The impact of foreign direct investment on China's carbon emission efficiency through energy intensity and low-carbon city pilot policy

Abstract

The rapid growth of foreign direct investment (FDI) has had a significant impact on rapid economic development and environmental pollution in China. Nonetheless, the underlying mechanism and empirical evidence of FDI impact on carbon emission efficiency have not been systematically explored. Therefore, this study investigates the impact of FDI on carbon emission efficiency via energy intensity, as well as the moderating role of the Low-Carbon City Pilot Policy (LCCP) in the process. We found that: (1) During the study sample period, the average carbon emission efficiency tends to rise, however, there remains a gap between the optimal carbon emission efficiency; (2) FDI is one of the key factors that inhibit the improvement of carbon emission efficiency, with a non-linear relationship between them; (3) FDI indirectly suppresses the improvement of carbon emission efficiency by promoting energy intensity. Nevertheless, the implementation of LCCP has a positive effect on carbon emission efficiency; (4) The implementation of LCCP has improved the negative impact of FDI on carbon emission efficiency, nonetheless, it cannot significantly influence the process that FDI affects carbon emission efficiency through energy intensity. Thus, we propose improvement measures from three aspects, i.e., increasing the introduction of foreign capital, adjusting the energy consumption structure, and expanding the scope of low-carbon cities.

Keywords

Foreign direct investment carbon emission efficiency energy intensity low-carbon city pilot policy China

Introduction

With the rapid development of industrialization, global carbon dioxide emissions have dramatically increased with a crucial impact on people's health and sustainable economic as well as social development.^1–3 President Xi announced stated at the general debate in the 75^th United Nations General Assembly that China aims to reach the peak of carbon dioxide emissions by 2030 and achieve carbon-neutral development by 2060. As the global largest carbon emitter and energy consumer, China faces severe problems in energy conservation and environmental protection.^4,⁵ Based on the China Energy Statistical Yearbook, China's total energy consumption increased from 1.362 billion tons of standard energy consumption to 4.87 billion tons between 1998 and 2019. According to the World Bank Data, China's carbon dioxide emissions account for approximately 20% of the world. The massive energy consumption and the continuous increase of carbon emissions have had a severe negative impact on the quality of the environment.⁶ As the primary embodiment of modern economic and social development, cities play a fundamental role in improving the ecological environment.^7,8 On the other hand, cities must consume a lot of energy to keep the economy running smoothly. Existing studies indicate that the energy consumption of Chinese cities accounts for 75% of the national total.⁹ Large-scale energy consumption and coal-based energy structure present severe challenges to the realization of emission reduction targets.¹⁰ Thus, the Chinese government is actively exploring a development path that will result in a win-win economic development and carbon emissions.¹¹

Existing studies have demonstrated that technological innovation plays an important role in improving energy efficiency and reducing carbon emissions.^12,13 As an approach to acquiring advanced technology, foreign direct investment (FDI) has typical technology and knowledge spillover effects, which promote technological innovation in host countries, with subsequently an important impact on energy efficiency and carbon emissions.^14,15 Since the reformation and opening up, China has become one of the most successful developing countries in attracting FDI due to its huge market potential.^16,17 Statistics from the Ministry of Commerce of China indicate that since joining the WTO in 2001, China's application of FDI has risen from US$60.63 billion in 2004 to US$134.97 billion in 2018. The rapid growth of FDI is a double-edged sword, i.e., it promotes the rapid development of China's economy, and exacerbates energy shortage as well as environmental pollution.¹⁸ Based on the existing research, FDI does not necessarily result in environmental problems to the host country, i.e., two hypotheses explain this, including “Pollution Halo” and “Pollution Refuge”.¹⁹ The Pollution Halo hypothesis is based on the positive impact of FDI on the environment, primarily through technology spillover effects that minimize the carbon emissions of the host country.²⁰ On the other hand, the Pollution Refuge hypothesis states that FDI will harm the environment of the host country;^15,21 specifically, as the intensity of environmental regulations continues to rise in developed countries, pollution-intensive industries are transferred to developing countries with relatively weak environmental regulations. This subsequently results in further environmental deterioration in the countries of transfer.²² However, there is no unanimity in the present research on the environmental impact of FDI on the host country.

The Chinese government has promulgated and implemented several emission reduction control measures to achieve emission reduction targets. The most typical approach is the Low-Carbon City Pilot Policy (LCCP) launched by the National Development and Reform Commission in 2010. The first batch of pilots included 8 cities in 5 provinces; this was further expanded to other regions in 2012 and 2017.²³ So far, the LCCP covers 6 provinces, 80 cities, and 1 region. The LCCP integrates innovative, coordinated, and green development concepts into modern urban construction. It is a “test field” for testing climate change policy, and injects new impetus into the improvement of urban energy efficiency. On the other hand, it also poses novel limitations to urban development. With the advancement of the LCCP, whether it promotes the sustainable development of the city has attracted widespread research attention.^24,25 Therefore, the impact of LCCP should not be neglected when discussing the impact of FDI on carbon emissions and its mechanisms. Carbon emission efficiency is a crucial indicator for evaluating carbon emission. Thus, reasonable emission reduction measures should focus on the amount and intensity of carbon emissions, as well as their efficiency.²⁶ Consequently, this work attempts to explore the actual association between FDI and carbon efficiency. Although numerous studies have explored the linear effect of FDI on CO2 emissions with different conclusions, only a few existing studies have estimated a nonlinear effect. Moreover, limited studies have used the LCCP to examine the impact of FDI on carbon emission efficiency. Inspired by the above, the possible contributions of this paper include: First, this paper uses the Meta-frontier and Nonradial Directional Distance Function (MNDDF) model to measure urban carbon emission efficiency. Secondly, we utilize the dynamic panel data and threshold model to empirically test the impact of FDI on urban carbon emission efficiency. Eventually, this work discusses the transmission mechanism of FDI on carbon emission efficiency and the moderating effect of LCCP.

As a result, we found that: (1) There is still potential for improvement in carbon emission efficiency; (2) FDI below a certain level has a negative effect on carbon emission efficiency, for instance, there is a nonlinear association between them; (3) FDI indirectly inhibits the improvement of carbon emission efficiency by promoting energy intensity. Nonetheless, the implementation of LCCP has a positive effect on carbon emission efficiency; (4) The implementation of LCCP has improved the negative impact of FDI on carbon emission efficiency, however, it significantly cannot influence the process that FDI affects carbon emission efficiency via energy intensity.

The rest of the paper is organized as follows: Section 2 reviews the relevant literature; Section 3 designs the research; Section 4 presents empirical analysis and results, then a summary of research conclusions and policy implications in Section 5.

Literature review

While FDI has a positive effect on the economic growth of the host country, it also influences the environment of the host country.^27,28 Existing studies indicate that cross-border capital flow has a negative impact on the environment of the host country, supporting the pollution refuge hypothesis.^6,^29,30 Notably, the pollution refuge hypothesis specifically states that as the intensity of environmental regulations continues to rise in developed countries, pollution-intensive industries are transferred to developing countries with relatively weak environmental regulations, resulting in further deterioration of the environment in the countries of transfer.²² Solarin et al.³¹ applied the autoregressive distributed lag (ARDL) boundary test method to confirm the existence of the pollution paradise hypothesis in Ghana. In China, Sun et al.³² used the ARDL lag model to analyze the impact of FDI on CO2 emissions to verify the efficacy of the pollution haven hypothesis. Elsewhere, Hamid et al.³³ and Mujtaba and Jena³⁴ also came to similar conclusions. On the other hand, FDI significantly improves the environmental quality of the host country, supporting the pollution halo hypothesis.^35,36 The hypothesis majorly believes that foreign direct investment triggers an advanced production technology to the host country, and a positive environmental impact on the host country via technology spillover.^20,³⁷ Zhu et al.³⁸ used Southeast Asian countries as a research sample to analyze the impact of FDI on carbon dioxide emissions and discovered that the hypothesis of the pollution halo effect exists in high-emission countries. Dasgupta et al.³⁹ suggested that developing countries mainly use the technology spillover effect of FDI to minimize their carbon emissions. Nevertheless, Cardoso and Rafaela²⁸ believe that the above two hypotheses exist due to the different resource endowments and economic development of each country, resulting in completely different effects of FDI on the host country's environment.

Existing research on carbon emission efficiency primarily focuses on influencing factors, regional differences, and measurement methods.^40–42 Wang et al.⁴³ utilized the dynamic panel threshold regression model to explore the impact of official development assistance (ODA) on carbon emissions in 59 countries. Meng and Niu⁴⁴ discovered that the effect of technological innovation on carbon emission efficiency is significantly greater than the upgrading of industrial structure. Lu et al.⁴⁵ believe that structural factors hinder carbon emission efficiency in China. Based on panel data of 134 countries between 1996 and 2015, Wang et al.⁴⁶ analyzed the impact mechanism of economic growth on carbon dioxide emissions and ecological footprint. Zhang and Zhang⁴⁷ used a spatial lag model to analyze the impact of foreign trade on carbon emission efficiency based on measuring the carbon emission efficiency in China. Zhang et al.⁴⁸ showed spatial differences in China's carbon emission efficiency, which is mainly reflected by the fact that the carbon emission efficiency in the eastern region is higher than that in other regions. Sun et al.⁴⁹ believe that there is a positive correlation between the global value chain (GVC) position index and carbon emission efficiency, i.e., if a country has a high position in the GVC, its carbon emission efficiency is high as well. Hu et al.⁴⁰ used the non-parametric DEA method to measure carbon emission efficiency. Consequently, the calculation outcomes showed that although China's carbon emission efficiency is annually improving, it is still at a relatively low level.

As a comprehensive environmental regulatory policy tool, LCCP is a fundamental strategy implemented by the central government of China to resolve the climate change crisis at the local level and plays a critical role in promoting green urban development.⁵⁰ The first batch of pilots included 8 cities in 5 provinces, and the scope was further expanded to other regions in 2012 and 2017.²³ So far, the pilot covers 6 provinces, 80 cities, and 1 region. The pilot cities were encouraged to implement comprehensive reforms to improve the efficiency of their production sectors and restructure their industries.⁵¹ The LCCP integrates innovative, coordinated, and green development concepts into modern urban construction. It is a “test field” for testing climate change policies, and introduces new impetus into the improvement of urban carbon emission efficiency. On the other hand, it also poses new challenges to urban development. With the progress of the national low-carbon city pilot policy, whether the policy promotes the sustainable development of cities has elicited global research attention.^24,25 By summarizing the existing researches, it can be seen that the research content mainly includes two aspects: On the one hand, it conducts research on the development status and path of LCCP in China. On the other hand, the performance evaluation of LCCP mainly includes the evaluation of carbon emission levels, technological innovation, and industrial structure upgrading and other aspects.⁵²^,53

Based on the summary of the existing literature, although the effect of FDI on the environment of the host country is relatively abundant, a consistent conclusion has not yet been reached; Secondly, although Shao⁵⁴ explored the mediating role of energy intensity in the process of FDI affecting emissions. Nevertheless, limited research exists on the impact of FDI on carbon emission efficiency as well as its internal mechanism; besides, the implementation of LCCP could have a critical impact on FDI and carbon emission efficiency. As such, the possible moderating role of LCCP in FDI on carbon emission efficiency should be further analyzed. Moreover, most of the existing studies only focus on the linear relationship between FDI and the environment, yet less on the nonlinear relationship between them. The aforementioned concerns also leave room for further probe on this paper. The innovations of this paper are as follows: First, most of the existing studies only focus on the linear relationship between FDI and the environment, but less on the nonlinear relationship between them. Therefore, the threshold model is used to empirically test the impact of FDI on carbon emission efficiency. Secondly, this paper examines the mechanism of FDI on carbon emission efficiency, and to further assesses the role of LCCP in the impact process.

Research design

Methods of measuring carbon emission efficiency

Based on the existing researches, this paper uses the MNDDF model to measure energy efficiency.^40,55 In this paper, each city is taken as a decision-making unit (DMU) to build the frontier of production. Each DMU uses capital (K), labor (L) and energy (E) as input indicators, and the city's GDP (Y) as expected output, and carbon dioxide emissions (CO2) as undesired output. Therefore, the technical production set (T) can be defined as the following formula:

T = {(K, L, Y, C O 2) : (K, L, E) c a n p r o d u c e (Y, C O 2)}

(1)

Färe et al.⁵⁶ pointed out that the technological production set (T) has four characteristics. One is weak disposability, that is, in the case of reducing pollution, the expected output is also reduced. The second is strong disposability, that is, when the input and pollution are maintained at a certain level, the expected output can be more or less, and the gap between them reflects the difference in technical efficiency under environmental regulations. The third is that the expected output and undesired output exist at the same time. If there is no undesired output, the expected output is zero. The fourth is that the input elements are freely available.

Based on the above characteristics, the technical production set containing undesired output can be expressed as follows:

\begin{aligned} T = {(K, L, E, Y, C O 2) : \sum_{n = 1}^{N} γ_{n} K_{n} \leq K, \sum_{n = 1}^{N} γ_{n} L_{n} \leq L, \\ \sum_{n = 1}^{N} γ_{n} E_{n} \leq E, \sum_{n = 1}^{N} γ_{n} Y_{n} \leq Y, \sum_{n = 1}^{N} γ_{n} C O 2_{n} \leq C O 2, n = 1, 2, \dots, N} \end{aligned}

(2)

Expressed in Eq. 2,

γ_{n}

is the weight of each cross-section. If

γ_{n} \geq 0, \sum_{n = 1}^{N} γ_{n} = 1

, it means that the variable scale return (VRS), otherwise constant scale return. In order to measure the carbon emission efficiency of each DMU, a non-radial distance function is constructed based on the practice of Zhang et al.⁵⁷

\vec{D} (K, L, E, Y, C O 2; g) = s u p {w^{T} β : [(K, L, E, Y, C O 2) + g * d i a g (β)] \in T}

(3)

In Eq. 3,

w^{T} = (w_{K}, w_{L}, w_{E}, w_{Y}, w_{C O 2})

represents the standardized weight vector. As there are three input variables, one desirable output, and one undesirable output, we set

w^{T} = (\frac{1}{9}, \frac{1}{9}, \frac{1}{9}, \frac{1}{3}, \frac{1}{3})

g = (- g_{K}, - g_{L}, - g_{E,} - g_{Y,} - g_{C O 2})

represents the direction vector, which is set as g = (-K, -L, -E, -Y, -CO2).

β = (β_{K}, β_{L}, β_{E}, β_{Y}, β_{C O 2}) \geq 0

indicates the relaxation vector, which needs to be obtained by solving the following linear program:

\vec{D} (K, L, E, Y, C O 2; g) = m a x \frac{β_{K}}{9} + \frac{β_{L}}{9} + \frac{β_{E}}{9} + \frac{β_{Y}}{3} + \frac{β_{C O 2}}{3}

(4)

\sum_{n = 1}^{N} γ_{n} K_{n} \leq (1 - β_{K}) K

\sum_{n = 1}^{N} γ_{n} L_{n} \leq (1 - β_{L}) L

s . t . \sum_{n = 1}^{N} γ_{n} E_{n} \leq (1 - β_{E}) E

\sum_{n = 1}^{N} γ_{n} Y_{n} \leq (1 - β_{Y}) Y

\sum_{n = 1}^{N} γ_{n} C O 2_{n} \leq (1 - β_{C O 2}) C O 2

γ_{n} \geq 0, n = 1, 2, \dots, N

β_{Y} \geq 0, 0 \leq β_{K}, β_{L}, β_{E}, β_{C O 2} \leq 1

According to the researches of Cheng et al.,⁵⁸ the carbon emission index (TCEI) can be expressed as:

TCEI = \frac{C O 2 - β_{C O 2} C O 2 / (Y + β_{Y} Y)}{C O 2 / Y} = \frac{(1 - β_{C O 2})}{(1 + β_{Y})}

However, the calculation of TCEI is based on the fact that all DMUs have similar production technologies and face the same production frontier, without considering the technological gap between different DMUs, which may lead to deviations in the calculation results. To overcome it, we follow Battese et al.⁵⁹ and O’Donnell et al.⁶⁰ construct three frontiers of the contemporaneous technology sets, the intertemporal technology sets, and the meta-frontier technology set.

Firstly, we classify all DUMs into H groups. The group h has N^h DMUs and $\sum_{h = 1}^{H} N^{h} = N$ . The input and output combinations of all DMUs in each group belong to the same production technology set R_h:

T_{R_{h}}^{C O 2} = {(K^{t}, L^{t}, Y^{t}, C O 2^{t}) : (K^{t}, L^{t}, E^{t}) c a n p r o d u c e (Y^{t}, C O 2^{t})

In Eq. 6,

T_{R_{h}}^{C O 2}

only contains all the DMUs of group R_h in period t. Furthermore, the intertemporal environmental technology set of groups R_h can be defined as

T_{R_{h}}^{I} = T_{R_{h}}^{1} \cup T_{R_{h}}^{2} \cup \dots T_{R_{h}}^{T}

. This set contains all the DUMs of group R_h at all times. The global environmental technology set can be defined as

T^{G} = T_{R_{1}}^{I} \cup T_{R_{2}}^{I} \cup \dots T_{R_{H}}^{I}

and contains all the DUMs of all groups at all times. Based on (4), we the construct the following contemporaneous NDFF:

{\vec{D}}^{C O 2} (K, L, E, Y, C O 2; g^{C O 2}) = m a x \frac{β_{K}^{C O 2}}{9} + \frac{β_{L}^{C O 2}}{9} + \frac{β_{E}^{C O 2}}{9} + \frac{β_{Y}^{C O 2}}{3} + \frac{β_{C O 2}^{C O 2}}{3}

\sum_{n = 1}^{N} γ_{n}^{t} K_{n}^{t} \leq (1 - β_{K}^{C O 2}) K

\sum_{n = 1}^{N} γ_{n}^{t} L_{n}^{t} \leq (1 - β_{L}^{C O 2}) L

s . t . \sum_{n = 1}^{N} γ_{n}^{t} E_{n}^{t} \leq (1 - β_{E}^{C O 2}) E

\sum_{n = 1}^{N} γ_{n}^{t} Y_{n}^{t} \leq (1 - β_{Y}^{C O 2}) Y

\sum_{n = 1}^{N} γ_{n}^{t} C O 2_{n}^{t} \leq (1 - β_{C O 2}^{C O 2}) C O 2

γ_{n}^{t} \geq 0, n = 1, 2, \dots, N^{h}

β_{Y}^{C O 2} \geq 0, 0 \leq β_{K}^{C O 2}, β_{L}^{C O 2}, β_{E}^{C O 2}, β_{C O 2}^{C O 2} \leq 1

According to (5), the contemporaneous-frontier TCEI can be defined as follows:

TCEI = = \frac{(1 - β_{C O 2}^{C O 2})}{(1 + β_{Y}^{C O 2})}

Empirical model

Panel data has both cross-sectional and time dimensions. There are a big number of observations and large sample size, which improve the efficacy of estimation.⁶¹ Thus, this paper considers FDI as the core explanatory variable and applies a panel data model to empirically evaluate the impact of FDI on carbon emission efficiency. We incorporate the lagging term of the explanatory variable into the empirical model to further investigate the impact of FDI on carbon emission efficiency under dynamic changes. Since the one-period lag term of the explanatory variable is added to the model, the model cannot meet the complete homogeneity assumption, resulting in biased and inconsistent estimates of the fixed effects or random effects of the ordinary panel. To resolve the potential endogenous issues, we apply the system GMM for quantitative analysis. Of note, the system GMM method proposed by Arellano and Bover⁶² and Blundell and Bond⁶³ properly addresses the endogeneity problem and effectively improves the estimation efficiency by introducing additional instrumental variables.⁶⁴ Based on existing studies, the following dynamic panel regression model is constructed:

T C E I_{i t} = α_{0} + α_{1} T C E I_{i t - 1} + α_{2} F D I_{i t} + \sum β_{i} c o n t r o l_{i t} + μ_{i} + ε_{i t}

In Eq. 9, i represents each city (1, 2, …, 260), t represents time (2003, 2004, …, 2016), the explained variable TCEI represents the carbon emission efficiency, and the core explanatory variable FDI represents the level of foreign direct investment by the ratio of actual foreign investment and fixed asset investment. control represents the control variable, α and β are the coefficient matrix of the core explanatory variable and the control variable, respectively, μ is the fixed effect coefficient, and ε_it is the random error term.

In order to further test the nonlinear relationship between FDI and carbon emission efficiency, this paper draws on the ideas of Hansen⁶⁵ threshold model to construct a threshold model with FDI as the threshold variable. The model is as follows:

T C E I_{i t} = α_{0} + α_{1} F D I_{i t} (q_{i} > γ) + α_{2} F D I_{i t} (q_{i} \leq γ) + \sum β_{i} c o n t r o l_{i t} + μ_{i} + ε_{i t}

In Eq. 10, q_i is the threshold variable, γ is the threshold value, and the remaining variables and regression coefficients are consistent with equation (9).

Data and variables

Data source

This paper uses panel data comprising 260 cities in China (some prefecture-level cities with newly added, adjusted, or missing data are excluded, and the statistical scope covers 260 prefecture-level cities in China) to empirically evaluate the impact of FDI on carbon emission efficiency between 2003 and 2016. The data primarily originates from the China City Statistical Yearbook, China Statistical Yearbook, China Energy Statistical Yearbook, and provincial statistical yearbooks. All nominal data were based on 2003, and the gross production and consumer price indices were used to deflate to the actual data. A small portion of missing data was estimated by trend fitting, at the same time, the data processing mostly adopted the ratio method given the possible heteroscedasticity.

Variable and definition

Input and output variables: Based on the existing research, when calculating carbon emission efficiency, the input factors majorly include capital, labor, and energy, whereas the output includes expected and undesired outputs.^40,66 Among them, capital input is represented by the capital stock. Because official survey data of the capital stock in China are currently unavailable, this work utilizes the perpetual inventory method to estimate the capital stock of cities.^67,68 The ideal index of labor input should include the time and quality of labor input, however, due to the lack of relevant statistical data. At the same time, considering the greater mobility of urban employees, labor input is measured by the average number of employees in the previous year and that in the present year.⁶⁹ Since China's city-level data does not include major energy consumption data such as coal and oil. Therefore, refer to Fu et al.⁵¹ to use urban electricity consumption instead of energy input indicator. The expected output selects the gross product value of the city to reflect the economic growth.⁷⁰ This paper measures the undesired output by CO2 emissions.⁷¹

Empirical variables: The explained variable and the core explanatory variable of this paper have been comprehensively illustrated in the previous paper. Based on existing studies, the urban economic development level (GDP), industrial structure (ST), industrialization level (IN), technological innovation (TE), and population density (DP) are the control variables.⁴⁰^,72 Among them, the level of urban economic development is measured by the natural logarithm of the urban per capita GDP. The industrial structure is measured by the industrial structure upgrading index, i.e., S = r1*1 + r2*2 + r3*3, where S represents the industrial structure upgrading index; r1, r2, and r3 are the proportion of the output value of the primary industry, the secondary industry and the tertiary industry in the urban GDP, respectively. The level of industrialization is expressed by the ratio of total industrial output value to the total output value of the city. Technological innovation adopts the scientific research, technical services, and geological prospecting practitioners of the city.⁷³ The population density uses the natural logarithm of the population per unit area of land in the city.

Data descriptive statistics

Table 1 details the descriptive statistics of the main variables. The average carbon emission efficiency is only 0.49, and there with a certain gap from the optimal carbon emission efficiency value. The maximum value is 1, whereas the minimum value is only 0.006, indicating a large difference in carbon emission efficiency between cities.

Table 1.

Descriptive statistics of the main variables.

Variable	Mean	Max	Min	Std. Dev	P25	P50	P75	P95	Obs.
TCEI	0.4884	1.0000	0.0060	0.3053	0.2126	0.4389	0.7322	1.000	3640
FDI	0.0428	0.4683	0.0020	0.0387	0.0104	0.0234	0.0532	0.1529	3640
GDP	10.0775	13.1349	7.5449	0.8883	9.4318	10.0601	10.6653	11.6269	3640
ST	2.2223	2.7972	1.8223	0.1409	2.1249	2.2074	2.3110	2.4736	3640
IN	1.2268	4.4200	0.0703	0.6128	0.7773	1.1482	1.5830	2.3358	3640
TE	1.0741	68.9751	0.0100	3.5486	0.1900	0.3400	0.6405	4.0975	3640
DP	5.7307	7.8867	1.5476	0.8903	5.2143	5.8279	6.4289	6.8652	3640

Source: Authors’ calculation. Notes: Obs. stands for the observations of the variables, Mean refers to the average value of the variables, Std. Dev. represents standard deviation, Min and Max indicate the minimum and maximum values of the variables, respectively.

Results

Carbon emission efficiency measurement results

This paper uses the MNDDF method to measure the carbon emission efficiency of 260 cities in China. Due to space limitations, we could not provide the annual carbon emission efficiency value of each city. Therefore, we drew the average change trend of China's overall and regional carbon emission efficiency between 2003 and 2016 is drawn (Figure 1). As shown in Figure 1, the carbon emission efficiency generally tends to increase, which is in line with the findings of Sun and Huang (2020). Further, we observed that the average carbon emission efficiency steadily increased at an average rate of 9.71% during the study period. The growth rate in 2008 was the largest, a 20.9% increase from the previous year. This is potentially attributed to the global economic crisis in 2008, which decreased the production scale and energy consumption, causing a relative increase in carbon emission efficiency. Although the average carbon emission efficiency improved during the study sample period, the average level in 2016 remained less than 1, with an optimal carbon emission efficiency gap of 17.6%. This also shows that the carbon emission efficiency in China has much room for improvement.

Figure 1.

Trends in carbon emission efficiency of China's overall and regional from 2003 to 2016. Source: Authors’ drawing.

There is significant heterogeneity when it comes to comparing different regions. Specifically, the Eastern region has the highest average carbon emission efficiency, whereas the Central region has the lowest average as also reported by Cai et al.⁷⁴ This could be because the Central and Western regions lag behind the Eastern regions based on the economic development level, advanced technology utilization, and industrial structure. Moreover, the development of the Central and Western regions is based on traditional industries consuming enormous fossil fuels, thus the carbon emission efficiency is at a relatively low level. The Eastern region is mainly dominated by the developed tertiary industries, with low energy consumption demand, and its technological level is relatively higher than that of the Central and Western regions. Thus, the carbon emission efficiency in the Eastern region is higher than that in the Central and Western regions.

Variable stationarity test

In order to avoid regression and ensure the unbiasedness and validity of the results, this paper conducts a stationarity test on the variables. From the test results, it can be seen that each variable is a horizontal series stationary, which can be perform regression analysis. The results of the test are shown in Table 2 below:

Table 2.

Results of unit root test.

Variables	LLC	HT	IPS
TCEI	−70.238***	−3.0e + 02***	−44.263***
FDI	−62.034***	−2.9e + 02***	−43.433***
GDP	−61.258***	−3.0e +−02***	−43.876***
ST	−66.609***	−2.8e + 02***	−42.008***
IN	−61.279***	−3.0e + 02***	−44.269***
TE	−73.846***	−3.1e + 02***	−44.683***
DP	−68.175***	−2.9e + 02***	−42.692***

Source: Authors’ calculation. Note: ***p < 0.01, **p < 0.05, *p < 0.1.

Empirical results

Dynamic panel data model estimation results

(1) Benchmark model regression

The system method proposed by Arellano and Bover⁶² and Blundell and Bond⁶³ properly addresses the endogeneity problem and effectively improves the estimation efficiency by introducing additional instrumental variables. Therefore, we used the system method to estimate the parameters of the constructed panel data model. At the same time, the robust standard error was used to eliminate the effect of heteroscedasticity on the model estimation results. Further considering the differences in carbon emission efficiency between different regions and the reasons for the robustness of the model estimation results, this study classifies and discusses the full sample and different regions, and selects the Stata15 measurement software to perform regression on the measurement model. Nevertheless, a few scholars have criticized the problem of weak instrumental variables in the systematic GMM method.⁷⁵ Therefore, it is crucial to evaluate the robustness of the estimation results. Thus, AR (1), AR (2), and Sargan tests were used for autocorrelation and over-identification tests to ensure the overall validity of the model. Based on the AR (1) and AR (2) test results, the disturbance items have a first-order serial correlation, but no second-order serial correlation; hence, the corresponding Sargan test P-value is high, and the over-identification test is passed. This suggests that the model regression results are reliable and effective. The regression results are shown in Table 3.

Table 3.

Estimation results of dynamic panel data model.

Variable	Total	East	Middle	West
L.TCEI	0.796*** (0.037)	0.729*** (0.039)	0.752*** (0.079)	0.652*** (0.066)
FDI	−0.815*** (0.119)	−0.764*** (0.278)	−1.241*** (0.279)	−0.654*** (0.058)
GDP	0.111*** (0.018)	0.110*** (0.028)	0.150*** (0.043)	0.184*** (0.026)
ST	0.479*** (0.120)	0.362*** (0.130)	0.436*** (0.096)	0.439 (0.278)
IN	−0.001 (0.011)	0.009 (0.013)	−0.011 (0.017)	−0.001 (0.024)
TE	0.017*** (0.003)	0.019*** (0.003)	0.007 (0.005)	0.004 (0.008)
DP	−0.015 (0.015)	−0.029 (0.021)	−0.024 (0.017)	0.031 (0.020)
_cons	−1.914*** (0.289)	−1.587*** (0.423)	−2.087*** (0.449)	−2.708*** (0.584)
Obs.	3379	1274	1131	974
Sargan Test	0.5037	0.6921	0.6479	0.6802
AR(1)	0.0000	0.0000	0.0003	0.0005
AR(2)	0.4702	0.5700	0.4183	0.3204

Source: Authors’ calculation. Note: The parenthesis are the robust standard error values, ***p < 0.01, **p < 0.05, *p < 0.1.

Based on the regression results of the model, the regression coefficient of L.TCEI is positive at a significant level of 1%. This indicates that the previous carbon emission efficiency has a certain promoting effect on the current carbon emission efficiency. That is, carbon emission efficiency has a certain cumulative effect and path dependence. From the regression coefficient of FDI that they are all negative at a significant level of 1%, i.e., FDI inhibits the improvement of carbon emission efficiency. This also indicates the existence of the Pollution Haven Hypothesis. The reason could be that cheap labor in the early stage provides good basic conditions for attracting foreign investment; besides, the existence of a “GDP competition” promotion mechanism in local governments often ignores the impact on the environment when introducing foreign investment. Consequently, some high-emission projects were introduced, which increased carbon emissions, and ultimately demonstrated a certain negative impact on carbon emission efficiency. The regression coefficients of economic development level and industrial structure transformation are both positive, nevertheless, the regression coefficients of industrial structure transformation in the Western region are insignificant. That is, economic development and industrial structure transformation are conducive to the improvement of carbon emission efficiency. This could be because, on one hand, continuous economic development provides a material basis for cleaner production; on the other hand, due to the improvement of economic development, there is an increased awareness of environmental protection. This has increased the demand for cleaner products, thus promoting cleaner production, ultimately improving energy efficiency and reducing carbon emissions. The process of transformation and upgrading of the industrial structure is continuously leaning towards the direction of the service industry dominated by the tertiary industry, hence reducing energy consumption and carbon emissions. The regression coefficient of technological innovation is positive at a significant level of 1% in the overall sample regression of the Eastern region; however, it is insignificant in the regression of the Central and Western regions. Hu et al.⁴⁰ also draw a similar conclusion. Furthermore, the level of industrialization and population density exerts an inhibitory effect on carbon emission efficiency but is statistically insignificant. With the continuous increase of energy consumption required by the improvement of the level of industrialization, a severe test has been suggested to improve the efficiency of carbon emission. At the same time, the increase in population density increases domestic energy consumption, particularly domestic power consumption. In China, power generation is primarily generated by coal, which is not conducive to the improvement of carbon emission efficiency.

(2) Endogeneity test

In order to further deal with the possible endogeneity problem and ensure the robustness and reliability of the empirical research conclusions in this paper, in Table 4, we use the instrumental variable method (IV) to regress the model. In this paper, the selection of instrumental variables is based on a relatively conventional practice, and the first-order lag term of endogenous variables is used as its own instrumental variable. The regression results show that the regression results of the impact of FDI on carbon emission efficiency are basically consistent with the conclusions of the benchmark regression, indicating that the core conclusions of this paper are still relatively robust after considering endogeneity.

(3) Robustness test

Table 4.

Estimation results of endogeneity test.

Variable	IV - 2SLS
	The first stage	The second stage
FDI		−1.044*** (0.108)
L.FDI	0.610*** (0.054)
Control variables	Yes	Yes
_cons	−0.059*** (0.013)	−1.736*** (0.068)
Shea's partial R²		0.5739
Cragg-Donald Wald F		4892.93
Obs.	3640	3640

Source: Authors’ calculation. Note: The parenthesis are the robust standard error values, ***p < 0.01, **p < 0.05, *p < 0.1.

To obtain more robust empirical results, we performed robustness tests from the following two aspects: First, to minimize the impact of sample time selection, this paper re-estimated the study period between 2004 and 2015 after excluding the two-year samples in 2003 and 2016. Secondly, we eliminated the effect of outliers, winsoring the highest and lowest 1% of all continuous variables, and re-estimated the benchmark model. Finally, there could be random error terms that do not meet the model assumptions in the above model estimation. The obtained results may be biased if the parameter estimation is directly performed. Generalized least squares (FGLS) effectively resolve the problems of serial correlation and heteroscedasticity caused by cross-sectional data and obtain more effective estimation results.⁷⁶ Nonetheless, due to the nature of the data in this paper, the standard deviation of the FGLS method may not effectively reflect its variation. In this case, panel correction standard error estimation (PCSE) should be used for correction to obtain a more accurate estimation result. The results of the above robustness test reveal that the signs and sizes of the core explanatory variable have not significantly changed, indicating that our research conclusions have desirable robustness Table 5.

Table 5.

Estimation results of robustness test.

Variable	(1)	(2)	(3)
L.TCEI	0.743*** (0.043)	0.792*** (0.036)
FDI	−0.875*** (0.138)	−0.963*** (0.234)	−0.876*** (0.116)
Control variables	Yes	Yes	Yes
_cons	−1.988*** (0.352)	−1.968*** (0.315)	−1.703*** (0.143)
Obs.	2895	3379	3640
Sargan Test	0.6041	0.5928
AR(1)	0.0000	0.0000
AR(2)	0.5305	0.5010
Wald			666.67
R-squared			0.4215

Source: Authors’ calculation. Note: The parenthesis are the robust standard error values, ***p < 0.01, **p < 0.05, *p < 0.1.

Threshold model estimation results

(1) Threshold effect test

Based on the construction of the panel threshold model, the single, double and triple threshold hypotheses were selected, respectively, and the Eq. 10 was self-sampled 300 times to obtain the threshold variable F statistics, and then the self-service method was selected to estimate the P value. Further, each threshold value and 95% confidence interval are calculated. The results are shown in Table 6 and Table 7. According to the test results, the single threshold and the double threshold are both significant at the 1% level, while the triple threshold is not significant, so the above threshold model has two threshold values. From the perspective of specific threshold values, the threshold values with FDI as the threshold variable are 0.0279 and 0.0813, respectively. It shows that during the sample period of the study, when FDI is at different levels, it has a heterogeneous impact on carbon emission efficiency.

Table 6.

Threshold effect self-sampling test.

Model	F value	P value	BS times
Single threshold	39.63***	0.0067	300
Double threshold	37.89***	0.0067	300
Triple threshold	12.53	0.8500	300

Source: Authors’ calculation. Note: ***p < 0.01, **p < 0.05, *p < 0.1.

Table 7.

Threshold estimation and 95% confidence interval.

Threshold estimate	95% confidence interval
0.0279	[0.0264, 0.0280]
0.0813	[0.0795, 0.0820]

Source: Authors’ calculation.

In order to make the test results more intuitive, the article draws the estimated likelihood ratio statistics of each threshold value according to the principle of the threshold model (Figures 2–4). The dotted line in the figure determines the LR test 95% confidence interval of the threshold. It can be seen from the figure below that there are multiple estimation results in Figure 4, so it is difficult to determine the actual existence of the estimated value. Combined with the P value of the joint test in Table 4, the performance of the double threshold model can be determined to be the best

(2) Regression analysis

Figure 2.

Confidence interval of a single threshold model.

Figure 3.

Confidence interval of the double threshold model.

Figure 4.

Confidence interval of the triple threshold model. Source: Authors’ drawing.

Table 8 reports the empirical regression results. As shown, there are significant differences in the impact of FDI on carbon emission efficiency at different levels. Specifically, as the level of FDI increases, its impact on carbon emission efficiency shifts from suppression to promotion. At present, the average FDI level in China is relatively low and has not yet played a role in promoting carbon emission efficiency. When FDI is at a low level, its inhibitory effect on carbon emission efficiency is remarkable; nevertheless, as long as it enters a certain stage, this negative hindrance will turn into a positive promotion. This is because when FDI is at a low level, its technology spillover effect is unapparent, primarily because the scale of foreign investment is small at this time, and there is no direct investment to build factories in China. Therefore, advanced technology has not been brought to China. Secondly, when FDI is at a low level, the host country is eager to increase the quantity of FDI without effectively screening its quality, causing an inflow of poorly qualified FDI. For instance, with the inflow of FDI, enterprises with high energy consumption and high pollution emerge, thereby suppressing the improvement of carbon emission efficiency. With the continuous improvement of FDI levels, on one hand, the technology spillover effect of FDI becomes prominent, thereby improving the technology and management level of the host country; this is conducive to the improvement of carbon emission efficiency. On the other hand, the screening of the FDI quality increases in the host country. For example, applying the implementation of environmental regulatory measures to screen the FDI quality, and prohibiting the entry of enterprises that do not meet the standards; also has a positive effect on the carbon emission efficiency of the host country.

Table 8.

Estimation results of panel threshold model.

Variable	TCEI
GDP	0.389*** (0.007)
ST	0.634*** (0.061)
IN	−0.024 (0.019)
TE	0.015*** (0.003)
DP	−0.035* (0.018)
FDI(FDI $\leq γ_{1}$ )	−3.594*** (0.501)
FDI( $γ_{1} < F D I \leq γ_{2}$ )	−0.988*** (0.178)
FDI( $F D I > γ_{2}$ )	0.034 (0.073)
_cons	−4.550*** (0.146)
Obs.	3640
F	829.26

Source: Authors’ calculation. Note: The parenthesis are the robust standard error values, ***p < 0.01, **p < 0.05, *p < 0.1.

Examination for mechanism

First, existing studies believe that fossil energy consumption is the major cause of environmental pollution. Regarding the relationship between energy intensity and environmental quality, existing research has also focused on empirical tests.^77,78 As an important factor affecting energy intensity, technological progress plays a critical role in reducing energy intensity.⁷⁹ For the vast number of developing countries and emerging economies, scholars have debated whether FDI, as a major type of international technology transfer and technology diffusion, will benefit the energy intensity of the region, however, no consensus has yet been reached.^69,80 Secondly, as a bottom-up emission reduction measure, the LCCP primarily aims at building a low-carbon industrial system by determining a sound low-carbon development mechanism and improving energy efficiency.⁵⁰ The LCCP mainly reduces pollution emissions in the production process by targeting the scale of production and improving the level of technological innovation of enterprises.⁸¹ Specifically, on one hand, the LCCP as a type of environmental regulation minimizes carbon emissions by squeezing out FDI with high energy consumption and high pollution. On the other hand, the implementation of LCCP improves energy efficiency by promoting the progress of green technology, ultimately contributing to the improvement of carbon emission efficiency (Song et al., 2020). Therefore, based on the existing literature, the mediating effect model with energy intensity as its variable and the moderating effect model with the implementation of LCCP as its variable has been constructed.⁶⁹^,82 The specific model is set as follows.

E I_{i t} = α_{0} + α_{1} F D I_{i t} + \sum β_{i} c o n t r o l_{i t} + μ_{i} + ε_{i t}

T C E I_{i t} = α_{0} + α_{1} T C E I_{i t - 1} + α_{2} F D I_{i t} + α_{3} E I_{i t} + \sum β_{i} c o n t r o l_{i t} + μ_{i} + ε_{i t}

T C E I_{i t} = α_{0} + α_{1} T C E I_{i t - 1} + α_{2} F D I_{i t} + α_{3} E I_{i t} + α_{4} L C C P_{i t}

+ α_{5} (L C C P \times F D I)_{i t} + \sum β_{i} c o n t r o l_{i t} + μ_{i} + ε_{i t}

T C E I_{i t} = α_{0} + α_{1} T C E I_{i t - 1} + α_{2} F D I_{i t} + α_{3} E I_{i t} + α_{4} L C C P_{i t}

+ α_{5} (L C C P \times E I)_{i t} + \sum β_{i} c o n t r o l_{i t} + μ_{i} + ε_{i t}

T C E I_{i t} = α_{0} + α_{1} T C E I_{i t - 1} + α_{2} F D I_{i t} + α_{3} E I_{i t} + α_{4} L C C P_{i t}

+ α_{5} (L C C P \times F D I)_{i t} + α_{6} (L C C P \times E I)_{i t} + \sum β_{i} c o n t r o l_{i t} + μ_{i} + ε_{i t}

In which, EI is energy intensity. At present, energy consumption in China is dominated by electricity.⁸³ Based on data availability, the natural logarithm of electricity consumption per unit of GDP is used as a measure of energy intensity.⁵¹ LCCP is the low-carbon city pilot policy. The National Development and Reform Commission officially issued the Notice on Carrying out Low-Carbon Provinces and Low-Carbon Cities Pilot Work on July 19, 2010, and launched the first batch of low-carbon provinces and low-carbon cities. The pilot areas included 5 provinces including Guangdong and 8 cities including Tianjin. Based on the experience of the first batch of pilot cities, the list of the second batch of low-carbon pilot cities was determined in 2012. The pilot areas were major cities, supplemented by provinces, including 1 province and 28 cities. In 2017, 45 cities (districts and counties) were identified as the third batch of pilot projects for low-carbon cities.²³ Since then, the LCCP has been fully implemented nationwide. Due to the serious lack of data in some cities, the small number of pilot cities in the first batch, and the short duration of the third batch of pilot cities, we considered 20 cities in the second batch of low-carbon pilot cities as the treatment group; the remaining 240 cities were considered the control group. Based on this, dummy variables were constructed for the LCCP.

Eq. 11-15 were regressed, respectively to further analyze the effect mechanism of FDI on carbon emission efficiency. The regression results in Table 9 show a significantly positive regression coefficient of FDI, indicating that FDI increased the energy intensity. This shows that the level of energy consumption increased with the initial increase of FDI; this is consistent with the study of Wang et al.⁶⁹. The regression coefficient of EI was significantly negative, indicating that energy intensity has a significant negative effect on carbon emission efficiency, which is consistent with general cognition. Nonetheless, the significance of the regression coefficient of FDI did not significantly change after adding the mediating variable energy intensity to the benchmark regression model. This demonstrates that energy intensity has a partial mediating effect in the process of FDI affecting carbon emission efficiency. Furthermore, based on the benchmark model, the dummy variables of LCCP and its interaction with FDI and EI were added. The study found that the regression coefficient of LCCP is positive, yet insignificant. This shows that the LCCP has a certain positive effect on carbon emission efficiency, however, from the current point of view, its promotion effect is not apparent. To a certain extent, the LCCP forces cities to conduct technological innovation to improve carbon emission efficiency; however, since the policy design needs further improvement, its role is not remarkable at present. The interaction coefficient between LCCP and FDI is positive, indicating that the implementation of LCCP improved the negative impact of FDI on carbon emission efficiency. The coefficient of the interaction term between the LCCP and EI is negative, i.e., the implementation of the LCCP will not influence carbon emission efficiency by reducing energy intensity.

Table 9.

Estimation results of mechanism test.

Variable	EI	TCEI	TCEI	TCEI	TCEI
L.TCEI		0.791*** (0.037)	0.792*** (0.037)	0.784*** (0.037)	0.784*** (0.037)
FDI	0.381** (0.155)	−0.802*** (0.110)	−0.710*** (0.118)	−0.714*** (0.118)	−0.796*** (0.126)
EI		−0.067** (0.031)	−0.065 (0.051)	−0.072 (0.050)	−0.072 (0.050)
LCCP			0.038 (0.026)	0.110 (0.092)	0.116 (0.089)
LCCP × FDI			0.054 (0.233)		0.110 (0.254)
LCCP × EI				−0.359 (0.445)	−0.360 (0.422)
Control	Yes	Yes	Yes	Yes	Yes
_cons	0.551 (0.513)	−1.890*** (0.294)	−1.858*** (0.290)	−1.824*** (0.288)	−1.827*** (0.291)
Obs.	3380	3380	3380	3380	3380
Sargan Test	0.7259	0.5931	0.7216	0.6025	0.7323
AR(1)	0.0808	0.0000	0.0000	0.0000	0.0000
AR(2)	0.9066	0.1406	0. 1410	0.1272	0.1183

Source: Authors’ calculation. Note: The parenthesis are the robust standard error values, ***p < 0.01, **p < 0.05, *p < 0.1.

Conclusions and policy implications

Since entering the 21st century, the use of FDI in China has gradually increased, from US$60.63 billion in 2004 to US$134.97 billion in 2018. Although the rapid growth of FDI regulates the rapid development of China's economy, it has a significant impact on the environment. Therefore, this paper uses prefecture-level city-level panel data and the MNDDF model to measure the urban carbon emission efficiency. Furthermore, we empirically tested the impact of FDI on carbon emission efficiency and its transmission mechanism. Based on the above research content, the following research conclusions are drawn: (1) The average carbon emission efficiency increased during the 2003–2016 period, however, there is still a gap between the optimal carbon emission efficiency; (2) FDI is one of the fundamental factors that inhibit the improvement of carbon emission efficiency, and there is a non-linear relationship between them; (3) FDI indirectly inhibits the improvement of carbon emission efficiency by promoting energy intensity. Nonetheless, LCCP implementation has a positive effect on carbon emission efficiency; (4) The implementation of LCCP has improved the negative impact of FDI on carbon emission efficiency, however, it could significantly influence the process that FDI affects carbon emission efficiency via energy intensity.

The findings of this paper provide insights helping the government in environmental policy formulation: (1) The introduction of FDI should be further increased, and the capacity to absorb the spillover effects of FDI technology should be improved. First, this paper shows that only when the introduction of FDI exceeds a certain threshold level, it gradually exerts a positive effect on carbon emission efficiency. Therefore, the government should continue to expand the field of FDI and improve the convenience of business. At the same time, attention should be channeled to the quality of FDI introduction and improve the accessibility standards for foreign investment projects in terms of energy consumption to attract and retain high-quality FDI. Secondly, all regions should further fully utilize the technology spillover effect of FDI. The full utilization of the technology spillover effect of FDI is based on the host country's desired absorption. Therefore, enterprises should focus on improving their technology while increasing investment in scientific research; they should perform independent innovation based on learning and imitating the advanced production technology of foreign enterprises, and gradually construct an efficient and low-carbon industrial system. (2) The energy consumption structure should be continuously adjusted to minimize energy intensity. On one hand, while vigorously advocating the use of new energy and renewable energy, China should strive to construct an efficient, economical, clean, and sustainable coal supply system that is in line with the requirements of a low-carbon economy. On the other hand, the government should increase its support for technological development and increase the level of technological innovation to achieve the goal of reducing energy intensity. (3) Besides, it should further expand the scope of the low-carbon city pilot project, and actively explore low-carbon city development models. On one hand, the government should promote the experience of successful pilot cities and encourage cities to optimize the design of policy combinations to achieve the goal of optimizing energy structure and improving carbon emission efficiency. On the other hand, government departments should provide LCCP with policy support and theoretical guidance to continuously explore innovative forms of low-carbon city development models. For instance, the policy evaluation mechanism for low-carbon pilot cities should be improved to guarantee the sustainability of policy effects.

Due to data limitations, more micro-data should be considered. Therefore, with the improvement of micro-data, further research should start from micro-enterprises to consider the impact of FDI on carbon emission efficiency.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Chongqing Municipal Education Commission Innovation Project, (grant number CYS18080).

ORCID iD

Lianghu Wang

Lianghu Wang is PhD candidate of Industrial Economics in Southeast University. He got master's degree about industrial economics at Southwest University in 2020. His research interests embrace environmental economics and industrial economics. During the postgraduate period, she has participated in many provincial and ministerial projects from supervisor to improve the scientific research ability.

Jun Shao is Full professor at the School of Economics and Management, Southeast University. He has published more than 50 papers in first-class academic journals at home and abroad, presided over a number of national social science fund projects, and related results have been adopted by government documents.

References

Seetanah

Sannassee

Fauzel

, et al. Impact of economic and financial development on environmental degradation: evidence from small island developing states (SIDS). Emerg Mark Financ Trade 2019; 55: 308–322.

Acheampong

. Modelling for insight: does financial development improve environmental quality? Energy Econ 2019; 83: 156–179.

Hussain

Sattar

, et al. Evaluating the impact of energy and environment on economic growth in BRI countries. Energy Environ 2022. https://doi.org/10.1177/0958305X211073805

Lin

. Impacts of urbanization and industrialization on energy consumption/CO2 emissions: does the level of development matter? Renew Sustain Energy Rev 2015; 52: 1107–1122.

Pan

and Wang

, et al. The effects of outward foreign direct investment and reverse technology spillover on China's carbon productivity. Energy Pol 2020; 145: 11730.

Wang

Liao

and Xu

, et al. The impact of foreign direct investment on China's carbon emissions through energy intensity and emissions trading system. Energy Econ 2021a; 97: 105212.

Lee

Erickson

. How does local economic development in cities affect global GHG emissions? Sustain Cities Soc 2017; 35: 626–636.

Ferreira

JJM

Fernandes

Ratten

. The effects of technology transfers and institutional factors on economic growth: evidence from Europe and oceania. J Technol Transf 2019; 44: 1–24.

Dhakal

. Urban energy use and carbon emissions from cities in China and policy implications. Energy Pol 2009; 37: 4208–4219.

10.

Guan

Liu

Geng

, et al. The gigatonne gap in China's Carbon dioxide inventories. Nat Clim Change 2012; 2: 672–675.

11.

Pan

Uddin

Han

, et al. Dynamics of financial development, trade openness, technological innovation and energy intensity: Evidence from Bangladesh. Energy 2019; 171: 456–464.

12.

Grossman

Helpman

. Quality ladders in the theory of growth. Rev Econ Stud 1991; 58: 43–61.

13.

Fisher-Vanden

Jefferson

Liu

, et al.

What is driving China's Decline in energy intensity?

Resour. Energy Econ 2004; 26: 77–97.

14.

Sapkota

Bastola

. Foreign direct investment, income, and environmental pollution in developing countries: panel data analysis of Latin America. Energy Econ 2017; 64: 206–212.

15.

Asumadu

Vladimir

. Effect of foreign direct investments, economic development and energy consumption on greenhouse gas emissions in developing countries. Sci Total Environ 2019; 646: 862–871.

16.

. Effect of environmental regulation policy tools on the quality of foreign direct investment: an empirical study of China. J Clean Prod 2020; 270: 122346.

17.

Zhang

. The impacts of GDP, trade structure, exchange rate, and FDI inflows on China’s carbon emissions. Energy Pol 2018; 120: 347–353.

18.

Zhang

Uddin

, et al. Environmental regulation, foreign investment behavior, and carbon emissions for 30 provinces in China. J Clean Prod 2020; 248: 119208.

19.

Balsalobre-Lorente

Gokmenoglu

Taspinar

, et al. An approach to the pollution haven and pollution halo hypotheses in MINT countries. Environ Sci Pollut Res 2019; 26: 23010–23026.

20.

Mielnik

Goldemberg

. Foreign direct investment and decoupling between energy and gross domestic product in developing countries. Energy Pol 2002; 30: 87–89.

21.

Hamid

Shabbir

Murshed

, et al. The roles of foreign direct investments, economic growth, and capital investments in decarbonizing the economy of Oman. Environ. Sci. Pollut. Res 2021. https://doi.org/10.1007/s11356-021-17246-3

22.

Markusen

Venables

. Foreign direct investment as a catalyst for industrial development. Eur Econ Rev 1999; 43: 335–356.

23.

Zhang

. Planning for sustainable cities? A comparative content analysis of the master plans of eco, low-carbon and conventional new towns in China. Habitat Int 2017; 63: 55–66.

24.

Wang

Yang

, et al. A holistic overview of the progress of China’s low-carbon city pilots. Sustain Cities Soc 2018; 42: 289–300.

25.

Song

Zhou

Liu

, et al. Investigation of a “coupling model” of coordination between low-carbon development and urbanization in China. Energy Pol 2018; 121: 346–354.

26.

Sun

Huang

. How does urbanization affect carbon emission efficiency? Evidence from China. J Clean Prod 2020; 272: 122828.

27.

Omri

Nguyen

Rault

. Causal interactions between CO2 emissions, FDI, and economic growth: evidence from dynamic simultaneous equation models. Econ Model 2014; 42: 382–389.

28.

Cardoso

Rafaela

. The impact of foreign direct investment on emission reduction targets: Evidence from high- and middle-income countries. Struct Change Econ Dynam 2020; 55: 107–118.

29.

Copeland

Taylor

. North-south trade and the environment. J Q Econ 1994; 109: 755–787.

30.

Rafindadi

Muye

Kaita

. The effects of FDI and energy consumption on environmental pollution in predominantly resource-based economies of the GCC. Sustain Energy Technol Assess 2018; 25: 126–137.

31.

Solarin

Al-Mulali

Musah

, et al. Investigating the pollution haven hypothesis in Ghana: an empirical investigation. Energy 2017; 124: 706–719.

32.

Sun

Zhang

. Investigation of pollution haven hypothesis for China: an ARDL approach with breakpoint unit root tests. J Clean Prod 2017; 161: 153–164.

33.

Hamid

Jena

Mukhopdhyay

. Should China prefer more foreign direct investment inflows to environmental change? J Pub Affairs. e 2020; 2464: 1–18.

34.

Mujtaba

Jena

. Analyzing asymmetric impact of economic growth, energy use, FDI inflows and oil prices on CO2 emissions through nardl approach. Environ. Sci. Pollut. Res 2021; 28: 30873–30886.

35.

Reppelin-Hill

. Trade and environment: an empirical analysis of the technology effect in the steel industry. J Environ Econ Manag 1999; 38: 283–301.

36.

Richard

Eric

. Do recipient country characteristics affect international spillovers of co-efficiency via trade and foreign direct investment. Clim Change 2012; 112: 469–491.

37.

Huang

Tao

. An analysis of technological factors and energy intensity in China. Energy Pol 2017; 109: 1–9.

38.

Zhu

Duan

Guo

, et al. The effects of FDI, economic growth and energy consumption on carbon emissions in ASEAN-5: evidence from panel quantile regression. Econ. Model 2016; 58: 237–248.

39.

Dasgupta

Laplante

Mamingi

. Pollution and capital markets in developing countries. J Environ Econ Manag 2001; 42: 310–335.

40.

Shi

Huang

, et al. The impacts of freight trade on carbon emission efficiency: evidence from the countries along the “belt and road”. Complexity 2020; 20: 2529718.

41.

Mujtaba

Jena

Mukhopadhyay

. Determinants of CO(2)emissions in upper middle-income group countries: an empirical investigation. Environ. Sci. Pollut. Res 2020; 27: 37745–37759.

42.

Wang

Zhang

. The effects of trade openness on decoupling carbon emissions from economic growth: Evidence from 182 countries. J Clean Prod 2022; 279: 123838.

43.

Wang

Guo

. Official development assistance and carbon emissions of recipient countries: a dynamic panel threshold analysis for low- and lower-middle-income countries. Sustain Prod Consump 2022a; 29: 158–170.

44.

Meng

Niu

. Three-dimensional decomposition models for carbon productivity. Energy 2012; 46: 179–187.

45.

Yang

Wang

. Factor decomposition of carbon productivity chang in China’s main industries: based on the Laspeyres Decomposition Method. International Conference on Applied Energy 2014; 61: 1893–1896.

46.

Wang

. Does urbanization redefine the environmental kuznets curve? An empirical analysis of 134 countries. Sust Cities Soc 2022b; 76: 103382.

47.

Zhang

. The impacts of GDP, trade structure, exchange rate, and FDI inflows on China’s carbon emissions. Energy Pol 2018; 120: 347–353.

48.

Zhang

Jia

Mou

, et al. Carbon productivity convergence club and its initial conditions: China’s construction industry. Chin. J. Popul. Resour. Environ 2019; 17: 12–24. (in Chinese).

49.

Sun

, et al. Carbon efficiency and international specialization position: evidence from global value chain position index of manufacture. Energy Pol 2019; 128: 235–242.

50.

Chen

. Climate experimentation and the limits of top-down control: local variation of climate pilots in China. J Environ Plann Manag 2020; 63: 109–126.

51.

Luo

. Does the low-carbon city policy make a difference? Empirical evidence of the pilot scheme in China with DEA and PSM-DID. Ecologic. Indic 2021; 122: 107238.

52.

Zheng

Shao

Liu

, et al. The impact of the pilot program on industrial structure upgrading in low-carbon cities. J Clean Prod 2021; 290: 125868.

53.

Song

Zhao

Shang

. The impact of low-carbon city construction on ecological efficiency: empirical evidence from quasi-natural experiments. Resour Conserv Recycl 2020a; 157: 10477.

54.

Shao

. Does FDI affect carbon intensity? New evidence from dynamic panel analysis. J Int Climate Change Strategies Manag 2018; 10: 27–42.

55.

Fei

Lin

. Energy efficiency and production technology heterogeneity in China's agricultural sector: a meta-frontier approach. Techno Fore So Change 2016; 109: 25–34.

56.

Färe

Grosskopf

Pasurkajr

. Environmental production functions and environmental directional distance functions. Energy 2007; 32: 1055–1066.

57.

Zhang

Zhou

Choi

. Energy efficiency, CO2 emission performance and technology gaps in fossil fuel electricity generation in Korea: a meta-frontier non-radial directional distance function analysis. Energy Pol 2013; 56: 653–662.

58.

Cheng

Liu

, et al. Total-factor carbon emission efficiency of China's provincial industrial sector and its dynamic evolution. Renew Sustain Energy Rev 2018; 94: 330–339.

59.

Battese

Rao

DSP

O’Donnell

. A metafrontier production function for estimation of technical efficiencies and technology gaps for firms operating under different technologies. J Prod Anal 2004; 21: 91–103.

60.

O’Donnell

Rao

DSP

Battese

. Metafrontier frameworks for the study of firm-level efficiencies and technology ratios. Empiric Econ 2008; 34: 231–255.

61.

Cheng

Yang

Wen

. The effect of technological factors on industrial energy intensity in China: new evidence from the technological diversification. Sustain. Prod. Consump 2021; 28: 775–785.

62.

Arellano

Bover

. Another look at the instrumental variable estimation of error-components models. J. Econ 1995; 68: 29–51.

63.

Blundell

Bond

. Initial conditions and moment restrictions in dynamic panel data models. J. Econ 1998; 87: 115–143.

64.

Bond

. Dynamic panel data models: a guide to micro data methods and practice. Por. Econ. J 2002; 1: 141–162.

65.

Hansen

. Threshold effects in non-dynamic panels: estimation, testing, and inference. J Econ 1999; 93: 345–368.

66.

Haider

Mishra

. Does innovative capability enhance the energy efficiency of Indian iron and steel firms? A Bayesian Stochastic Frontier Analysis. Energy Econ 2021; 95: 105128.

67.

. Financial agglomeration, technological innovation, and green total factor energy efficiency. Alex. Eng. J 2021; 60: 4085–4095.

68.

Lin

Tan

. Ecological total-factor energy efficiency of China’s energy intensive industries. Ecol. Ind 2016; 270: 480–497.

69.

Wang

Yao

. Evaluation and difference analysis of regional energy efficiency in China under the carbon neutrality targets: insights from DEA and theil models. J Environ Manage 2021b; 293: 112958.

70.

Wang

. Effects of technological innovation on energy efficiency in China: evidence from dynamic panel of 284 cities. Energy Econ 2020; 26: 123–134.

71.

Hussain

Sattar

. Efects of urbanization and nonrenewable energy on carbon emission in Africa. Environ. Sci. Pollut. Res 2021. https://doi.org/10.1007/s11356-021-17738-2

72.

Gao

. Government cooperation, market integration, and energy efficiency in urban gglomerations-Based on the quasi-natural experiment of the Yangtze River Delta Urban Economic Coordination Committee. Energy Environ 2021; 10: 1–16.

73.

Shao

Shi

. Does financial agglomeration promote the increase of energy efficiency in China? Energy Policy 2020; 146: 111810.

74.

Cai

Guo

, et al. Benchmarking carbon emissions efficiency in Chinese cities: a comparative study based on high resolution gridded data. Appl Energy 2019; 242: 994–1009.

75.

Bazzi

Clemens

. Blunt instruments: avoiding common pitfalls in identifying the causes of economic growth. Am. Econ. J. Macroecon 2013; 5: 152–186.

76.

Sun

Liu

. Environmental regulation，clean - technology innovation and China's industrial green transformation. Sci Rese Manage 2021; 42: 54–61.

77.

Chuai

Huang

Wang

, et al. Spatial econometric analysis of carbon emissions from energy consumption in China. J Geo Sci 2012; 22: 630–642.

78.

Poon

JPH

Casas

. The impact of energy，transport and trade on air pollution in China. Eurasian Geo Econ 2006; 47: 568–584.

79.

Akram

Chen

Khalid

, et al. Heterogeneous effects of energy efficiency and renewable energy on carbon emissions: evidence from developing countries. J Clean Prod 2020; 247: 119122.

80.

Paramati

Ummalla

Apergis

. The effect of foreign direct investment and stock market growth on clean energy use across a panel of emerging market economies. Energy Econ 2016; 56: 29–41.

81.

Song

Qin

Wang

, et al. How does the nested structure affect policy innovation?: empirical research on China’s low carbon pilot cities. Energy Pol 2020b; 144: 111695.

82.

Baron

Kenny

. The moderator-mediator variable distinction in social psychological research: conceptual, strategic, and statistical considerations. J Pers Soc Psychol 1986; 51: 1173–1182.

83.

Zhou

Liu

. Impact of China's carbon emission trading policy on the performance of urban carbon emission and its mechanism. China Environ Series 2020; 40: 453–464. (in Chinese).