ þ f ðH t −zt Þg ðzt Þf ðHt −zt Þ ¼ ð1 þ ηÞEt f ðH t −zt Þ ð24Þ

Eqs. (22) and (24) jointly determine the optimal level of abatement investment z⁎(Ht). At the solutions, Propositions 1 and 2 and
Lemma 1 still hold, which is shown in the proofs of Lemma 2 and Proposition 4. In addition, the effects of population growth on some
key variables are shown in Propositions 3 and 4. Eq. (24) defines the optimal level of abatement investment. Its LHS is the marginal
benefit of an increase in the abatement investment, and its RHS is its marginal cost. Population growth rate, η, shows in both sides
of the equation. A higher η increases both the marginal benefit and the marginal cost of abatement investment, so how does η affect
the optimal level of investment in pollution abatement? Proposition 3 provides an answer for this question.

Proposition 3. The optimal level of investment in pollution abatement increases in the rate of population growth. That is, zt/dη N 0 .

Proof. Proposition 3 can be proved by reshaping Eq. (24) and differentiating the reshaped equation with respect to η. The detailed
proof is shown in the Appendix C. qed.
Proposition 3 implies that if two economies are identical except the population growth rates, then the economy with higher pop-
ulation growth rate will invest more in pollution abatement. But higher pollution abatement investment does not necessarily imply
lower pollution emission in country with higher population growth rate.

Lemma 2. Agents invest more in the pollution abatement technology as their human capital accumulates. That is, 0 b dzt/dHt b 1.

Proof. Totally differentiate Eqs.(22) and (24) and solve the equation system of the two differentiation equations for dzt/dHt. Under
the assumption that f ''(⋅) b 0 b f ' (⋅) and g'(⋅) b 0 b g ' '(⋅), we have 0 b dzt/dHt b 1. The detailed proof is shown in the Appendix D. qed.
Proposition 3 and Lemma 2 together suggest that a less developed country with a higher rate of population growth may have sim-
ilar level of investment in pollution abatement per worker as a more developed country with a lower rate of population growth.
154 S.X. Wang et al. / China Economic Review 36 (2015) 146–165

Proposition 4. With positive population growth, the pollution–income relationship is still inverted U-shaped. Higher rate of population
growth makes the EKC steeper.

Proof. The proof is shown in the Appendix E. qed.


Though proposition 3 shows that higher rate of population growth gives agents higher incentive to invest in pollution abatement,
this effect is outweighed by its positive effect on production and thus pollution emission. As a result, the overall effect of population
growth on the environmental quality is negative, as shown in Propositions 4. However, population growth does not fundamentally
change the pollution–income relationship, as the EKC hypothesis still holds, which implies that the level of pollution emission will
fall after some income turning point. Therefore, the effect of population growth on environment is limited, unlike the doomsday pre-
diction by the neo-Malthusians.

4. Simulation

In this section, we will analyze the EKC by simulation. We replace the general functional forms in our model with specific forms
that satisfy our assumptions. We assume that each generation is 20 years and the dynamics occur on yearly basis. The production
and pollution abatement technology functions are assumed to be:

θ
F ðH t  zt Þ ¼ ðHt  zt Þ

Fig. 1. The time paths of key variables.


S.X. Wang et al. / China Economic Review 36 (2015) 146–165 155

1
g ðzt Þ ¼
ðs þ zt Þγ þ K

with 0 b θ b γ b 1. A small positive number s is added to prevent the nonexistence of a real solution to g ' (z t ). Otherwise,
when zt = 0, g ðzt Þ ¼  γ γ 2 1γ with the denominator value of zero.
0

ðzt þKÞ zt
The parameters are assumed as follows: θ = 1/3 , s= 0.01,γ = 0.5,and K =1. The human capital growth rate is h= 0.05. The equi-
librium level of the environmental quality is E⁎ = 20. The environmental self-adjustment parameter is b =0.02, and its pollution self-
absorbing parameter is P ¼ 1.
We start with the case of zero population growth and use MATLAB to find the solutions to the dynamic system of Eqs. (15) and
(17). Fig. 1 shows the dynamics of the investment in pollution abatement per worker, income per worker, pollution emission and en-
vironmental quality. According to our model, the level of environmental quality stays at the equilibrium level at stage 1 and it starts to
deteriorate at the beginning of stage 2. Therefore, Panel D in Fig. 1 shows that stage 2 starts at period 6. Also, according to our model,
the level of pollution emission reach P and stay at this level at stage 3. Therefore, Panel C in Fig. 1 shows that stage 3 starts at period
180. Together, stage 1 is from period 1 to period 5, stage 2 is from period 6 to period 179, and stage 3 is from period 180 onwards.
In Fig. 1, Panel A illustrates the time path of the investment in pollution abatement per worker. As discussed in Section 3.2, the level
of abatement investment stays at zero at stage 1 and the beginning of stage 2. From period T = 17 onwards, the level of abatement
investment is positive and increasing over time. Panel B shows the time path of income per worker. The level of income per worker
increases over time, representing economic growth. Panel C shows the time path of pollution emission. The level of pollution emission
increases at stage 1; at stage 2, it first increases and then decreases, as stated in Lemma 1; and it reaches the level of P and stay there at
stage 3. It takes about 30 periods for the level of pollution emission to increase from level of P to its peak, but it takes much longer time
for it to decrease back to the level of P from its peak. Panel D shows the time path of environmental quality. The level of environmental
 then it deteriorates rather fast to its minimum, and after that it takes a
quality stays at its equilibrium level at E⁎ at stage 1 where P t ≤ P,
long time to gradually converge back to its equilibrium level at E⁎ at stage 3. Fig. 2 shows the EKC, the relationship between pollution
emission and income. The EKC is inverted U-shaped, but the shape is not symmetric. As the economy grows, the level of pollution
emission rises very fast and reaches its peak in period 43, after which it falls slowly.
The effects of population growth on the environment and the EKC are shown in Figs. 3–4. The population growth rates are assumed
to be 0% (the baseline), 0.2% (low) and 0.4% (high), and the values of other parameters are the same as described above. First of all,
higher rate of population growth extends the number of periods at stage 2. Stages 2 and 3 start in periods 5 and 180, 5 and 246,
and 4 and 373 when η = 1, η = 1.002, and η = 1.004, respectively. In Fig. 3, Panel A shows that higher rates of population growth
shift up the time path of abatement investment, which illustrates that agents invest more in pollution abatement as population
growth rate increases, as stated in Proposition 3. Panel B shows that higher rates of population growth shift up the time path of pol-
lution emission and make the time for the pollution emission to fall back to the level of P from its peak much longer. Panel C shows that
higher rates of population growth shift down the time path of environmental quality everywhere except at stage 1 when the environ-
mental quality stays at its equilibrium level, and the higher rates also make the environmental recovery much slower. Fig. 4 shows that
higher rates of population growth shift up the EKC and make its slope steeper and its peak higher. Especially, as the economy grows,
the pollution emission rises at a faster rate to a higher peak if the population growth rate is higher.

5. Empirical evidence

To extent the literature on the effect of population on the environment, we examine the effect of population growth on the EKC for
two aspects of environmental degradation other than deforestation: industrial waste gas emission and industrial solid wastes. Our

Fig. 2. The EKC.


156 S.X. Wang et al. / China Economic Review 36 (2015) 146–165

Fig. 3. The time paths of key variables at different population growth rates.

Fig. 4. The EKCs.


S.X. Wang et al. / China Economic Review 36 (2015) 146–165 157

main regression model is a reduced-form polynomial function as follows

k k
X it ¼ ∑ βk Y it þ ∑ α k Y it Popit þ μ i þ φt þ εit ð25Þ
k k

where Xit is the yearly per capita emission of the pollutant in province i at time t, Yit is the local GDP per capita, Popit is the local pop-
ulation growth rate, μi is the provincial fixed effects which account for time-invariant province-specific shocks to the pollution emis-
sion, φtis the year fixed effects which account for time-varying shocks common to all provinces, and εit is the error term.8
Before we estimate our main model, we first investigate the relationship between the pollution emission and income and test the
EKC hypothesis, by estimating the following polynomial function

k
X it ¼ ∑ βk Y it þ μ i þ φt þ ε it ð26Þ
k

The data are from the Chinese Statistical Yearbook (2001–2010) and the Chinese Environmental Statistical Yearbook (2001–2010).
Table 1 lists the descriptive statistics of the key variables. It is a balanced panel of 30 provinces in 10 years.9 The measures of environ-
mental degradation are the annual per capita emissions of industrial waste gases and industrial solid wastes. There are substantial
variations in the levels of pollutants in the sample, with provinces specializing in heavy industry such as Inner Mongolia, Shanxi
and Hebei the most polluted areas. The within-province variation in real GDP per capita over time is consistent with the rapid
economic growth during this time period. The between-province variation in real GDP per capita is more striking, reflecting unequal
regional development in China. The richest areas are the three municipalities, Shanghai, Beijing and Tianjin, followed by provinces of
Jiangsu, Zhejiang and Guangdong. Population growth rates have been low over years at around 0.5% on average.
Eqs. (25) and (26) are estimated using both the fixed effects model and the random effects model. But Hausman (1978) tests show
that only the fixed effects models can be estimated consistently. In addition, F-tests show that quartic functions best fit our data. The
results of estimating our two-way fixed effects models are shown in Tables 2 and 3. Table 2 shows the estimating results of the EKCs
for industrial waste gases and industrial solid wastes. Columns 2 and 3 are estimations of quartic functions, and to show the compar-
ison, columns 4 and 5 are estimations of quadratic functions. The coefficients of the cubic and quartic terms in the quartic functions are
jointly significant, though the magnitudes are small. Actually the coefficients are so small that they change the curvature of the pol-
lution–income curves without adding additional turning points, as shown in Fig. 5. The turning points of the estimated EKC for indus-
trial waste gases in quartic function and quadratic function are $8320 and $5624, respectively (both measured in 1985 USD, for easy
comparison with previous literature). And the turning points of the estimated EKC for industrial solid wastes are $8137 and $4783 (in
1985 USD), using quartic function and quadratic function. The comparison between the results from the two functional forms indi-
cates that using quadratic function will result in underestimations of the turning points of the EKCs. According to the estimating re-
sults from the quartic function, the increase in these industrial pollutants has been slowing down in more developed areas in more
recent years, but the levels of emission are still rising. The average marginal effects imply that an increase in 1000 yuan (in 1997
RMB) in per capita income is associated with an increase in industrial waste gas emission of 3827 cubic meters per capita and an in-
crease in industrial solid wastes of 0.192 t per capita, on average.
Table 3 shows the estimating results of the effect of population growth rate on the EKCs for industrial waste gas emission and in-
dustrial solid wastes, and Fig. 6 shows the results graphically. In Table 3, all of the coefficients of the polynomial terms are significant,
and for both industrial waste gases and industrial solid wastes, the interaction terms of population growth rate with the quadratic in-
come, cubic income and quartic income are significant. However, the average marginal effects of population growth in both equations
are insignificant, which offers evidence against neo-Malthusian school of thought, in favor of “cornucopia”. Fig. 6 shows that the EKCs
with population growth are on or above the baseline EKCs with zero population growth, and they are steeper and have higher peaks,
as predicted by our theoretical model. Also, the effects are non-linear, with larger effects in higher income areas. At the income levels
in the sample, the magnitudes of the effects of population growth on pollution emissions are small and insignificant.

6. Conclusions

This paper provides a specific application of the environmental Kuznets curve (EKC) in order to explain the effect of population
growth on the environment. We develop an overlapping generations (OLG) model in which the altruistic agents have preferences
over consumption and environmental quality. As the economy grows, pollution emission increases until the level of investments in
pollution abatement catch up with the level of production. Thus, the relationship between pollution emission and income exhibits
an inverted U-shape, as described by the EKC hypothesis. Then, we examine the effect of population growth on this relationship. Sim-
ulations illustrate the model's predictions that higher population growth makes the EKC steeper and have higher peak. But population
growth does not fundamentally change the pollution–income relationship, unlike the doomsday prediction by the neo-Malthusians.
Using province level data from China, we test the EKC hypothesis and examining the effect of population growth on the pollution–

8
We also consider including population density in our model. But its coefficient is insignificant, and therefore, we do not report these results. As a robustness check,
we add population growth as a regressor in the model, and get similar result. The average marginal effect of population growth is insignificant as in our main findings.
The results are available upon request.
9
We exclude Tibet because of missing data. We understand that the time span of 10 year may be too short to test a growth model. But our focus here is to test the
effect of population growth rate on the EKC, rather than to test the validity of the OLG model.
158 S.X. Wang et al. / China Economic Review 36 (2015) 146–165

Table 1
Descriptive statistics of key variables.

Variable Mean Observations Variations

Std. Dev. Min Max

Waste gas per capita (1000 cubic meters) 23.801 300 Overall 16.881 5.330 101.074
Between 13.060 9.728 54.104
Within 10.934 −10.202 71.730
Solid waste per capita (ton) 1.110 300 Overall 0.834 0.094 4.926
Between 0.708 0.159 3.289
Within 0.458 −0.616 3.309
Real GDP per capita (1000RMB, base year = 1997) 14.293 300 Overall 9.742 2.745 56.948
Between 8.513 4.642 42.208
Within 4.961 0.816 31.544
Population growth rate (‰) 5.485 300 Overall 3.069 −1.900 13.100
Between 2.979 0.626 11.151
Within 0.900 2.959 9.754

Source: the Chinese Statistical Yearbook (2001–2010) and the Chinese Environmental Statistical Yearbook (2001–2010).

income relationship. We find that the relationship for industrial waste gas and industrial solid waste is inverted U-shaped. At the in-
come levels in the sample, the pollution levels are still rising, though at slower rates in more developed areas. Population growth has
limited effects on pollution emissions.
China has experienced a growth miracle over the past three decades, but it comes at a cost to the environment. China's current
national income level is far from the income turning point of the EKC for industrial waste gases, yet most areas in the country have
been severely affected by toxic smog for years. According to Chen, Wang, Ma, and Zhang (2013), air pollution causes 350,000 to
500,000 premature deaths in China each year, and air pollution has become the fourth biggest threat to the health of Chinese people.
However, the governments have done far from enough to protect the environment. The increase in industrial waste gas emission as-
sociated with economic growth is large, but it would be inappropriate to conclude that China should call a halt to economic growth. In
addition, our findings suggest that controlling population growth is not a cure for pollution problem, and relaxing the one-child policy
in recent years probably have neglectable effect on the environment. At current low fertility rate in China, children are deeply
cherished by their parents, especially in urban area. Given the fact that children live in developing countries are the most vulnerable
population to pollution, increasing fertility rate by relaxing the one-child policy may raise the voice in pollution control and help to
tighten pollution regulations. Ultimately, lack of environmental regulation and government accountability are the root of the environ-
mental degradation problem. Proper government policies are indispensable to guide firms to adopt such production and abatement
technologies that are in accordance with environmental protection.

Appendices

The appendices provide the mathematical proofs of the theoretical results in the paper.

Table 2
Two-way fixed effects models of the EKCs for industrial waste gases and solid wastes.
Source: the Chinese Statistical Yearbook (2001–2010) and the Chinese Environmental Statistical Yearbook (2001–2010).
Waste gas Solid waste Waste gas Solid waste

Y 8.384*** 0.478*** 2.397*** 0.085***


(1.952) (0.084) (0.497) (0.022)
Y2 −0.253*** −0.017*** −0.029*** −0.001***
(0.092) (0.004) (0.006) (0.0002)
Y3 0.004* 0.0003***
(0.002) (0.0001)
Y4 −0.00002* −0.000002**
(0.00002) (0.000001)
Constant −41.185*** −2.213*** −4.794 0.125
(11.320) (0.485) (4.315) (0.190)
R-Squared 0.652 0.633 0.627 0.583
F-Statistics 36.982 34.054 39.526 32.890
P-value 0.000 0.000 0.000 0.000
Observations 300 300 300 300
Income turning point (1997 RMB) 61,340 59,991 41,464 35,264
Income turning point (1985 USD, PPP) 8320 8137 5624 4783

* p b 0.10, ** p b 0.05, *** p b 0.01.


Note: Y is real GDP per capita (1000 RMB), with base year of 1997. Y2, Y3 and Y4 are its quadratic, cubic and quartic terms. Waste gas is industrial waste gas emission per
capita measured in 1000 cubic meters, and solid waste is the per capita generation of industrial solid wastes measured in ton.
S.X. Wang et al. / China Economic Review 36 (2015) 146–165 159

Table 3
The effect of population growth on the EKCs.Source: the Chinese Statistical Yearbook (2001–2010) and the Chinese Environmental Statistical Yearbook (2001–2010).
Waste gas Solid waste

Y 8.193*** 0.471***
(1.996) (0.086)
Y2 −0.304*** −0.019***
(0.103) (0.004)
Y3 0.007** 0.0004***
(0.003) (0.0001)
Y4 −0.0001** −0.000004***
(0.00003) (0.000001)
Population Growth Rate*Y −0.158 −0.008
(0.139) (0.006)
Population Growth Rate*Y2 0.033* 0.001*
(0.017) (0.001)
Population Growth Rate*Y3 −0.002** −0.0001*
(0.001) (0.00003)
Population Growth Rate*Y4 0.00002** 0.000001*
(0.00001) (0.0000004)
Constant −38.453*** −2.096***
(11.326) (0.487)
R-Squared 0.663 0.643
F-Statistics 29.343 26.791
P-value 0.000 0.000
Observations 300 300

* p b 0.10, ** p b 0.05, *** p b 0.01.


Note: Y is real GDP per capita (1000 RMB), with base year of 1997. Y2, Y3 and Y4 are its quadratic, cubic and quartic terms. Waste gas is industrial waste gas emission per
capita measured in 1000 cubic meters, and solid waste is the per capita generation of industrial solid wastes measured in ton.

Acknowledgments

We would like to thank Terry Heaps, Alexander Karaivanov, Kenneth Kasa and the anonymous reviewer for the constructive com-
ments and suggestions, and thank the seminar participants at the 27th Chinese Economics Society Australia (CESA) Annual Confer-
ence, Wollongong, Australia, 2015 for their helpful comments. Part of this work is supported by the Program for Innovation
Research at the Central University of Finance and Economics. Any errors or omission are solely our own.

Fig. 5. The EKCs for industrial waste gas emission and industrial solid wastes. Source: Authors' calculations using the Chinese Statistical Yearbook (2001–2010) and the
Chinese Environmental Statistical Yearbook (2001–2010).
160 S.X. Wang et al. / China Economic Review 36 (2015) 146–165

Fig. 6. The effects of population growth on the EKCs. Source: Authors' calculations using the Chinese Statistical Yearbook (2001–2010) and the Chinese Environmental
Statistical Yearbook (2001–2010).

Appendix A. Proof of Proposition 1

Proof. Totally differentiating Eq. (17) gives

∂F ðH t ; zt ; Et Þ ∂F ðHt ; zt ; Et Þ
dHt þ dzt  2f 0 ðH t  zt ÞdEt ¼ 0 ðA1Þ
∂Ht ∂zt

where

2 0
F ðH t ; zt ; Et Þ ¼ ½ f ðH t  zt Þ g 0 ðzt Þ  ½2Et  f ðH t  zt Þg ðzt Þf ðH t  zt Þ

2
∂F ðH t ; zt ; Et Þ 0
2 2 ½ f ðÞ g 0 ðÞ 0 0
¼ 2f ðÞf ðÞg 0 ðÞ þ ½ f 0 ðÞ g ðÞ þ ½ f ðÞg ðÞ  2Et f 0 0 ðÞ ¼ 2f ðÞf 0 ðÞg 0 ðÞ þ ½ f 0 ðÞ g ðÞ þ f ðÞ
∂Ht f 0 ðÞ

∂F ðH t ; zt ; Et Þ ∂F ðH t ; zt ; Et Þ 2
¼  ½ f ðHt  zt Þ g 0 0 ðzt Þ þ f ðH t  zt Þf 0 ðHt  zt Þg 0 ðzt Þ:
∂zt ∂H t

Each term in ∂F(Ht, zt, Et)/∂Ht is positive and each term in ∂F(Ht, zt, Et)/∂zt is negative, under the assumption that f ''(⋅) b 0 b f ' (⋅) and
g'(⋅) b 0 b g ' '(⋅); therefore, ∂F(Ht, zt, Et)/∂Ht N 0, and ∂F(Ht,zt,Et)/∂zt b 0.
Totally differentiating Eq. (15) gives

g ðzt Þf 0 ðHt  zt ÞdHt  ½g 0 ðzt Þf ðHt  zt Þ  g ðzt Þf 0 ðH t  zt Þdzt ¼ dEt : ðA2Þ

Solving the equation system of (A1) and (A2) by eliminating dEt gives

∂F ðH t ; zt ; Et Þ 2
þ 2½ f 0 ðHt  zt Þ g ðzt Þ
∂zt ∂H t
¼
∂Ht ∂F ðH t ; zt ; Et Þ 2
 þ 2½ f 0 ðHt  zt Þ g ðzt Þ  2f ðH t  zt Þf 0 ðH t  zt Þg 0 ðzt Þ
∂zt

or

∂F ð∙Þ 2
þ 2½ f 0 ð∙Þ gð∙Þ
∂zt ∂H t
¼ :
∂Ht ∂F ð∙Þ 2 2
þ 2½ f 0 ð∙Þ g ð∙Þ þ ½ f ð∙Þ g 0 0 ð∙Þ  3f ð∙Þf 0 ð∙Þg 0 ð∙Þ
∂Ht
S.X. Wang et al. / China Economic Review 36 (2015) 146–165 161

Each term in the numerator and denominator of the equation above is positive, under the assumption that f ''(⋅) b 0 b f ' (⋅) and
g'(⋅) b 0 b g ' '(⋅). Thus, 0 b dzt/dHt b 1. qed.

Appendix B. Proof of Proposition 2

Proof: If the EKC hypothesis holds here, environmental degradation measured by pollution emission per worker and income mea-
sured by output per worker should be first positively related, and then after some turning point, negatively related. Thus, we will ex-
amine the sign of dpt/dyt.

dpt dpt dpt


dpt dHt dHt dHt
¼ ¼ ¼

dyt dyt df ðH t  zt Þ dz
f 0 ðH t  zt Þ 1  t
dHt dH t dHt

Since the denominator is positive from Proposition 1, dpt/dyt and dpt/dHt have the same sign. Therefore, we will examine the sign of
dpt/dHt, instead. In this subsection, as we assume zero population growth, pt = Pt = g(zt)f(Ht - zt). Differentiate Eq. (10) with respect to
Ht, and substitute dzt/dHt from the proof of Proposition 1 in the Appendix A into it, we have

dpt d½g ðzt Þf ðH t  zt Þ dz


¼ ¼ g ðzt Þf 0 ðH t  zt Þ þ ½g 0 ðzt Þf ðHt  zt Þ  g ðzt Þf 0 ðH t  zt Þ t
dHt dHt dHt
( ) g ðzt Þ f ðH  zt Þ
2 0 2 2 2 2 0 2d  0 þ 0 t
½ f ð∙Þ f ð∙Þ½g ð∙Þ gð∙Þg ð∙Þ  ½g ð∙Þ
0 00 0
f ð∙Þf ð∙Þ  ½ f ð∙Þ
00 0
½ f ðHt  zt Þ f ðH t  zt Þ½g ðzt Þ
0
g ðzt Þ f ðH t  zt Þ
¼  0 2 þ  0 2 ¼
D g ð∙Þ f ð∙Þ D dzt

where D ¼ ∂ FðH∂H t ;zt ;Et Þ 2 2


t
þ 2½f 0 ðHt  zt Þ gðzt Þ þ ½f ðHt  zt Þ g 0 0 ðzt Þ  3 f ðHt  zt Þf 0 ðH t  zt Þg 0 ðzt ÞN0, and F(Ht, zt, Et) is defined in the proof
of Proposition 1 in the Appendix A.
Because the first term, [f(Ht - zt)]2f '(Ht - zt)[g'(zt)]2/D, is positive, dpt/dHt has the same sign as the second term, d[-g(zt)/
g ' (zt) + f(Ht - zt)/f ' (Ht - zt)]/dzt. d[f(Ht - zt)/f '(Ht - zt)]/dzt b 0 for any value of zt, while the sign of d[-g(zt)/g ' (zt)]/dzt is uncertain, as
-[g'(zt)]2 b 0 and g(zt)g''(zt) N 0. But, as zt increases, g(zt)g''(zt) decreases; thus, there exists some zt such that d[-g(zt)/g ' (zt) + f(Ht -
zt)/f ' (Ht - zt)]/dzt b 0, and therefore, the EKC has a turning point before which the EKC is increasing and after which it is decreasing.
At the turning point, the level of pollution abatement investment, z, satisfies
 gðz ̅Þ f ðH −z ̅Þ 
d −g0 ðz ̅Þþ t
f 0 ðHt −z ̅Þ
dz
¼0
qed.

Appendix C. Proof of Proposition 3

Proof. Reshape Eq. (24) to get

g 0 ðzt Þ½ f ðHt  zt Þ2 þ g ðzt Þf ðHt  zt Þf 0 ðH t  zt Þ 1 þ η


¼
f 0 ðH t  zt ÞEt ðzt Þ ηt

Differentiating both sides of the equation above with respect to η gives

dLHS dzt t þ ðt  1Þη


¼
dzt dη ηtþ1

or

t þ ðt  1Þη

dzt ηtþ1
¼ :
dη dLHS
dzt

The numerator of the equation above is negative for t ≥1. Its denominator is

dQ dE
E  tQ
dLHS dzt t dzt
¼
dzt E2t
162 S.X. Wang et al. / China Economic Review 36 (2015) 146–165

0 2 0
where Q ¼ g ðzt Þ½ f ðHt zt Þf 0þgðz t Þ f ðH t zt Þ f
ðH t zt Þ
ðH t zt Þ
N0, dQ/dzt b 0, and dEt/dzt N 0, as shown below. Thus, dLHS/dzt b 0, and therefore, dzt/
dη N 0.

dQ f ð∙Þ2 f 0 ð∙Þg 0 0 ð∙Þ þ 2 f ð∙Þ½ f 0 ð∙Þ2 g 0 ð∙Þ  ½ f ð∙Þ2 f 0 0 ð∙Þg 0 ð∙Þ


¼ þ f ð∙Þg 0 ð∙Þ  g ð∙Þf 0 ð∙Þ b 0
dzt ½ f 0 ð∙Þ2

dEt t1
¼ η ½g ðzt Þf 0 ðH t  zt Þ  g 0 ðzt Þf ðHt  zt Þ N 0
dzt

qed.

Appendix D. Proof of Lemma 2

Proof. Totally differentiating Eq. (24) gives

∂GðHt ; zt ; Et Þ ∂GðHt ; zt ; Et Þ 0
dHt þ dzt  ð1 þ ηÞf ðHt  zt ÞdEt ¼ 0 ðD1Þ
∂Ht ∂zt

where
n o
t 2 0
GðHt ; zt ; Et Þ ¼ η ½ f ðH t  zt Þ g ðzt Þ þ f ðHt  zt Þg ðzt Þf ðHt  zt Þ  ð1 þ ηÞEt f 0 ðH t  zt Þ
0

∂GðHt ; zt ; Et Þ n o
t 2 00
¼ η 2 f ð∙Þf 0 ð∙Þg 0 ð∙Þ þ g ð∙Þ½ f 0 ð∙Þ þ g ð∙Þf ð∙Þf ð∙Þ  ð1 þ ηÞf 0 0 ð∙ÞEt
∂Ht ( )
t 0 0 0 2 ½ f ð∙Þ2 g 0 ð∙Þ f 00 ð∙Þ t ∂F ðH t ; zt ; Et Þ
¼ η 2f ð∙Þf ð∙Þg ð∙Þ þ g ð∙Þ½ f ð∙Þ þ ¼η
f0 ∂H t

∂GðHt ; zt ; Et Þ n o
t 2 2 00
¼ η ½ f ð∙Þ g 0 0 ð∙Þ þ 2 f ð∙Þf 0 ð∙Þg 0 ð∙Þ þ f ð∙Þf 0 ð∙Þg 0 ð∙Þ  g ð∙Þ½ f 0 ð∙Þ  g ð∙Þf ð∙Þf ð∙Þ þ ð1 þ ηÞf 0 0 ð∙ÞEt
∂zt ( )
2
t ∂F ðHt ; zt ; Et Þ
¼η   ½ f ð∙Þ g0 0 ð∙Þ þ f ð∙Þ f 0 ð∙Þg 0 ð∙Þ
∂H t

and ∂F(Ht, zt,Et)/∂Ht N 0 is defined in the proof of Proposition 1 in the Appendix A. Thus, ∂G(Ht, zt, Et)/∂Ht N 0, and ∂G(Ht, zt, Et)/∂zt b 0.
Totally differentiating Eq. (22) gives

t1 t1
η g ðzt Þf 0 ðH t  zt ÞdHt  η ½g 0 ðzt Þ f ðH t  zt Þ  g ðzt Þ f 0 ðH t  zt Þdzt ¼ dEt ðD2Þ

Solving the equation system of (D1) and (D2) by eliminating dEt gives

∂GðH t ; zt ; Et Þ t1 2
þ ð1 þ ηÞη ½ f 0 ðH t  zt Þ gðzt Þ
∂zt ∂H t
¼ n o
∂Ht ∂GðHt ; zt ; Et Þ t1 2
 þ ð1 þ ηÞη ½ f 0 ðH t  zt Þ g ðzt Þ  f ðH t  zt Þf 0 ðHt  zt Þg 0 ðzt Þ
∂zt
∂F ð∙Þ 2
η þ ð1 þ ηÞ½ f 0 ð∙Þ g ð∙Þ
∂H t
¼
∂F ð∙Þ 0 2 2 00 0 0
η þ ð1 þ ηÞ½ f ð∙Þ g ð∙Þ þ η½ f ð∙Þ g ð∙Þ  ð1 þ 2ηÞf ð∙Þf ð∙Þg ð∙Þ
∂Ht

or

∂zt 1
¼ :
∂Ht 2
η½ f ð∙Þ g 0 0 ð∙Þ  ð1 þ 2ηÞf ð∙Þf 0 ð∙Þg 0 ð∙Þ

∂F ð∙Þ 2
η þ ð1 þ ηÞ½ f 0 ð∙Þ g ð∙Þ
∂Ht
S.X. Wang et al. / China Economic Review 36 (2015) 146–165 163

Define

η½ f ð∙Þ2 g 0 0 ð∙Þ  ð1 þ 2ηÞf ð∙Þf 0 ð∙Þg 0 ð∙Þ L


X¼ ¼ :
∂F ð∙Þ 0 2 M
η þ ð1 þ ηÞ½ f ð∙Þ g ð∙Þ
∂Ht

Note that X N 0, as L N 0 and M N 0. Thus,

∂zt 1
¼ ∈ð0; ; ; 1Þ:
∂Ht 1 þ X

qed.

Appendix E. Proof of Proposition 4

Proof. The slope of the pollution–income relationship is given by

dpt dpt dpt


dpt dHt dHt dHt
¼ ¼ ¼
:
dyt dyt df ðHt  zt Þ 0 dz
f ðHt  zt Þ 1  t
dHt dHt dHt

Since the denominator is positive from Lemma 2, dpt/dyt and dpt/dHt have the same sign. Therefore, we will examine the sign of
dpt/dHt, instead. Differentiating Eq. (10) with respect to Ht gives

dpt d½g ðzt Þf ðH t  zt Þ dz


¼ g ðzt Þf ðH t  zt Þ þ ½g ðzt Þf ðHt  zt Þ  g ðzt Þf ðH t  zt Þ t :
0 0 0
¼
dHt dHt dHt

Substituting dzt/dHt from the proof of Lemma 2 in the Appendix D into the equation above gives

 
dpt η 2 2 ∂F ðHt ; zt ; Et Þ
¼ ½ f ð∙Þ g ð∙Þf 0 ð∙Þg0 0 ð∙Þ  f ð∙Þg ð∙Þ½ f 0 ð∙Þ g 0 ð∙Þ þ f ð∙Þg 0 ð∙Þ
dHt M þ L ∂H t

where M N 0 and L N 0 are defined in the proof of Lemma 2 in the Appendix D. Substituting ∂ F(Ht, zt, Et)/∂ Ht from the proof of
Proposition 1 in the Appendix A into the equation above gives


( ) g ðzt Þ f ðH  zt Þ
2 2 2 2 2 0 2d  0 þ 0 t
dpt η½ f ð∙Þ f ð∙Þ½g ð∙Þ g ð∙Þg ð∙Þ  ½g ð∙Þ
0 0 00 0
f ð∙Þf ð∙Þ  ½ f ð∙Þ
00 0
η½ f ðH t  zt Þ f ðH t  zt Þ½g ðzt Þ
0
g ðzt Þ f ðHt  zt Þ
¼ þ ¼ :
dHt MþL ½g 0 ð∙Þ2 ½ f 0 ð∙Þ2 MþL dzt

Because the first term, η[f(Ht -zt)]2f '(Ht -zt)[g'(zt)]2/(M+L)N 0, dpt/dHt has the same sign as the second term, d[-g(zt)/g'(zt)+f(Ht -zt)/
f'(Ht -zt)]/dzt. d[f(Ht -zt)/f '(Ht -zt)]/dzt b 0 for any value of zt, while the sign of d[-g(zt)/g ' (zt)]/dzt is uncertain, as -[g'(zt)]2 b 0 and
g(zt)g''(zt) N 0. But, as zt increases, g(zt)g''(zt) decreases; thus, there exists some zt such that d[-g(zt)/g'(zt)+f(Ht -zt)/f '(Ht -zt)]/dzt b 0,
and therefore, the pollution–income relationship has a turning point before which the level of pollution is increasing and after which
it is decreasing. Therefore, the EKC hypothesis still holds with positive population growth. At the turning point, the level of pollution
abatement, z, satisfies


gðzÞ f ðH  zÞ
d  0 þ 0 t
g ðzÞ f ðHt  zÞ
¼ 0:
d
z
164 S.X. Wang et al. / China Economic Review 36 (2015) 146–165

Next, we will exam the effect of population growth on the slope of the EKC. Substituting dpt/dHt and dzt/dHt into the slope of the
EKC gives

g ðz Þ f ðH  zt Þ
2 2 d  0 t þ 0 t
dpt η½ f ðH t  zt Þ f ðHt  zt Þ½g ðzt Þ
0 0
g ðzt Þ f ðHt  zt Þ
dpt dHt MþL dzt
¼
¼
dyt dzt L
f ðHt  zt Þ 1 
0
f ðH t  zt Þ
0

dHt MþL
g ðzt Þ f ðH t  zt Þ g ð∙Þ f ð∙Þ
d  0 þ 0 2 d  0 þ 0
η 2 2 g ðzt Þ f ðHt  zt Þ f ð∙Þ½g ð∙Þ
0
g ð∙Þ f ð∙Þ
¼ ½ f ðH t  zt Þ ½g 0 ðzt Þ ¼
:
L dzt 00 1 dzt
f ð∙Þg ð∙Þ  þ 2 f ð∙Þg ð∙Þ
0 0

Note that the first term of the equation above is positive and increases in η, and therefore, the absolute value of the slope of the EKC
increases in η. Thus, higher rate of population growth makes the EKC steeper. qed.

References

Arrow, K., Bolin, B., Costanza, R., Dasgupta, P., Folke, C., Holling, C.S., ... Pimentel, D. (1995). Economic growth, carrying capacity, and the environment. Science, 268,
520–521.
Baek, J., & Kim, H.S. (2013). Is economic growth good or bad for the environment? Empirical evidence from Korea. Energy Economics, 36, 744–749.
Barbier, E.B., & Burgess, J.C. (2001). The economics of tropical deforestation. Journal of Economic Surveys, 15(3), 413–433.
Bhattarai, M., & Hammig, M. (2001). Institutions and the environmental Kuznets curve for deforestation: a cross country analysis for Latin America, Africa and Asia.
World Development, 29(6), 995–1010.
Boserup, E. (1965). The conditions of agricultural growth: the economics of agrarian change under population pressure. Chicago: Aldine.
Boserup, E. (1976). Environment, population and technology in primitive societies. Population and Development Review, 2(1), 21–36.
Boserup, E. (1981). Population and technological change. Chicago: University of Chicago Press.
Brajer, V., Mead, R.W., & Xiao, F. (2011). Searching for an environmental Kuznets curve in China's air pollution. China Economic Review, 22, 383–397.
Cao, D., Wang, L., & Wang, Y. (2011). Endogenous fluctuations induced by nonlinear pollution accumulation in an OLG economy and the bifurcation control. Economic
Modelling, 28, 2528–2531.
Carson, R., Jeon, Y., & McCubbin, D.R. (1997). The relationship between air pollution emissions and income: US data. Environment and Development Economics, 2,
433–450.
Caviglia-Harris, J., Chambers, D., & Kahn, J. (2009). Taking the "U" out of Kuznets: A comprehensive analysis of the EKC and environmental degradation. Ecological
Economics, 68, 1145–1155.
Chen, Z., Wang, J., Ma, G., & Zhang, Y. (2013). China tackles the health effects of air pollution. The Lancet, 382(9909), 1959–1960.
Chimeli, A.B., & Braden, J.B. (2005). Total factor productivity and the environmental Kuznets curve. Journal of Environmental Economics and Management, 49(2),
366–380.
Cole, M.A., Rayner, A.J., & Bates, J.M. (1997). The environmental Kuznets curve: An empirical analysis. Environment and Development Economics, 2(4), 401–416.
Cropper, M.L., & Griffiths, C. (1994). The interaction of population growth and environmental quality. American Economic Review, 84(2), 250–254.
Dinda, S. (2004). Environmental Kuznets curve hypothesis: A survey. Ecological Economics, 49, 431–455.
Dinda, S. (2005). A theoretical basis for the Environmental Kuznets Curve. Ecological Economics, 53, 403–413.
Grossman, G.M., & Krueger, A.B. (1991). Environmental impacts of a North American Free Trade Agreement. NBER, working paper 3914. Cambridge: MA.
Grossman, G.M., & Krueger, A.B. (1995). Economic growth and the environment. Quarterly Journal of Economics, 110(2), 353–377.
Hausman, J.A. (1978). Specification tests in econometrics. Econometrica, 46, 1251–1271.
Hettige, H., Lucas, R.E.B., & Wheeler, D. (1992). The toxic intensity of industrial production: Global patterns, trends, and trade policy. American Economic Review, 82(2),
478–481.
Hilton, F.G.H., & Levinson, A. (1998). Factoring the environmental Kuznets curve: Evidence from automotive lead emissions. Journal of Environmental Economics and
Management, 35, 126–141.
Holtz-Eakin, D., & Selden, T.M. (1995). Stoking the fries? CO2 emissions and economic growth. Journal of Public Economics, 57, 85–101.
John, A., & Pecchenino, R. (1994). An overlapping generations model of growth and the development. The Economic Journal, 104, 1393–1410.
Kaufmann, R.K., Davidsdottir, B., Pauly, P., & Garnham, D. (1998). The determinants of atmospheric SO2 concentrations: Reconsidering the environmental Kuznets
curve. Ecological Economics, 25(2), 209–220.
Koop, C., & Tole, L. (1999). Is there an environmental Kuznets curve for deforestation? Journal of Development Economics, 58, 231–244.
Koren'kov, V.N. (1991). Biological purification of sewage from chemical plants. Russian Chemical Reviews, 60, 273.
López, R. (1994). The environment as a factor of production: the effects of economic growth and trade liberalization. Journal of Environmental Economics and
Management, 27(2), 163–184.
Malthus, T. (1798). An essay on the principle of population, as it affects the future improvement of society. New York: Modern Library and Random House.
Martinez-Zarzoso, I., & Maruotti, A. (2011). The impact of urbanization on CO2 emissions: Evidence from developing countries. Ecological Economics, 70(7), 1344–1353.
Munasinghe, M. (1999). Is environmental degradation an inevitable consequence of economic growth: Tunneling through the environmental Kuznets curve. Ecological
Economics, 29(1), 89–109.
Onafowora, O.A., & Owoye, O. (2014). Bounds testing approach to analysis of the environment Kuznets curve hypothesis. Energy Economics, 44, 47–62.
Panayotou, T. (1993). Empirical tests and policy analysis of environmental degradation at different stages of economic development. International labor organization
working paper 238. Geneva: Switzerland.
Panayotou, T. (1997). Demystifying the environmental Kuznets curve: Turing a black box into a policy tool. Environment and Development Economics, 2, 465–484.
Roy, N., & van Kooten, G.C. (2004). Another look at the income elasticity of non-point source pollutants: A semi parametric approach. Economic Letters, 85(1), 17–22.
Scrieciu, S.S. (2007). Can economic causes of tropical deforestation be identified at a global level? Ecological Economics, 62(3–4), 603–612.
Selden, T.M., & Song, D. (1994). Environmental air quality and development: Is there a Kuznets curve for air pollution emissions. Journal of Environmental Economics
and Management, 27, 147–162.
Selden, T.M., & Song, D. (1995). Neoclassical growth, the J curve for abatement, and the inverted U curve for pollution. Journal of Environmental Economics and
Management, 29, 162–168.
Shafik, N. (1994). Economic development and environmental quality: An econometric analysis. Oxford Economic Papers, 46, 757–777.
Shafik, N., & Bandyopadhyay, S. (1992). Economic growth and environmental quality: Time series and cross section evidence. Working paper. Washington, DC.: World
Bank.
Shen, J. (2006). A simultaneous estimation of environmental Kuznets curve: Evidence from China. China Economic Review, 17, 383–394.
S.X. Wang et al. / China Economic Review 36 (2015) 146–165 165

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.
Song, T., Zheng, T., & Tong, L. (2008). An empirical test of the environmental Kuznets curve in China: A panel cointegration approach. China Economic Review, 19,
381–392.
Stokey, N. (1998). Are there limits to growth? International Economic Review, 39(1), 1–31.
Vincent, J. (1997). Testing for environmental Kuznets curve within a developing country. Environment and Development Economics, 2(4), 417–431.
Wagner, M. (2008). The carbon Kuznets curve: A cloudy picture emitted by bad econometrics? Resource and Energy Economics, 30, 388–408.
Wang, C., Fang, Y., Wu, X., Hu, D., Zuo, Z., Li, C., ... Xiao, G. (2006). Research on the ability of twenty-one ornamental plants to absorb and purify environmental pol-
lutants. Journal of Shandong Forestry Science and Technology, 6, 11–13.
Yang, H., He, J., & Chen, S. (2015). The fragility of the environmental Kuznets curve: Revisiting the hypothesis with Chinese data via an “extreme bound analysis”.
Ecological Economics, 109, 41–58.

You might also like