Empirics on linkages among industrialization, urbanization, energy consumption, CO2 emissions and economic growth: a heterogeneous panel study of China

Abstract

This is the first attempt to explore linkages among industrialization, urbanization, energy consumption, CO2 emissions, and economic growth based on estimations in simultaneous equations framework. An economic growth model is extended to incorporate industrialization as shift factor and pollutant emissions as determinant of total factor productivity. A country panel of 30 Chinese provinces/cities and three regional panels, for periods 2000–2016, are estimated employing Augmented Mean Group (AMG) estimator as well as Common Correlated Effects Mean Group (CCEMG) estimator that are robust to both cross-sectional dependence and cointegration. The empirical findings are as follows: (i) the impact of urbanization on economic growth varied from negative to neutral to positive for western-, intermediate-, and eastern-economic zone, respectively, and is known as “urbanization ladder effect,” (ii) moving from western to eastern economic zone, as regions develop, industry expansion has more powerful impact on economic growth, and thus, we define it as “industry expansion effect,” (iii) industrialization promotes rapid urbanization in less developed and less employment saturated regions; hence, we name it as “employment saturation effect,” (iv) economic growth in more developed regions pulled the CO2 emissions downward, while pushed it upward in case of less developed regions; therefore, we call it “growth pull/push effect,” and (v) in bidirectional causality between energy consumption and economic growth, the “feedback effect” of economic growth remained dominant for all four panels. Based on empirical findings, policies are suggested for Chinese economy. Further, these policies have potential to extract implications for the rest of the world.

Introduction

The birth of great industrial revolution led to most prominent and long-lasting societal transformations termed as urbanization and industrialization. Since the birth of industrial revolution in the west, the conception on linkage between urbanization and industrialization has been predominantly accepted worldwide. This linkage was based on the unambiguous motive of economies to achieve high economic growth and hence economic welfare (Sadorsky 2013). This inter-linkage, on one hand, proved to be no less than blessing through enhancing the scale of production and thus fostering economic growth in the developed as well as emerging economies. On the other hand, massive energy use at industrial and household levels gave birth to phenomenon of pollutant emissions including CO2 and other greenhouse gases (Nasreen et al. 2018) which has been no less than curse for human beings. In this way, urbanization and industrialization equipped the society with modernization and elevated living standards accompanying drastic consequences in the form of inevitable health issues.

Nomenclature
GRP	Gross regional product	CD	Cross-section dependence
Urban	Urbanization	AAC	Averaged absolute correlation
Industry	Industrialization	CIPS	Cross-section augmented IPS
Energy	Energy consumption	RMSE	Root mean squared error
Coal	Coal consumption	χ 2	Chi-square
CO2	Carbon dioxide emissions	EE zone	Eastern economic zone
K	Physical capital	IE zone	Intermediate economic zone
AMG	Augmented Mean Group	WE zone	Western economic zone
CCEMG	Common Correlated Effects Mean Group

In the first decade of twenty-first century, approximately half of the world’s population has transformed into urban society. In this context, China, the largest population and second largest economy worldwide, has urbanized more than 50% of its population. As evident from urban and rural population profile for China in Fig. 1, the urban–rural population ratio has been increased from 0.56:1 in 2000 to 1.31:1 in 2016. Moreover, in the last two decades, the urbanization rate increased from 35.9% in 2000 to 56.8% in 2016. During this period, GDP per capita of China has shown increasing trend with urbanization (Fig. 2). About this scenario, China has witnessed miraculously high economic growth rate of approximately 10% during the last 3 decades (World Development Indicators 2017). Those high growth rates were largely attributed to industrial sector of the economy (Cherniwchan 2012). In view of this, the share of secondary and tertiary industry as percentage of GDP varied from 45.5% and 39.8% in 2000 to 39.9% and 51.6% in 2016, respectively. However, the contribution of secondary industry to GDP remained dominant until 2012 (Fig. 3). The overall profile of industry value-added, as in Fig. 4, showed increasing trend along with energy consumption from year 2000 to 2016.

Fig. 1

Changes in urban and rural population profile for China from 2000 to 2016

Fig. 2

Profile of urban and rural population and GDP per capita for China between 2000 and 2016

Fig. 3

Profile of industry value-added (primary, secondary, and tertiary) and GDP per capita for China between 2000 and 2016

Fig. 4

Changes in total energy consumption, coal consumption, and industry value-added for China between 2000 and 2016

Considering industrialization, it is explicit that China’s economy has been largely dependent on fossil fuel-based energy industry, hence leading to tremendous energy consumption in terms of fossil fuels. As a matter of fact, the energy use increased from 146,964 Mtsce¹ in 2000 to 436,000 Mtsce in 2016 (Fig. 5). Among major fossil fuels, coal has been the leading contributor towards industrial energy consumption. Because of gigantic energy use, an upsurge has been recorded in CO2 emissions from 3405.2 Mt.² in 2000 to 10,247.6 Mt. in 2016 (Fig. 6). Meanwhile, the behavior of energy consumption (aggregate and individual sources, i.e., coal, coke, crude oil) accompanied by CO2 emissions exhibited increasing trend (Figs. 5 and 6).

Fig. 5

Profile of total energy consumption, coal consumption, and CO2 emissions for China between 2000 and 2016

Fig. 6

Profile of industrial energy consumption (i.e., coal, crude oil, coke) and CO2 emissions for China from 2000 to 2016

It is viewed that industrialization has been the prime factor to boost economic growth in developed and developing economies. This process, primarily, works through the advancement of non-agriculture sectors of those economies in urban settings (Chen et al. 2014). In this way, this is directly coincided with urbanization which proliferates the access of labor to industrial sector (Frick, Rodríguez-Pose 2018). As to China, this enhancement of industrial sector led China to surpass (43% higher) the set national standards of pollutant emissions in 2016 (Chen 2016). The story, starting from industrialization and urbanization ending up with high economic growth accompanied by alarming levels of pollutant emissions, attracted the researchers around the globe considering it vital policy concern.

In this regard, number of studies focused the relationship among urbanization, energy consumption, pollutant emissions, and economic growth (Destek et al. 2016; Pao and Yu 2011; Shahbaz et al. 2014; Raggad 2018; Wang et al. 2018). These studies can be categorized based on type of analysis. First category conducted correlation studies (Ewing and Rong 2008; Jones 1991; Parikh and Shukla 1995; Zhou et al. 2013) among which some studies found positive correlation between urbanization and energy consumption (Jones 1991; Parikh and Shukla 1995), while some studies found negative correlation between the two (Destek et al. 2016; Shahbaz et al. 2014). The second category adopted the approach of causality among urbanization, energy consumption, and economic growth (Fan and Xia 2011; Mishra et al. 2009; Shahbaz et al. 2014). The findings of these studies varied from positive to neutral to negative causal relationships among those variables. The third category tested the long-run relationship among energy use, urbanization, and economic growth (Cetin and Ecevit 2015; Azam et al. 2016) and found results varying from existence to non-existence of long-run relationship among those variables.

In this study, we conduct longitudinal data analysis at both aggregated and disaggregated levels. For this purpose, we analyze a country panel as well as three regional panels of China classified into three economic zones, i.e., eastern economic zone, intermediate economic zone, and western economic zone. The selection of this classification is based on development levels of those regions as used by Zhou et al. (2015). They viewed that the eastern region was the most developed region of China, the intermediate region being next to the eastern region, while the western region was the least developed one among the three.

The existing body of knowledge offers considerable gap in following ways. First, previous studies on urbanization-energy-emissions-growth nexus did not incorporate industrialization (see Parikh and Shukla 1995; Shahbaz et al. 2014). Nevertheless, industrialization has been the core factor to explain the process of urbanization, energy consumption, emissions, and economic growth (see Cherniwchan 2012; Gungor and Simon 2017). Along these lines, those studies disregarding industrialization missed the vital channel to explain those processes. Second, those studies utilized estimators implicitly assuming cross sectional independence or homogeneity (e.g., Liu et al. 2013; Wang 2014; Zhou et al. 2015). However, the current study proves the existence of cross sectional dependence or heterogeneity, in all the variables of this study, by employing second generation Pesaran’s (2004) CD test. And thus, in the presence of heterogeneity, the empirical results based on those estimators may contain heterogeneity bias problem. Third, the previous studies, using production function-based economic growth model, did not incorporate industrialization and pollutant emissions in their models (e.g., Lee et al. 2008; Narayan and Smyth 2008). Fourth, none of the previous studies addressed the causal connections among industrialization, urbanization, energy consumption, CO2 emissions, and economic growth in simultaneous equations framework. Alternatively, those studies either focused one-way impact analysis of those variables, with exception of industrialization, on energy consumption and economic growth (see, inter alia Parikh and Shukla 1995; Liddle 2013; Sadorsky 2013; Shahbaz et al. 2014; Tzeremes 2018) or merely established pairwise bidirectional causality between energy consumption and economic growth (e.g., Mishra et al. 2009; Hanif 2018), energy consumption and pollutant emissions (e.g., Zhou et al. 2015; Nasreen et al. 2018), urbanization and pollutant emissions (e.g., Wang et al. 2016), and urbanization and energy consumption (e.g., Jones 1991; Liu et al. 2013).

The purpose of this study is to examine causal linkages among industrialization, urbanization, energy consumption, CO2 emissions, and economic growth for 30 Chinese provinces and cities for the periods 2000–2016. In technical terms, this study makes several contributions to the new body of knowledge. First, we extend economic growth model by Lee et al. (2008), to include industrialization as shift factor and pollutant emissions as determinant of total factor productivity. Second, we establish theoretical linkages among industrialization, urbanization, energy consumption, CO2 emissions, and economic growth. Third, we set up a system of five simultaneous structural equations to analyze the five-ways causal linkages among study variables. Fourth, we allow for possibility of panel heterogeneity, to this end, we employ second generation Augmented Mean Group (AMG) estimator by Eberhardt and Teal (2010) and Common Correlated Effects Mean Group (CCEMG) estimator by Pesaran (2006) to obtain reliable estimates robust to cross sectional dependence and cointegration issues. Finally, the findings of the study uncover stylized empirics along with important policy suggestions.

The organization of the remaining study is given as follows: “Modeling and estimation method” section develops model and explains estimation method; “Data” section is based on data description; “Major findings and discussion” section reports empirical results and, finally, “Conclusions and policy implications” section contains conclusion and policy suggestions.

Modeling and estimation method

Developing model of economic growth

We start augmenting a model of economic growth based on Cobb–Douglas production function following constant returns to scale used by Lee et al. (2008), with total factor productivity (TFP) as addition:

$$ {Y}_{i,t}={A}_{i,t}{K}_{i,t}^{\alpha }{E}_{i,t}^{\beta }\ {N}_{i,t}^{1-\alpha -\beta }{\left({U}_{i,t}\right)}^{\varphi } $$

(1)

where Y is the aggregate output, K is the physical capital, N is the labor input, E is the energy consumption, and A is total factor productivity. Moreover, i = 1, 2, 3, …, N is the index of provinces and cities; t = 1, 2, 3, …, T is the time index.

We augment the Eq. (1) considering industrialization as a shift factor of output:

$$ {Y}_{i,t}={A}_{i,t}{K}_{i,t}^{\alpha }{E}_{i,t}^{\beta }\ {N}_{i,t}^{1-\alpha -\beta }{\left({U}_{i,t}\right)}^{\varphi }{\left({I}_{i,t}\right)}^{\psi }. $$

(2)

Next, according to Rusiawan et al. (2015), pollutant emissions significantly contribute to TFP; thus, it can be considered as determinant of TFP. Hence for our purpose, assuming all else in TFP equal, pollutant emissions can be substituted for A as shift factor in Eq. (2):

$$ {Y}_{i,t}={K}_{i,t}^{\alpha }{E}_{i,t}^{\beta }\ {N}_{i,t}^{1-\alpha -\beta }{\left({U}_{i,t}\right)}^{\varphi }{\left({I}_{i,t}\right)}^{\psi }{\left({P}_{i,t}\right)}^{\lambda } $$

(3)

where I denotes industrialization and P denotes pollutant emissions. Following Lee et al. (2008) and Narayan and Smyth (2008), Eq. (3) is normalized by labor force. Further, natural log transformation is applied:

$$ \mathit{\ln}{\overset{\sim }{Y}}_{i,t}=\alpha ln{\overset{\sim }{K}}_{i,t}+\beta ln{\overset{\sim }{E}}_{i,t}+{\varphi ln U}_{i,t}+{\psi ln I}_{i,t}+{\lambda ln P}_{i,t} $$

(4)

where ~ superscript represents the variables in per-worker form. The Eq. (4) is further transformed into econometric model.

Econometric models

We have Eq. (4) as a new extended version of model of economic growth including urbanization, industrialization, energy consumption, and pollutant emissions. We rewrite Eq. (4) in econometric form:

$$ \mathit{\ln}{\overset{\sim }{Y}}_{i,t}=\alpha ln{\overset{\sim }{K}}_{i,t}+\beta ln{\overset{\sim }{E}}_{i,t}+{\varphi ln U}_{i,t}+{\psi ln I}_{i,t}+{\lambda ln P}_{i,t}+{\epsilon}_{i,t}. $$

(5)

From Eq. (5), we have constructed a system of five simultaneous equations to treat economic growth, energy consumption, urbanization, industrialization, and pollutant emissions simultaneously:

$$ \mathit{\ln}{\overset{\sim }{Y}}_{i,t}={\alpha}_1\mathit{\ln}{\overset{\sim }{K}}_{i,t}+{\alpha}_2\mathit{\ln}{\overset{\sim }{E}}_{i,t}+{\alpha}_3{lnU}_{i,t}+{\alpha}_4{lnI}_{i,t}+{\alpha}_5{lnP}_{i,t}+{\epsilon}_{i,t,1} $$

(6)

$$ \mathit{\ln}{\overset{\sim }{E}}_{i,t}={\beta}_1\mathit{\ln}{\overset{\sim }{Y}}_{i,t}+{\beta}_2{lnU}_{i,t}+{\beta}_3{lnI}_{i,t}+{\beta}_4{lnP}_{i,t}+{\epsilon}_{i,t,2} $$

(7)

$$ {lnU}_{i,t}={\gamma}_1\mathit{\ln}{\overset{\sim }{E}}_{i,t}+{\gamma}_2\mathit{\ln}{\overset{\sim }{Y}}_{i,t}+{\gamma}_3{lnI}_{i,t}+{\gamma}_4{lnP}_{i,t}+{\epsilon}_{i,t,3} $$

(8)

$$ {lnI}_{i,t}={\theta}_1\mathit{\ln}{\overset{\sim }{E}}_{i,t}+{\theta}_2{lnU}_{i,t}+{\theta}_3\mathit{\ln}{\overset{\sim }{Y}}_{i,t}+{\theta}_4{lnP}_{i,t}+{\epsilon}_{i,t,4} $$

(9)

$$ {lnP}_{i,t}={\xi}_1\mathit{\ln}{\overset{\sim }{E}}_{i,t}+{\xi}_2{lnU}_{i,t}+{\xi}_3{lnI}_{i,t}+{\xi}_4\mathit{\ln}{\overset{\sim }{Y}}_{i,t}+{\epsilon}_{i,t,5} $$

(10)

where ϵ_{i, t}s’ are the error terms of the models; the α₁, α₂, α₃, α₄, and α₅ capture the impact of physical capital, energy consumption, urbanization, industrialization, and pollutant emissions on economic growth, respectively; β₁, β₂, β₃, and β₄ capture the impact of economic growth, urbanization, industrialization, and pollutant emissions on energy consumption, respectively; γ₁, γ₂, γ₃, and γ₄ capture the impact of energy consumption, economic growth, industrialization, and pollutant emissions on urbanization, respectively; θ₁, θ₂, θ₃, and θ₄ capture the impact of energy consumption, urbanization, economic growth, and pollutant emissions on industrialization, respectively; and, ξ₁, ξ₂, ξ₃, and ξ₄ capture the influence of energy consumption, urbanization, industrialization, and economic growth on pollutant emissions, respectively.

Linkages among study variables and hypotheses formulation

The shift of labor force from rural agriculture sector to urban industrial sector has been posited to be the key factor responsible for accelerated urbanization and enhanced economic growth. Similarly, economic growth is also thought to be driven by urbanization via benefits in economies of scale in infrastructure and institutional development (Liddle 2013). The high economic growth associated with urbanization has been viewed to contribute to significant enhancement in industrial and urban household energy consumption (Sadorsky 2013; Wang 2014). Besides, high economic growth via advantages of employment creation attracts the labor force to urban settings, hence accelerating urbanization. Since energy use is normal good, therefore, urban concentration may drive high level of energy consumption, as argued by Liddle (2014) in China’s perspective and Bakirtas and Akpolat (2018) in new emerging-market countries. This process is likely to work in a fashion that extensive urbanization generates enormous demand for energy that steps up the energy consumption (Liu et al. 2013). The resulting rise in energy consumption adds to the industry output promoting high economic growth. The increment in economic growth creates further energy demand both at industry and household levels leading to further addition in energy use (Gungor and Simon 2017).

However, urban concentration may bring about negative externality of high-level pollutant emissions attributed to rapid industrialization, and household energy consumption (Gul et al. 2015; Bulut 2017; Nasreen et al. 2018). Because of these pollutant emissions, human capital may be susceptible to negative health impacts diminishing its efficiency and hence deteriorating economic growth (Destek et al. 2016). As economies transform from agriculture to industry-based framework, this process leads to capital accumulation, which in turn shifts the economy’s output towards pollution intensive industry output, hence driving pronounced levels of pollutant emissions (Cherniwchan 2012). Meanwhile, as economies end up with achieving high level of economic development, it prompts improvement in production technology and techniques, and thus leading to lower level of pollutant emissions (Environmental Kuznets Curve). The summary of these channels is presented in Fig. 7. Based on theoretical grounds explained in these channels, the hypotheses tested in this study are as follows: (1) industrialization is positively associated with energy consumption, (2) energy consumption is positively associated with CO2 emissions, (3) CO2 emissions is negatively associated with economic growth, (4) economic growth is positively associated with urbanization, (5) urbanization is positively associated with industrialization, (6) industrialization is positively associated with economic growth, (7) economic growth is positively associated with energy consumption, (8) energy consumption is positively associated with urbanization, (9) urbanization is positively associated with CO2 emissions, and finally (10) CO2 emissions is positively associated with industrialization.

Fig. 7

Linkages among industrialization, urbanization, energy consumption, CO2 emissions, and economic growth

Estimation method

The multivariate longitudinal data models in the form of five structural equations are used to estimate five-ways linkages among industrialization, urbanization, energy consumption, CO2 emissions, and economic growth. This study has employed the “second-generation” state of art panel techniques known as Augmented Means Group (AMG) estimator developed by Eberhardt and Teal (2010) and Common Correlated Effects Mean Group (CCEMG) estimator by Pesaran (2006). The justification of using AMG and CCEMG estimators over others are described by the following arguments: (i) the conventional longitudinal data techniques in practice to overcome endogeneity problem (like instrumentation techniques) are merely applicable in stationary state of variables. However, AMG and CCEMG estimators deal with this problem even in non-stationary framework, hence being robust to non-stationary property whether the panel time series are cointegrated or not (Pesaran 2006; Eberhardt and Teal 2010), (ii) AMG estimator incorporates a term of “common dynamic process” derived from first differenced time dummy coefficients of a pooled regression, which yields the property to account for cross-sectional dependence, (iii) CCEMG yields consistent estimates regardless of degree of dependence among cross-sectional errors, (iv) CCEMG retains the capability to work fine even for panels with short time dimension, and finally (v) both of AMG and CCEMG estimators offer prospect to apply post-estimation tests (using estimated residuals) called Pesaran’s CD and CIPS tests for cross-sectional dependence and stationary property, respectively, to validate the viability of the estimated elasticities. The log-linearized form models (6–10) to be estimated by AMG and CCEMG estimators can be rewritten as:

$$ \mathit{\ln}{\overset{\sim }{Y}}_{i,t}={\alpha}_{i1}\mathit{\ln}{\overset{\sim }{K}}_{i,t}+{\alpha}_{i2}\mathit{\ln}{\overset{\sim }{E}}_{i,t}+{\alpha}_{i3}{lnU}_{i,t}+{\alpha}_{i4}{lnI}_{i,t}+{\alpha}_{i5}{lnP}_{i,t}+{\epsilon}_{i,t,1} $$

(11)

$$ \mathit{\ln}{\overset{\sim }{E}}_{i,t}={\beta}_{i1}\mathit{\ln}{\overset{\sim }{Y}}_{i,t}+{\beta}_{i2}{lnU}_{i,t}+{\beta}_{i3}{lnI}_{i,t}+{\beta}_{i4}{lnP}_{i,t}+{\epsilon}_{i,t,2} $$

(12)

$$ {lnU}_{i,t}={\gamma}_{i1}\mathit{\ln}{\overset{\sim }{E}}_{i,t}+{\gamma}_{i2}\mathit{\ln}{\overset{\sim }{Y}}_{i,t}+{\gamma}_{i3}{lnI}_{i,t}+{\gamma}_{\mathrm{i}4}{lnP}_{i,t}+{\epsilon}_{i,t,3} $$

(13)

$$ {lnI}_{i,t}={\theta}_{i1}\mathit{\ln}{\overset{\sim }{E}}_{i,t}+{\theta}_{i2}{lnU}_{i,t}+{\theta}_{i3}\mathit{\ln}{\overset{\sim }{Y}}_{i,t}+{\theta}_{i4}{lnP}_{i,t}+{\epsilon}_{i,t,4} $$

(14)

$$ {lnP}_{i,t}={\xi}_{i1}\mathit{\ln}{\overset{\sim }{E}}_{i,t}+{\xi}_{i2}{lnU}_{i,t}+{\xi}_{i3}{lnI}_{i,t}+{\xi}_{i4}\mathit{\ln}{\overset{\sim }{Y}}_{i,t}+{\epsilon}_{i,t,5} $$

(15)

where i s’ as subscripts of elasticity coefficients indicate that the elasticities are heterogenous across provinces and cities of China.

Data

The study data comprising 30 Chinese provinces and cities, for time periods of 2000 to 2016 are compiled from National Bureau of Statistics (2016, 2017). The data include gross regional product (GRP) which is GDP at province and city level (100 million Chinese yuan), utilized as proxy for economic growth; urban population (percent of total population) to measure urbanization; total energy consumption (10,000 tons); consumption of coal (10,000 tons); carbon dioxide emissions (million tons) calculated by using formula by Intergovernmental Panel on Climate Change (IPCC) (2014); and gross fixed capital formation (100 million Chinese yuan) used as proxy for physical capital. Further, all the data of variables are transformed into per capita form except urbanization and industrialization as those are used in percentage form. The description of variables is reported in Table 1.

Table 1 Description of variables used in estimation process

Pre-estimation testing

Prior to employing main estimation method, it is appropriate to test the time series and cross-sectional property of longitudinal data. The choice or decision to opt estimation technique is crucial. This decision-making along with flow of analysis is presented in Fig. 8. Application of conventional unit root tests in presence of cross-sectional dependence may yield misleading test outcomes. Therefore, before testing for stationary property, we tested the cross-sectional dependence via using Pesaran’s (2004) CD test. The test results declared all the series to be cross-sectionally dependent. Consequently, we employed the second generation Pesaran’s (2007) CIPS unit root test, because it yields correct outcomes even in the presence of cross-sectional dependence. The test results declared all series, except GRP, to be non-stationary at level. Further, all series turned stationary at first difference. It proved presence of cross-sectional dependence and unit root in longitudinal data as recorded in Table 2. In order to tackle this peculiar situation, AMG estimator is employed to estimate all four panels of longitudinal data.

Fig. 8

Choice of estimation technique and flow of analysis

Table 2 Second generation tests of cross-section dependence/unit root for longitudinal data

Major findings and discussion

Panel-specific estimation results and discussion

The long-run elasticity estimates (significant) based on AMG estimator are reported in Table 3, whereas, those based on CCEMG estimator are documented in Table 4. Since the elasticity magnitudes of all the study variables estimated employing AMG estimator are not much different from those estimated via CCEMG, and further, the signs of the elasticities are also highly consistent across the two methods, therefore, we have merely discussed the elasticity magnitudes based on AMG estimator. Additionally, the detailed results of each model estimated via both methods are recorded in their respective tables provided in appendix (Tables A.1–A.10).

Table 3 Long-run elasticity estimates (significant) based on Augmented Mean Group estimator

Table 4 Long-run elasticity estimates (significant) based on Common Correlated Effects Mean Group (CCEMG) estimator

Model of GRP

In model of GRP (proxied for economic growth), in both cases, either with energy or with coal, urbanization demonstrated very striking and interesting behavior in terms of difference in its impact on economic growth across regions. It negatively influenced economic growth in case of WE zone which is least developed among the three regions. Further, its effect is neutral for moderately developed region, i.e., IE zone. And finally, its impact is positive in case of EE zone that is most developed among the three regions. This phenomenon is termed as “urbanization ladder effect.” It states that as entities (countries/regions) develop, the contribution of urbanization to economic growth varies from negative to neutral to positive. This is consistent with the findings of Liddle (2013). Based on AMG estimator, the elasticity magnitudes of urbanization are 0.210, 0.234, and − 0.192 for country panel, EE zone, and WE zone, respectively. Next, industrialization and physical capital have positive and statistically significant impact on economic growth for all the panels, with highest elasticity magnitudes for EE zone and lowest for WE zone. It has confirmed the findings of Gungor and Simon (2017). For industrialization, it implies that as regions develop, industry expansion has more powerful and positive contribution to economic growth; we call it “industry expansion effect.” The elasticity magnitudes of industrialization are 0.225, 0.258, 0.252, and 0.219 for country panel, EE zone, IE zone, and WE zone, respectively. It revealed that industrialization has the strongest impact on economic growth in EE zone, while this effect is least strong in case of WE zone. The elasticities of energy consumption and coal consumption are positive and statistically significant, with magnitudes highest for EE zone (i.e., 0.179 and 0.168, respectively) and lowest for WE zone (i.e., 0.168 and 0.143, respectively). It has validated the results of Tzeremes (2018) for US states. However, CO2 emissions depicted negative and statistically significant impact on economic growth, with strong effect for EE zone and less strong effect for WE zone. The nature of relationship in this finding is consistent with Azam et al. (2016) and Raggad (2018). These results are reported in Tables 3 and 4.

Model of energy/coal consumption

Turning to model of energy/coal consumption, the variables of economic growth, urbanization, and industrialization exhibited positive and statistically significant impact on energy/coal consumption for all the panels. Their elasticity magnitudes are highest for EE zone and lowest for WE zone. Overall, this behavior is consistent across both types of models, i.e., model with energy consumption and model with coal consumption. However, there is slight difference between the elasticity values of the two models. The elasticity magnitudes of economic growth in terms of its impact on energy consumption are 0.582, 0.601, 0.596, and 0.501 for country panel, EE zone, IE zone, and WE zone, respectively. Whereas, its elasticity magnitudes in terms of its impact on coal consumption are 0.510, 0.577, 0.563, and 0.488, respectively. In terms of regional differences, urbanization showed more vital results. For more developed region like EE zone, increase in urbanization is expected to significantly raise energy/coal consumption more than that in less developed region like WE zone. These results are similar to those of Liddle (2014). Moreover, these results verified the findings of Wang (2014) and Wang et al. (2018). The elasticity magnitudes of urbanization, in case of model with energy consumption, are 0.132, 0.141, 0.135, and 0.100, respectively. However, its elasticity magnitudes in model with coal consumption are 0.110, 0.123, 0.118, and 0.080, respectively. As far as industrialization is concerned, its elasticity magnitudes are 0.188, 0.193, 0.182, and 0.163, respectively, in case of model with energy consumption. In the similar fashion, its elasticity values in model with coal consumption are 0.159, 0.162, 0.151, and 0.126, respectively. In addition, CO2 emissions demonstrated statistically insignificant contribution to energy/coal consumption for all panels. Moreover, the nature of associations between variables is consistent across the two cases of models (i.e., model with energy consumption and model with coal consumption). These results are recorded in Tables 3 and 4.

Model of urbanization

In model of urbanization, economic growth has positive and statistically significant contribution to urbanization, with stronger effect for EE zone and weaker for WE zone (in case of both model with energy consumption and coal consumption). In this regard, the elasticity values of economic growth are 1.398, 1.420, 1.396, and 1.318 for country panel, EE zone, IE zone, and WE zone, respectively (for model with energy consumption). Whereas, these elasticity magnitudes in model with coal consumption are 1.016, 1.139, 1.096, and 1.072 for country panel, EE zone, IE zone, and WE zone, respectively. The results of economic growth verified the findings of Shahbaz et al. (2014). The case of industrialization is very unusual in terms of different impact across regions. It showed positive and statistically significant impact on urbanization, with stronger effect for WE zone, while less stronger effect for EE zone. It implies that industrialization is expected to promote rapid urbanization in less developed as well as less employment saturated region in contrast to more employment saturated region. We define this phenomenon as “employment saturation effect.” The elasticity magnitudes of industrialization in case of model with energy consumption are 0.097, 0.089, 0.101, and 0.109, for country panel, EE zone, IE zone, and WE zone, respectively. Besides this, in case of model with coal consumption, its elasticity magnitudes are 0.093, 0.076, 0.095, and 0.114, for all the four panels, respectively. This finding confirmed that of Gungor and Simon (2017) for South Africa. The nature of linkages between variables is consistent across the two types of models (i.e., model with energy consumption and model with coal consumption). Moreover, energy/coal consumption expressed statistically insignificant impact on urbanization. These results are presented in Tables 3 and 4.

Model of industrialization

In this model, urbanization and economic growth exhibited positive and statistically significant contribution to industrialization. The elasticity values of urbanization, in case of model with energy consumption, are 0.019, 0.027, 0.025, and 0.018, for country panel, EE zone, IE zone, and WE zone, respectively. In the same way, its elasticity values for model with coal consumption are 0.021, 0.025, 0.021, and 0.013, for all the four panels, respectively. This finding is consistent with that of Sadorsky (2013) found for developing countries. Similarly, the elasticity magnitudes of economic growth are 0.043, 0.048, 0.040, and 0.035, for all the four panels, respectively. Further, its elasticity magnitudes for model with coal consumption are 0.058, 0.062, 0.051, and 0.028, for all the four panels, respectively. In this way, the impact of urbanization and economic growth is strong in case of EE zone, while turned less strong for IE zone and WE zone, respectively. It implies that at high level of development opposite to low level of development, increase in urbanization and economic growth induces rapid industrialization. Further, energy/coal consumption and CO2 emissions revealed neutral influence on industrialization. The empirical findings showed that the nature of links between dependent and independent variables is consistent across the two cases (i.e., model with energy consumption and model with coal consumption). These results are documented in Tables 3 and 4.

Model of CO2 emissions

In case of model of CO2 emissions, energy/coal consumption, urbanization, and industrialization disclosed positive and statistically significant influence on CO2 emissions. In this regard, energy/coal consumption massively contributed to CO2 emissions with strong impact for country panel and EE zone, while weaker effect for WE zone. The elasticity magnitudes of energy consumption in terms of its effect on CO2 emissions are 0.374, 0.391, 0.383, and 0.368, for country panel, EE zone, IE zone, and WE zone, respectively. In the similar way, the elasticity values of coal consumption in terms of its influence on CO2 emissions are 0.322, 0.340, 0.327, and 0.289, for country panel, EE zone, IE zone, and WE zone, respectively. These findings confirmed those of Gul et al. (2015) in Malaysian context and Bulut (2017) in case of Turkey. Industrialization, in this context, is significant contributor to CO2 emissions. Its elasticity values in case of model with energy consumption are 0.232, 0.239, 0.234, and 0.227, for all the four panels, respectively. However, the elasticities in case of model with coal consumption are 0.216, 0.230, 0.221, and 0.189, for all the four panels, respectively. This contribution is prominent in case of EE zone, but relatively less intensive in case of WE zone. Urbanization, next to industrialization, heavily polluted the environment with CO2 emissions. This is inconsistent with the findings of Wang et al. (2016). Like industrialization, its impact is dominant for EE zone, but less intensive for IE zone and WE zone, respectively. The elasticity values of urbanization are 0.121, 0.128, 0.123, and 0.116, for country panel, EE zone, IE zone, and WE zone, respectively (model with energy consumption). Similarly, in case of model with coal consumption, the elasticity magnitudes of urbanization are 0.101, 0.112, 0.094, and 0.079, for all the four panels respectively. However, the elasticity values of economic growth are − 1.109, − 1.130, − 1.000, and 0.989, for country panel, EE zone, IE zone, and WE zone, respectively (in case of model with energy consumption). Moreover, in case of model with coal consumption, those elasticities are − 1.127, − 1.138, − 0.990, and 0.902, for all the four panels, respectively. In this way, economic growth revealed negative and statistically significant impact for country panel, EE zone, and IE zone, while positive impact for WE zone. The findings of this linkage verified those of Tzeremes (2018) for US states, and Hanif (2018) in Sub Sahara African perspective. It implicates that economic growth promotes CO2 emissions for less developed regions like it did for WE zone, curtails CO2 emissions with small magnitude for moderately developed regions like IE zone. While, for most developed regions like EE zone, increase in economic growth significantly mitigates CO2 emissions. We term this phenomenon as “growth-push/pull effect.” For less developed regions, growth is expected to push the CO2 emissions upward, while for more developed regions, it is expected to pull the CO2 emissions downward. In terms of conventional mainstream theory, this phenomenon is also termed as Environmental Kuznets Curve. These results are displayed in Tables 3 and 4.

Results summary

In order to make the comparisons more visible and clear, the summary of empirical findings demonstrating five-ways causal linkages among study variables is presented in Figs. 9 and 10. In brief, an unidirectional positive causal linkage is found from industrialization to energy/coal consumption for all study panels. Similarly, energy/coal consumption is found to have positive impact on CO2 emissions, and not the vice versa, for all four panels. Furthermore, bidirectional negative causal linkage is revealed between CO2 emissions and economic growth for all panels, except WE zone which demonstrated positive impact of economic growth on CO2 emissions. Next, economic growth and energy/coal consumption are exposed to have bilateral causal linkage for all panels; however, the impact of the former is more powerful than the latter, as opposed to vice versa. It signifies that energy consumption promotes economic growth that in turn has stronger “feedback effect” on energy consumption. This effect is strong for EE zone, while less strong for IE zone and WE zone, respectively. Then, urbanization led energy/coal consumption and CO2 emissions, but is not led by the same for all the panels. In contrast to this, economic growth and industrialization are driven by each other for all panels. In turn, industrialization led CO2 emissions and is not led by the same for all panels. Moreover, economic growth and urbanization disclosed more versatile results. A bidirectional positive causal linkage is found between the two in case of country panel and EE zone, and unidirectional positive linkage is found running from economic growth to urbanization for IE zone. However, for WE zone, these variables exhibited bidirectional causal linkage with mixed signs; positive causality from economic growth to urbanization and negative causality from urbanization to economic growth. Finally, bilateral positive causal linkage is revealed between urbanization and industrialization for all study panels.

Fig. 9

Empirical summary of five-ways causal linkages among industrialization, urbanization, energy/coal consumption, CO2 emissions, and economic growth (estimations based on AMG estimator). Notes: (a1, a2), (b1, b2), (c1, c2), and (d1, d2) refer to country panel, eastern economic zone, intermediate economic zone, and western economic zone, respectively. Whereas, I, U, E/C, P, and Y denote industrialization, urbanization, energy/coal consumption, CO2 emissions, and economic growth, respectively. Single-headed and double-headed arrows indicate unidirectional and bidirectional causality, respectively. Magnitudes express strength of linkages between variables.

Fig. 10

Empirical summary of five-ways causal linkages among industrialization, urbanization, energy/coal consumption, CO2 emissions, and economic growth (estimations based on CCEMG estimator). Notes: See notes of Fig. 9.

Post-estimation

The χ² probability values are significant indicating that all the models are good fit. Moreover, Pesaran’s (2004) CD test and Pesaran’s (2007) CIPS test confirmed that residuals of estimated models are cross-sectionally independent and stationary. These outcomes of diagnostic tests confirm the robustness of estimation method, and hence reliability and validity of the estimated elasticities of study variables. These results are reported in appendix (Tables A.1–A.10).

Variation in individual province/city-specific estimations

The AMG estimator, in first step, involves estimating individual province/city-based regressions and then averages the estimated elasticities across province/cities. In case of current study, the examination of province/city-specific regressions exposed ample variations within different panels (i.e., country panel, EE zone, IE zone, and WE zone). For province/city-specific regression estimates, the summary is provided in Table 5, listing (i) range of significant elasticities and (ii) standard deviation of significant elasticities for all four panels. The detailed results are recorded in appendix (Tables B.1–B.5).

Table 5 Variation in individual province/city long-run estimations by panel

In model of GRP, the measures of dispersion (range, standard deviation) revealed the highest variation in elasticity estimates of urbanization. For all panels, several provinces/cities have negative and statistically significant estimates of urbanization. It exhibited the diversified impact of urbanization on economic growth which is also verified from the averaged elasticities of urbanization for all panels. Like urbanization, industrialization also demonstrated considerable variations across provinces/cities. Industrialization, next to urbanization, exposed the second highest range and standard deviation. Similarly, CO2 emissions displayed third highest range and standard deviation. Interestingly, CO2 emissions indicated more variation for EE zone that is more developed region, while moderate to less variation for IE zone and WE zone, respectively.

In energy/coal consumption model, economic growth demonstrated highest variations across all four panels. It entails that economic growth influenced energy/coal consumption with diverse effects in case of different provinces and cities. In the like manner, urbanization has shown the second highest measures of dispersion. It is notable that variations in economic growth and urbanization declined from EE zone to IE zone to WE zone, with decline in level of regional development. Next, in model of urbanization, too, economic growth revealed highest range and standard deviation. It implies that impact of economic growth on urbanization is subject to high variation across provinces/cities. In the same way, industrialization indicated second highest variation for all panels. Then, in industrialization model, urbanization demonstrated relatively high variations among other variables. Finally, in model of CO2 emissions, economic growth displayed highest variation across provinces/cities.

Conclusions and policy implications

Main conclusions

The current study explored and found five-ways linkages among industrialization, urbanization, energy/coal consumption, CO2 emissions, and economic growth through system of simultaneous equations. Based on longitudinal data estimations, employing Augmented Mean Group (AMG) estimator and Common Correlated Effects Mean Group (CCEMG) estimator, the stylized empirical findings include the statements that follow.

First, overall, a bidirectional positive causal linkage is found between economic growth and energy/coal consumption, economic growth and industrialization, economic growth and urbanization, and urbanization and industrialization. However, unidirectional positive causal linkage is revealed from industrialization to energy/coal consumption, energy/coal consumption to CO2 emissions, urbanization to energy/coal consumption, urbanization to CO2 emissions, and industrialization to CO2 emissions. Furthermore, bidirectional negative causal linkage is exposed between economic growth and CO2 emissions. These findings varied across regions. Particularly, the linkages between economic growth and urbanization, CO2 emissions and economic growth, and industrialization and economic growth, demonstrated versatile and varied findings across different panels.

Second, urbanization exhibited a very intuitive behavior in terms of difference in its impact on economic growth that varied from negative to neutral to positive for WE zone, IE zone, and EE zone, respectively. This finding is particularly vital in face of difference in development levels of these regions. This phenomenon is called as “urbanization ladder effect.” It implies that as provinces/cities develop, urbanization has potential to exhibit positive impact on economic growth. Furthermore, the province/city-specific estimates of urbanization revealed highest variations in its impact on economic growth.

Third, the long-run elasticity estimates for industrialization indicated that its impact on economic growth varied from weaker to moderate to stronger from WE zone to IE zone to EE zone. It implies that as regions develop, industry expansion has more powerful and positive impact on economic growth. We say it “industry expansion effect.” Based on province/city-specific estimates, industrialization demonstrated second highest variations, next to urbanization, concerning its impact on economic growth.

Fourth, economic growth disclosed very peculiar behavior in terms of difference in its impact on urbanization across different regions in that it varied from strong positive impact for WE zone to less strong impact for EE zone. It implied that economic growth promoted rapid urbanization in less urbanized and less employment saturated regions as compared to the opposite situation. We define this phenomenon as “employment saturation effect.”

Fifth, energy/coal consumption led to massive CO2 emissions. In this regard, industrialization remained the major role player, while urbanization is next to industrialization in driving massive CO2 emissions.

Sixth, economic growth and energy consumption drive each other; however, the impact of former is dominant on latter, as opposed to the other way around. It conveys that energy consumption promotes economic growth that in turn has stronger “feedback effect” on energy consumption. This effect is strong for EE zone, while less to moderately strong for IE zone and WE zone, respectively.

Finally, the impact of economic growth on CO2 emissions varied from negative for country panel, EE zone, and IE zone, while positive for WE zone. It signifies that economic growth for country panel, EE zone, and IE zone curtailed CO2 emissions, while CO2 emissions is promoted along with economic growth for WE zone. We term this phenomenon as “growth-push/pull effect” of economic growth. It directs that economic growth in more developed regions, like EE zone, pulled the CO2 emissions downward, whereas, CO2 emissions is pushed upward by economic growth in less developed regions of China like WE zone.

Policy implications

Based on heterogenous empirical findings, it is worthy to suggest that province/city-specific policies should be designed and implemented for China. Following policies are advised based on our empirical results: (i) the negative contribution of urbanization to economic growth for WE zone implies lack of employment in the region; therefore, it is suggested that urban migration should be accompanied with employment creation in this region. Moreover, there is found high variation in impact of urbanization on economic growth at province/city level, suggesting that policies at province/city level will be more effective than at aggregate level, (ii) the industry expansion showed powerful impact on economic growth for more developed region, that is, EE zone, while less powerful impact for WE zone. It suggests that industry expansion may boost economic growth in WE zone if development is focused in this region. This process of development in WE zone may reduce the disparity among different regions of China, (iii) as EE zone is found the highest CO2 emitter as a result of energy consumption, therefore, it is recommended to use energy efficient means to mitigate CO2 emissions, particularly, in EE zone. In this regard, the energy can be conserved in transport system in two ways: (a) in big cities like Beijing and Shanghai, public transport development should be focused to reduce usage of personal vehicles and hence may be helpful to curtail CO2 emissions, (b) a useful alternative could be to promote hybrid energy vehicles that should be able to run both on fossil-fuels as well as green energy (i.e., solar energy). In this way, the possibility to save CO2 emissions will be enhanced.

As mentioned earlier, China has already surpassed the set national standards of pollutant emissions. In current study, economic growth is found to significantly mitigate CO2 emissions in developed region that is eastern economic zone. However, it promoted CO2 emissions in less developed region, that is, western economic zone. Therefore, it may be implied that up to certain level of development, CO2 emissions are expected to increase along with economic growth in less developed regions. Based on diverse findings across development levels, this study on China may provide opportunity to derive lessons for both developing and developed world.

There are various avenues and dimensions of further research in this arena. A comparative analysis between non-renewable energy consumption and renewable energy consumption would be useful rather than using energy consumption in terms of its linkages with other variables. Moreover, urbanization can further be used as comprehensive variable hybrid of several factors like urban population density, and urban industrial employment. Similarly, industrialization can be further disaggregated into primary, secondary, and tertiary levels.