Journal of Cleaner Production 214 (2019) 848e859

Contents lists available at ScienceDirect

Journal of Cleaner Production

journal homepage: www.elsevier.com/locate/jclepro

Relationship between forest resources and economic growth:

Empirical evidence from China
Yu Hao a, b, c, d, e, *, 1, Yalin Xu b, Jingjing Zhang a, b, 1, Xinlei Hu b, Junbing Huang f,
Chun-Ping Chang g, h, 1, Yaqi Guo b
a
Center for Energy and Environmental Policy Research, Beijing Institute of Technology, Beijing, 100081, China
b
School of Management and Economics, Beijing Institute of Technology, Beijing, 100081, China
c
Sustainable Development Research Institute for Economy and Society of Beijing, Beijing, 100081, China
d
Collaborative Innovation Center of Electric Vehicles in Beijing, Beijing, 100081, China
e
Beijing Key Lab of Energy Economics and Environmental Management, Beijing, 100081, China
f
School of Economics, Southwestern University of Finance and Economics, Chengdu 611130, China
g
Department of Marketing Management, Shih Chien University at Kaohsiung, Kaohsiung, 845, Taiwan
h
School of Economics and Finance, Xi'an Jiaotong University, Xi'an, 710049, Shaanxi, China

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

Article history: Natural resources, especially forests, are critical to China's sustainable development. In recent years,
Received 8 May 2018 effective utilization of forest resources has become a significant concern in China. In this study, on the
Received in revised form basis of the environmental Kuznets curve (EKC) hypothesis, the relationship between forest resources
30 November 2018
and economic growth is examined using panel data from 30 provinces in China for the period 2002
Accepted 30 December 2018
Available online 4 January 2019
e2015 and the generalized method of moments (GMM) approach to control for potential endogeneity.
The empirical results indicate that with continuous economic growth, the timber output and area of
afforestation would at first increase and then decrease after reaching the corresponding turning points.
Keywords:
Environmental Kuznets curve (EKC)
Given the considerable regional gap in economic and social development, three regions of China were
Forest resources investigated to perform a more in-depth analysis. Evidence observed for the country and the three re-
Economic growth gions suggests an EKC for forest resources. These findings indicate positive effects of China's pursuit of a
China more balanced growth path where forest resources would be less consumed and more actively protected.
© 2019 Elsevier Ltd. All rights reserved.

1. Introduction global context, economic growth was accompanied by a substantial

decrease in forest resource in the long run, although there was
During the past few decades, China has consistently improved signs that deforestation stabilized more recently. Naidoo (2004)
its economic growth and standard of living. These improvements and Zambrano-Monserrate et al. (2018) found similar results.
have been termed as the Chinese economic miracle (Cai and Population growth has led to an increasing demand for food,
Treisman, 2006). The consumption of natural resources, especially fuel, fiber, and other natural resources. Forest resources have
forest resources, has played a main role in this miracle. According to decreased rapidly due to the expansion of cultivated areas and
the literature, the changes in the forestry sector and economy are excessive commercial logging for fuel and infrastructure materials,
closely linked. For example, Stern et al. (1996) found that, in the thereby accelerating the rate of forest extinction. Despite the
increasing awareness of how forests protect the environment and
public health, the resource remains scarce. From 1990 to 2015, the
* Corresponding author. School of Management and Economics, Beijing Institute global forested area decreased by 19.35 million hectares. Serious
of Technology, Beijing, 100081, China.
consequences from reduced forest resources have began to be
E-mail addresses: [email protected] (Y. Hao), [email protected] (Y. Xu),
[email protected] (J. Zhang), [email protected] (X. Hu), [email protected].
observed, such as global warming, soil degradation, and biodiver-
cn (J. Huang), [email protected] (C.-P. Chang), [email protected] sity losses (Damette and Delacote, 2012).
(Y. Guo). Forest resources are a critical natural resource that provide a
1
Present address: School of Management and Economics, Beijing Institute of wide range of essential services for humans and are irreplaceable as
Technology, 5 Zhongguancun South Street, Haidian District, Beijing 100081, China
a component of social services. Furthermore, the ecological value of
https://ideas.repec.org/d/cebitcn.html, http://www.ceep.net.cn/english/index.htm.

https://doi.org/10.1016/j.jclepro.2018.12.314
0959-6526/© 2019 Elsevier Ltd. All rights reserved.
 Y. Hao et al. / Journal of Cleaner Production 214 (2019) 848e859 849

forests is vital. For example, forests can mitigate the effects of global a variable in this study would be problematic (Wang et al., 2007).
warming by reducing greenhouse gases in the atmosphere to Thus, timber output (Turner et al., 2006) and afforestation area are
improve the earth's climate and environment. Forests also maintain the dependent variables to reflect forest resources and forest
the soil, water, and biodiversity as well as improve the regional management performance.
ecological environment (Führer, 2000). Thus, policymakers must In summary, the contribution of this study is threefold. First, in
delineate the appropriate use of forest resources by determining addition to conventional static regression with fixed effects (FEs),
the nexus between the economy and forests. the GMM method is employed to control for potential endogeneity
Previous studies have investigated whether the environmental and allow dynamic effects. Second, the influences of several in-
Kuznets curve (EKC) hypothesis exists for forest resources (Foster dicators for economic and social development, including industrial
and Rosenzweig, 2003). The EKC hypothesizes an inverted U-sha- structure, urbanization rate, and trade openness, on forest re-
ped nexus between environmental indicators, including natural sources are investigated. Third, as stressed by Wang and Liu (2003),
resources and the economy level (Grossman and Krueger, 1991; China's forest resources present regional characteristics. Hence,
Caviglia-Harris et al., 2009). More specifically, the EKC framework national and regional panel data (eastern, central, and western
implies that in the early stages of economic growth, the con- regions) are used to explore whether the nexus of forest and eco-
sumption of forest resources may at first increase and then nomic growth varied at different geographical locations and levels
decrease with economic growth. of economic development. We observed that choosing appropriate
According to the literature that has used diverse indicators and estimation methods and specifications improves the reliability of
methods, distinct conclusions have been observed for forest re- this study. The conclusions of this study provide valuable and
sources and economic growth: the results vary significantly reasonable policy implications for environmental protection in
because of the variables selected. Notably, these studies have China. We found an EKC relationship between forest resources and
mainly been conducted in foreign countries, and investigations on economic development at China's regional level.
the relationship between forest resources and economic develop- To demonstrate the spatial correlation between forest condi-
ment in China remain scarce. Given the importance of forest re- tions and economic growth in an intuitive manner, the timber
sources in China's sustainable development and that output, afforestation area, and per capital income of all 30
environmental protection is higher priority than before, a more in-
depth study on the nexus of forest resources and economic growth
in China is necessary.
Since the 1990s, China's substantial changes in forest resources
have been accompanied by its rapid economic growth. According to
the data of Global Forest Resources Assessment, China's forest area
increased from 157.1 million hectares in 1990 to 208.3 million
hectares in 20152; during the same period, the percentage of forest
cover increased from 16.3% in 1990 to 21.6% in 2015. Despite the
increase, China's forest cover remains considerably lower than the
global average level (30.8% in 2015).
For a developing country like China that is pursuing sustainable
development that balances economic growth and environmental
protection, the exploration of how forest resources affect economic
growth is of great importance. The EKC hypothesis is a useful
framework to test this relationship. Following the EKC hypothesis,
an assumption is made that in the early stages of economic growth,
the government prioritizes the economy, increasing the demand for
forest products such as timber (Bhattarai and Hammig, 2001). Next,
the importance of environmental protection along with economic
growth is realized to a certain level, thus contributing to an increase
in forested lands (Van and Azomahou, 2007). During this process,
environmental policy, industrial upgrading, and other socioeco-
nomic factors may be an impetus for forest resource protection.
Because of potential bilateral causality between forest resources
and economic development, and the possible omitted variables
that affect forest resources, endogeneity problem may be observed
in the estimation; this problem has generally not been considered
in the literature and would have thus resulted in biased and
incorrect estimation results. In this regard, the present study em-
ploys the generalized method of moments (GMM) method to
address the possible endogenous problem.
The main task of this paper is to empirically explore the rela-
tionship between forest resources and economic growth in China
over the period 2002e2015 at the national and regional levels.
Forest cover is a relatively common dependent variable, but in
China, forest cover is a policy indicator, and the use of this figure as

Fig. 1. Distribution of timber output, afforestation area, and GDP per capita across
Chinese provinces in 2002 and 2015.
2
FRA country report for China (2015) is available for download: http://www.fao. Data source: China Statistical Yearbook 2003 and 2016 and China Forestry Statistics
org/3/a-az186e.pdf. Yearbook 2003 and 2016.
850 Y. Hao et al. / Journal of Cleaner Production 214 (2019) 848e859

provinces under investigation in 2002 and 2015 are presented in proposed by Grossman and Krueger (1991) first used the empirical
Fig. 1.3 Two main features of the data can be observed in Fig. 1. The evidence of atmospheric quality data to demonstrate an inverted U-
first relates to the production of timber output over time. The shaped relationship between environmental quality and per capita
provinces with the highest timber output have shifted from income, which was named the EKC because the shape of the rela-
northern to southern China. For instance, during the sample period, tionship is similar to the original Kuznets curve found by Kuznets
the timber output in the northern provinces such as Inner Mongolia (1955) to describe the inverted U-shaped relationship between
and Heilongjiang plummeted, whereas the production of timber in economic growth and income inequality. Many scholars have
the southern province of Guangdong increased remarkably. investigated the relationship between the change in forests and
The second feature of the data in Fig. 1 relates to the changes in economic growth and their inflection points (Cropper and Griffiths,
afforestation area. Although the afforestation intensity increased in 1994; Wang et al., 2007). Notably, because the indicator of defor-
the southeast coastal provinces, no significant change was observed estation is usually used in these studies on EKC, the inverted U-
in the remaining provinces. Notably, some southern provinces with shaped relationship means the relationship between forest re-
higher wood production, such as Guangdong and Guangxi, have sources (contrary to degradation measures) is U shaped. That is, as
also correspondingly increased the areas of afforestation. These two per capita income increases, forest resources (like forest cover)
observations to some extent reflect the basic characteristics of the would initially decrease along with economic growth and then
change in forest resource in China shown in the seven major forest subsequently reverses after a certain level of income threshold is
shelter projects d the wood production area has gradually shifted reached. However, although a growing body of literature has veri-
from the north to the south of China. In the meanwhile, several fied the existence of EKC in China (e.g., Kang et al., 2016; Li et al.,
provinces in the regions in the northern, western, and central 2016; Wu et al., 2018), the research on the EKC for China's forest
plains strengthened in afforestation intensity. resources is still scarce.
The remainder of this paper is organized as follows. Section 2 Forest resources have been measured by various methods, and
reviews the literature on the EKC hypothesis and related forest these studies can generally be divided into two groups based on
resources. Section 3 presents the data sources and specifies the their dependent variables. One group has investigated the influence
empirical models. Section 4 provides the estimated results and of economic growth on forest resources by adopting stock variables,
corresponding analysis. Section 5 concludes and provides relevant including the deforestation rate, afforestation, and timber output.5
policy suggestions. Allen and Barnes (1985) first proved the existence of an EKC be-
tween the reduction in forest resources and per capita gross do-
mestic product (GDP) by using cross-sectional data in developed
2. Literature review counties from 1968 to 1978. Other scholars have preferred using
flow variables, such as forest cover or forest surface, to measure the
Due to the pressing environmental problems in China, the change in forests. Zhang et al. (2006) concluded that China is in the
highly efficient utilization of natural resources, especially forests, latter stages of a U-shaped forested curve based on its national,
have attracted great concern of the public and the academia. It is regional, and provincial models of forested areas in mainland China
generally acknowledged that forest resources provide many pro- from 1999 to 2010.
duction materials which are critical for economic growth (e.g., Notably, the conclusions from the literature on forest EKC are
Narita et al., 2018), and that forest is vital to the health of local usually controversial. For example, Cropper and Griffiths (1994)
ecosystem (e.g., Montagne -Huck and Brunette, 2018; Nguyen et al., empirically verified an EKC relationship between the decrease in
2018). However, deforestation has long been a global problem. forest resources and economic growth in Africa and Latin America.
According to the estimates of the Food and Agriculture Organiza- Bhattarai and Hammig (2001) studied a sample of 66 countries
tion of the United Nations (FAO), after the industrial revolution, the from 1972 to 1991 and observed a forest EKC in Latin America,
earth's forest resources decreased by approximately one-third.4 A Africa, and Asia. By contrast, Koop and Tole (1999) investigated 66
series of literature has stressed the implications of deforestation on countries' panel data from 1962 to 1986 and found no evidence of
economic development. For instance, using a dataset of 70 coun- this nexus between the decline in forest resources and economic
tries, Naidoo (2004) found that the countries with rapidly growth in Asia. Van and Azomahou (2007) focused on parametric
decreasing forest resources had faster economic growth than those and semiparametric models and used 59 developing countries’
with higher forest conservation rates. Thus, an assumption could be panel data over the period from 1972 to 1994 and also found no
made that the consumption of forest resources is usually accom- evidence of a nexus between these variables. In a recent study,
panied with fast socioeconomic development. In a more recent Waluyo and Terawaki (2016) found empirical evidence for a long-
study, Zambrano-Monserrate et al. (2018) found evidence for a run inverted-U relationship for deforestation in Indonesia.
long-run relationship between agricultural exports, GDP per capita Zambrano-Monserrate et al. (2018) found an inverted U-shaped
and deforestation in five selected European countries. relationship between the reduction of forest resources and GDP per
Given the importance of forest resources in the economic capita in France, Germany, Portugal, and Turkey during the period
development, there has been a large body of literature that in- 1974e2013. Additionally, Joshi and Beck (2016) used GMM to
vestigates the relationship between forest resources and economic explore the relationship for OECD and non-OECD regions and found
growth (e.g., Stern et al., 1996; Angelsen and Kaimowitz, 1999; that OECD countries had an N-shaped curve, whereas an income-
Naidoo, 2004; Choumert et al., 2013; Zambrano-Monserrate et al., based EKC pattern was observed in only the African region.
2018). A powerful empirical framework frequently used in previ- In addition to forest-related indicators, when using the EKC
ous literature on the nexus of forest resources and economic
growth is environmental Kuznets curve (EKC). EKC was at first
5
Deforestation is the removal of a forest where the land is converted to a non-
forest use. By contrast, afforestation means the establishment of a forest or stand
3
Currently, there are 22 provinces, four Centrally Administered Municipalities, of trees (in the form of forestation) in an area with no tree cover previously.
and five autonomous regions in the mainland of China. Because these entities are Therefore, afforestation is just the antonym of deforestation. Moreover, timber
administratively equal, the term “province” is utilized throughout the paper. output is the product of timber during a given time period. In general, timber
4
For more information, please refer to http://www.fao.org/publications/sofa/ output comes from deforestation, while afforestation provides potential resources
2012/en/. of timber production in the future.
 Y. Hao et al. / Journal of Cleaner Production 214 (2019) 848e859 851

framework to study the relationship between forest resources and population, per capita GDP, secondary industry output value, urban
economic growth, many scholars have also measured several so- population, and import and export trade volume, are from the
cioeconomic factors that may affect forests, such as technology China Statistical Yearbook. Notably, all the GDP values, used or
development (Bhattarai and Hammig, 2001; Culas, 2007), popula- calculated, are real GDP per capita at constant 2000 prices. As a
tion (Panayotou, 1993; Koop and Tole, 1999; Barbier and Burgess, summary, the descriptive statistics of all the variables used in this
2001), geophysical location (Liu et al., 2012), and political in- study are presented in Table 1.
stitutions (Zhang et al., 2006; Culas, 2007). These studies imply that To show the relationship between forest resources and eco-
factors affecting changes in forest resources are highly heteroge- nomic development intuitively and ensure the transparency of the
neous. Indarto and Mutaqin (2016) reviewed the literature on the data used, the scatter plots of two indicators for forest resources
relationship between deforestation (or the use of forest resources) against GDP per capita for three geographical regions are depicted
and economic development from theoretical and empirical re- in Fig. 2.7 Panel A of Fig. 2 shows the relationship between per
spects. As mentioned by Damette and Delacote (2012), the tradi- capita timber output and GDP per capita in the eastern, central and
tional ordinary least squares method would lead to a biased western regions of China; while panel B represents the nexus of per
outcome. capita afforestation and GDP per capita in three regions. The scat-
With regard to the methodology in the forest resources litera- ters in both graphs do not exhibit clear patterns in all three regions
ture, in the 1990s, most empirical studies focused on the cross- as well as for the whole country. Therefore, at the first glance, it is
sectional data to examine the relationship between forest re- difficult to determine the specific relationship between forest re-
sources and economic growth (Allen and Barnes, 1985; Tole, 1998); sources and economic growth from the scatterplots. In this regard,
however, the approach neglected the time dimension and caused precise quantitative estimation is required to investigate this
problems. Since the 2000s, panel data analysis has become relationship.
increasingly popular because it overcomes the limitation of the On the basis of the theoretical analysis, socioeconomic factors
reduced-form model. In addition, on the basis of panel data, more may also correlate with forests. Therefore, consistent with the
suitable methods have been used, such as the effects model and FE literature, this study included the following control variables:
model (Koop and Tole, 1999; Bhattarai and Hammig, 2001), autor-
egressive distributed lag model (Ahmed et al., 2015), and panel
3.1.1. Industrial structure
smoothing conversion model (Chiu, 2012).
Industrial structure is the composition and proportion of various
industries, and the economic development of most developing
3. Data and Methods countries mainly rely on primary and secondary industries. High
resource consumption and an extensive economic growth model
3.1. Data and variables stresses forest resources; thus, the demand for wood and timber
output increases. In addition, more trees must be planted to supply
In this paper, we estimate the relationship between forest re- timber output.
sources and economic growth by using the panel data of 30 prov- The proportion of tertiary industries in the national economy in
inces in China from 2002 to 2015. Specifically, timber output China has steadily increased since mid-1980s, and these industries
(Turner et al., 2006) and afforestation area (Wang and Xu, 2015) have gradually become an important pillar of national economic
were chosen as dependent variables to measure forest resources.6 development. Notably, the structure of the tertiary industry is un-
In addition, per capita timber output and per capita afforestation balanced. Compared with the number of resource shortage in-
area, instead of total timer output and total afforestation area, were dustries such as transportation, the number of industries that are
used to mitigate the scale effect. The data on timber output were high-value added and less resource consuming remains small
collected from the China Forestry Statistics Yearbook and China (Yang, 2016). Meanwhile, the proportion of the secondary industry
Statistical Yearbook. Urbanization rate, trade openness, and second remains large, and the process of industrialization and urbanization
industry proportion were selected as control variables. Following in China has imposed substantial pressure on forest resources.
the literature, including Du et al. (2012) and Hao and Wei (2015), Hence, the industrial structure, calculated by the proportion of
the urbanization rate in this study is calculated as the ratio of urban value added by secondary industries to the GDP, is incorporated
population to total population; trade openness is the ratio of im- into this paper.
ports and exports to GDP; and second industry proportion is the
ratio of secondary industry value added to GDP. 3.1.2. Urbanization rate
In addition, the data on total timber output afforestation area Urbanization rate is a typical demographic dynamics indicator
are obtained from the China Statistical Yearbook (2003e2015) and that generally requires a large area and leads to extensive damage
China Forestry Statistics Yearbook (2002). Other data, such as to various plants, forests, and grasslands (Cropper and Griffiths,
1994). Many empirical studies have observed that urbanization
rates are highly correlated with the consumption of forest re-
6
As pointed out by an anonymous reviewer, forest cover is also an important
sources. Ehrhardt-Martinez (1998) found that urbanization and the
indicator for forest resources. However, the available data for provincial forest cover
from the National Bureau of Statistics (NBS) of China are unavailable for many reduction of forest resources form an inverted U-shaped relation-
provinces in various years. The data from FAO are only for the national level, while ship. The early urbanization process mainly relies on resource
in this study, the provincial panel data are used. Therefore, mainly because of data extraction (e.g., timber for construction. Nevertheless, countries
availability, this study chooses the timber output and afforestation area as the main with higher level urbanization rates also use energy, increasing the
indicators for forest resources. The provincial data for these two indicators are
inaccessible before 2002; hence, the sample period begins in 2002. With regard to
the sample size, because the provincial panel data is used, the number of total
7
observations is 14*31 ¼ 434. Therefore, the sample size of the panel data is suffi- Following the classification of the National Bureau of Statistics (NBS) of China,
ciently large to ensure the estimations are reasonable and meaningful (Wooldridge, the eastern region includes Beijing, Tianjin, Hebei, Liaoning, Shanghai, Jiangsu,
2016). Additionally, in the literature on environmental issues in China, provincial Zhejiang, Fujian, Shandong, Guangdong, Guangxi, and Hainan; the central region
data with a similar or even shorter sample period has been commonly used (e.g., includes Shanxi, Inner Mongolia, Jilin, Heilongjiang, Anhui, Jiangxi, Henan, Hubei,
Shen, 2006; Li et al., 2016). In this regard, we posit that the data in this study has an and Hunan; and the western region consists of Chongqing, Sichuan, Guizhou,
appropriate sample size. Yunnan, Tibet, Shanxi, Gansu, Qinghai, Ningxia, and Xinjiang.
852 Y. Hao et al. / Journal of Cleaner Production 214 (2019) 848e859

Table 1
Descriptive statistics of the variables used in this study.

Variables Explanation (unit) N Mean Std. dev Minimum Maximum

Timber Timber output per capita (cubic metres/ person) 427 0.05 0.07 0.00 0.48
Afforestation Afforestation area per capita (hectare/person) 434 0.006 0.007 0.0003 0.06
Y Per capita GDP (yuan, at constant 2000 prices) 434 21849.41 15223.35 3185.91 85036.41
P Urbanization rate (%) 434 48.3 15.0 14.6 89.6
E Trade openness (%) 434 30.7 39.2 0.1 182.8
Z Second industry proportion (%) 434 46.8 8.2 19.7 61.5

Source:China Forestry Statistics Yearbook and China Statistical Yearbook.

a) Per capita timber output vs. GDP per capita

b) Per capita afforestation area vs. GDP per capita

Fig. 2. Scatter plots of timber output per capita (upper panel a) and afforestation area per capita (lower panel b) against GDP per capita.
Notes: The three regions are distinguished by the colored and shaped points: red squares represent eastern region, black triangles denote central region, and blue dots stand for the
western region.

use of oil, coal, and alternative energy rather than forests. As a 1978 to 2012 in China (National Bureau of Statistics of China, 2013).
result, such a shift contributes to less pressure on forests. Therefore, this study includes the urbanization rate as an influential
Furthermore, the urbanization rate increased from 18% to 53% from factor in the modeling.
 Y. Hao et al. / Journal of Cleaner Production 214 (2019) 848e859 853

3.1.3. Trade openness the estimations is the EKC hypothesis, which states a nonlinear
Trade openness has often been used as an influential factor in relationship between forest resources and economic development;
forest EKC studies. The trade theory implies that developed coun- next, the U-shaped and N-shaped curves are estimated. Second, the
tries normally establish capital intensive production, whereas linear specification is essentially a reduced form of the regression
production in developing countries is labor and resource intensive equation (1) under certain conditions. For instance, in Eq. (1), if the
(Van and Azomahou, 2007). More specifically, developed countries coefficient of GDP per capita in logarithmic form (b1) is statistically
tend to transfer resource-intensive industries to developing coun- significant and the coefficient of its squared term (b2) is statistically
tries either physically or through commercial investment. More- insignificant, the relationship between forest resources and eco-
over, developing countries normally export low technology nomic development is linear. Similar models have been frequently
products at a lower price to take advantage of the international used in the empirical studies on EKC (e.g., Song et al., 2008; Hao and
market. How international trade affects forest resources is contro- Wei, 2015; Kang et al., 2016). Song et al. (2008) also discussed
versial. Kallbekken (2000) found the final effect of international various possible shapes of the EKC relationship, (e.g., linear, U,
trade was negative, that is, the increase in trade exacerbated the inverted-U, N, inverted-N, level) that could be displayed by the
diminution of forest resources, whereas Ehrhardt-Martinez et al. estimation results of Eqs. (1) and (2).
(2002) found that trade may to some extent contribute to the Notably, a bilateral causal relationship probably exists between
reduction in forest resources, although the estimation results were forest resources and economic development. On the one hand,
not statistically significant. Hence, introducing trade openness to previous studies revealed that forest resources play an important
the estimated relationship is reasonable. role in economic growth (e.g., Chen et al., 2015; Iacob, 2015; Dobra
In the present study, how economic growth affects forest timber et al., 2018); on the other hand, a large body of literature has
production and afforestation area is expected to be nonlinear. In the explored and verified that economic development affects forest
early stages of social development, economic and industrial resources (e.g., Bhattarai and Hammig, 2001; Singh et al., 2017;
development are high priorities that increase the demand for wood Zambrano-Monserrate et al., 2018). In addition to the aforemen-
and timber production. Additionally, more trees must be planted to tioned variables, other factors could affect forest growth but cannot
ensure a sufficient supply of wood. Subsequent to these early states, be observed directly or have no statistics. The possible bilateral
timber output gradually decreases when the level of economic causality between forest resources and economic growth and the
development is sufficiently high. More specifically, when the pro- omitted variables may lead to endogeneity (Wooldridge, 2016).
portion of high-tech and service industries increase, citizens’ Therefore, consideration of the endogeneity problems is necessary.
environmental awareness gradually increases, and this phenome- This study uses the first-difference GMM model to address the
non lead to a decreased demand for wood. endogeneity problem.
Afforestation may also decrease because of the limited land area The GMM estimation was originally proposed by Arellano and
and lower timber consumption, and urbanization may be positively Bover (1995) and is a special form of the instrumental variable
related with timber consumption as the demand for wood would (IV) method that uses lagged terms of predetermined variables or
probably increase during urbanization. Notably, how secondary exogenous variables as instruments. In this regard, compared with
industry proportion and trade openness affect timber output and conventional IV method, GMM has the advantage of obtaining in-
afforestation rate are ambiguous because no theory has argued the struments more easily. There are two types of commonly used
directions of the influence of these factors. GMM approaches: system GMM developed by Arellano and Bover
(1995) and Blundell and Bond (1998) and first-difference GMM
posited by Arellano and Bond (1991). The main difference between
3.2. Econometric method these two GMM approaches is how the instruments are chosen.
Moreover, similar to the conventional FEs’ FE estimator, the
This study uses the FE method to determine preliminary results. first-difference GMM approach could remove the unobserved in-
The specifications of the regression model are set as follows. dividual by taking the first-order differences. Notably, no consensus
has been reached regarding which approach is superior in practice
lnFit ¼ b1 lnYit þ b2 ðlnYit Þ2 þ b4 Zit þ b5 Pit þ b6 Eit þ εit (1) (Hayakawa, 2009), and the choice should be made based on the
research topic and features of the data. Although the system GMM
is more efficient than the first-difference GMM, the asymptotic
lnFit ¼ b1 lnYit þ b2 ðlnYit Þ2 þ b3 ðlnYit Þ3 þ b4 Zit þ b5 Pit þ b6 Eit
properties of the system GMM are guaranteed only when the N (i.e.,
þ εit the number of cross-section units) is sufficiently large; otherwise,
(2) the weak finite-sample bias may be considerably large (Huang,
2010). Because T (i.e., the number of time periods) of the sample
where subscripts i and t denote the province index and time index, is comparable with N (i.e., the number of provinces) in this study,
respectively. Fit is timber output or afforestation area. Yit is real per the first-difference GMM is employed as the benchmark estimation
capita GDP at the 2000 constant price. The variables Zit , Pit , and Eit technique.8 This approach has been widely used to assess the ex-
represent the second industry proportion, urbanization rate, and istence of EKC (e.g., Hao et al., 2016b; Li et al., 2016). Following the
trade openness, respectively. εit is a random error representing all literature on EKC, such as Halkos and Paizanos (2013), Hao et al.
other factors that may influence forest resources. Because last (2015) and Zhang et al. (2018), to highlight the reasonability of
years' level affects current timber output or afforestation, the lag- utilizing GMM as the benchmark estimation approach, the
ged terms of the dependent variables as an explanatory variable
were incorporated. The reason the two models are used is that the
EKC relationship between the forest resource and economic
8
development level in each region may be Ne or U-shaped. Notably, system GMM and first-difference GMM methods are appropriate for
The choice of appropriate regression models is vital for the the panel data with T < N, because the large-sample properties of the two GMM
methods are based on the conditions of holding T constant while increasing N
reasonability and reliability of the estimation results. This study toward infinity (Roodman, 2009). Moreover, there might be a weak instrument
chooses Eqs. (1) and (2) as the benchmark specifications for the problem for the first-difference GMM when T is large, because the number of the
empirical study for two reasons. First, the empirical framework for instruments is approximately a quadratic function of T (Windmeijer, 2005).
854 Y. Hao et al. / Journal of Cleaner Production 214 (2019) 848e859

Table 2
Estimated results of timber output per capita as the dependent variable with fixed effects.

Dependent variable Whole sample Eastern Central Western

ln Y 10.241*** (3.263) 203.143** (97.309) 12.111*** (3.537) 15.192* (8.905)

ln2Y 0.051*** (0.170) 22.182** (9.811) 0.650*** (0.179) 0.865* (0.485)
ln3Y 0.798** (0.329)
P 0.014* (0.010) 0.020* (0.011) 0.038*** (0.021) 0.113** (0.051)
E 0.006 (0.004) 0.002 (0.005) 0.016 (0.015) 0.070** (0.029)
Z 0.022 (0.018) 0.134*** (0.041) 0.015 (0.010) 0.027 (0.045)
Adjusted R2 0.0487 0.179 0.191 0.929
AIC 1381.890 549.658 132.2531 508.439
BIC 1406.329 571.526 149.2708 526.088
Shape Inverted U Inverted N Inverted U Inverted U
Turning point(yuan) 4.02  1043 3830.33, 29429.35 11196.15 6493.99

Notes: ln, ln,2 and ln3 represent taking the logarithmic form, its square, and its cube, respectively. The numbers in the parentheses beneath the estimated coefficients are the
corresponding standard errors. *, **, and *** indicate parameters have statistical significance at the 10%, 5%, and 1% levels, respectively. Following the literature on EKC like
Song et al. (2008), the turning points of the inverted Ue and inverted N-shaped relationships are calculated using corresponding formulas.

Table 3
GMM estimation results with the timber output per capita as the dependent variable.

Dependent variable Whole sample Eastern Central Western

L1. Timber 0.467*** (0.020) 0.102* (0.053) 0.406 (0.298) 0.555** (0.226)
ln Y 1712.59*** (68.877) 22.43*** (6.930) 19.37** (7.862) 115.02*** (26.859)
ln2Y 176.97*** (6.994) 1.043*** (0.356) 0.934* (0.492) 6.312*** (1.553)
ln3Y 6.079*** (0.235)
P 0.139* (0.078) 0.010 (0.030) 0.099 (0.224) 0.497 (0.578)
E 0.021** (0.011) 0.000 (0.006) 0.080*** (0.026) 0.131* (0.079)
Z 0.128*** (0.042) 0.000* (0.052) 0.092*** (0.029) 0.089 (0.100)
AR(2) 0.24 1.06 1.07 1.06
P value 0.812 0.289 0.285 0.291
Hansen 27.37 6.73 8.29 1.71
P value 0.038 0.987 0.960 1.000
Shape N Inverted U Inverted U Inverted U
Turning point (yuan) 6677.09, 40199.02 46973.99 29911.40 9052.22

Notes: L1. denotes taking the first-order difference. ln, ln,2 and ln3 represent taking logarithmic form, its square, and its cube, respectively. The numbers in the parentheses
beneath the estimated coefficients are the corresponding standard errors. *, **, and *** indicate parameters have statistical significance at the 10%, 5%, and 1% levels,
respectively. AR(2) represents the Arellano-Bond (A-B) test for the second-order lags of the residuals. Hansen stands for Hansen tests to examine the overall validity of the
instruments. Following the literature on EKC such as Song et al. (2008), the turning points of the inverted Ue and inverted N-shaped relationships are calculated using
corresponding formulas.

estimation results calculated by the conventional panel data FE estimated to be significantly positive and negative, respectively.9
method and GMM approach are compared and discussed. This result is in line with the expectations that in the early stages
of social development, compared with environmental protection,
4. Estimation results and discussions the development of the economy and industry has relatively high
priority. During the sample period, the demand for wood and
Forests are affected by many factors such as local climatic con- timber output continued to increase. Notably, according to the
ditions, topography, and policies for harvesting trees. Tables 2 and 3 estimation results, timber output would decrease when economic
show the results of the natural logarithm of the timber output as growth increases beyond a certain level, especially when the pro-
the dependent variable. Tables 4 and 5 report the results of the portion of high-tech and service industries rises, and the awareness
natural logarithm of the afforestation area as the dependent vari- of environmental protection gradually increases.
able. The results calculated by the FE method are shown in Tables 2 The estimation results for the three geographical regions indi-
and 4, and the GMM estimation results are presented in Tables 3 cate that there were inverted U-shaped relationships in the central
and 5 and western regions and an inverted N-shaped relationship in the
As aforementioned, because of the regional gaps in economic eastern region.10 Because the levels of GDP per capita for the three
growth and the forestry sector, this study explores the nexus be-
tween these variables for the whole country and individually for
the eastern, central, and western regions. Notably, the estimation 9
Notably, the adjusted R2 are not too high for the fixed effects estimations. A
results are not consistent for the whole country and the three possible reason might be that our sample T (the number of time periods) is smaller
geographical regions: for instance, Ue and N-shaped relationships than N (the number of cross-section units) (Wooldridge, 2016). Moreover, as
are observed in the regions. In the following sections, the differ- mentioned in Data and Methods, because of potential endogeneity, first-difference
GMM estimations are the benchmark estimations, and conventional FE estimations
ences in the estimation results are explained in detail.
are provided for comparison.
Table 2 shows the estimated results for timber output per capita 10
The regression specifications with cubic GDP per capita [i.e., Eq. (3)] have also
as the dependent variable by FE. The estimation results for the been tested for the central and western regions. However, the estimated co-
whole nation are depicted in the first column, while the results for efficients of the cubic terms were statistically insignificant. Therefore, the specifi-
three geographical regions are presented in the last three columns. cations for the central and western regions are essentially Eq. (2), and the
estimation results without the cubic terms are reported. For more detailed dis-
At the national level, there is clear evidence for the existence of an cussions, refer to Lieb (2003) and Song et al. (2008). For similar reasons, in Tables 3
inverted U-shaped relationship between timber output and eco- and 4, the estimation results with cubic GDP per capita are reported for some re-
nomic development, because the coefficients of lnYit and ln2 Yit are gions where the coefficients of the cubic term are significant.
 Y. Hao et al. / Journal of Cleaner Production 214 (2019) 848e859 855

Table 4
Results with the afforestation area per capita as the dependent variable and by using the fixed effect method.

Dependent variable Whole sample Eastern Central Western

ln Y 8.210*** (1.540) 118.365*** (41.120) 363.903*** (91.316) 8.324** (3.255)

ln2Y 0.450*** (0.080) 11.037*** (4.146) 36.546*** (9.430) 0.482*** (0.177)
ln3Y 0.338** (0.139) 1.223*** (0.324)
P 0.034*** (0.005) 0.038*** (0.005) 0.089*** (0.028) 0.056*** (0.018)
E 0.007*** (0.020) 0.008*** (0.002) 0.002 (0.020) 0.005 (0.010)
Z 0.020** (0.008) 0.050*** (0.017) 0.041*** (0.013) 0.010 (0.016)
Adjusted R2 0.164 0.382 0.268 0.097
AIC 729.951 260.229 210.392 226.630
BIC 754.390 282.097 230.246 244.279
Shape U inverted N inverted N U
Turning point (yuan) 9156.53 14242.39, 199836.07 17639.85, 25427.56 5624.32

Notes: ln, ln,2 and ln3 represent taking the logarithmic form, its square, and its cube, respectively. The numbers in the parentheses are standard errors, and *, **, and *** indicate
parameters have a statistical significance at the 10%, 5%, and 1% levels, respectively. Following the literature on EKC such as Song et al. (2008), the turning points of the inverted
Ue and inverted N-shaped relationships are calculated using corresponding formulas.

Table 5 economic development, potential endogeneity problem is

GMM estimation results with the afforestation area per capita as the dependent observed, which should be managed with the GMM method.
variable.
To deal with the potential problem of endogeneity and enhance
Dependent variable Whole sample Eastern Central Western the persuasiveness of the empirical results, the first-difference
L1. Afforestation 0.104** 0.0238 0.231 0.233 GMM method is employed to estimate the benchmark regression
0.046 0.20 0.250 0.159 Eqs. (1) and (2), of which the results for timber output are shown in
ln Y 937.052*** 441.809** 1993.098** 880.029* Table 3. Similar to Table 2, the first column reports the results for
136.248 216.605 821.911 503.646
the whole nation, while the last three columns for the three re-
ln2Y 96.835*** 44.268** 205.126** 92.689*
14.014 21.402 85.484 53.498 gions. According to the test results, the p values for the Hansen
Ln3Y 3.319*** 1.473** 6.964** 3.246* statistics are all higher than 0.1, and the A-B tests for AR (2) are also
0.479 0.703 2.939 1.888 greater than 0.1, suggesting that the instrumental variables chosen
P 0.082*** 0.033* 1.312** 0.006
in the GMM estimations are suitable and reliable (Roodman, 2009).
0.022 0.018 0.659 0.112
E 0.051*** 0.020** 0.374* 0.081
For the three regions of China, the coefficients of GDP per capita are
0.008 0.009 0.197 0.050 estimated to be significantly positive, while the coefficients of
Z 0.059** 0.719 0.066 0.033 squared GDP per capita turn out to be significantly negative, These
0.028 0.057 0.086 0.053 estimation results are basically in line with the results shown in
AR(2) 1.62 0.03 1.21 1.10
Table 2 obtained by FE and indicate that there exists inverted U-
P value 0.105 0.954 0.226 0.271
Hansen 22.09 5.25 0.60 7.22 shaped relationship between timber output per capita and GDP per
P value 0.140 0.994 1.000 0.969 capita in the three regions. In contrast, for the whole country, the
Shape inverted N inverted N inverted N inverted N GMM estimation results suggest that the relationship between
Turning point (yuan) 8419.69 12163.03 6754.36 8449.04 timber output and GDP per capita is N shaped (Song et al., 2008).
33266.57 30611.94 49953.78 21911.09
However, because the values of GDP per capita in most provinces
Notes: L1. denotes taking the first-order difference. ln, ln,2 and ln3 represent taking during the sample period are within the range of the two estimated
the logarithmic form, its square, and its cube, respectively. The numbers in the
turning points (6677 and 40199 yuan, respectively), the actual
parentheses beneath the estimated coefficients are the corresponding standard
errors. *, **, and *** indicate parameters have statistical significance at the 10%, 5%, relationship is inverted U-shaped (Lieb, 2003; Kaika and Zervas,
and 1% levels, respectively. AR(2) represents the Arellano-Bond (A-B) test for the 2013).
second-order lags of the residuals. Hansen stands for Hansen tests to examine the Notably, after controlling for the potential endogeneity, the
overall validity of the instruments. Following the literature on EKC such as Song
estimated levels of GDP per capita corresponding to the turning
et al. (2008), the turning points of the inverted Ue and inverted N-shaped re-
lationships are calculated using corresponding formulas.
points of timber output are more reasonable. specifically, by FE
estimates, the estimated GDP per capita at the turning point for the
whole nation is 4.02  1043, which is unreasonably high and far
regions at the end of the sample period have already been higher beyond the current level of China's economic development. In
than the estimated inflection points, the results suggest that timber contrast, according to the GMM estimation results, the inflection of
output would gradually decrease with the growth of the economy timber output is expected to be obtained when GDP per capita is
in all these regions. 40,199 yuan. Given that China's average GDP per capita in 2015 was
Although the effects of economic and social indicators (e.g., 36,500 yuan, the turning point would soon be reached at the na-
urbanization rate, trade factors, and proportion of secondary in- tional level.11
dustry) on timber production have been considered, the influence For the eastern region, the turning point of per capita GDP is
of other socioeconomic factors remains, such as the various climate 46,973.99 yuan. Compared with real per capita GDP in the regions
conditions in the provinces. For instance, although the provinces in in 2015, a conclusion could be made that the eastern region has
the same geographical region have similar levels of economic and already arrived at the phase of decreasing forest resource use.
social development, the natural conditions vary significantly across Similarly, the turning points in the central and western regions in
the provinces in the same region (e.g., Hao et al., 2016a). These
factors cannot be measured well but affect forest conservation and
timber harvesting, thus making the results of the FE model in the 11
Notably, a similar upward estimation bias for the level of GDP per capita at the
previous example less reliable than expected. Moreover, because of turning point has also been found in EKC-related studies for China by using FE
the possible bilateral causality between forest resources and estimations that do not account for potential endogeneity (e.g., Hao et al., 2015;
Zhang et al., 2018).
856 Y. Hao et al. / Journal of Cleaner Production 214 (2019) 848e859

2015 are 29,911.40 yuan and 9052.22 yuan, respectively. Therefore, change the sign of the estimated coefficient (Wooldridge, 2016).
the western region has also entered the stage where timber output The coefficient of the urbanization rate in the western region is
decreases with economic growth, while the turning point for the also positive, no matter FE or GMM is employed (although the
central region is about to come soon. In summary, the three regions estimated coefficient by GMM is insignificant). This result indicates
are in a similar phase in terms of the relationship between timber urbanization in the western region bought higher demand for
output and GDP per capita, as timber output is expected to decrease wood, and timber output correspondingly increased. In the eastern
with further economic growth. and central regions, the coefficients are negative after endogeneity
According to the empirical results, the relationship between is controlled for (as shown in Table 3). During urbanization, the
timber output and GDP per capita is N-shaped, and the two in- demand for timber shifted from local areas to other regions or to
flection points are 6677 yuan and 40,199 yuan, respectively. For foreign countries. For example, in 2000, China became the world's
China, per capita GDP in 2015 (the end of the sample period) is in second-largest timber importer, and timber imports continue to
the phase between the two inflection points, and the majority of increase.
the whole sample lies before the second inflection point. In this No significant effect in the eastern region is observed in the
regard, there is essentially an inverted U-shaped nexus of timber outcome of trade openness in the FE method, but a positive effect
output and GDP per capita for the whole country (Lieb, 2003; Song was observed in the central and western provinces when using the
et al., 2008). Moreover, as the level of GDP per capita corresponding GMM method. Taking the regional export trade structure into ac-
to second inflection point is expected to be reached soon, the count, a reasonable explanation is that the main export products in
timber output would decrease with further economic growth. the eastern region are in the machinery manufacturing category,
These results suggest that the Chinese government must continue and the proportion of forest products is relatively small; thus, the
to strengthen environmental regulations and protection of the impact of trade is considerably small.12 Compared with the eastern
environment and natural resources; failing to implement the region, in the central and western regions, the proportion of import
appropriate policies would cause the demand of resources, repre- and export forest products is larger; therefore, when trade volume
sented by timber output, to surge again; this is because economic increases, timber output would also increase.
development and quality of life have increased to a relatively high As shown in Table 3, after controlling for the endogeneity, the
level. second industry proportion has positive effects for the whole
For the control variables, the impact of the urbanization rate at country and in three regions except less developed western region.
the national level is significantly positive. One possible reason for This finding is consistent with the expectation that the amount of
this finding is that urbanization increases the construction of res- wood required increases in the industrial production process when
idences, infrastructure, and public places, which increases the de- the proportion of the secondary industries increases. Because the
mand for timber and timber output. This finding is basically in line eastern region leads the other regions in industrial structural
with some extant studies such as Shen et al. (2005) and Wang upgrading, its economic structure is relatively more reasonable and
(2013) who found that timber consumption would increase as ur- sustainable because it has a relatively high proportion of high-tech
banization increased in China and other countries, respectively. In and service industries (e.g., Zhou et al., 2018; Xu and Lin, 2018).
particular, Shen et al. (2005) investigated the utilization of energy Notably, the impact of the secondary production proportion in the
and mineral resources during the process of urbanization in China. west is negative (although not statistically significant); that is,
They found that the demand of a series of resources, including when the secondary industry booms, the local production of wood
timber, would increase rapidly at the early stage of China's rapid decreases. The Chinese government could control the amount of
urbanization. Some recent literature has also verified the positive timber harvested in a local area after recognizing the importance of
relationship between urbanization and natural resources or energy protecting the environment and forest ecology. In addition, when
in China (e.g., Hao et al., 2015; Fan et al., 2017; Cui et al., 2019). the amount of local production cannot fulfill the demand, the
Compared with the results by FE shown in Table 2, the estimated amount of imported timber will greatly increase. As aforemen-
coefficient of urbanization rate for the whole country after the tioned, China became the world's second largest timber importer in
endogeneity is controlled shown in Table 3 turns out to be signif- 2000.
icantly positive (from negative). Once again, this difference in- In Table 4, the estimation results by FE when the afforestation
dicates that the endogeneity bias may be sufficiently large to even area per capita is used as the dependent variable to characterize the
forest resources are reported. The estimation results by conven-
tional FE method that do not account for endogeneity appear to be
12
inconsistent as the results for the whole country and the three
As an anonymous reviewer asserted, import and export may have different
geographical regions differ remarkably. For instance, because the
influences on timber output. For instance, domestic timber output can decrease as
timber imports increase, whereas domestic timber output can increase as timber estimated coefficients of GDP per capita, its square and cube are
exports increase. In this regard, the effects of imports and exports have also been significantly negative, positive and negative, respectively, the
estimated separately by replacing the variable of trade openness with the ratios of eastern and central regions have an inverted N-shaped estimated
imports and exports to GDP, respectively. However, the estimation results indicate EKC (Lieb, 2003; Kaika and Zervas, 2013). In contrast, for the whole
that neither imports nor exports affect timber output significantly. One possible
reason for these results is that during the sample period of 2002e2015, partly
sample and western regions, there is evidence for the existence of
because of the adjustments of economic and industrial structures and the changes U-shaped relationship between afforestation area per capita and
in relevant policies and regulations, the imports, exports, and output of timber in GDP per capita, as the coefficients of GDP per capita and its square
many provinces changed remarkably. For example, in some northern provinces are negative and positive, respectively. Therefore, these results
with rich endowment of primordial forests like Inner Mongolia and Heilongjiang,
suggest no evidence for conventional EKC nationwide or in western
timber output decreased dramatically due to restrictive policies and tighter regu-
lations as the central government realized the key role forests play in environ- China. As aforementioned in the discussion of the estimation re-
mental protection and air quality improvement in the North China Plain. In sults for timber output by FEs shown in Table 2, the FE estimation
contrast, partly because of wide use of artificial afforestation technique, in some results are probably biased because of potential endogeneity. In this
rich southern provinces that are traditionally net timber importers like Guangdong, regard, first-difference GMM estimations have also been conduct-
the timber output has increased rapidly. As such, imports and exports did not have
the significant impacts on timber output as expected. Because of space limitations,
ed, and the results are shown in Table 5.
these complementary estimation results are not provided but are available upon As shown in Table 5, according to the first-difference GMM
request. estimation results, which have controlled for the potential
 Y. Hao et al. / Journal of Cleaner Production 214 (2019) 848e859 857

endogeneity problem, an inverted N-shaped nexus of afforestation land available for afforestation (Liu et al., 2014). The estimated
area per capita and GDP per capita is observed for the whole coefficients of the secondary industry ratio are largely negative
country and the three regions. Because the levels of GDP per capita (although insignificant for the east and central regions and insig-
for the majority of the sample are higher than the estimated first nificantly positive for the western region), suggesting that the ef-
inflection point, the EKC relationship is in fact inverted U-shaped fect of secondary industry development on the afforestation area is
(Lieb, 2003; Song et al., 2008). Given that the per capita GDP level of overall negative. Therefore, the reduction in the proportion of the
most of China's provinces in 2015 was higher than the second secondary industry may help increase the area of afforestation.
turning point of the inverted N-shaped relationship for the whole
nation and three regions, the afforestation area would likely 5. Conclusion and policy suggestions
decrease with economic growth for the country as a whole.
Since 1980s, China began to realize the ecological value of forest 5.1. Conclusion
resources and planting efforts; afforestation areas thus increased
with economic development (Wang et al., 2007). However, as On the basis of the EKC hypothesis, this study investigates the
economy continued to grow, the demand for forest resources rose relationship between forest resources and economic growth by
significantly, and afforestation began to decrease with economic using panel data from 30 provinces in China for the period of
growth. Notably, such downward trend of afforestation area may 2002e2015. The preliminary results were obtained using the FE
not be consistent with the demand for sustainable development method. Additionally, taking the endogeneity problem and dy-
given the importance of forests in improving environmental qual- namic effects into account, the first-difference GMM method was
ity. Therefore, policymakers should draft, and the Chinese gov- used as benchmark model to test the relationship between forest
ernment should implement relevant policies that encourage resources and economic growth. According to the estimation re-
afforestation. sults, the EKC relationship between forest resources (represented
The comparison of the estimation results by FE and GMM is by timber output per capita and afforestation area per capita) and
presented in Tables 4 and 5 and indicates that the endogeneity economic development is essentially inverted U-shaped at both
biases are sufficiently large to change the shape of the estimated national and regional levels of China. Specifically, when using
EKC relationship. Therefore, the use of the GMM approach as the timber output and afforestation area as dependent variables to
benchmark estimation method is necessary and reasonable. characterize forest resources, the estimation results indicate that
Notably, the benchmark GMM estimation results with timber China is currently at the stage where the amount of timber output
output per capita and afforestation area per capita as dependent and afforestation area decreases with economic growth (i.e., at the
variables (Tables 3 and 5) are roughly consistent, suggesting that right part of the inverted U-shaped EKC), because the national GDP
the consumption of forest resources would at first increase and per capita now has already passed the levels corresponding to the
then decrease as China's economy continually grows. Therefore, the peaks of the two EKC curves for the two indicators of forest re-
estimation results for timber output and the afforestation indicate sources (6677 yuan for timber output per capita and 33,266 yuan
that the peak of demand for China's forest resources would be for afforestation area per capita). These results are basically
reached soon (after the ending year of the sample period, i.e., 2015). consistent with the expectations and previous relevant studies.
This finding is in line with the existent relevant literature,
including Liu and Lv (2010) and Zambrano-Monserrate et al. (2018), 5.2. Suggestions
which focused on the nexus of forest resources and economic
growth in China and European countries. At present, the con- Some important and straightforward policy implications could
sumption of forest resources has already begun to decrease with be drawn based on the empirical results of this study. First of all, the
economic growth in China. Notably, a caveat remains and has been estimation results indicate that it is necessary to accelerate indus-
indicated in the literature. For instance, in a series of recent studies, trial structural upgrades because the development of secondary
Stern (2017) and Gill et al. (2018) warned that policymakers should industry is positively related with the consumption of forest re-
not be completely dependent on the predictions of EKC studies to sources at the national level. As the economy becomes less
formulate relevant policies, because the environmental quality may dependent on the secondary industry, the decreased demand
not automatically improve as economy continues to grow even if would significantly affect forest resources. Second, land-use effi-
there is statistical evidence for the existence of EKC. As a result, ciency should be improved, and the government should continue to
policy and legal interventions remain necessary to achieve the release land as forested land. Given the important role forests play
political goals of protecting forest resources in China. Given that the in the improvement of environmental quality, Chinese government
Chinese government has restricted the logging of natural forests for at various levels should strongly encourage the protection of forests
commercial purposes, the increasing demand for timber output and afforestation. Third, considering the fact that the relationship
could only be fulfilled with increased afforestation.13 At the same between forest resources and economic growth differ across re-
time, to protect the forest ecology and environment, a simulta- gions in China and that there is remarkable regional gap in eco-
neous increase in financial support for afforestation would be nomic and social development, the relevant policies for the
required to achieve the goal of balancing economic and ecological economic development and forest protection should be more tar-
relations. geted and on the basis of differentiated characteristics of different
Regarding the control variables, as shown in the Table 5 provinces and regions.
(benchmark GMM estimation results), the effects of the urbaniza-
tion rate on the afforestation area is in general negative. This result Acknowledgements
is partly because the expansion of cities is at the cost of woodlands
and arable land; therefore, a higher urbanization rate reduces the The authors acknowledge financial support from the National
Natural Science Foundation of China (71761137001, 71403015,
71521002), the Beijing Natural Science Foundation (9162013), the
13
For more information of China's ban on logging natural forests, please refer to
key research program of Beijing Social Science Foundation
http://english.mep.gov.cn/News_service/media_news/201703/t20170317_408329. (17JDYJA009), the National Key Research and Development Pro-
shtml. gram of China (2016YFA0602801, 2016YFA0602603), and the
858 Y. Hao et al. / Journal of Cleaner Production 214 (2019) 848e859

Special Fund for Joint Development Program of Beijing Municipal (carbon dioxide) emissions: an empirical estimation based on Chinese pro-
vincial panel data. J. Clean. Prod. 131, 667e677.
Commission of Education. The authors are also very grateful to the
Hao, Y., Chen, H., Zhang, Q., 2016b. Will income inequality affect environmental
three anonymous reviewers, Associate Editor Prof. Dr. Xin Tong, quality? Analysis based on China's provincial panel data. Ecol. Indicat. 67,
Associate Editor Prof. Dr. Giovanni Baiocchi, Editor-in-Chief Jiri 533e542.
Jaromir Klemes, and Editor-in-Chief Prof. Dr. Yutao Wang for their Hayakawa, K., 2009. First difference or forward orthogonal deviation-Which
transformation should be used in dynamic panel data models?: a simulation
insightful comments that helped us sufficiently improve the quality study. Econ. Bull. 29 (3), 2008e2017.
of this paper. The usual disclaimer applies. Huang, Y., 2010. Political institutions and financial development: an empirical
study. World Dev. 38 (12), 1667e1677.
Iacob, S.E., 2015. The role of the forest resources in the socioeconomic development
References of the Rural areas. Proc. Econom. Finance 23, 1578e1583.
Indarto, J., Mutaqin, D., 2016. An overview of theoretical and empirical studies on
Ahmed, K., Shahbaz, M., Qasim, A., Long, W., 2015. The linkages between defores- deforestation. J. Int. Develop. Cooperat. 22, 107e120.
tation, energy and growth for environmental degradation in Pakistan. Ecol. Joshi, P., Beck, K., 2016. Environmental Kuznets curve for deforestation: evidence
Indicat. 49, 95e103. using GMM estimation for OECD and non-OECD regions. iFor. Biogeosci. For. 10
Allen, J.C., Barnes, D.F., 1985. The causes of deforestation in developing countries. (1), 196.
Ann. Assoc. Am. Geogr. 75 (2), 163e184. Kaika, D., Zervas, E., 2013. The Environmental Kuznets Curve (EKC) theorydPart A:
Angelsen, A., Kaimowitz, D., 1999. Rethinking the causes of deforestation: lessons Concept, causes and the CO2 emissions case. Energy Pol. 62, 1392e1402.
from economic models. World Bank Res. Obs. 14 (1), 73e98. Kallbekken, S., 2000. An Alternative Environmental Kuznets Curve Approach to
Arellano, M., Bond, S., 1991. Some tests of specification for panel data: Monte Carlo Deforestation. University of York, UK (Unpublished Master of Science
evidence and an application to employment equations. Rev. Econ. Stud. 58 (2), dissertation).
277e297. Kang, Y.Q., Zhao, T., Yang, Y.Y., 2016. Environmental Kuznets curve for CO2 emissions
Arellano, M., Bover, O., 1995. Another look at the instrumental variable estimation in China: a spatial panel data approach. Ecol. Indicat. 63, 231e239.
of error-components models. J. Econom. 68 (1), 29e51. Koop, G., Tole, L., 1999. Is there an environmental Kuznets curve for deforestation?
Barbier, E.B., Burgess, J.C., 2001. The economics of tropical deforestation and land J. Dev. Econ. 58 (1), 231e244.
use: an introduction to the special issue. Land Econ. 77 (2), 155e171. Kuznets, S., 1955. Economic growth and income inequality. Am. Econ. Rev. 45 (1),
Bhattarai, M., Hammig, M., 2001. Institutions and the environmental Kuznets curve 1e28.
for deforestation: a cross-country analysis for Latin America, Africa and Asia. Li, T., Wang, Y., Zhao, D., 2016. Environmental Kuznets curve in China: new evidence
World Dev. 29 (6), 995e1010. from dynamic panel analysis. Energy Pol. 91, 138e147.
Blundell, R., Bond, S., 1998. Initial conditions and moment restrictions in dynamic Lieb, C.M., 2003. The Environmental Kuznets Curve: A Survey of the Empirical
panel data models. J. Econom. 87 (1), 115e143. Evidence and of Possible Causes (No. 391). Discussion. Paper Series, University
Cai, H., Treisman, D., 2006. Did government decentralization cause China's eco- of Heidelberg, Department of Economics.
nomic miracle? World Polit. 58 (4), 505e535. Liu, C., Lv, J., 2010. On China's forest resource environmental Kuznets curve. Res.
Caviglia-Harris, J.L., Chambers, D., Kahn, J.R., 2009. Taking the “U” out of Kuznets: a Institut. Econom. 2010 (2), 138e161 (in Chinese).
comprehensive analysis of the EKC and environmental degradation. Ecol. Econ. Liu, S., Zhou, T., Wei, L., Shu, Y., 2012. The spatial distribution of forest carbon sinks
68 (4), 1149e1159. and sources in China. Chin. Sci. Bull. 57 (14), 1699e1707.
Chen, W., Xu, D., Liu, J., 2015. The forest resources inputeoutput model: an appli- Liu, J., Kuang, W., Zhang, Z., Xu, X., Qin, Y., Ning, J., Zhou, W., Zhang, S., Li, R., Yan, C.,
cation in China. Ecol. Indicat. 51, 87e97. Wu, S., Shi, X., Jiang, N., Xu, D., Pan, X., Chi, W., 2014. Spatiotemporal charac-
Chiu, Y.B., 2012. Deforestation and the environmental Kuznets curve in developing teristics, patterns, and causes of land-use changes in China since the late 1980s.
countries: a panel smooth transition regression approach. Can. J. Agric. J. Geogr. Sci. 24 (2), 195e210.
Econom./Revue canadienne d'agroeconomie 60 (2), 177e194. Montagne -Huck, C., Brunette, M., 2018. Economic analysis of natural forest distur-
Choumert, J., Motel, P.C., Dakpo, H.K., 2013. Is the environmental Kuznets curve for bances: a century of research. J. For. Econ. 32, 42e71.
deforestation a threatened theory? A meta-analysis of the literature. Ecol. Econ. Naidoo, R., 2004. Economic growth and liquidation of natural capital: the case of
90, 19e28. forest clearance. Land Econ. 80 (2), 194e208.
Cropper, M., Griffiths, C., 1994. The interaction of population growth and environ- Narita, D., Lemenih, M., Shimoda, Y., Ayana, A.N., 2018. Economic accounting of
mental quality. Am. Econ. Rev. 84 (2), 250e254. ethiopian forests: a natural capital approach. For. Policy Econ. 97, 189e200.
Cui, X., Fang, C., Liu, H., Liu, X., 2019. Assessing sustainability of urbanization by a National Bureau of Statistics of China, 2013. Statistical Yearbook of China. China
coordinated development index for an Urbanization-Resources-Environment Statistical Publisher House, Beijing.
complex system: a case study of Jing-Jin-Ji region, China. Ecol. Indicat. 96, Nguyen, M.D., Ancev, T., Randall, A., 2018. Forest governance and economic values of
383e391. forest ecosystem services in Vietnam. Land Use Pol. https://doi.org/10.1016/
Culas, R.J., 2007. Deforestation and the environmental Kuznets curve: an institu- j.landusepol.2018.03.028 (in press).
tional perspective. Ecol. Econ. 61 (2), 429e437. Panayotou, T., 1993. Empirical Tests and Policy Analysis of Environmental Degra-
Damette, O., Delacote, P., 2012. On the economic factors of deforestation: what can dation at Different Stages of Economic Development (No. 992927783402676).
we learn from quantile analysis? Econ. Modell. 29 (6), 2427e2434. International Labour Organization.
Dobra, J., Dobra, M., Ouedraogo, A., 2018. Does mineral development provide a basis Roodman, D., 2009. How to do xtabond2: an introduction to difference and system
for sustainable economic development? Resour. Pol. https://doi.org/10.1016/ GMM in Stata. STATA J. 9 (1), 86e136.
j.resourpol.2018.03.013 (in press). Shen, L., Cheng, S., Gunson, A.J., Wan, H., 2005. Urbanization, sustainability and the
Du, L., Wei, C., Cai, S., 2012. Economic development and carbon dioxide emissions in utilization of energy and mineral resources in China. Cities 22 (4), 287e302.
China: provincial panel data analysis. China Econ. Rev. 23 (2), 371e384. Shen, J., 2006. A simultaneous estimation of environmental Kuznets curve: evi-
Ehrhardt-Martinez, K., 1998. Social determinants of deforestation in developing dence from China. China Econ. Rev. 17 (4), 383e394.
countries: a cross-national study. Soc. Forces 77 (2), 567e586. Singh, M.P., Bhojvaid, P.P., de Jong, W., Ashraf, J., Reddy, S.R., 2017. Forest transition
Ehrhardt-Martinez, K., Crenshaw, E.M., Jenkins, J.C., 2002. Deforestation and the and socio-economic development in India and their implications for forest
environmental Kuznets curve: a cross-national investigation of intervening transition theory. For. Pol. Econ. 76, 65e71.
mechanisms. Soc. Sci. Q. 83 (1), 226e243. Stern, D.I., Common, M.S., Barbier, E.B., 1996. Economic growth and environmental
Fan, J.L., Zhang, Y.J., Wang, B., 2017. The impact of urbanization on residential energy degradation: the environmental Kuznets curve and sustainable development.
consumption in China: an aggregated and disaggregated analysis. Renew. World Dev. 24 (7), 1151e1160.
Sustain. Energy Rev. 75, 220e233. Song, T., Zheng, T., Tong, L., 2008. An empirical test of the environmental Kuznets
Foster, A.D., Rosenzweig, M.R., 2003. Economic growth and the rise of forests. Q. J. curve in China: a panel cointegration approach. China Econ. Rev. 19 (3),
Econ. 118 (2), 601e637. 381e392.
Führer, E., 2000. Forest functions, ecosystem stability and management. For. Ecol. Stern, D.I., 2017. The environmental Kuznets curve after 25 years. J. Bioecon. 19 (1),
Manag. 132 (1), 29e38. 7e28.
Gill, A.R., Viswanathan, K.K., Hassan, S., 2018. The Environmental Kuznets Curve Tole, L., 1998. Sources of deforestation in tropical developing countries. Environ.
(EKC) and the environmental problem of the day. Renew. Sustain. Energy Rev. Manag. 22 (1), 19e33.
81, 1636e1642. Turner, J.A., Buongiorno, J., Zhu, S., 2006. An economic model of international wood
Grossman, G.M., Krueger, A.B., 1991. Environmental Impacts of a North American supply, forest stock and forest area change. Scand. J. For. Res. 21 (1), 73e86.
Free Trade Agreement (No. W3914). National Bureau of Economic Research. Van, P.N., Azomahou, T., 2007. Nonlinearities and heterogeneity in environmental
Halkos, G.E., Paizanos, E.А., 2013. The effect of government expenditure on the quality: an empirical analysis of deforestation. J. Dev. Econ. 84 (1), 291e309.
environment: an empirical investigation. Ecol. Econ. 91, 48e56. Waluyo, E.A., Terawaki, T., 2016. Environmental Kuznets curve for deforestation in
Hao, Y., Wei, Y.M., 2015. When does the turning point in China's CO2 emissions Indonesia: an ARDL bounds testing approach. J. Econom. Cooperat. Develop. 37
occur? Results based on the Green Solow model. Environ. Dev. Econ. 20 (6), (3), 87.
723e745. Wang, S., 2013. Forest economics in an increasingly urbanized society: the next
Hao, Y., Zhang, Z.Y., Liao, H., Wei, Y.M., 2015. China's farewell to coal: a forecast of frontier. For. Pol. Econ. 35, 45e49.
coal consumption through 2020. Energy Pol. 86, 444e455. Wang, L., Liu, J., 2003. An analysis on the factors affecting forest coverage (1978-
Hao, Y., Chen, H., Wei, Y.M., Li, Y.M., 2016a. The influence of climate change on CO2 1998). J. Beijing For. Univ. 2 (1), 33e36 (in Chinese).
 Y. Hao et al. / Journal of Cleaner Production 214 (2019) 848e859 859

Wang, P., Xu, C., 2015. Research on the comprehensive evaluation of China's envi- Yang, Y., 2016. An analysis of the relationship between the development of the
ronmental health, 2003e2010. Soc. Indicat. Res. 122 (3), 709e721. tertiary industry and employment in China. Modern Econom. Informat. 11,
Wang, S., Liu, C., Wilson, B., 2007. Is China in a later stage of a U-shaped forest 383e384 (in Chinese).
resource curve? A re-examination of empirical evidence. For. Pol. Econ. 10 (1), Zambrano-Monserrate, M.A., Carvajal-Lara, C., Urgile s-Sanchez, R., Ruano, M.A.,
1e6. 2018. Deforestation as an indicator of environmental degradation: analysis of
Windmeijer, F., 2005. A finite sample correction for the variance of linear efficient five european countries. Ecol. Indicat. 90, 1e8.
two-step GMM estimators. J. Econom. 126 (1), 25e51. Zhang, Y., Tachibana, S., Nagata, S., 2006. Impact of socio-economic factors on the
Wooldridge, J.M., 2016. Introductory Econometrics: A Modern Approach (6e). changes in forest areas in China. For. Pol. Econ. 9 (1), 63e76.
South-Western Cengage Learning, Mason, USA. Zhang, Z., Hao, Y., Lu, Z.N., Deng, Y., 2018. How does demographic structure affect
Wu, J., Zheng, H., Zhe, F., Xie, W., Song, J., 2018. Study on the relationship between environmental quality? Empirical evidence from China. Resour. Conserv. Recycl.
urbanization and fine particulate matter (PM2.5) concentration and its impli- 133, 242e249.
cation in China. J. Clean. Prod. 182, 872e882. Zhou, Y., Guo, Y., Liu, Y., 2018. High-level talent flow and its influence on regional
Xu, B., Lin, B., 2018. Investigating the role of high-tech industry in reducing China's unbalanced development in China. Appl. Geogr. 91, 89e98.
CO2 emissions: a regional perspective. J. Clean. Prod. 177, 169e177.