Are urbanization, industrialization and CO2 emissions cointegrated?

Komivi AFAWUBOa
a
CEREFIGE-University of Lorraine-France, CRESE-University of Franche-Comté-France
Email : komivi.afawubo@univ-lorraine.fr
and
Clarisse NGUEDAM NTOUKOb*
b
Post-Doctoral Researcher, EconomiX, UMR 7235, Université Paris Ouest Nanterre La
Défense, and Consultant at The World Bank, Washington, DC, USA
Email: cnguedam@gmail.com
* The views expressed herein are those of the author and should not be attributed to the World Bank, its
Executive Board or its management; or to the University Paris Ouest, Nanterre La Défense.

(Preliminary version, do not quote without permission)

1
Abstract

This paper analyses the relationship between urbanization, industrialization and CO2

emissions in a panel of 142 countries over the period 1960-2014. Using the Autoregressive

Distributed Lag (ARDL) model, the empirical results reveal a robust cointegration

relationship with heterogeneity depending on countries’ income level and the consideration of

long-run and short-run perspectives. In low-income countries, there is no significant

relationship between industrialization, urbanization and CO2 emissions in short-run, while in

long-run, urbanization and industrialization are positively correlated with CO2 emissions. In

upper-middle-income countries, only urbanization has a significant correlation in short-run,

but in long-run, both industrialization and urbanization have significant effects on CO2

emissions. In lower-middle-income countries, only industrialization is lightly correlated with

CO2 emissions in long-run. Industrialization and urbanization in 1-period lag have significant

correlation with CO2 emissions in high-income non-OECD countries in short-run.

Concerning OECD countries, urbanization in 1-period lag and industrialization are correlated

with CO2 emissions in short-run, while in long-run only industrialization has a significant

correlation. Furthermore, the granger-causality between urbanization, industrialization and

CO2 emissions varies across income groups. Our findings open up new insights for

urbanization and industrialization policies while considering economic development stage and

their short-run and long-run effects on CO2 emissions.

Keywords: Urbanization, Industrialization, CO2 emissions, Panel Cointegration

JEL Classification: Q43, R11, C1

2
 1. Introduction

The process of urbanization and industrialization is a large driving force of economic

growth. There is a worldwide rapid spread of the process imposing additional infrastructure

and production which increase energy consumption. The resulting carbon dioxide emissions

(henceforth, CO2) cause a global warming and consequently attract widespread concerns

from the international community in promoting sustainable development and environment

protection, especially in terms of reduction of CO2 emissions. CO2 remains the major

anthropogenic greenhouse gas (GHG) which accounts for 76% of total emissions in 2010.

Total annual anthropogenic GHG emissions have increased by about 10 gigatonnes of CO2-

equivalent between 2000 and 2010. This increase mainly came from energy (47%), industry

(30%), transport (11%) and the building sector (3%). Economic and population growth is the

most important drivers of increases in CO2 emissions from fossil fuel combustion (IPCC

(2014). Globally, 54% of the world’s population resides in urban areas in 2014 and urban

population is projected to reach 66% by 2050 (United Nations, 2014). However, levels of

urbanization and industrialization vary greatly across regions1. CO2 emissions in the world

have increased from 4.6 to 4.9 metric tons per capita over the period 2006-2011, with income-

level heterogeneity: 11.1 for high-income countries; 3.4 for middle-income and 0.3 for low-

income (World Development Indicators, 2015). This trend will intensify over the next

decades as a consequence of high urbanization growth in all regions, especially in Africa and

Asia.

The analysis of the relationship between urbanization, industrialization and CO2

emissions is important as the results can have useful implications for sustainable development

1
Urbanization rate is about 82% in Northern America, 80% in Latin America and the Caribbean; 73% in Europe.

Africa and Asia, in contrast, remain mostly rural, with 40% and 48% of their respective populations living in

urban areas (United Nations, 2014).

3
and climate change policies. Many empirical studies have analyzed this relationship. Some of

them [e.g. York (2007); Cole and Neumayer (2004); York et al. (2003a); Parikh and Shukla

(1995)] have generally found a positive relationship between urbanization and CO2

emissions, while conversely, others have found that urbanization and urban density improve

the efficiency of public infrastructure use such as public transport and other utilities, lowering

energy consumption and emissions [e.g. Chen et al. (2008); Liddle (2004); Newman and

Kenworthy (1989)]. Many authors have made country-level analysis [e.g. Lee and Lee (2014)

for the U.S.; Zhou et al. (2013); Li et al. (2012), for China; Alam et al. (2007) for Pakistan;

Ang (2007), for France]. However, in a number of these studies, little attention has been paid

to income levels heterogeneity as well as dynamic relationships; assuming implicitly that the

relationship between urbanization, industrialization and CO2 emissions is homogenous across

countries regardless their development stages and without distinction between short-run and

long-run effects. This assumption is contrary to the theory of ecological modernization and

the theory of urban environmental transition which argue that the impact of urbanization and

population pressure on environment varies at each stage of economic development. Moreover,

some studies have been made using static models which cannot capture dynamic

relationships. Thus, a number of previous studies have provided inconsistent results, probably

because of differences in methods, sample data, and the absence of consideration of income

level heterogeneity.

In this paper, we investigate the cointegration relationship and assess the long-run and

short-run impact of urbanization and industrialization on CO2 emissions. Our empirical

analysis has several contributions to the literature. First, we use a large panel data set of 142

countries including all income level categories over the period 1960-2014. Secondly, we take

into account income heterogeneity by dividing our sample into 5 groups according to

countries’ income levels. Our empirical analysis relies on the dynamic Autoregressive

4
Distributed Lag (ARDL) model (Pesaran et al., 2001). This approach has several advantages

compared to other traditional cointegration techniques. It allows for smaller sample sizes and

can be used regardless of whether the variables are purely I(0), purely I(1), or mutually

cointegrated. Moreover, it provides unbiased long-run estimates and valid t-statistics. Finally,

the ARDL model is advantageous over static models because it provides a method of

assessing short-run and long-run relationships. In order to analyze the causal relationship

between our variables, we apply the Granger causality test (Granger 1969).

The rest of this paper is structured as follows: in the second part, we present the

relevant literature; the third part presents data and descriptive statistics. The fourth part is

devoted to panel data cointegration procedure. The fifth part presents the empirical results and

discussions and the sixth part concludes.

2. Literature review

From a theoretical point of view, the existing literature points out three theories in

understanding the effects of urbanization on environment: the compact city theory, the

ecological modernization theory, and the urban environmental transition theory

(Poumanyvong and Kaneko, 2010).

The compact city theory relies on the positive effects of urbanization. Higher urban

density increases economies of scale for public infrastructure (hospitals, schools, public

transportation, water supply and electricity production) which tend to lower environmental

damages [Burton (2000); Capello and Camagni (2000); Jenks et al. (1996); Newman and

Kenworthy (1989)].

According to the theory of ecological modernization, urbanization is a process of

social transformation and modernization. Environmental problems may arise as societies

move from low to middle stages of development given the fact that in these stages of

5
development, economic growth prioritized over environmental issues. Environmental

damages become more important as countries evolve to higher stages of development, leading

the society to seek environmental sustainability. Factors such as urbanization, a shift from a

manufacturing to a service based economy and technological innovation can reduce the

negative impact of economic growth on the environment [Mol and Spaargaren (2000);

Gouldson and Murphy (1997); Crenshaw and Jenkins (1996)].

The theory of urban environmental transition (McGranahan et al., 2001) suggests that

cities often become wealthier by increasing their manufacturing activities and this can lead to

industrial pollution. Wealthier cities have wealthier residents. This situation increases demand

for high energy intensive products and puts further stress on the environment. In this context,

industrial pollution may be reduced through environmental regulations, changes in economic

sector composition or technological innovation.

Some recent empirical studies support the idea that the impact of economic growth on

emissions varies with countries’ income levels. Asane-Otoo (2015) used a multi-region input–

output modeling framework to show that industrialization exert positive and significant

effects on CO2 emissions in middle-income countries. However, the impact of urbanization is

generally insignificant across income groups of his sample. Madlener and Sunak (2011) found

that urbanization led to a substantial increase in energy consumption, and that the relevance of

the effect varied considerably between developing and developed countries. Sharma (2011)

argued that urbanization has a negative and significant impact on carbon emissions for a panel

of 69 countries, but this impact is insignificant when he considers income-level groups.

Some studies used the STIRPAT model (Stochastic Impacts by Regression on

Population, Affluence and Technology) of York et al. (2003b) based on the Influence,

Population, Affluence and Technology (IPAT) model (Ehrlich and Holdren, 1971). Sadorsky

(2015) uses a STIRPAT model to explore the impact of urbanization on CO2 emissions in

6
emerging economies. His estimations results from the ARDL model show positive long-run

elasticities which indicate that increases in affluence, population, or energy intensity increase

CO2 emissions in the long-run. Using the STIRPAT framework and a panel dataset of 73

countries classified into four income-levels groups over the period of 1971–2010, Li and Lin

(2015) found that in the low-income group, urbanization increases CO2 emissions; in the

middle-/low-income and high-income groups, both urbanization and industrialization increase

CO2 emissions; for the middle-/high-income group, urbanization hinder the growth of CO2

emissions; while industrialization have an insignificant impact. Martinez-Zarzoso and

Maruotti (2011) used a STIRPAT model and a panel of 88 developing countries over the

period 1975-2003. Their results support an inverted U-shaped relationship between

urbanization and CO2 emissions. Poumanyvong and Kaneko (2010) analyzed the impact of

urbanization on CO2 emissions in 99 countries over the period 1975-2005. Using the

STIRPAT model and a variety of static panel regression techniques, they found a positive

impact of urbanization on emissions for all income groups, but the impact was greatest in the

middle-income group than in the other groups. In the framework of the STIRPAT model Lin

et al. (2009) show that urbanization and industrialization have a potential effect on GHG

emissions in China over the period 1978–2006. Fan et al. (2006) analyzed the impact of

population and technology on CO2 emissions of countries at different income-levels over the

period 1975–2000 using the STIRPAT model. For developing countries, they found that the

impact of population, affluence and technology on CO2 emissions varies at different levels of

development. York et al. (2003a) found evidence that high level of urbanization leads to

increase CO2 emissions in a cross-section of 137 countries in the framework of a STIRPAT

model.

Xu and Lin (2015) examined the impact of industrialization and urbanization on

CO2 emissions in China using nonparametric additive regression models and provincial panel

7
data from 1990 to 2011. They found a nonlinear effect of urbanization on CO2 emissions

across regions, while industrialization follows an inverted U-shaped link with CO2 emissions.

Nejat et al. (2015) believed that urbanization and industrialization were the main driving

factors of the growth of CO2 emissions. Tian et al. (2014) argue that the disparity in regional

industrial structure substantially impacts regional CO2 emissions in China. Salim and Shafiei

(2014) support the evidence of an Environmental Kuznets Curve (EKC) between urbanization

and CO2 emissions in 9 industrialized countries, while industrialization has a positive impact.

Shahbaz et al (2014) found an EKC between industrialization and CO2 emissions using

quarter frequency data in Bangladesh over the period of 1975–2010. Li et al. (2012) found

that in most regions in China, urbanization had a bigger impact on CO2 emissions than other

factors such as GDP per capita, industrial structure, population and technology level.

Cherniwchan (2012) estimated a two-sector model of neoclassical growth using sulfur

emissions data for 157 countries over the period 1970–2000. He found that an increase in

industry's share of total output is associated with an increase in the level of emissions per

capita.

Other recent studies analyzed the long-run and short-run relationships. Al-Mulali and

Ozturk (2015) analyzed the determinants of environmental degradation in the Middle East and

North African countries over the period 1996-2012. They used the Pedroni cointegration test

and found that ecological footprint, energy consumption, urbanization, trade openness,

industrial development and political stability are cointegrated. In addition, the Granger

causality test revealed short-run and long-run causal relationships between the used variables

and the ecological footprint. Kasman and Duman (2015) estimated a panel error correction

model and found a short-run unidirectional panel causality running from urbanization to CO2

emissions for new European Union members and candidate countries over the period 1992–

2010. Al-mulali et al. (2012) investigate the long-run relationship between urbanization,

8
energy consumption and CO2 emissions in seven regions over the period 1980-2008. Using

the Fully Modified Ordinary Least Square model, they found a positive long-run relationship

between urbanization and CO2 emissions in 84% of countries, while 16% of countries had

mixed results. Some countries of their sample have a negative long-run relationship and

others, especially low-income countries have no relationship between urbanization and CO2

emissions. Sharif-Hossain (2011) studied 9 new industrialized countries over the period 1971-

2007 and found the existence of a positive long-run cointegrating vector between CO2

emissions, urbanization, energy consumption and trade openness.

3. Data and descriptive statistics

This paper uses an unbalanced panel dataset of 142 countries classified into five

different income levels (Low-income, Lower-middle-income, Upper-middle-income, High

income non-OECD, High-income OECD). Data on CO2 emissions, urbanization, and

industrialization are obtained from the World Development Indicators database (WDI).

Industrialization is measured as the industry value added (% of GDP). CO2 emissions2 are

measured in metric tons per capita, urbanization is measured by the population in urban

agglomerations of more than 1 million (% of total population). Appendix A provides a

detailed description of the variables and data sources used in this study.

Appendix B presents descriptive statistics for the three variables in the different

income level groups. The statistics show that the high-income non-OECD countries present

the highest standard deviation for the three variables respectively with 14.968 metric tons per

capita for CO2 emissions, 28.987% for urbanization and 21.142% for industrialization.

2
CO2 emissions are those stemming from the burning of fossil fuels and the manufacture of cement,

consumption of solid, liquid, and gas fuels and gas flaring.

9
Furthermore this group of countries has the highest average CO2 emissions (11.294 metric

tons per capita), highest average for urbanization (71.807%) and the highest average for

industrialization (30.002%). The second group of countries that emit more CO2 are OECD

countries with 8.497 metric tons per capita followed by the upper-middle-income countries.

We remind that China is one of the most polluting countries in the world that belongs to

upper-middle-income countries. It is important to note that the group of countries that emit

less CO2 (0.161 metric tons per capita) is low-income countries. They have the lowest level

of urbanization (7.408%) and industrialization (18.842%).

Appendix C displays Pearson’s correlation matrix between variables. OECD countries

show the highest CO2 emissions correlation (76%) with industrialization followed by the

high-income non-OECD countries.

4. Empirical strategy

This paper essentially uses the newly developed panel cointegration technique to

estimate the relationship between urbanization ( Urban ), industrialization ( Industry ) and

CO2 emissions ( CO2 ) in different income-level countries. Panel cointegration analysis is

equivalent to the steps of time series analysis: (i) unit root testing, (ii) cointegration testing,

and (iii) estimation of the long-run and short-run relationships.

4.1. Panel Unit Root Test (PURT)

The first step in panel cointegration analysis requires testing the stationarity properties

of the variables since most macroeconomic variables exhibit trends in them. Unit root tests

seek to know if there are restrictions on the autoregressive process across cross-sections or

series. Consider the following AR(1) process for panel data:

10
 yit = ρi yit −1 + X itδ i + ε it (1)
Where i = 1, 2,..., N are cross-section units or series observed over periods t = 1, 2,..., T .

X it represents the exogenous variables in the model, including fixed effects or individual

trends, ρi are the autoregressive coefficients, and the errors ε it are assumed to be mutually

independent idiosyncratic disturbance. If ρ i < 1 , yit is said to be weakly (trend) stationary.

Furthermore, if ρ i = 1 , then yit contains a unit root.

Panel unit root tests are grouped into two main categories: (i) first-generation tests,

which assume cross-sectional independence [e.g., Maddala and Wu (1999); Choi (2001),

Levin, Lin and Chu (2002); Im, Pesaran, and Shin (2003)] and (ii) second generation tests,

which explicitly allow for some form of cross-sectional dependence [e.g. Pesaran (2004,

2007)]. This article relies on the Im, Pesaran, and Shin (IPS) and the Levin, Lin, and Chiu

(LLC) tests for panel unit roots since they require N to be small enough relative to T , while

the LLC test also requires a strongly unbalanced panel (Baltagi, 2008).

4.2. Cointegration test

If the variables under consideration are found to be integrated of order 1, then one

should use panel cointegration tests to determine whether a long-run equilibrium relationship

exists among the nonstationary variables in their level form. The existence of a cointegrated

relationship between the series allows us to assume a potential long-run relationship between

them. To examine the possible existence of one or more cointegrated relationships among the

series considered, we first perform Pedroni’s well-known test. The method of Pedroni (1999,

2004) is analogous to the Engle and Granger (1987) test which is usually computed in time

series studies and tests the presence of a unit root of the residuals from the following form:

yit = α i + wit' β1 + uit β 2 + ε it (2)

11
where yit denotes the endogenous variable (CO2 emissions in this case), α is a fixed effect

dealing with the unobserved heterogeneity between the groups of countries considered; u and

w are vectors of urbanization and industrialization, respectively, and ε it the error term. The

main advantage of the Pedroni method, unlike Kao’s (1999), is to take into account the

heterogeneity under the alternative hypothesis for the three between statistics by assuming

parameters that can differ across individuals (i.e across different income level of countries in

this study).

4.3. Estimation of the short-run and long-run relationship

Once cointegration is ascertained, it becomes interesting to estimate efficiently the

long-run economic relationships amongst the variables. Thus, a number of panel estimators

have been suggested in the literature. However, this article uses the Pooled Mean Group

(PMG) estimator (Pesaran et al., 1999) for dynamic heterogeneous panels. The test procedure

fits an autoregressive distributed ARDL model to the data, which can be respecified as an

error correction equation to facilitate economic interpretation. Consider the following error

correction representation of a lag ARDL( p, q, q,..., q ) model:

q −1 p −1
'
yit = φi ECit + ∑ ∆X it − j βi , j + ∑ λ *i , j ∆yit − j + ε it (3)
j =0 j =1

Where X ' is a vector of explanatory variables, βi contains information about the long-run

impacts, φi is the error correction term (due to normalization), and λi*, j incorporates short-run

information. Equation (3) can be written in this form:

p −1 q =1
∆CO2it = φi ECit + ∑ Γi ∆CO2it−k + ∑ λij ∆industryit − j + ∑ δ ij' ∆urbanit − j + ε it (4)
j =1 j =0

Where Γi are the coefficients of lagged CO2, λij and δ ij' are respectively the

coefficients of industrialization and urbanization. The PMG estimator can be seen as an

12
intermediate procedure somewhere between the mean group (MG) estimator and the dynamic

fixed-effects (DFE) approach. The MG estimator is obtained by estimating N independent

regressions and then averaging the (unweighted) coefficients, while the DFE requires pooling

the data and assuming that the slope coefficients and error variances are identical. However,

the PMG restricts the long-run coefficients to be the same ( β = βi for all i ) but allows the

short-run coefficients and error variances to vary across countries (Pesaran et al., 1999). This

approach can be used whether the regressors are I(0) or I(1) (Pesaran et al., 1999).

Estimating the model in Equation (4) is not without problems, notably, that of reverse

causality from urbanization, industrialization to CO2 emissions. However, as argued by

Pesaran (1997) and Pesaran et al. (1999), the long-run coefficient matrix can be consistently

estimated by the pooled mean group estimator for heterogeneous panels; but consistent

estimation of the short-run coefficients requires that the regressors and the error terms be

independent of each other. Therefore, the short-run results reported in this article are to be

interpreted with caution.

5. Empirical results and discussion

5.1. Panel Unit Root Test Results

We begin the empirical analysis with the application of panel unit root tests to verify

whether the variables are nonstationary. As explained in section 4, we rely on Levin, Lin &

Chu (LLC), Breitung t-stat (B), Im, Pesaran and Shin W-stat (IPS), ADF - Fisher Chi-square

(ADF), and PP - Fisher Chi-square(PP) panel unit root tests on the variables over 1960–2014.

The LLC test is based on the common unit root process assumption that the autocorrelation

coefficients of the tested variables across cross sections are identical. However, the IPS test

relies on the individual unit root process assumption that the autocorrelation coefficients vary

across cross sections.

13
 In each PURT, with the aim of identify the deterministic components, we include

individual intercept, individual intercept and trend, no intercept and no trend in order to

minimize problems arising from cross-sectional dependence. The Schwarz-Bayesian

Information Criterion (BIC) is used to determine the country-specific lag length for the ADF

regressions, with a maximum lag of 3. Further, the Bartlett kernel is used to estimate the long-

run variance in the LLC test, with the maximum lags determined by the Newey-West

bandwidth selection algorithm. The panel unit root test results3 presented in Tables 1 to 3

show stationarity for certain variables in level like Industry for low-income. Overall, panel

unit root test shows evidence of stationarity especially in first difference for Industry , CO2 ,

and Urban for each countries’ group.

Table 1. Panel Unit Root Tests Results for Industry with Individual Intercept and Trend

Level
Low-
Test Lower-middle-Income Upper-middle-income High-income non-OECD OECD
income
Levin, Lin & Chu t* -2.25*** -7.43*** 1.21 -0.56 -0.55
Breitung t-stat -2.67*** -2.75*** -0.97 -1.20 -0.85
Im, Pesaran and
-1.81*** -4.31*** 1.71 -0.35 -1.14
Shin W-stat
ADF - Fisher Chi-
77.08*** 152.31*** 37.01 55.97 60.95
square
PP - Fisher Chi-
77.91*** 196.49*** 54.98 65.01 61.33
square
First difference
Low-
Test Lower-middle-income Upper-middle-income High-income non-OECD OECD
income
Levin, Lin & Chu t* -28.44*** -9.81*** -14.42*** -17.99*** -18.96***
Breitung t-stat -14.52 -1.10 -8.55 -12.84*** -13.65***
Im, Pesaran and
-25.84*** -15.91*** -15.02*** -17.66*** -15.78***
Shin W-stat
ADF - Fisher Chi-
575.23*** 417.85*** 323.41*** 392.73*** 325.24
square
PP - Fisher Chi-
829.18*** 463.33*** 693.45*** 1412.71*** 648.85***
square
Automatic selection of lags based upon SIC: 0 to 6 (maximum lags). Newey –West bandwidth selection using Bartlett
Kernel. *, ** and *** denote significance at 10, 5 and 1% levels, respectively.

3
We have reported in Tables 1 to 3 only individual intercept and trend Unit Root results for the 5 groups of

countries. The results for individual intercept and no intercept no trend could be provided on request.

14
Table 2. Panel Unit Root Tests Results for CO2 with Individual Intercept and Trend

Level
Low- Lower-middle- Upper-middle- High-income non-
Test OECD
income income income OECD
Levin, Lin & Chu t* 0.07 0.59 0.92 3.17 -0.36
Breitung t-stat 0.65 -0.09 1.21 4.12 6.41
Im, Pesaran and Shin
-0.05 0.36 1.77 2.10 1.73
W-stat
ADF - Fisher Chi-
72.16 77.63 40.83 47.04 43.63
square
PP - Fisher Chi-
73.41 92.62** 41.19 90.42*** 47.75
square
First difference
Low- Lower-middle- Upper-middle- High-income non-
Test OECD
income income income OECD
Levin, Lin & Chu t* -38.46*** -23.07*** -15.54*** -14.14*** -12.53***
Breitung t-stat -18.56*** -9.88*** -7.51*** -8.55*** -9.38***
Im, Pesaran and Shin
-36.66*** -22.78*** -16.91*** -18.39*** -15.40***
W-stat
ADF - Fisher Chi-
889.51*** 545.63*** 356.65*** 408.52*** 316.46***
square
PP - Fisher Chi-
914.56*** 887.45*** 756.93*** 1243.22*** 693.23***
square
Automatic selection of lags based upon SIC: 0 to 6 (maximum lags). Newey –West bandwidth selection using Bartlett
Kernel. *, ** and *** denote significance at 10, 5 and 1% levels, respectively.

Table 3. Panel Unit Root Tests Results for Urbanization with Individual Intercept and
Trend

Level
Low- Lower-middle- Upper-middle- High-income non-
Test OECD
income income income OECD
Levin, Lin & Chu t* 7.12 -6.99*** -4.86*** -5.11*** -12.72***
Breitung t-stat 24.61 -1.04 -1.12 -3.96*** 3.93
Im, Pesaran and Shin 11.12 -5.11*** -0.29 0.46 -2.82***
ADF - Fisher Chi-
22.92 135.36*** 65.77* 63.02 231.52***
square
PP - Fisher Chi-square 17.71 31.79 23.74 24.03 67.13**
First difference
Low- Lower-middle- Upper-middle- High-income non-
Test OECD
income income income OECD
Levin, Lin & Chu t* -2.14*** -5.01*** -1.23* -1.69** -2.13***
Breitung t-stat -6.04*** -6.91*** -4.60*** -4.42***
Im, Pesaran and Shin -3.92*** -7.22*** -2.54*** -3.01*** -9.29***
W-stat
ADF - Fisher Chi-
86.88*** 164.49*** 79.18*** 91.05*** 243.21***
square
PP - Fisher Chi-square 113.81*** 227.86*** 73.77*** 77.35*** 102.39***
Automatic selection of lags based upon SIC: 0 to 6 (maximum lags). Newey –West bandwidth selection using Bartlett
Kernel. *, ** and *** denote significance at 10, 5 and 1% levels, respectively.

15
 5.2. Panel cointegration tests results

To preclude spurious regression, we perform a cointegration test to ascertain a stable

equilibrium relationship. Table 4 points out that all the Pedroni’s cointegration4 results and

all statistics clearly reject the null hypothesis of no cointegration. Overall, we could accept the

existence of a cointegrated relationship between variables because most of the statistics are in

favor of cointegration and the two Pedroni’s tests applying the Augmented Dickey–Fuller

(ADF) principle are very robust and outperform the others (Wagner and Hlouskova, 2007).

Moreover, the panel ADF test has the best size and size-adjusted power properties among all

the Pedroni statistics (Örsal, 2008).

The main advantage of the Pedroni’s method, unlike Kao’s (1999), is to take into

account the heterogeneity under the alternative hypothesis for the three between statistics by

assuming parameters that can differ across individuals (Countries in each countries’ income-

level groups). According to our cointegration result, we can assume the existence of a long-

run equilibrium relationship between Industry , CO2 , and Urban in Pedroni’s cointegration

model with individual intercept and individual trend.

In order to test the presence of a long-run relationship between CO2 emissions,

industrialization and urbanization, it is important to ensure that no restrictions on the

coefficients of the lagged terms are imposed. Otherwise, the test results can be subject to a

pretesting problem. However, to estimate the level effects and short-run dynamics between

CO2 emissions, industrialization and urbanization adjustments, the use of a parsimonious

specification is more desirable. First, the lag orders of the ARDL with the variables ( CO2

emissions, industrialization and urbanization) were selected by searching across the 43 = 64

4
Only individual intercept and individual trend results are reported in table 4. Individual intercept test, no

intercept and no trend results could be provided on request.

16
ARDL, spanned by pi = 1 to 4, i = 0,1, 2, Akaike Information Criterion (AIC). Such a

general model is then tested down by dropping statistically insignificant terms [Konya (2000);

Wolde-Rufael (2004, 2005)]. However, given the uncertainty for the lag length and that

ARDL regression is sensitive to lag length and in order to the test the robustness of our

results, tests were carried out by varying lag from 4 to 1. For the optimal lag selected, the final

parsimonious model must satisfy the residuals that checked for white noise using AIC, the

Schwarz Criterion (BIC), or the Hannan-Quinn (HQ) criterion. This is justified by the fact

that we have different optimal lag ARDL models for each country group.

Table 4. Panel cointegration testing

Test Lower-middle- Upper-middle- High-income non-

Low-income income income OECD OECD
Panel v-Statistic -5.12 -8.41 1.85*** -7.05 -4.54
Panel rho-Statistic -14.11*** -3.76*** -17.14*** -9.60*** -11.54***
Panel PP-Statistic -45.26*** -39.58*** -24.25*** -20.92*** -33.05***
Panel ADF- -2.25*** -21.49*** -22.12*** -15.86*** -13.15***
Statistic
Group rho-Statistic -9.54 -7.69*** -15.08*** -13.49*** -7.15***
Group PP-Statistic -37.44 -42.64*** -28.63*** -53.49*** -33.52***
Group ADF- -13.45 -28.41*** -22.47*** -19.96*** -10.66***
Statistic
*, ** and *** denote significance at 10, 5 and 1% levels, respectively.

Having established a unique cointegrating relationship among the variables of interest,

we apply the PMG estimator which uses the panel extension of the single equation ARDL

model to investigate the short-run and long-run relationship between CO2 emissions,

industrialization and urbanization, and to ensure robustness of the analysis. This resulted in

the choice of ARDL specification reported in Table 5.

The results show a negative and significant coefficient of the ECM for all groups of

countries. This indicates that any short-term fluctuations between CO2 emissions,

urbanization and industrialization will give rise to a stable long-run relationship between the

all the variables.

17
 In low-income countries and in short-run, there is no significant impact of

industrialization and urbanization on CO2 emissions. However, in long-run, the PMG

estimation shows a positive impact of urbanization and industrialization on CO2 emissions for

that group of countries. The coefficient of urbanization is 0.01 with significance at the 5%

level and the coefficient of industrialization is very weak and tends to zero with 1%

significance. According to Jones (1991) and Parikh and Shukla (1995), the processes of

industrialization and urbanization accompany each other and are part of economic

development. These processes affect emissions through several channels such as scale of

production, food delivery, infrastructure, and change in domestic activity. In low-income

countries, food processing, especially non-exported food is traditional and mainly uses human

or animal energy. Urbanization leads to larger-scale and more efficient food processing for

urban markets requiring additional energy-using which contributes to the emission of

greenhouse gases. The concentration of people in cities requires construction and maintenance

of transportation, sanitation and water infrastructure. The construction materials, particularly

structural steel and cement are energy intensive and contribute to emissions. Thus, in low-

income countries, changes in domestic activity can have important impact on energy use and

therefore, on CO2 emissions in long-run.

In lower-middle income countries, only industrialization is lightly and positively

correlated with CO2 emissions in long-run with 10% level significant. India which is ranked

4th World countries most polluter belongs to this group.

In upper-middle income countries and in short-term, only urbanization has a

significant relationship with CO2 emissions. However, in long-run industrialization and

urbanization affect positively and significantly CO2 emissions. The coefficients are 0.07 and

0.01 respectively for urbanization and industrialization with 1% and 5% significance and the

18
magnitude of their coefficients is larger than that of the low-income countries. China which is

one of belonging countries in upper-middle income plays a dominant role in international

climate policy since it is not only the most populous country in the world (with more than 1.3

billion inhabitants) but has also overtaken the United States as the worldwide biggest

producer of CO2 emissions and even has surpassed the European Union in per-capita

emissions (Le Quéré et al., 2014).

In high-income non-OECD countries, industrialization and urbanization have

significant correlation with CO2 emissions only in short-run with coefficients (1.51; 0.17) at

10% and 5% significance respectively.

Concerning OECD countries, our results show that in long-run they exhibit the

highest coefficient (0.04) of correlation between industrialization and CO2 emissions at 5%

level compared to other group of countries. This result is similar to that of Salim and Shafiei

(2014) who have found that industrialization increases CO2 emissions in OECD countries.

However, urbanization does not cause pollutant emissions in long-run. In short-run,

industrialization and urbanization significantly give rise to CO2 emissions in OECD

countries. It should be noted that urbanization in 1-period lag affects positively CO2

emissions and the coefficient is 1.70 at 1% level significance. Urbanization leads to relative

concentration of population as well as economic activities in OECD countries’ urban areas.

As a result of migration from rural to urban areas, the labor force is transferred from the

agricultural sector in the rural areas to the industrial and service sectors in the urban areas.

This structural transformation of the OECD’s economy causes many fundamental changes in

natural resources and energy use as well. In OECD’s countries, the transformation of

production from the low-energy intensive agricultural sector to the high-energy intensive

industrial sectors could lead to the introduction of new industrial technologies, thus

19
contributing to CO2 emissions. Moreover, urban living generally requires more energy

consumption for travelling to work by fuel-using vehicles, maintaining urban infrastructure

and services including housing, water supply, roads and bridges [Jones (2004); Madlener and

Sunak (2011); Parikh and Shukla (1995)] that increase CO2 emissions (Salim and Shafiei,

2014). Our results also show that in short-run, industrialization without lag and with one lag

positively affect emissions. A 1% increase in the proportion of industry added value (% GDP)

without lag and with 1-period lag increases per capita emissions by 7% and 12% respectively.

The process of industrialization increases energy consumption, denoting that industrialization

is an important factor behind increased emissions in OECD countries. Therefore, combating

climate change caused by emissions from OECD countries, securing and diversifying the

supply of energy mix have been of heightened interest in promoting renewable energy sources

in OECD countries in recent years. This growing interest could be supported by various

government incentive policies such as feed-in tariff, subsidies for renewable technologies and

tax rebate.

20
 Table 5. Lag ARDL Estimation Results for CO2 emissions (dependent variable)

High-income
Groups of countries Low-Income Lower-middle income Upper-middle income non-OECD OECD
Variables Long-Run
0.07(2.36)**
∆ Urban 0.01(2.38)** -0.00(-0.18) 0.02(0.78) -0.09(-0.81)

∆ Industry 0.00(3.12)*** 0.00(1.85)* 0.01(3.59)*** -0.01(-0.62) 0.04(2.01)**

Short-Run
-1.01(-8.14)*** -1.02(-16.08)***
ECM -1.31(-11.67)*** -1.19(-23.56)*** -1.35(-19.22)***
0.01(0.13)
∆ CO2 (-1) 0.18(3.27)*** -0.52(-1.06)
0.03(0.82)
∆ CO2 (-2) 0.04(0.67) -0.11(-0.58)
0.53(1.67)*
∆ Urban -0.05(-0.68) 0.00(0.166) -0.31(-1.03) -0.24(-0.17)
-0.28(-0.99) 1.51(1.68)*
∆ Urban (-1) 1.70(1.86)***
0.37(2.31)**
∆ Urban (-2) -1.78(-1.59)
-0.01(-1.47)
∆ Industry -0.00(-0.29) 0.00(1.52) 0.10(1.66) 0.07(1.78)*
0.02(1.03) 0.17(1.96)**
∆ Industry (-1) 0.12(2.22)**
0.02(0.92) -0.01(-0.082)
∆ Industry (-2) 0.07(1.00)
C -0.03(-0.93) -0.08(-1.13) 0.02(1.15) -4.16(-0.81) 0.63(1.72)
0.25(0.80)
Trend 0.00(0.96) 0.00(1.41) 0.04(0.29) -0.02(-1.95)**
Log likelihood 1853.53 1388.80 417.59 -1229.11 -82.97
∆ denotes the first order difference operator. All of the models are based on AIC Criterion. The lag-length is based on AIC
Criterion. *, ** and *** denote significance at 10, 5 and 1% levels, respectively.

5.3. Granger causality test

In order to analyze the causal relationship, we apply the Granger causality test

(Granger 1969). The results reported in Table 6 show that there is no granger causality

between CO2 emissions, industrialization and urbanization in low-income and in lower-

middle-income countries. Nevertheless, concerning upper-middle-income countries, there is a

strong bidirectional causal relationship between industrialization and CO2 emissions at 1%

significance. Industrialization in these countries generates CO2 emissions and reversely. In

High income non-OECD, there is significant feedback causality between industrialization and

urbanization. Urbanization leads to CO2 emissions at 5% level significance and

21
industrialization generates CO2 emissions. In OECD countries, urbanization granger-causes

CO2 emissions and urbanization instigates industrialization. These results highlight the

heterogeneity of the causal relationship between urbanization, industrialization and CO2

emissions depending on economic development level. Thus, Granger causal relationship is

significant especially in upper-middle, high-income non-OECD and OECD countries.

Table 6. Granger causality between variables CO2 emissions, urbanization and

industrialization

Groups of D(Urban) D(CO2) D(Industry) D(CO2) D(Industry) D(Urban)

Variables
countries ->D(CO2) ->D(Urban) ->D(CO2) ->D(Industry) ->D(Urban) ->D(Industry)
Low- Observations 990 973 654
income F-Statistic 0.023 0.063 0.833 1.460 0.042 0.205
P-Value 0.878 0.801 0.362 0.227 0.838 0.651
Lower- Observations 1207 942 1045
middle- F-Statistic 0.831 1.260 1.076 1.233 0.37798 0.556
income P-Value 0.527 0.278 0.372 0.291 0.8640 0.733
Upper- Observations 1215 930 969
middle- F-Statistic 1.009 1.012 0.134 0.974 0.831 1.260
income P-Value 0.365 0.364 0.000*** 0.051** 0.874 0.377
High-income Observations 1207 942 1045
non-OECD F-Statistic 4.295 1.729 3.113 0.412 2.42891 7.951
P-Value 0.013** 0.178 0.044** 0.662 0.088* 0.001***
OECD Observations 1162 622 671
F-Statistic 7.642 0.792 0.861 1.098 0.259 5.318
P-Value 0.001*** 0.454 0.423 0.334 0.771 0.005***
*, ** and *** denote significance at 10, 5 and 1% levels, respectively.

6. Conclusion

CO2 emissions are an important concern of global warming and climate change. As

part of economic growth and development process, urbanization and industrialization are

generally energy intensive and may result in CO2 emissions. Thus, it is crucial to understand

how these factors can affect countries’ CO2 emissions.

22
 This paper has been structured around two main questions: is the relationship between

urbanization, industrialization and CO2 emissions homogeneous for all countries at all stages

of economic development? Is the relationship varies throughout the time? To answer these

questions, we have used a large panel data set covering 142 countries over the period 1960-

2014 to analyze the cointegration relationship between industrialization, urbanization and

CO2 emissions. Our empirical analysis has focused on heterogeneity effects in term of

countries’ income level as well as long-run and short run effects. In order to take into account

income level heterogeneity effects, the sample has been divided into five income levels

groups. After panel unit root tests, we used the PMG estimator (Pesaran et al., 1999) for

dynamic heterogeneous panels to fit an ARDL model.

Our empirical analysis brings out three main interesting results. The first important

result is that the impact of industrialization and urbanization on CO2 emissions varies

according to countries’ income levels. The second result is the heterogeneity of short-run and

long-run effects. More specifically, in low-income countries, we found no significant

relationship between industrialization and urbanization on CO2 emissions in short-run; while

in long-run, urbanization and industrialization positively correlated with CO2 emissions but

the effect of industrialization is very weak. In lower-middle-income countries, only

industrialization is lightly and positively correlated with CO2 emissions in long-run. Our

results for upper-middle income countries show that only urbanization has a significant

correlation with CO2 emissions in short-run, but in long-run, both industrialization and

urbanization are positively and significantly correlated with CO2 emissions. In high-income

non-OECD countries, industrialization and urbanization in 1-period lag have significant

correlation with CO2 emissions in short-run. In OECD countries, only industrialization has a

positive and significant effect on CO2 emissions in long-run with a higher coefficient

compared to others groups of countries. Urbanization doesn’t increase CO2 emissions in long-

23
run for OECD countries. In short-run, urbanization in 1-period lag and industrialization

positively affects CO2 emissions. The third result of this paper is that the granger-causal

relationship between urbanization, industrialization and CO2 emissions varies across

countries having different economic development level.

These findings have important policy implication. Urbanization and industrialization

strategies must take into account countries’ income levels at each stage of development as

well as long-run and short-run effects to mitigate CO2 emissions in environmental

management and sustainable development perspective.

24
7. Appendices

Appendix A: Description of variables and data sources

Variables Variables Name Source
Industry Industry, value added (% of GDP) WDI
CO2 emissions CO2 emissions (metric tons per capita) WDI
Urbanization Population in urban agglomerations of more than 1 million (% of total population) WDI

Appendix B: Descriptive statistics

Groups of countries Statistics CO2 Urban Industry
Mean 0.161 7.408 18.842
Low-income Std. Dev. 0.30 5.031 8.231
Observations 648 648 648
Mean 0.876 14.889 27.791
Lower-middle income Std. Dev. 1.232 7.757 10.569
Observations 1029 1029 1029
Mean 2.811 22.538 33.717
Upper-middle income Std. Dev. 2.158 9.452 10.189
Observations 976 976 976
Mean 11.294 71.807 30.002
High-income non-OECD Std. Dev. 14.968 28.987 21.142
Observations 1050 1050 1050
Mean 8.497 27.046 30.898
OCDE Std. Dev. 3.588 15.948 5.556
Observations 683 683 683

Appendix C: Pearson’s correlation between industrialization, urbanization and CO2

Groups of countries Variables Industry Urban CO2
Industry 1
Low-income Urban 0.577** 1
CO2 0.388** 0.087** 1
Industry 1
Lower-middle-income Urban 0.351** 1
CO2 0.234** 0.190** 1
Industry 1
Upper-middle-income Urban -0.327** 1
CO2 0.260** 0.001 1
High-income non-OECD Industry 1
Urban -0.158** 1
CO2 0.419** -0.459** 1
Industry 1
OECD Urban 0.139** 1
CO2 0.761** 0.252** 1
*, ** and *** denote significance at 10, 5 and 1% levels, respectively.

25
Appendix D: List of countries by income level
Lower-Middle Upper-Middle High-Income Non
Low-Income OECD
Income Income OECD
Afghanistan Bangladesh Algeria Andorra Australia
Benin Bolivia Angola Antigua and Barbuda Austria
Burkina Faso Cameroon Belize Argentina Belgium
Burundi Congo. Rep. Brazil Aruba Canada
Cambodia Cote d'Ivoire Bulgaria Bahamas. The Chile
Central African Republic Egypt. Arab Rep. China Bahrain Czech Republic
Chad El Salvador Colombia Barbados Denmark
Comoros Georgia Costa Rica Bermuda Estonia
Congo. Dem. Rep. Ghana Cuba Brunei Darussalam Finland
Eritrea Guatemala Dominican Republic Cayman Islands France
Ethiopia Honduras Ecuador Croatia Germany
Gambia. The India Iran. Islamic Rep. Cyprus Greece
Guinea Indonesia Iraq Equatorial Guinea Hungary
Guinea-Bissau Kenya Jordan Faeroe Islands Ireland
Haiti Morocco Lebanon French Polynesia Italy
Korea. Dem. Rep. Myanmar Malaysia Greenland Japan
Liberia Nigeria Mexico Hong Kong SAR. China Korea. Rep.
Madagascar Pakistan Mongolia Kuwait Netherlands
Malawi Philippines Panama Macao SAR. China New Zealand
Mali Senegal Paraguay Malta Poland
Mozambique Sudan Peru Oman Portugal
Nepal Syrian Arab Republic Romania Puerto Rico Spain
Niger Ukraine South Africa Qatar Sweden
Rwanda Uzbekistan Thailand Russian Federation Switzerland
Sierra Leone Vietnam Tunisia Saudi Arabia United Kingdom
Somalia Yemen. Rep. Turkey Seychelles United States
South Sudan Zambia Singapore
Tanzania St. Kitts and Nevis
Togo Trinidad and Tobago
Uganda United Arab Emirates
Zimbabwe Uruguay
Venezuela. RB

26
References

1. Al-Mulali U., Ozturk I., 2015. The effect of energy consumption, urbanization, trade
openness, industrial output, and the political stability on the environmental
degradation in the MENA (Middle East and North African) region. Energy 84, 382-
389.
2. Al-Mulali U., Binti Che Sab C.N., Fereidouni HG., 2012. Exploring the bi-directional
long run relationship between urbanization, energy consumption, and carbon dioxide
emission. Energy 46, 156-167.
3. Alam S., Fatima A., Butt M.S., 2007. Sustainable development in Pakistan in the
context of energy consumption demand and environmental degradation. Journal of
Asian Economics 18(5), 825-837.
4. Ang, J.B., 2007. CO2 emissions, energy consumption, and output in France. Energy
Policy 35 (10), 4772-4778.
5. Asane-Otoo E., 2015. Carbon footprint and emission determinants in Africa. Energy
82, 426-435.
6. Baltagi, B. H., 2008. Econometric Analysis of Panel Data. (Fourth Edition), John
Wiley & Sons, Chichester.
7. Burton, E., 2000. The compact city: just or just compact? A preliminary analysis.
Urban Studies. 37, 1969–2001.
8. Capello, R., Camagni, R., 2000. Beyond optimal city size: an evaluation of alternative
urban growth patterns. Urban Studies. 37, 1479–1496.
9. Chen, H., Jia, B., Lau, S.S.Y., 2008. Sustainable urban form for Chinese compact
cities: challenges of a rapid urbanized economy. Habitat International 32, 28–40.
10. Cherniwchan, J., 2012 Economic growth, industrialization, and the environment.
Resource and Energy Economics 34(4), 442–467.
11. Choi, I., 2001. Unit Root Tests for Panel Data. Journal of International Money and
Finance 20, 249-272.
12. Cole M.A, Neumayer E., 2004. Examining the impact of demographic factors on air
pollution. Population and Environment 26(1), 5–21.
13. Crenshaw, E.M., Jenkins, J.C., 1996. Social structure and global climate change:
sociological propositions concerning the greenhouse effect. Sociological Focus 29,
341–358.
14. Ehrlich Paul R.; Holdren John P., 1971. Impact of Population Growth Ehrlich Science.
New Series 171(3977), 1212-1217.
15. Engle, R. and Granger, C. 1987. Co-integration and Error-Correction: Representation,
Estimation and Testing. Econometrica 55, 251-276.
16. Fan, Y., Liu, L., Wu, G., Wei, Y., 2006. Analyzing impact factors of CO2 emissions
using the STIRPAT model. Environmental Impact Assessment Review 26, 377–395.
17. Granger, C. W., 1969. Investigating Causal Relations by Econometric Models and
Cross-Spectral Methods. Econometrica 37(3), 424-438.
18. Gouldson, A.P., Murphy, J., 1997. Ecological modernization: economic restructuring
and the environment. The Political Quarterly 68 (5), 74–86.
27
19. IPCC, 2014. Climate Change 2014 Mitigation of Climate Change Working Group III
Contribution to the Fifth Assessment Report of the Intergovernmental Panel on
Climate Change. https://www.ipcc.ch/pdf/assessment-
report/ar5/wg3/ipcc_wg3_ar5_full.pdf
20. Im, K.S., Pesaran, M.H. and Shin, Y., 2003. Testing for Unit Roots in Heterogenous
Panels. Journal of Econometrics 115, 53-74.
21. Jenks, M., Burton, E., Williams, K. (Eds.), 1996. The Compact City: A Sustainable
Urban Form? E & FN Spon, New York.
22. Jones D.W., 1991. How urbanization affects energy-use in developing countries.
Energy Policy, 621–630.
23. Kao, C., 1999. Spurious Regression and Residual-Based Tests for Cointegration in
Panel Data. Journal of Econometrics 90, 1–44.
24. Kasman A, Duman Y.S. 2015. CO2 emissions, economic growth, energy
consumption, trade and urbanization in new EU member and candidate countries: a
panel data analysis. Economic Modelling 44, 97–103.
25. Kónya L. (2000). Bivariate causality between immigration and long-term
unemployment in Australia, 1981-1998, Working Paper,Victoria University, School of
Applied Economics, 18,00-37.
26. Le Quéré C. et al. (2014). Global carbon budget, Earth System Science Data, 7, 521-
610
27. Lee S, Lee B., 2014. The influence of urban form on GHG emissions in the U.S.
household sector. Energy Policy 68, 534–549.
28. Levin, A., C. Lin, and Chu, C.J., 2002. Unit Root Tests in Panel Data: Asymptotic
and Finite-sample Properties. Journal of Econometrics 108, 1-24.
29. Li, K.; Lin, B., 2015. Impacts of urbanization and industrialization on energy
consumption/CO2 emissions: Does the level of development matter? Renewable and
Sustainable Energy Reviews 52, 1107–1122.
30. Li H., Mu H., Zhang M., Gui S., 2012. Analysis of regional difference on impact
factors of China's energy – related CO2 emissions. Energy 39(1), 319–326.
31. Liddle, B., 2004. Demographic dynamics and per capita environmental impact: using
panel regressions and household decompositions to examine population and transport.
Population and Environment 26, 23–39.
32. Maddala, G.S. Wu, S., 1999. A Comparative Study of Unit Root Tests with Panel
Data and a new simple test, Oxford Bulletin of Economics and Statistics 61, 631-652.
33. Madlener R, Sunak Y., 2011. Impacts of urbanization on urban structures and energy
demand: what can we learn for urban energy planning and urbanization management?
Sustainable Cities and Society 1(1), 45–53.
34. Martínez-Zarzoso I., Maruotti A., 2011. The impact of urbanization on CO2
emissions: evidence from developing countries. Ecological Economics 70, 1344–1353.
35. McGranahan, G., Jacobi, P., Songsore, J., Surjadi, C., Kjellen, M., 2001. The Citizen
at Risk: From Urban Sanitation to Sustainable Cities. Earthscan, London.
36. Mol, A.P.J., Spaargaren, G., 2000. Ecological modernization theory in debate: a
review. Environmental politics 9(1), 17–49.
28
37. Nejat P., Jomehzadeh F., Taheri M. M., Gohari M., Abd, Majid M.Z., 2015 A global
review of energy consumption, CO2 emissions and policy in the residential sector
(with an overview of the top ten CO2 emitting countries). Renewable and Sustainable
Energy Reviews 43, 843–862.
38. Newman, P.W.G., Kenworthy, J.R., 1989. Cities and Automobile Dependence: An
International Sourcebook. Aldershot, UK: Gower.
39. Örsal, D. K., 2008. Comparison of panel cointegration tests. Economics Bulletin, 3, 1–
20.
40. Parikh, J., Shukla, V., 1995. Urbanization, energy use and greenhouse effects in
economic development: results from a cross-national study of developing countries.
Global Environmental Change 5 (2), 87–103.
41. Pedroni, P., 2004. Panel Cointegration; Asymptotic and Finite Sample Properties of
Pooled Time Series Tests with an Application to the PPP Hypothesis. Econometric
Theory 20, 597- 625.
42. Pedroni, P. 1999. Critical Values for Cointegration Tests in Heterogeneous Panels
with Multiple Regressors. Oxford Bulletin of Economics and Statistics 61, 653–670.
43. Pesaran, H., 2007. A Simple Panel Unit Root Test in the Presence of Cross-Section
Dependence. Journal of Applied Econometrics 22(2), 265-312.
44. Pesaran, H. 2004. General Diagnostic Tests for Cross Section Dependence in Panels.
Cambridge Working Papers in Economics 0435, Faculty of Economics, University of
Cambridge.
45. Pesaran M.H., Shin Y., Smith R.J., 2001. Bounds testing approaches to the analysis of
level relationships. Journal of Applied Econometrics 16, 289–326.
46. Pesaran, M.H., Shin, Y., Smith, R. P., 1999. Pooled Mean Group Estimation and
Dynamic Heterogeneous Panels. Journal of the American Statistical Association
94(446), 621-634.
47. Poumanyvong P., Kaneko S., 2010. Does urbanization lead to less energy use and
lower CO2 emissions? A cross-country analysis. Ecological Economics 70, 434–444.
48. Sadorsky, P., 2014. The effect of urbanization on CO2 emissions in emerging
economies. Energy Economics 41, 147–153.
49. Salim RA, Shafiei S., 2014. Urbanization and renewable and non-renewable energy
consumption in OECD countries: an empirical analysis. Economic Modelling 38, 581–
591.
50. Shahbaz, M.; Salah Uddin, G.; Ur Rehman, I; Imran, K., 2014. Industrialization,
electricity consumption and CO2 emissions in Bangladesh. Renewable and Sustainable
Energy Reviews 31, 575–586.
51. Sharif-Hossain M., 2011. Panel estimation for CO2 emissions, energy consumption,
economic growth, trade openness and urbanization of newly industrialized countries.
Energy Policy 39, 6991–6999.
52. Sharma S.S., 2011. Determinants of carbon dioxide emissions: empirical evidence
from 69 countries. Applied Energy 88, 376–382.

29
53. Tian X., Chang M., Shi F., Tanikawa H., 2014. How does industrial structure change
impact carbon dioxide emissions? A comparative analysis focusing on nine provincial
regions in China. Environmental Science & Policy 37, 243–254.
54. United Nations, 2014. World Urbanization Prospects The 2014 Revision Highlights.
New York. http://esa.un.org/unpd/wup/FinalReport/WUP2014-Report.pdf
55. Wagner, M., Hlouskova, J., 2007. The performance of panel cointegration methods:
results from a large scale simulation study. IHS Economics Series 210.
56. Wolde-Rufael, Y., 2005. Energy Demand and Economic Growth. Journal of Policy
Modeling 27, 891-903.
57. Wolde-Rufael, Y., 2004. Disaggregated industrial energy consumption and GDP: the
case of Shanghai, 1952–1999. Energy Economics 26 (1), 69–75.
58. Xu B., Lin B., 2015. How industrialization and urbanization process impacts on CO2
emissions in China: evidence from nonparametric additive regression models. Energy
Economics 48, 188–202.
59. York, R. 2007. Demographic trends and energy consumption in European Union
Nations, 1960-2025. Social Science Research 36(3), 855-872.
60. York R., Rosa E., Dietz T., 2003a. A rift in modernity? Assessing the anthropogenic
sources of global climate change with the STIRPAT Model . International Journal of
Sociology and Social Policy, 23 (10), 31–51.
61. York R., Rosa E., Dietz T., 2003b. STIRPAT, IPAT and ImPACT: Analytic tools for
unpacking the driving forces of environmental impact. Ecological Economics 46, 351–
365.
62. Zhou X., Zhang J., Li J., 2013. Industrial structural transformation and carbon dioxide
emissions in China. Energy Policy 57, 43–51.

30