-
Yangtze River Delta, located in the alluvial plain before Yangtze River enters the Pacific Ocean, includes Shanghai, Nanjing, Wuxi, Changzhou, Nantong, Yancheng, Yangzhou, Zhenjiang, Taizhou (Jiangsu), Suzhou, Hangzhou, Ningbo, Jiaxing, Huzhou, Shaoxing, Jinhua, Zhoushan, Taizhou (Zhejiang), Hefei, Wuhu, Ma’anshan, Tongling, Anqing, Chuzhou, Chizhou, and Xuancheng (Fig. 1). Data obtained in 2017 revealed that this region accounted for 2.21% of China’s land, 9.39% of China’s population, and more than 21.24% of the country’s electricity consumption, and yielded 21.31% of the GDP (NBSC, 2005–2018). According to the air quality status report of China’s key regions and 74 cities in 2018, the Yangtze River Delta is among the regions with the highest concentration of air pollution in China (Xiao et al., 2020).
-
(1) Global spatial autocorrelation
Spatial autocorrelation refers to the correlation of the same variable in different spatial positions, which can be used to determine the spatial correlation between different regional economic indicators and to determine the spatial heterogeneity of economic indicators in different regions, such as the gap in economic development, energy efficiency, and environmental regulations, to reveal the regional structure shape of spatial variables (Yang et al., 2019). Previous studies have shown that SO2 emissions tend to have strong cross-regional and agglomeration characteristics, so the method applies to the spatial correlation test of SO2 emissions. In this paper, we use the Global Moran’s Index (GMI) to test the spatial correlation between FDI and SO2 emissions, which is calculated as follows (Liu et al., 2018):
$$GMI = \frac{n}{{\displaystyle\sum\nolimits_i {{{({x_i} - \overline x)}^2}} }}\frac{{\displaystyle\sum\nolimits_i {\sum\nolimits_{i \ne j} {{W_{ij}}({x_i} - \overline x)({x_j} - \overline x)} } }}{{\displaystyle\sum\nolimits_i {\sum\nolimits_{i \ne j} {{W_{ij}}} } }}$$ (1) where n represents the number of cities in the study area, xi and xj represent the values of the tested variable of city i and city j respectively;
$\overline x $ represents the average values of the tested variable, and Wij indicates the spatial weight matrix between city i and city j. When the value of GMI equals –1, it means that the variable is completely negatively spatial correlated. When the value of GMI equals 0, it means that the variable has no spatial correlation. When the value of GMI equals 1, it means that the variable is fully positively spatial correlated.(2) Local spatial autocorrelation
Global spatial autocorrelation can only be used to describe the overall spatial correlation characteristics of FDI and SO2 emissions in the Yangtze River Delta, which may ignore some atypical features in some areas. Local spatial autocorrelation can reveal the spatial agglomeration of FDI and SO2 emissions in the attribute values between each unit and its surrounding units, which can demonstrate spatial correlation patterns in different locations (Yu, 2012). Therefore, Getis-Ord
$G_i^* $ statistics was used to measure the local spatial dependence and heterogeneity of FDI and SO2 emissions, which is calculated as follows:$$ G_i^* = \sum\nolimits_{j = 1}^{26} {{{{W_{ij}}{x_j}} / {\sum\nolimits_{j = 1}^{26} {{x_j}} }}} $$ (2) where n, xj and Wij have the same meaning as in Equation (1). The Z (
$G_i^* $ ) value is divided into hot spots, sub-hot spots, cold spots, and sub-cold spots by using the nature break classification method.(3) Spatial econometrics model
The influence factors of air pollution such as SO2 emissions were often quantitatively analyzed by establishing regression models. Ordinary least squares (OLS) are one of the most traditional and simplest regression models, which are calculated as follows:
$$Y{\rm{ = }}X\alpha {\rm{ + }}\varepsilon $$ (3) where Y represents the dependent variable, X represents the independent variable, α represents the regression coefficient, and ε represents the random error.
The premise of the regression analysis of influencing factors with the OLS method is that each dependent variable is independent in space. However, SO2 emissions have a significant spatial correlation and dependence, which violates the premise of the OLS method and leads to a bias of regression results. The spatial econometric model can effectively solve this spatial correlation and spatial dependence.
Among common spatial regression models, spatial lag model (SLM) can explain the endogenous dependence of dependent variables, spatial error model (SEM) can explain the interaction effect of error terms, and spatial Durbin model (SDM) can investigate the endogenous dependence of dependent variables and detect the direct and interaction effect of external factors (Zhou et al., 2017). Previous studies have proved that SO2 emissions have a significant spatial dependence and a certain impact on the environmental level of neighboring regions (Hu et al., 2016). Therefore, the spatial econometrics model can be used to more accurately estimate the effect of various factors on SO2 emissions, which is calculated as follows:
The SLM is defined as:
$$\begin{split} \ln {{\rm{SO}}_2}_{it} =& {\alpha _0} + \rho \sum\nolimits_{j = 1}^{26} {{W_{ij}}} \ln {{\rm{SO}}_2}_{it} + \lambda FD{I_{it}} + \alpha \ln GD{P_{it}} + \\ &\beta \ln I{S_{it}} +\gamma \ln R{D_{it}} + \delta \ln P{D_{it}} + \varepsilon \ln E{I_{it}} + {\theta _{it}} \end{split}$$ (4) The SEM is defined as:
$$\begin{split} \ln {{\rm{SO}}_2}_{it} =& {\alpha _0} + \lambda FD{I_{it}} + \alpha \ln GD{P_{it}} + \beta \ln I{S_{it}} + \gamma \ln R{D_{it}} + \\ &\delta \ln P{D_{it}} + \varepsilon \ln E{I_{it}} + {\xi _{it}},{\xi _{it}} = \xi {W_{ij}}{\xi _{it}} + {\theta _{it}} \end{split}$$ (5) The SDM is defined as:
$$ \begin{split} \ln {{\rm{SO}}_2}_{it} = &{\alpha _0} + \rho \displaystyle\sum\nolimits_{j = 1}^{26} {{W_{ij}}} \ln s{o_2}_{it}{\rm{ + }}\lambda FD{I_{it}} + \alpha \ln GD{P_{it}} +\\ &\beta \ln I{S_{it}} + \gamma \ln R{D_{it}} + \delta \ln P{D_{it}} + \varepsilon \ln E{I_{it}} {\rm{ + }}\\ &\omega \displaystyle\sum\nolimits_{j = 1}^{26} {{W_{ij}}} \ln FD{I_{it}}{\rm{ + }}\varphi \sum\nolimits_{j = 1}^{26} {{W_{ij}}} \ln GD{P_{it}}{\rm{ + }}\\ &\tau \sum\nolimits_{j = 1}^{26} {{W_{ij}}} \ln I{S_{it}}{\rm{ + }}\sigma \sum\nolimits_{j = 1}^{26} {{W_{ij}}} \ln R{D_{it}} {\rm{ + }}\\ &\eta \displaystyle\sum\nolimits_{j = 1}^{26} {{W_{ij}}} \ln P{D_{it}}{\rm{ + }}\psi \sum\nolimits_{j = 1}^{26} {{W_{ij}}} \ln E{I_{it}} + {\theta _{{\rm{it}}}} \\ \end{split} $$ (6) where i and t represent cities and research years, respectively; θ represents random disturbance term; ρ represents spatial auto regression coefficient; ζ represents spatial error coefficient; α0, ω, φ, τ, σ, η, ψ represent spatial regression coefficient of independent variables; and Wij is an element in the adjacency space weight matrix W, representing the spatial correlation between city i and j.
-
In view of the existing literature on FDI and air pollution and considering the availability of data, we selected 26 cities in this region, using China’s City Statistical Yearbook from 2005 to 2018 as the database to characterize the pollution level of SO2 emissions (NBSC, 2005–2018). We took FDI as the explanatory variable of SO2 emissions and focused on the role of FDI in SO2 emissions. To minimize the influence of other factors on the dependent variable, we added several control variables to the model, including: industrial structure, research and development investment, population size, and energy intensity. All the indicators are shown in Table 1. The specific definitions of the selected indicators are as follows.
Table 1. Statistical description of the variables
Variable Unit Obs. Min. Max. Mean SD SO2 104 t 364 0.19 51.28 6.98 7.56 FDI 106 yuan (RMB) 364 22.57 125001.34 9098.45 16850.26 IS % 364 27.64 76.00 52.41 8.87 RD % 364 0.01 12.79 2.75 2.21 POP 104 person 364 39.02 1450.00 201.26 258.87 EI kWh/104 yuan (RMB) 364 25.00 29645.48 495.93 1562.44 Foreign direct investment (FDI): FDI inflow may have two effects on the environment of the host country. On the one hand, it may improve the situation of environmental pollution of the host country by promoting technological progress and environmental standards. On the other hand, it may further degrade the local environment by transferring domestic polluting enterprises to the host country. The FDI inflow is represented by the amount of FDI utilized.
Industrial structure (IS): As an important indicator to measure the degree of industrialization, the secondary industry has an important impact on environmental pollution. In the early stages of economic development, the rapid development of secondary industries such as heavy chemical industries greatly increases SO2 emissions and other pollutants, which then deteriorate the environmental quality. With further development of the economy, the proportion of the output value of secondary industries in the national economy declines, while the proportion of the tertiary industries rises. Therefore, the proportion of the output value of secondary industries in the gross domestic product was selected to represent the industrial structure, to illustrate the impact of the industrial structure on SO2 emissions.
Research and development investment (RD): The progress of research and development levels can reduce the intensity of comprehensive energy utilization, playing an important role in improving the levels of environmental pollution control and hence reducing emissions of air pollutants. The RD level is expressed by the proportion of science and technology expenditure to public finance expenditure.
Population size (POP): Previous studies have confirmed that population agglomeration leads to the expansion of production scale and industrial specialization agglomeration. On the one hand, the expansion of production scale leads to the increase in resource consumption and pollution. On the other hand, industrial specialization agglomeration first promotes and later restrains pollution. Population size is expressed by the total population of municipal districts at the end of the year.
Energy intensity (EI): Energy intensity is the reflection of the production process and technology level. Higher energy intensity often means more energy input and pollutant emission. Energy intensity is expressed as the proportion of electricity consumption and total industrial output value.
-
Fig. 2 shows the GMI values and the overall change trend of FDI and SO2 emissions in the Yangtze River Delta from 2004 to 2017, which well reflects the spatial autocorrelation of the two indexes. As shown in Fig. 2, the GMI values of FDI and SO2 emissions in the Yangtze River Delta were greater than 0 during the study period, indicating that the spatial distribution of FDI and SO2 emissions in the Yangtze River Delta has a completely positive spatial autocorrelation. Namely, the spatial distribution of FDI and SO2 emissions expressed spatial agglomeration characteristics. From an overall perspective, the spatial autocorrelation level of FDI showed a rapid decline. The GMI values for FDI decreased from 0.165 in 2004 to 0.014 in 2017. This shows a trend of decreasing spatial concentration of cities in the Yangtze River Delta during the study period. This may be attributed to the fact that FDI first invests in the coastal and riverside cities with convenient transportation and developed economy. Superior physical geography and social and economic advantages are often important factors that attract FDI. In addition, cities in these areas have a higher exposure to the outside world and preferential investment policies, which greatly reduce industrial restrictions on FDI. Inland cities in the Yangtze River Delta have no advantages of physical geography or social and economic development. At the same time, low technological levels, labor productivity, and rate of return on investment have become obstacles to FDI. While with the comprehensive development and increasing openness of the Yangtze River Delta economy, as well as the industrial policy guidance of inland cities, preferential policies for basic industries, and lower labor costs, FDI is no longer limited to a few developed cities, but has expanded to other regions of the Yangtze River Delta. Therefore, some clusters have disappeared. In contrast to FDI, the spatial autocorrelation level of SO2 emissions generally presents an upward trend, increasing from 0.133 in 2004 to 0.393 in 2017. Global spatial autocorrelation provides a more holistic understanding of the spatial effects of the relationship between FDI and SO2 emissions in the Yangtze River Delta from 2004 to 2017. In addition, the average GMI values of FDI and SO2 emissions during the study period were 0.086 and 0.191, respectively: which indicates that spatial agglomeration was more significant in SO2 emissions than in FDI. However, global spatial autocorrelation did not show a continuous trajectory of divergence or convergence, although these changes lead to different stages of development.
-
We introduced Z(
$G_i^* $ ) statistics to measure the dependence and heterogeneity of FDI and SO2 in local space, and divided them into hot spots, sub-hot spots, sub-cold spots, and cold spots according to the results of Z($G_i^* $ ) (Figs. 3, 4).Figure 3. Spatial distribution of ‘hot’ and ‘cold’ spots of FDI of the Yangtze River Delta in 2004, 2010, and 2017
Figure 4. Spatial distribution of ‘hot’ and ‘cold’ spots of SO2 emissions of the Yangtze River Delta in 2004, 2011, and 2017
In Fig. 3, we can see that the spatial distribution of FDI revealed four different clusters. The hot spots remained stable and were always located in Shanghai, Suzhou, and Jiaxing during the study period. The sub-hot spots expanded significantly, with an increase of 120%, among which the stable cities accounted for 45.45%. The sub-cold spots and the cold spots showed fluctuation and decrease from southeast to northwest by 20% and 50%, respectively. From a holistic perspective, the cities represented by Shanghai, Suzhou, and Jiaxing are areas where FDI exhibited hot spots clustering. These cities, as diffusion centers of FDI during the study period have a strong ability to attract FDI. Through exchange and cooperation with surrounding cities, these cities drive the improvement of investment levels in surrounding cities. In the meantime, we can see that Anqing, Chizhou, Tongling, and Taizhou (Jiangsu) are areas where FDI exhibited cold spots clustering. This is mainly because the levels of FDI, demonstration effects, and regional investment are all relatively low in Anqing, Chizhou, and Taizhou (Jiangsu) at present. From these results, we find that the effect of a given city’s FDI is closely related to its location and the use of FDI in surrounding areas.
Similarly, we can see from Fig. 4 that the spatial distribution of SO2 emissions formed four distinct clusters too. The hot spots expanded and were mainly distributed in the estuary area of the Yangtze River Delta, with a growth rate of 133.33%. The sub-hotspots gradually shrank from northwest to southeast, decreasing by 50%. Both the cold spots and sub-cold spots showed a slight expansion trend and were mainly distributed in the central and southwest parts of the Yangtze River Delta. In addition, as the main field of FDI input, the total industrial output value of hot spots in the Yangtze River Delta accounted for 46.96% in 2017, while that of cold spots only accounted for 11.50%. It is clear that the cause of the difference in the spatial distribution of SO2 emissions may be the impact of the imbalance of the amount of FDI flowing into the industrial sectors of cities in the Yangtze River Delta.
From the preceding spatial distribution analysis and clusters test of FDI and SO2 emissions, we can conclude that significant spatial autocorrelation existed in FDI and SO2 emission levels in the Yangtze River Delta cities during the study period. FDI and SO2 emissions have obvious path dependence characteristics in line with the Yangtze River Delta geography and form distinct agglomeration areas. Considering the spatial dependence or spatial heterogeneity of geographic data, traditional non-spatial regression methods lead to regression bias (Liu et al., 2013). Therefore, we use a spatial econometric model to estimate the impact of FDI on SO2 emissions that can effectively avoid such bias.
-
First, the Hausman test was conducted on the OLS model regression to determine whether a fixed effect or a random effect model should be established. The statistical value obtained was 21.17, and the 1% significance test was passed. Therefore, a fixed effect model was established. Meanwhile, the OLS model was compared with the SLM, SEM, and SDM model. We found that the R2 and adjusted R2 values of SDM were significantly larger, and LM and R-LM both passed the 1% significance test, which further indicated that an econometric model including spatial interaction should be established. In addition, the Wald test and LR test both passed the 1% significance test, indicating that the original hypothesis that SDM can be simplified as SLM and SEM must be rejected, and SDM be selected as the optimal model. Since SDM has three models of time fixed effects, spatial fixed effects, and spatial-time fixed effects, the LR test is adopted to select the model. The results show that the model of spatial-time fixed effects is accepted. Therefore, the SDM with spatial-time fixed effects is the best model, and its parameter estimation results are discussed (Table 2).
Table 2. Estimation results of the FDI on SO2 emissions based on the spatial regression model
Determinants OLS SLM SEM SDM C 4.508*** (10.393) – – – lnFDIit 0.213*** (5.910) 0.248*** (7.565) 0.259*** (8.074) 0.225*** (6.723) lnISit 1.220*** (5.696) 0.814*** (3.815) 0.752*** (3.576) 1.034*** (4.487) lnRD it 0.037 (1.260) 0.136*** (2.916) 0.116** (2.411) 0.092** (1.965) LnPOPit 0.490*** (6.728) 0.357*** (5.133) 0.299*** (4.638) 0.397 *** (4.909) lnEIit 0.312*** (6.579) 0.242*** (4.929) 0.225*** (4.639) 0.176*** (3.607) W×lnFDIit – – – 0.071 (0.845) W×ISit – – – 1.628*** (3.444) W×RDit – – – 0.218 ** (2.405) W×POPit – – – 0.099 (0.514) W×EIit – – – 0.353 *** (3.284) R2 0.608 0.708 0.689 0.733 Adj-R2 0.602 0.666 0.655 0.704 Sigma2 0.291 0.213 0.215 0.194 LogL 288.606 236.420 239.279 218.511 LMlag 66.681*** – – – R-LMlag 12.511*** – – – L-Merr 64.271*** – – – R-LMerr 10.102*** – – – Wald test spatial lag – – – 41.081*** LR test spatial lag – – – 35.818*** Wald test spatial error – – – 48.340*** LR test spatial error – – – 41.536*** Note: Numbers in the parentheses represent t-values; *, **, and *** indicate the significance level at 10%, 5%, and 1%, respectively Table 2 shows that FDI has a positive impact on SO2 emissions in the SDM with spatial-time fixed effects, and the regression coefficient indicates that there is a significantly positive correlation between FDI and SO2 emissions: an increase in FDI of 1% triggers a 0.225% increase in SO2 emissions. Meanwhile, spatial spillover effects of adjacent regions on FDI are positive but not significant: an increase in FDI of 1% should lead to a 0.071% increase in SO2 emissions. It is obvious that the entry of FDI aggravates regional SO2 emissions, and PHH exists in the Yangtze River Delta.
Among other regression coefficients, the coefficient of IS shows a positive effect on SO2 emissions, an increase in IS of 1% promotes a 1.034% increase in SO2 emissions correspondingly. This demonstrates that IS has the greatest impact on SO2 emissions in the Yangtze River Delta. Spatial spillover effects on adjacent regions on IS shows a significantly positive correlation as well, a 1% increase in IS accompanied by a 1.628% increase in adjacent regions. This is related to the fact that the economic development of the Yangtze River Delta is shaped by energy-intensive industries, which puts considerable pressure on the reduction of SO2 emissions. It is worth noting that the impact of RD on SO2 emissions is significantly positive, and the coefficient implies that an increase in RD results in a 0.092% increase in SO2 emissions. The spatial spillover effects on adjacent regions on RD are positive too: a 1% increase in RD of adjacent regions leads to a 0.218% increase in SO2 emissions in adjacent regions. The coefficient of POP is significantly positive for SO2 emissions; this implies that a 1% increase in POP is accompanied by a 0.397% increase in SO2 emissions. The spatial spillover effects of POP on adjacent regions are positive but not significant, indicating that a 1% increase in POP is conducive to increase SO2 emissions by 0.099%. The coefficient of EI was significantly positive for SO2 emissions: a 1% increases in EI triggers a 0.176% increase in SO2 emissions. Furthermore, spatial spillover effects on adjacent regions on EI are significantly positive, and a 1% increase in EI of adjacent regions leads to a 0.353% increase in SO2 emissions in local regions.
Does Foreign Direct Investment Affect SO2 Emissions in the Yangtze River Delta? A Spatial Econometric Analysis
-
Abstract: As the major source of air pollution, sulfur dioxide (SO2) emissions have become the focus of global attention. However, existing studies rarely consider spatial effects when discussing the relationship between foreign direct investment (FDI) and SO2 emissions. This study took the Yangtze River Delta as the research area and used the spatial panel data of 26 cities in this region for 2004–2017. The study investigated the spatial agglomeration effects and dynamics at work in FDI and SO2 emissions by using global and local measures of spatial autocorrelation. Then, based on regression analysis using a results of traditional ordinary least squares (OLS) model and a spatial econometric model, the spatial Durbin model (SDM) with spatial-time effects was adopted to quantify the impact of FDI on SO2 emissions, so as to avoid the regression results bias caused by ignoring the spatial effects. The results revealed a significant spatial autocorrelation between FDI and SO2 emissions, both of which displayed obvious path dependence characteristics in their geographical distribution. A series of agglomeration regions were observed on the spatial scale. The estimation results of the SDM showed that FDI inflow promoted SO2 emissions, which supports the pollution haven hypothesis. The findings of this study are significant in the prevention and control of air pollution in the Yangtze River Delta.
-
Table 1. Statistical description of the variables
Variable Unit Obs. Min. Max. Mean SD SO2 104 t 364 0.19 51.28 6.98 7.56 FDI 106 yuan (RMB) 364 22.57 125001.34 9098.45 16850.26 IS % 364 27.64 76.00 52.41 8.87 RD % 364 0.01 12.79 2.75 2.21 POP 104 person 364 39.02 1450.00 201.26 258.87 EI kWh/104 yuan (RMB) 364 25.00 29645.48 495.93 1562.44 Table 2. Estimation results of the FDI on SO2 emissions based on the spatial regression model
Determinants OLS SLM SEM SDM C 4.508*** (10.393) – – – lnFDIit 0.213*** (5.910) 0.248*** (7.565) 0.259*** (8.074) 0.225*** (6.723) lnISit 1.220*** (5.696) 0.814*** (3.815) 0.752*** (3.576) 1.034*** (4.487) lnRD it 0.037 (1.260) 0.136*** (2.916) 0.116** (2.411) 0.092** (1.965) LnPOPit 0.490*** (6.728) 0.357*** (5.133) 0.299*** (4.638) 0.397 *** (4.909) lnEIit 0.312*** (6.579) 0.242*** (4.929) 0.225*** (4.639) 0.176*** (3.607) W×lnFDIit – – – 0.071 (0.845) W×ISit – – – 1.628*** (3.444) W×RDit – – – 0.218 ** (2.405) W×POPit – – – 0.099 (0.514) W×EIit – – – 0.353 *** (3.284) R2 0.608 0.708 0.689 0.733 Adj-R2 0.602 0.666 0.655 0.704 Sigma2 0.291 0.213 0.215 0.194 LogL 288.606 236.420 239.279 218.511 LMlag 66.681*** – – – R-LMlag 12.511*** – – – L-Merr 64.271*** – – – R-LMerr 10.102*** – – – Wald test spatial lag – – – 41.081*** LR test spatial lag – – – 35.818*** Wald test spatial error – – – 48.340*** LR test spatial error – – – 41.536*** Note: Numbers in the parentheses represent t-values; *, **, and *** indicate the significance level at 10%, 5%, and 1%, respectively -
[1] Akbostanc E, Türüt-Aşık S, Tunç G İ, 2009. The relationship between income and environment in Turkey: is there an environmental Kuznets curve? Energy Policy, 37(3): 861–867. doi: 10.1016/j.enpol.2008.09.088 [2] Albornoz F, Cole M A, Elliott R J R et al., 2014. The environmental actions of firms: examining the role of spillovers, networks and absorptive capacity. Journal of Environmental Management, 146: 150–163. doi: 10.1016/j.jenvman.2014.07.011 [3] Asghari M, 2013. Does FDI promote MENA region’s environment quality? Pollution halo or pollution haven hypothesis. International Journal of Scientific Research in Environmental Sciences, 1(6): 92–100. doi: 10.12983/ijsres-2013-p092-100 [4] Baek J, 2016. A new look at the FDI-income-energy-environment nexus: dynamic panel data analysis of ASEAN. Energy Policy, 91: 22–27. doi: 10.1016/j.enpol.2015.12.045 [5] Bakhsh K, Rose S, Ali M F et al., 2017. Economic growth, CO2 emissions, renewable waste and FDI relation in Pakistan: new evidences from 3SLS. Journal of Environmental Management, 196: 627–632. doi: 10.1016/j.jenvman.2017.03.029 [6] Birdsall N, Wheeler D, 1993. Trade policy and industrial pollution in Latin America: where are the pollution havens? The Journal of Environment & Development, 2(1): 137–149. doi: 10.1177/107049659300200107 [7] Chen S M, Zhang Y, Zhang Y B et al., 2019. The relationship between industrial restructuring and China’s regional haze pollution: a spatial spillover perspective. Journal of Cleaner Production, 239: 115808. doi: 10.1016/j.jclepro.2019.02.078 [8] Chen X Y, Shao S, Tian Z H et al., 2017. Impacts of air pollution and its spatial spillover effect on public health based on China’s big data sample. Journal of Cleaner Production, 142: 915–925. doi: 10.1016/j.jclepro.2016.02.119 [9] Chudnovsky D, López A, 2007. Foreign direct investment and development: the MERCOSUR experience. CEPAL Review, 92: 7–23. doi: 10.18356/121766cf-en [10] Eskeland G S, Harrison A E, 2003. Moving to greener pastures? Multinationals and the pollution haven hypothesis. Journal of Development Economics, 70(1): 1–23. doi: 10.1016/S0304-3878(02)00084-6 [11] He J, 2006. Pollution haven hypothesis and environmental impacts of foreign direct investment: the case of industrial emission of sulfur dioxide (SO2) in Chinese provinces. Ecological Economics, 60(1): 228–245. doi: 10.1016/j.ecolecon.2005.12.008 [12] Hoffmann R, Lee C G, Ramasamy B et al., 2005. FDI and pollution: a granger causality test using panel data. Journal of International Development, 17(3): 311–317. doi: 10.1002/jid.1196 [13] Hu Zhiqiang, Miao Jianming, Miao Changhong, 2016. Agglomeration characteristics of industrial pollution and their influencing factors on the scale of cities in China. Geographical Research, 35(8): 1470–1482. (in Chinese) [14] Jiang L, Zhou H F, Bai L et al., 2018. Does foreign direct investment drive environmental degradation in China? An empirical study based on air quality index from a spatial perspective. Journal of Cleaner Production, 176: 864–872. doi: 10.1016/j.jclepro.2017.12.048 [15] Letchumanan R, Kodama F, 2000. Reconciling the conflict between the ‘pollution-haven’ hypothesis and an emerging trajectory of international technology transfer. Research Policy, 29(1): 59–79. doi: 10.1016/S0048-7333(99)00033-5 [16] Levinson A, Taylor M S, 2008. Unmasking the pollution haven effect. International Economic Review, 49(1): 223–254. doi: 10.3386/w10629 [17] Liang F H, 2017. Does foreign direct investment improve the productivity of domestic firms? Technology spillovers, industry linkages, and firm capabilities. Research Policy, 46(1): 138–159. doi: 10.1016/j.respol.2016.08.007 [18] Liu Y S, Lu S S, Chen Y F, 2013. Spatio-temporal change of urban-rural equalized development patterns in China and its driving factors. Journal of Rural Studies, 32: 320–330. doi: 10.1016/j.jrurstud.2013.08.004 [19] Liu Q Q, Wang S J, Zhang W Z et al., 2018. Does foreign direct investment affect environmental pollution in China’s cities? A spatial econometric perspective. Science of the Total Environment, 613: 521–529. [20] Manderson E, Kneller R, 2012. Environmental regulations, outward FDI and heterogeneous firms: are countries used as pollution havens? Environmental and Resource Economics, 51(3): 317–352. doi: 10.1007/s10640-011-9500-z [21] Mielnik O, Goldemberg J, 2002. Foreign direct investment and decoupling between energy and gross domestic product in developing countries. Energy Policy, 30(2): 87–89. doi: 10.1016/S0301-4215(01)00080-5 [22] NBSC (National Bureau of Statistics of China), 2005–2018. China City Statistical Yearbook. Beijing: China Statistics Press. (in Chinese) [23] Poon J P H, Casas I, He C F, 2006. The impact of energy, transport, and trade on air pollution in China. Eurasian Geography and Economics, 47(5): 568–584. doi: 10.2747/1538-7216.47.5.568 [24] Shen J, Wei Y D, Yang Z, 2017. The impact of environmental regulations on the location of pollution-intensive industries in China. Journal of Cleaner Production, 148: 785–794. doi: 10.1016/j.jclepro.2017.02.050 [25] Sheng B, Lü Y, 2012. Impact of foreign direct investment on China’s environment: an empirical study based on industrial panel data. Social Sciences in China, 33(4): 89–107. doi: 10.1080/02529203.2012.735899 [26] UNCTAD, 2019. World Investment Report 2019. Available at: https://unctad.org/en/pages/PublicationWebflyer.aspx?publicationid=2460. Cited 12 June 2019. [27] Walter I, Ugelow J L, 1979. Environmental policies in developing countries. AMBIO, 8(2–3): 102–109. [28] Wang Guoxia, Shi Xiaowei, Cui Haiyan et al., 2020. Impacts of migration on urban environmental pollutant emissions in China: a comparative perspective. Chinese Geographical Science, 30(1): 45–58. doi: 10.1007/s11769-020-1096-1 [29] Wang H, Dong C G, Liu Y, 2019a. Beijing direct investment to its neighbors: a pollution haven or pollution halo effect? Journal of Cleaner Production, 239: 118062. doi: 10.1016/j.jclepro.2019.118062 [30] Wang S J, Zhou C S, Wang Z B et al., 2017. The characteristics and drivers of fine particulate matter (PM2.5) distribution in China. Journal of Cleaner Production, 142: 1800–1809. doi: 10.1016/j.jclepro.2016.11.104 [31] Wang X Y, Zhang C T, Zhang Z J, 2019b. Pollution haven or porter? The impact of environmental regulation on location choices of pollution-intensive firms in China. Journal of Environmental Management, 248: 109248. doi: 10.1016/j.jenvman.2019.07.019 [32] Wu Xing, 2016. The new changes of China-Japan relations after China’s GDP surpassed Japan. Academic Exploration, 2: 25–31. (in Chinese) [33] Xiao Q Y, Geng G N, Liang F C et al., 2020. Changes in spatial patterns of PM2.5 pollution in China 2000-2018: impact of clean air policies. Environment International, 141: 105776. doi: 10.1016/j.envint.2020.105776 [34] Xu F, Xiang N, Higano Y, 2017. How to reach haze control targets by air pollutants emission reduction in the Beijing-Tianjin-Hebei region of China? PLoS One, 12(3): e0173612. doi: 10.1371/journal.pone.0173612 [35] Xu W X, Sun J Q, Liu Y X et al., 2019. Spatiotemporal variation and socioeconomic drivers of air pollution in China during 2005−2016. Journal of Environmental Management, 245: 66–75. doi: 10.1016/j.jenvman.2019.05.041 [36] Yan M, An Z, 2017. Foreign direct investment and environmental pollution: new evidence from China. Econometrics Letters, 4(1): 1–17. doi: 10.5455/ELET.2017.4.1.1 [37] Yang W Y, Li T, Cao X S, 2015. Examining the impacts of socio-economic factors, urban form and transportation development on CO2 emissions from transportation in China: a panel data analysis of China’s provinces. Habitat International, 49: 212–220. doi: 10.1016/j.habitatint.2015.05.030 [38] Yang Zhen, Zhang Xiaolei, Lei Jun et al., 2019. Spatio-temporal pattern characteristics of relationship between urbanization and economic development at county level in China. Chinese Geographical Science, 29(4): 553–567. doi: 10.1007/s11769-019-1053-z [39] Yu H Y, 2012. The influential factors of China’s regional energy intensity and its spatial linkages: 1988-2007. Energy Policy, 45: 583–593. doi: 10.1016/j.enpol.2012.03.009 [40] Yu S W, Zhang J J, Zheng S H et al., 2015. Provincial carbon intensity abatement potential estimation in China: a PSO-GA-optimized multi-factor environmental learning curve method. Energy Policy, 77: 46–55. doi: 10.1016/j.enpol.2014.11.035 [41] Zhang Y J, Peng Y L, Ma C Q et al., 2017. Can environmental innovation facilitate carbon emissions reduction? Evidence from China. Energy Policy, 100: 18–28. doi: 10.1016/j.enpol.2016.10.005 [42] Zhang S Y, Li H X, Zhang Q et al., 2019. Uncovering the impacts of industrial transformation on low-carbon development in the Yangtze River Delta. Resources, Conservation and Recycling, 150: 104442. doi: 10.1016/j.resconrec.2019.104442 [43] Zhou C S, Wang S J, Wang J Y, 2019. Examining the influences of urbanization on carbon dioxide emissions in the Yangtze River Delta, China: Kuznets curve relationship. Science of the Total Environment, 675: 472–482. doi: 10.1016/j.scitotenv.2019.04.269 [44] Zhou Liang, Zhou Chenhu, Yang Fan et al., 2017. Spatio-temporal evolution and the influencing factors of PM2.5 in China between 2000 and 2011. Acta Geographica Sinica, 72(11): 2079–2092. (in Chinese) [45] Zhu L, Gan Q M, Liu Y et al., 2017. The impact of foreign direct investment on SO2 emissions in the Beijing-Tianjin-Hebei region: a spatial econometric analysis. Journal of Cleaner Production, 166: 189–196. doi: 10.1016/j.jclepro.2017.08.032