Environmental Regulation, Foreign Direct Investment, and Green Total Factor Productivity: An Empirical Test Based on Chinese City-Level Panel Data

Abstract

Environmental regulation is a crucial tool for government intervention in the field of green technology innovation. It can boost an enterprise’s competitiveness and encourage green technology innovation, both of which have a major effect on luring foreign investment. This paper first systematically elaborates on the relationship between environmental regulation, foreign direct investment (FDI), and green total factor productivity (GTFP) and then combines panel data from Chinese cities to empirically test these relationships using various methods, such as the mediation effect model, two-stage least squares, and difference-in-differences method. The study found that environmental regulation significantly boosts FDI and GTFP. FDI helps to improve GTFP, and environmental regulation can impact GTFP indirectly through FDI. The way that FDI and environmental regulations affect GTFP demonstrates regional variation. Large cities with high economic growth gain more from environmental regulation. FDI has a stronger promotion effect on GTFP in medium- and small-sized cities than in large-sized cities, and it does not significantly impact GTFP in cities with high levels of economic development or in the eastern region.

1. Introduction

The Industrial Revolution that emerged in the 1760s greatly promoted the development of productivity, driving rapid economic growth worldwide. However, the extensive growth characterized by high energy consumption, pollution, and emissions has also triggered global issues, including ecological degradation, climate change, and resource depletion, posing serious challenges to human survival and development. Countries around the world have begun to explore development approaches that balance economic growth with environmental protection [1,2]. Green technology innovation, building upon traditional technological advancements, integrates ecological principles to guide technological innovation to resource conservation and environmental protection and promote a virtuous cycle of economic, resource, environmental, and social systems. Against the backdrop of accelerating the transformation of economic development modes, there is an inevitable reliance on green technology innovation to mitigate excessive reliance on resources and environmental degradation caused by production activities, enhance factor productivity, reduce environmental burdens, and thereby achieve green growth [3,4,5].

Since green technology innovation aims at environmental protection, and environmental pollution entails negative externalities, the lack of a sound pollution trading rights system and market pricing mechanism for pollution emissions leads to a serious underestimation of the cost of pollutant emissions. The private cost of pollution for enterprises is lower than the social cost, leading to excessive pollution discharge by enterprises and exacerbating environmental pollution. Externalities prevent market mechanisms from allocating resources optimally, hence requiring government intervention [6,7,8]. Environmental regulation is a significant measure for the government to influence the progress of green technology innovation. Raising environmental control standards and increasing pollution control costs for enterprises not only internalizes the externalities of environmental pollution but also incentivizes enterprises to accelerate research and development (R&D) of green technology, promote green technology levels, and improve GTFP [9,10,11].

The three primary categories of research on how environmental regulations affect GTFP are as follows. First, some researchers maintained that environmental regulation improves green innovation efficiency. With the “Porter hypothesis” as the core, it is believed that firm innovation can be prompted by adequate environmental regulation [12], and the advantages of innovation can offset the increased cost of regulations. Recent research found that environmental regulations enhance the level of technical advancement [13] and have beneficial effects on the growth of plant’s GTFP [14]. Second, the detrimental impact of environmental regulations on green innovation efficiency is considered. Represented by the neoclassical economics school, it argues that environmental regulations have “following the cost” effects, which raises expenses and insufficient investment in R&D, hence impeding the improvement of GTFP. Greenstone et al. [15] discovered that the implementation of the US Clean Air Act led to a 2.6% decrease in the overall factor productivity of enterprises in a quasi-natural experiment. Lanoie et al. [16], Tang et al. [17], and Cai and Ye [18] obtained similar conclusions. Third, it is unclear how environmental regulations would affect the total factor productivity. There may not be a linear relationship between environmental regulation and GTFP. The GTFP shows an inverted “U” shaped curve, rising at first and subsequently declining as a result of the increased severity of environmental control [19,20].

Most of the existing studies believe that environmental regulation requires enterprises to conduct pollution control, which drives up the cost of enterprises and is not conducive to attracting foreign investment [21,22], while there are also studies that believe that environmental regulation motivates enterprises to perform green technology innovation, can boost enterprises’ competitiveness, and attract foreign enterprises [23,24].

The effects of FDI on GTFP have not yet reached a consensus. Some scholars [25] contend that FDI inflows can introduce clean technology and advanced production, thereby improving the technological level of the inflow area through spillover effects, reducing environmental pollution in the recipient area, and promoting GTFP. This argument is based on the “pollution halo” hypothesis. Using Chinese data, some scholars found that FDI has positive productivity spillover effects [26] and can increase the efficiency of city green development [27]. However, other researchers, based on the “pollution haven” theory, argue that multinational firms tend to transfer polluting industries abroad through direct investment to reduce environmental compensation costs, exacerbating environmental pollution in the recipient area [28], which is detrimental to the improvement of GTFP. Rafindadi et al. [29] pointed out that developing countries, in order to develop their economies, have reduced their environmental protection requirements. This has led developed nations to shift highly polluting industries to these developing nations, worsening environmental degradation in the host nation and undermining the advancement of GTFP. Influenced by the dual effects of FDI, some scholars believe that FDI has heterogeneous effects on GTFP. Qiu et al. [30] discovered, using provincial-level data from China, that the effect of FDI on GTFP differs by area, with the “pollution halo” effect on GTFP observed in western areas and the “pollution haven” effect on GTFP in central and eastern areas.

Panel data from 288 Chinese cities from 2008 to 2020 are used to examine the connection between environmental regulation, FDI, and GTFP. It does this by employing a variety of methods, including the mediation effect model, two-stage least squares, and difference-in-differences method. Compared with previous literature, this research makes three primary contributions. First, it offers fresh insights into the relationship between foreign investment and environmental regulations. While existing research mostly suggests that environmental regulations increase firms’ environmental governance costs, thereby adversely affecting foreign investment attraction, this study examined and verified the favorable effect of environmental regulation on FDI in terms of production costs and innovation benefits. Second, it utilized a more advanced method to calculate GTFP. Existing studies often overlook issues such as unexpected outputs in innovation activities and the relaxation or crowding-out of input factors, leading to certain biases in the calculation results. This study incorporated environmental benefits into the measurement system based on economic and social benefits, using super efficiency slack-based measure (SBM) data envelopment analysis (DEA) models, which include unexpected outputs and the Malmquist–Luenberger index [31] to calculate GTFP for various cities. Third, plentiful heterogeneous tests were applied to unlocking the causal nexus of environmental regulation and FDI on GTFP, yielding new findings. Existing research often overlooks differences in innovation capabilities among regions of different types, leading to controversy. This study disaggregates sample data based on economic development levels, city sizes, and geographical locations to examine the heterogeneous impacts of environmental regulation and FDI on GTFP.

2. Theoretical Framework and Research Hypotheses

2.1. The Theoretical Mechanism of Environmental Regulation on GTFP

Short-term effects of more rigorous environmental regulations include an increase in environmental governance costs [32], lower profit margins, increased costs associated with environmental governance for businesses, and a “crowding out” effect on firms’ available resources for innovation. These effects are harmful to technological innovation and impede the advancement of innovation efficiency [33,34]. However, over time, environmental regulations have a substantial beneficial impact on GTFP [35,36,37], which mainly manifests in the following aspects.

First, environmental regulations motivate firms to engage in technological innovation. Under lower levels of environmental regulations, firms face lower pollution control costs and lack the motivation to engage in green technology innovation. There is room for further improvement in factors input, energy utilization, and other aspects, leading to inefficient production. The increase in environmental regulation intensity can make firms realize their inefficiencies in factor allocation, energy utilization, energy conservation, and emission reduction, motivating them to adopt more optimal production strategies to achieve higher production efficiency at the current technological level [38,39]. Moreover, the rise in pollution control costs due to environmental regulations compels firms to intensify R&D of green technology to reduce environmental costs, enhance green technology levels, and promote the improvement of GTFP.

Second, factor structure upgrading is encouraged by environmental regulations. Low levels of environmental regulation lead to a significant undervaluation of the expenses associated with pollution emissions, energy and environmental resource pricing, and other associated costs. Enterprises lack the motivation to decrease environmental emissions and increase energy efficiency, gradually forming a development model of “high energy consumption, high pollution”, resulting in increasingly poor environmental performance. The marketization of resource and energy prices is encouraged by the rise in environmental regulation, particularly the adoption of market-incentive regulatory instruments like resource taxes, sewage treatment fees, and ecological compensation mechanisms. Through market mechanisms, price signals reflect the characteristics of resources and energy, environmental costs, and supply and demand conditions, effectively avoiding phenomena such as excessive pollutant emissions and excessive resource consumption due to distortion of resource prices and low usage costs. This compels enterprises to change their development approach, increase R&D of green technology, lower their energy usage and emissions of pollutants, change the proportion of factor inputs, and use other environmentally friendly factors to replace energy. This shift from relying on energy to relying on green technology, human capital, etc., encourages the improvement of GTFP by upgrading factor structure and increasing factor allocation and utilization efficiency. Pei et al. [40] opined that environmental regulation restricts how enterprises use resources. Enterprises can lower the carbon intensity of economic output by investing more in R&D, as well as improving the efficiency of production technology and resource utilization.

Third, environmental regulations promote industrial structure adjustment. Currently, the guidance and intervention of industrial policies serve as the primary driving force of China’s industrial structure adjustment, transmitted from the central government to local governments and then to enterprises, which have strong characteristics of a planned economy and lack internal incentives. However, the increase in the level of environmental regulation provides precisely the intrinsic incentive for structural adjustment by imposing environmental pressure on enterprises. Environmental regulations compel enterprises to increase their internal costs. In a competitive market structure, enterprises must adjust their product structure and technological level to offset the increased costs in order to gain a competitive advantage and avoid being eliminated by the market. Therefore, increasing environmental regulation is akin to a mandatory “cleansing” of enterprises in the market, eliminating backward production technologies or enterprises with higher pollution and energy consumption, thereby promoting industrial structural adjustment [41,42,43]. Moreover, industrial structural adjustment can boost the share of knowledge- and technology-intensive industries, encourage the development of green technologies, and concurrently decrease the share of energy-intensive and polluting industries. This will help to control pollution generation and emissions at the source and will boost GTFP [44].

To intuitively illustrate the effect of environmental regulation on GTFP, this paper draws a diagram of the mechanism of environmental regulation on GTFP. Figure 1a,b, respectively, provide an explanation of the short- and long-term effects of environmental legislation on GTFP. In these figures, $\mathrm{X}$ and $\mathrm{Y}$ represent the undesirable output and desirable output in enterprise production activities, respectively. $\mathrm{f}$ represents the technological level of enterprises, that is, the production possibility frontier of enterprises, $f_l<f_m<f_h$. With lax environmental regulation intensity, the production function of enterprises is $F_m=f_m\left(x\right)$. Enterprise production is first assumed to be efficient and located at point $A$ on the curve $f_m$ in order to simplify the study. In the short term, as depicted in Figure 1a, enterprises inevitably adopt a series of production decisions to meet regulatory requirements. This process raises enterprise costs in the short term, exerting a “crowding out” effect on innovation activities, as well as productive investment [45,46]. As a result, enterprises’ technological level declines, and the production possibility frontier decreases to $F_l=f_l\left(x\right)$. The equilibrium point moves from point $A$ to the lower left, ultimately achieving a new equilibrium at point $B$. Although there is a certain degree of reduction in undesired output at point $B$ compared with point $A$, the reduction in desired output is larger. Hence, in the short run, GTFP is negatively impacted by environmental regulation.

In the long term, firms are compelled to carry out green technology innovation, which raises their technological level in order to acquire new competitive advantages and reduce the expenses associated with environmental regulations, as shown in Figure 1b. The production possibility curve shifts upward to F_h = f_h(x). At the new technological level, the same amount of input can generate more desired output without increasing non-desired output. The emergence of new technology implies that points on the production possibility frontier become inefficient production methods. Enterprises adjust factor inputs, improve production processes, and move production activities upward and to the left. Ultimately, a new equilibrium is formed at point C on the curve F_h. Compared with both the initial equilibrium point A and the short-term equilibrium point B, the production method at point C not only yields less non-desired output but also generates more desired output. Consequently, environmental regulation benefits GTFP over the long run. In light of this, we put out the following hypothesis:

Hypothesis 1.

Environmental regulation has a promotion impact on GTFP.

2.2. The Theoretical Mechanism of Environmental Regulation on FDI

According to the pollution haven theory, differences in environmental standards result in cost differentials, which is a significant factor influencing the location distribution of foreign investment. To reduce environmental costs and maintain market competitiveness, foreign enterprises often invest in areas with less environmental control. The Porter hypothesis, on the other hand, contends that environmental regulations can encourage corporate innovation and that the advantages of this innovation can outweigh the drawbacks of these regulations. According to the Porter hypothesis, drawing in foreign investment is facilitated by suitably tightening environmental regulations. Empirically, there is insufficient evidence to suggest that environmental regulation hinders FDI. For example, Friedman et al. [23] found that environmental regulation did not have an adverse effect on FDI; instead, it had a stimulating and promoting effect. Eskeland and Harrison [47] tested FDI in different industrialized model countries, such as Mexico, Venezuela, and Morocco, and found that there was no correlation between foreign capital inflow and environmental governance cost, thus not supporting the pollution haven hypothesis. Javorcik and Wei [48] analyzed investment choices made by multinational enterprises in some Eastern European and former Soviet countries and found no necessary connection between general environmental standards and FDI.

In China, low labor costs, favorable policies, and stable political relations are the main reasons foreign enterprises invest in China rather than lower environmental standards. Therefore, increasing environmental regulation will not cause a large-scale shift of foreign investment outward; instead, the innovation compensation effect can increase attractiveness to foreign investment. Using joint ventures in China’s provincial administrative regions as the research subject, Dean et al. [24] investigated the effect of environmental regulation on FDI and discovered that while environmental control did not impede FDI, it did encourage the entrance of non-Chinese capital.

In fact, environmental costs constitute only a small portion of a firm’s total cost function. Compared with fixed production costs and other variable costs, such as land transfer fees, labor costs, and raw material procurement costs, the proportion of environmental costs paid by enterprises is minimal. Relative to the differences in production costs, the cost increase brought about by environmental regulation is almost negligible. Moreover, firms are encouraged to participate in green innovation initiatives and create green products, green technology, and green processes through the improvement of environmental standards. The advantages of innovation can not only offset the cost increases caused by environmental regulations but can also boost an enterprise’s competitiveness, thereby increasing the attractiveness of foreign investment enterprises. Additionally, to attract foreign investment, China has implemented a series of preferential policies while enhancing environmental regulation. These policies include providing direct or indirect subsidies to foreign-funded enterprises through investment subsidies, discounted loans, etc., greatly increasing the enthusiasm of foreign-funded enterprises to invest. The second hypothesis is put forward as follows based on this premise:

Hypothesis 2.

Environmental regulation can directly promote FDI.

2.3. The Theoretical Mechanism of FDI on GTFP

2.3.1. The Promotion Effect of FDI on GTFP

FDI plays a crucial role in promoting green technology innovation, primarily manifested in the funding support effect, competitive effect, and technology spillover effect.

First, the funding support effect is considered. The firm’s self-owned funds and capital market financing are the main sources of investment in R&D of green technology innovation. However, on the one hand, the limited availability of enterprises’ own funds makes it difficult to meet the needs of green technology innovation. On the other hand, due to the long investment return cycle and high risk of green technology innovation, domestic investors have low enthusiasm for making R&D investments in green technology. Moreover, due to the imperfect capital market, enterprises find it challenging to obtain the funds needed for green technology R&D at a low cost, which hinders the development of green technology innovation. The inflow of foreign capital provides financial support for green technology innovation, improves capital stock in the inflow area, alleviates the dilemma of insufficient capital in the process of green technology innovation, and effectively enhances the GTFP of the inflow area.

Second, the competitive effect is considered. Firms are motivated to innovate in green technologies by competitiveness. Before the entry of foreign-funded enterprises, the market only consisted of domestic enterprises, resulting in relatively small differences in green technology levels among enterprises and low market competitiveness. This lack of competition hindered enterprises from having the motivation to innovate in green technology. However, after the entry of foreign investment, the market equilibrium was disrupted. The increase in the number of competitors diversified the competition from solely between domestic enterprises to competition involving domestic enterprises, joint ventures, and foreign-funded enterprises. Foreign-funded enterprises are more competitive due to their comparative advantages in the research and application of green technologies, as well as China’s vigorous promotion of green innovation development. To establish a foothold in the intense market competition, domestic firms are forced to perform green technology innovation, increase R&D in this area, and speed up the transformation of innovation outcomes. This drives the improvement of GTFP for the entire industry and region.

Third, the technology spillover effect is considered. The technology spillover from FDI mainly manifests in training effects, demonstration and learning effects, and industry linkage effects [49,50]. The training effect refers to the utilization of local labor resources by foreign direct invested enterprises in the inflow area when labor costs and geographical restrictions are present. Compared with foreign-funded enterprises, the local labor force has relatively lower levels of green technology. To alleviate the restriction of the technology gap on enterprise development, foreign-funded enterprises tend to train and guide local labor through the introduction of technical personnel from their home countries, thereby improving the technical level of local labor. Through labor mobility between enterprises, technology spillover is achieved [51].

The demonstration and learning effects stem from the differences in technological levels among enterprises and market competition. Foreign-funded enterprises possess more advanced green technology and serve as examples for local enterprises. Although the inflow of foreign investment intensifies competition among enterprises, it indirectly facilitates technical exchanges between local and foreign-funded enterprises, giving local enterprises chances to learn and absorb advanced green technology through joint ventures, personnel mobility, and product research, which improves their technological levels.

The industry correlation effect refers to the technology spillover generated by cooperation along the industrial chain by foreign-funded enterprises. Foreign-funded enterprises have higher quality requirements for the products provided by the upstream industry, which can encourage the upstream industry to develop technological innovation and improve the green technology level of the upstream industry through production equipment support, training, and guidance. The green technology progress of foreign-funded enterprises can also be transferred to downstream industries through R&D spillover effects, increasing green technology innovation efficiency in downstream industries.

2.3.2. The Inhibition Effect of FDI on GTFP

Compared with industrialized countries, developing countries have laxer environmental restrictions because of disparities in economic development levels. Enterprises with high pollutant features incur higher environmental expenses as environmental restrictions tighten in developed nations. To reduce production costs and gain a competitive advantage, these enterprises tend to transfer some or all of their manufacturing activities to developing nations with lower environmental standards. Meanwhile, developing countries, in pursuit of technological progress and economic catch-up, vigorously introduce foreign investment by formulating various preferential policies, providing objective conditions for developed nations to send high-polluting industries to developing nations [52]. In this scenario, as foreign investment flows, pollution-intensive companies from affluent countries would relocate to nations or regions with laxer environmental regulations, worsening pollution in the inflow areas [29]. Moreover, green technology innovation is also influenced by the quality of foreign investment. The rapid industrialization mindset and inter-regional competition for investment incentives among local officials may lower the investment threshold and draw in some foreign capital that is out of step with the local economic development model and has excessive energy and pollution consumption. Although the quantity of foreign investment continues to increase, the low quality of foreign investment may not necessarily lead to technology spillovers and may even have the opposite impact, suppressing green technology innovation efficiency in inflow areas.

To visually represent the impact of FDI, this paper constructs a mechanism diagram illustrating the effect of FDI on GTFP, as shown in Figure 2. For simplification purposes, we assume that enterprises only produce one type of product, and the more they produce, the higher the pollution emissions. Y and X represent the expected and non-expected outputs in green technology innovation activities. F refers to different levels of green technology innovation performance, where F_H2 > F_H1 > F_M > F_L1 > F_L2. Initially, the level of green technology innovation performance is assumed to be low, located at point A on the curve F_M = f_m(x). At this time, the expected and non-expected outputs are represented by Y₀ and X₀. After the FDI influx, the degree of green technology in the inflow area is increased, reducing non-expected output while increasing expected output through the funding support effect, competitive effect, and technology spillover effect. This shift moves the GTFP from point A to point B on F_H2 = f_h2(x). On the other hand, high-energy-consuming and high-polluting industries are transferred to the inflow area with the entry of foreign capital, leading to increases in both expected and non-expected outputs. In the absence of technological progress, the increase in non-expected output exceeds that of expected output, shifting the GTFP from point A to point C on the curve F_l2 = f_l2(x).

According to the principle of vector addition, the respective magnitudes of the two effects that FDI brings about determine how the GTFP changes. If the promoting effect of FDI is greater than the hindering effect, as shown in Figure 2a, GTFP ultimately moves toward point $D$ on curve $F_{H1}=f_{h1}\left(x\right)$, where the expected and undesirable outputs are $Y_1$ and $X_1$, respectively. Compared with the initial state, the desirable output increases $Y_1{Y}_0$, while the undesirable output decreases $X_1{X}_0$. The inflow of foreign capital promotes green technology innovation efficiency in the inflow area. If the promoting effect of FDI is equal to the hindering effect, as shown in Figure 2b, the green technology innovation performance ultimately moves toward point $E$ on the curve $F_M=f_m\left(x\right)$, where the expected and non-expected outputs are $Y_2$ and $X_2$, respectively. Compared with the initial state, the expected output increases $Y_2{Y}_0$ while the non-expected output also increases $X_2{X}_0$. GTFP in the inflow area does not change with the inflow of foreign capital. If the promoting effect of FDI is smaller than the hindering effect, as shown in Figure 2c, GTFP ultimately moves toward point $F$ on the curve $F_{l1}=f_{l1}\left(x\right)$. Compared with the initial state, the expected output increases $Y_3{Y}_0$ but is not sufficient to offset the increase in non-expected output $Y_3{Y}_0$. The inflow of foreign capital reduces GTFP in the inflow area.

Based on the aforementioned analysis, the relative magnitude of the effects of technological innovation and pollution transfer determines how much FDI affects GTFP. Overall, China’s capacity for green technology innovation still lags behind that of developed nations, and foreign capital’s technological spillover is a major driver in the advancement of green technology. The government has increased the strictness of environmental rules while aggressively luring foreign investment, which has lessened the impact of FDI on pollutant transfer. Consequently, it makes sense to believe that FDI can encourage an overall improvement in GTFP. However, there is a glaring regional mismatch and polarization in China’s distribution of innovation resources, which is highly uneven. Higher levels of economic growth, larger urban scales, and the eastern coastal regions are associated with relatively high levels of autonomous innovation efficiency and capacity. There is little variation in green technology between domestic and foreign firms and less reliance on foreign investment. This leads to a smaller marginal improvement effect of FDI on green technology innovation efficiency. The entry of low-quality foreign capital, to some extent, may even squeeze out innovation resources in cities and may even inhibit the improvement of GTFP.

In conclusion, we posit the subsequent hypothesis:

Hypothesis 3.

In terms of the total effect, FDI can promote the improvement of GTFP.

Hypothesis 4.

Environmental regulation can promote GTFP through FDI.

Hypothesis 5.

In regions with high innovation capacity, FDI has a dampening effect on GTFP.

3. Research Design

3.1. Model Construction

3.1.1. Baseline Regression Model

To examine the effect of environmental regulations on GTFP, the paper presents the baseline regression model as follows:

$$GTF{P}_{it}={\beta }_0+cE{R}_{it}+\gamma X_{it}+f_i+f_t+{\epsilon }_{it}$$

(1)

In Equation (1), $i$ represents the region index, and $t$ is the period index. $GTF{P}_{it}$ and $E{R}_{it}$ represent GTFP and environmental regulation intensity in region $i$ at time $t$. $X_{it}$ are control variables. $f_t$ and $f_i$ represent time and region fixed effects, respectively, and ${\epsilon }_{it}$ is the random error term. If the coefficient $c$ of $E{R}_{it}$ is significantly positive at the 10% level, it reveals that environmental regulation benefits GTFP, which supports Hypothesis 1; otherwise, it is not supported.

3.1.2. Mediation Effects Model

This research builds a mediation model based on theoretical analysis and research assumptions to examine how environmental regulations and FDI affect GTFP. The specific econometric model is set as follows:

$$FD{I}_{it}={\beta }_0+aE{R}_{it}+\gamma X_{it}+f_i+f_t+{\epsilon }_{it}$$

(2)

$$GTF{P}_{it}={\beta }_0+c^{\prime }E{R}_{it}+bFD{I}_{it}+\gamma X_{it}+f_i+f_t+{\epsilon }_{it}$$

(3)

Equation (2) explores how environmental regulations affect FDI, where $FD{I}_{it}$ reflects foreign direct investment. If the estimated value of coefficient $a$ is statistically positive at the 10% level, it indicates that environmental regulations are favorable to FDI, and Hypothesis 2 is confirmed.

The direct impact of environmental regulations on GTFP is shown in Equation (3), as is the indirect impact of FDI on GTFP. If the coefficient $b$ is statistically positive at the 10% level, it indicates that FDI promotes GTFP, supporting Hypothesis 3. Moreover, when coefficient $c^{\prime }$ is statistically positive at the 10% level and is less than $c$ in Equation (1), it reveals that environmental regulation promotes GTFP through FDI, and Hypothesis 4 is verified.

3.2. Variable Description

3.2.1. Dependent Variable

The paper measures the efficiency of green technology innovation in each region by utilizing the super-efficiency SBM model [53] under undesirable outputs. The super-efficiency SBM model, in contrast to the conventional DEA model, has the ability to compare the effective Decision-Making Units (DMUs) in addition to handling undesirable outputs. The specific model is constructed as follows.

$$\begin{gathered} Min \rho =\frac{1+\frac{1}{m}{{\displaystyle \sum }}_{i=1}^m\raisebox{1ex}{${s_i}^{-}$}\!\left/ \!\raisebox{-1ex}{$x_{ik}$}\right.}{1-\frac{1}{q_1+q_2}({{\displaystyle \sum }}_{r=1}^{q_1}\raisebox{1ex}{${s_r}^{+}$}\!\left/ \!\raisebox{-1ex}{$y_{rk}$}\right.+{{\displaystyle \sum }}_{t=1}^{q_2}\raisebox{1ex}{${s_t}^{b-}$}\!\left/ \!\raisebox{-1ex}{$b_{tk}$}\right.)}, \\ s.t. {{\displaystyle \sum }}_{j=1,j\ne k}^n{x}_{ij}{\lambda }_j-{s_i}^{-}\le x_{ik}, \\ {{\displaystyle \sum }}_{j=1,j\ne k}^n{y}_{rj}{\lambda }_j+{s_r}^{+}\ge y_{rk}, \\ {{\displaystyle \sum }}_{j=1,j\ne k}^n{b}_{tj}{\lambda }_j-{s_t}^{b-}\le b_{tk}, \\ 1-\frac{1}{q_1+q_2}({{\displaystyle \sum }}_{r=1}^{q_1}\raisebox{1ex}{${s_r}^{+}$}\!\left/ \!\raisebox{-1ex}{$y_{rk}$}\right.+{{\displaystyle \sum }}_{t=1}^{q_2}\raisebox{1ex}{${s_t}^{b-}$}\!\left/ \!\raisebox{-1ex}{$b_{tk}$}\right.)>0, \\ \lambda ,s^{-},s^{+}\ge 0;i=1,2\cdots m;r=1,2\cdots q;j=1,2\cdots n\left(r\ne k\right), \end{gathered}$$

(4)

where $\rho$ represents the green technology innovation efficiency of the evaluated DMUs. $x_{ik}$, $y_{rk}$, and $b_{tk}$ are inputs, desirable outputs, and undesirable outputs of DMU. $q_1$ and $q_2$ represent a number of factors for expected and unexpected outputs. ${s_i}^{-}$, ${s_r}^{+}$, ${s_t}^{b-}$ donate slack factors of inputs, expected outputs, and unexpected outputs, respectively.

To reflect the dynamic characteristics of green technology innovation efficiency, we combined the results of Equation (4) with the Malmquist–Luenberger index (MLI) approach [31] to measure GTFP. Since the MLI represents the rate of change in productivity rather than the value of productivity itself. Therefore, in the actual analysis, it is necessary to set the productivity value of the base period and then multiply it by the index of the second period. Referring to previous research, the total factor productivity of each DMU was set to 1 in the first year, and the value of the first year multiplied by the index of the second year defined the total factor productivity in the second year. Similarly, the GTFP of each DMU at different periods was obtained and denoted as $GTFP$.

In selecting input and output indicators, we followed the existing literature [54,55] and used labor, capital, and energy as primary input indicators based on the availability of urban data, as shown in

Table 1. The descriptions of input and output indicators.

First Grade Indicators	Second Grade Indicators	Indicator Description
	Labor force input	Total employment at the end of the year in the region
Input index	Capital input	Estimation of regional capital stock using the perpetual inventory method
	Energy input	Regional total electricity consumption
Desirable outputs	Economic output	Regional Real GDP
	R&D output	Number of invention patents filed in the region
Undesirable outputs	Pollutant emissions	Regional industrial wastewater discharge
		Regional industrial sulfur dioxide emissions
		Regional industrial smoke (dust) emissions

. Specifically, the quantity of end-of-year employment in urban areas serves as a proxy for labor input. Total electricity consumption was used to calculate the energy input. The capital stock (deflated at the base year 2008) was evaluated using the perpetual inventory method, with a 9.6% depreciation rate [56,57]. We used fixed assets investment to represent capital input. Regarding the desirable output indicators, green technology innovation is adopted as an output indicator from two aspects: economic output and R&D output. The real gross domestic product (GDP) for the region was used to calculate economic output. We utilized the number of invention patents applied in the region to proxy R&D output. To reflect the green aspect, from the perspective of pollutant emissions, industrial wastewater discharge, SO₂ emissions, and industrial smoke (dust) emissions were utilized as indicators of undesirable outputs.

3.2.2. Core Independent Variable

We created a composite index to gauge the level of environmental regulation based on pollutant emissions [58,59]. Here is the expression for the equation.

$$E{R}_i=\raisebox{1ex}{$1$}\!\left/ \!\raisebox{-1ex}{${{\displaystyle \sum }}_j^m\frac{\raisebox{1ex}{$P_{ij}$}\!\left/ \!\raisebox{-1ex}{$Gd{p}_i$}\right.}{{{\displaystyle \sum }}_{i=1}^n\frac{\raisebox{1ex}{$P_{ij}$}\!\left/ \!\raisebox{-1ex}{$Gd{p}_i$}\right.}{n}}$}\right.,$$

(5)

where $E{R}_i$ denotes environmental regulation intensity, and $P_{ij}$ represents the $j$ pollutant emissions in region $i$. The ratio of $P_{ij}$ to $Gd{p}_i$ denotes pollutant emissions of type $j$ per unit GDP in region $i$. ${{\displaystyle \sum }}_{i=1}^n\frac{\raisebox{1ex}{$P_{ij}$}\!\left/ \!\raisebox{-1ex}{$Gd{p}_i$}\right.}{n}$ represents the average level of emissions of the $j$ pollutant per unit of national output value, where $m$ represents the number of pollutant types. The intensity of environmental regulation was measured by the ratio of the $j$ pollutant’s emissions per unit of output value in region $i$ to the average of $j$ pollutant’s emissions per unit of output value across the country. A greater degree of environmental regulation was indicated by the smaller ratio. For convenience, we took the reciprocal of this ratio.

3.2.3. Mediation Variable

Based on theoretical analysis, ${\mathrm{FDI}}_{\mathrm{it}}$ served as the mediation variable in this study. The ratio of the actual amount of FDI in each region to GDP was used as a proxy indicator for ${\mathrm{FDI}}_{\mathrm{it}}$. We first converted the amounts using the annual average exchange rate from US dollars to Chinese yuan. Then, we removed the impact of price fluctuations by adjusting the figures using the Fixed Asset Investment Price Index for each region, with 2007 serving as the base year.

3.2.4. Control Variables

In addition to government subsidies and environmental regulations, existing research indicates that GTFP is also influenced by factors such as human capital, firm size, informationalization level, degree of trade openness, and ownership structure. Therefore, this study incorporated these factors as control variables. The following are the specific methods of measurement. The degree of trade openness $\left(\mathrm{Tdo}\right)$ was determined by dividing total regional imports and exports by GDP. The informationalization level ($\mathrm{Inf}$) was proxied using the percentage of mobile phone subscribers to permanent residents. As a measure of human capital $(\mathrm{Hc}$), we utilized the percentage of college students among permanent residents. The logarithm of the average output of companies larger than the designated size was used to calculate the firm scale $\left(\mathrm{Scale}\right)$. Industrial structure $\left(\mathrm{Ic}\right)$ was calculated as the ratio of the tertiary industry’s added value to the secondary industry’s total added value. Transport infrastructure level $\left(\mathrm{Tran}\right)$ was proxied using the ratio of highway mileage to area. The ownership structure variable ($\mathrm{Es}$) was quantified using the ratio of the number of employees in urban non-private units at year-end to total employment. Finally, we defined government intervention ($\mathrm{Gov}$) as the percentage of the entire amount of fiscal expenditure to GDP.

3.3. Data Source

We utilized panel data from 288 Chinese cities at the prefecture level and above from 2008 to 2020 to examine how environmental regulation and FDI affect GTFP. These cities span across 31 provinces and municipalities (excluding county-level cities and autonomous prefectures). The year 2008 was chosen as the starting point because research indicates that the 2007 United Nations Conference on Climate Change spurred the development of green technology innovation in China [60] while avoiding the impact of the 2007–2008 global financial crisis. Our data were derived from the China Statistical Yearbook and the 2009–2021 provincial and municipality bulletins. In data processing, cities with severe data deficiencies were excluded. For cities with minor missing data, various methods, such as averaging and exponential smoothing, were employed to fill in the gaps based on the characteristics of the indicators. The descriptive statistics for pertinent variables are shown in

Table 2. Summary statistics of main variables.

Variables	Observation	Mean	Std. Dev.	Min	Max
$GTFP$	3744	0.6163	0.2100	0.1210	2.2754
$ER$	3744	2.0301	3.3747	0.2836	91.7963
$FDI$	3744	0.0290	0.0330	0.0000	0.4262
$Tdo$	3744	0.1957	0.3292	0.0000	3.2573
$Hc$	3744	0.0167	0.0197	0.0000	0.1276
$Scale$	3744	0.8435	0.6254	−1.3911	3.2929
$Ic$	3744	0.9898	0.5620	0.0943	5.3500
$Inf$	3744	0.9176	0.3641	0.0510	3.1330
$Tran$	3744	1.0582	0.5026	0.0476	2.6278
$Es$	3744	0.1934	0.1101	0.037	0.9397
$Gov$	3744	0.1984	0.1064	0.0426	1.0268

.

4. Empirical Results

4.1. Baseline Results

To verify the research hypotheses, this study combined panel data from 288 Chinese cities that were prefecture-level or above from 2008 to 2020. To explore the connection between environmental regulations, FDI, and GTFP, we employed a mediation effect model. Regression results are presented in

Table 3. Baseline results.

Variables	$\mathit{G}\mathit{T}\mathit{F}\mathit{P}$	$\mathit{F}\mathit{D}\mathit{I}$	$\mathit{G}\mathit{T}\mathit{F}\mathit{P}$	$\mathit{G}\mathit{T}\mathit{F}\mathit{P}$	$\mathit{F}\mathit{D}\mathit{I}$	$\mathit{G}\mathit{T}\mathit{F}\mathit{P}$
	(1)	(2)	(3)	(4)	(5)	(6)
$ER$	0.0075 ***	0.0009 ***	0.0073 ***	0.0072 ***	0.0008 ***	0.0071 ***
	(0.0009)	(0.0002)	(0.0009)	(0.0008)	(0.0002)	(0.0008)
$FDI$			0.1987 **			0.2274 ***
			(0.0902)			(0.0841)
$Tdo$				−0.1151 ***	0.0060 **	−0.1165 ***
				(0.0130)	(0.0026)	(0.0130)
$Hc$				−0.0685	−0.2168 **	−0.0192
				(0.4607)	(0.0934)	(0.4607)
$Scale$				0.0740 ***	−0.0044 ***	0.0750 ***
				(0.0082)	(0.0017)	(0.0082)
$Ic$				−0.0080	−0.0063 ***	−0.0066
				(0.0085)	(0.0017)	(0.0085)
$Inf$				0.0631 ***	0.0096 ***	0.0609 ***
				(0.0135)	(0.0027)	(0.0135)
$Tran$				−0.0296 *	0.0111 ***	−0.0322 *
				(0.0167)	(0.0034)	(0.0167)
$Es$				0.4395 ***	−0.0038	0.4403 ***
				(0.0416)	(0.0084)	(0.0415)
$Gov$				−0.5927 ***	−0.0067	−0.5912 ***
				(0.0484)	(0.0098)	(0.0484)
$Constant$	0.9834 ***	0.0199 ***	0.9795 ***	1.0081 ***	0.0139 ***	1.0050 ***
	(0.0066)	(0.0013)	(0.0069)	(0.0221)	(0.0045)	(0.0221)
Year FE	Yes	Yes	Yes	Yes	Yes	Yes
City FE	Yes	Yes	Yes	Yes	Yes	Yes
Observations	3744	3744	3744	3744	3744	3744
R-squared	0.6243	0.0406	0.6248	0.6781	0.0542	0.6788
City	288	288	288	288	288	288

Standard errors in parentheses, * p < 0.1, ** p < 0.05, *** p < 0.01

. The regression results with time and regional fixed effects, devoid of control variables, are shown in columns (1)–(3). Columns (4)–(6) present the results after including control variables such as trade openness, informationalization level, enterprise size, and industrial structure.

Environmental regulations greatly increased both GTFP and FDI, as shown by the significantly positive coefficients of environmental regulations in Table 3’s columns (1) and (2) at the 1% level. Thus, Hypothesis 1 and Hypothesis 2 were validated. In column (3), the coefficients of environmental regulations and FDI are both positive and significant, and the coefficient of environmental regulations is smaller than that in column (1). This suggests that environmental regulations can promote GTFP through FDI, and the mediation effect is significant. Therefore, Hypothesis 3 and Hypothesis 4 were proved. The model’s goodness of fit increased with control variables added while leaving the sign and significance of environmental regulations and FDI remain unchanged. This indicates the stability of the regression results, suggesting that environmental regulations reinforce GTFP both directly and indirectly through FDI.

Regarding control variables, we obtained that in columns (4) and (6), the coefficient of trade openness, transportation infrastructure, and the degree of government intervention is significantly negative; the coefficient of human capital and industrial structure is negative but not significant; and the coefficient of firm scale, informatization level, and ownership structure are significantly positive, which may be related to China’s comparative advantage. The coefficient of government intervention was notably negative because excessive government intervention will distorts the market mechanism to a certain extent, resulting in resource mismatch and efficiency loss [61,62]. This can impede the advancement of green technology innovation efficiency. Our study found that human capital did not have a significant impact on GTFP. In fact, prior research demonstrated that human capital has a non-promoting impact on green technology innovation efficiency [63], which is mainly related to the lagged effect of human capital [64,65]. The current human capital investment does not have an immediate impact on the efficiency of green technology innovation. Resources and labor are relatively cheap in China, where labor- and resource-intensive products make up the majority of the export structure. The increase in exports leads to changes in labor and capital input, resulting in a decline in factors use efficiency. To some extent, transportation infrastructure leads to resource waste and environmental pollution, and excessive government intervention in economic activities will distort the market mechanism, resulting in resource mismatch and efficiency loss. Meanwhile, green technology innovation mainly focuses on capital-intensive strategic emerging industries with high industry concentration, where non-private entities have higher efficiency and stronger market power compared with private units.

4.2. Robustness Checks

4.2.1. Replacing Dependent Variable

To mitigate the impact of measurement bias in explanatory variables on the estimation outcomes, based on innovation achievements, we utilized the number of applied green invention patents ($Gupa$) [66,67], as well as the number of granted green invention patents lagged by one period ($Gipa$) [68,69] as alternative indicators of GTFP. Robustness tests on the baseline regression results were conducted.

Table 4. Replacement dependent variable regression analysis.

Variables	$\mathit{G}\mathit{u}\mathit{p}\mathit{a}$	$\mathit{F}\mathit{D}\mathit{I}$	$\mathit{G}\mathit{u}\mathit{p}\mathit{a}$	$\mathit{G}\mathit{i}\mathit{p}\mathit{a}$	$\mathit{F}\mathit{D}\mathit{I}$	$\mathit{G}\mathit{i}\mathit{p}\mathit{a}$
	(1)	(2)	(3)	(4)	(5)	(6)
$ER$	0.1045 **	0.0008 ***	0.1010 **	0.0685 ***	0.0008 ***	0.0639 ***
	(0.0439)	(0.0002)	(0.0441)	(0.0170)	(0.0002)	(0.0170)
$FDI$			4.0972			5.1196 ***
			(4.4001)			(1.6897)
Year FE	Yes	Yes	Yes	Yes	Yes	Yes
City FE	Yes	Yes	Yes	Yes	Yes	Yes
Controls	Yes	Yes	Yes	Yes	Yes	Yes
Observations	3744	3744	3744	3456	3744	3456
R-squared	0.2820	0.0542	0.2822	0.1723	0.0542	0.1747
City	288	288	288	288	288	288

Standard errors in parentheses, ** p < 0.05, *** p < 0.01

displays the corresponding outcomes.

Table 4 shows that the regression results remain largely unchanged when the measurement indication for GTFP is replaced. The environmental regulation coefficients are notably positive at the 5% level in all specifications. Additionally, column (3) has a lesser coefficient than column (1), and column (6) has a smaller coefficient than column (4). This suggests that environmental regulation can, directly and indirectly, boost GTFP through FDI, providing further validation for research Hypothesis 1 to Hypothesis 4. The control variable’s coefficients and significance levels are generally consistent with the baseline results, which show that the benchmark regression findings are reasonably robust. The control variable regression results are available upon request; however, they are not presented due to space constraints.

4.2.2. Replacing Core Explanatory Variable

To mitigate potential measurement bias and guarantee the robustness of the regression results, the utilization rate of general industrial solid waste was employed as a proxy environmental regulation indicator [58], reflecting its regulatory effectiveness.

Table 5. Replacement explanatory variable regression analysis.

Variables	$\mathit{G}\mathit{T}\mathit{F}\mathit{P}$	$\mathit{F}\mathit{D}\mathit{I}$	$\mathit{G}\mathit{T}\mathit{F}\mathit{P}$
	(1)	(2)	(3)
$ER$	0.0468 ***	0.0076 ***	0.0447 ***
	(0.0135)	(0.0027)	(0.0135)
$FDI$			0.2729 ***
			(0.0846)
Year FE	Yes	Yes	Yes
City FE	Yes	Yes	Yes
Controls	Yes	Yes	Yes
Observations	3744	3744	3744
R-squared	0.6723	0.0497	0.6733
City	288	288	288

Standard errors in parentheses, *** p < 0.01.

displays the robustness of the regression results.

As indicated in Table 5, the coefficients associated with environmental regulation and FDI remain significantly positive at the 1% level even after applying alternative measurement methods of environmental regulation. Furthermore, the coefficient of environmental regulation in column (3) is smaller than the coefficient in column (1), suggesting that FDI and environmental regulation both significantly increase GTFP. The reliability of the benchmark regression results is confirmed by these conclusions.

4.2.3. Winsorization

This study used winsorization to lessen the possible impact of extreme outliers in FDI and environmental regulation on the regression findings. Taking environmental regulation as an example, the 1st and 99th percentiles were identified as the lower and upper thresholds, denoted as $l$ and $h$, respectively. $l$ was used to represent values below the 1st percentile, while $h$ was used to represent values above the 99th percentile. Re-conducting the empirical analysis yielded the regression findings, which are shown in

Table 6. The result of winsorization.

Variables	$\mathit{G}\mathit{T}\mathit{F}\mathit{P}$	$\mathit{F}\mathit{D}\mathit{I}$	$\mathit{G}\mathit{T}\mathit{F}\mathit{P}$
	(1)	(2)	(3)
$ER$	0.0280 ***	0.0023 ***	0.0275 ***
	(0.0019)	(0.0003)	(0.0019)
$FDI$			0.2100 **
			(0.1002)
Year FE	Yes	Yes	Yes
City FE	Yes	Yes	Yes
Controls	Yes	Yes	Yes
Observations	3744	3744	3744
R-squared	0.6916	0.0720	0.6920
City	288	288	288

Standard errors in parentheses, ** p < 0.05, *** p < 0.01.

.

The regression findings in Table 6 are in line with those in Table 3 after bilateral trimming on FDI and environmental regulation. The coefficients for environmental regulation and FDI are both notably positive at the 10% level; the coefficient for environmental regulation in column (3) is smaller than that in column (1). These results imply that environmental regulation directly promotes GTFP and indirectly stimulates it through FDI. Hence, the baseline regression results are deemed reliable.

4.3. Endogeneity Test

4.3.1. Instrumental Variable Method

To address potential endogeneity issues in model specification, this study employed instrumental variable estimation for control. Regarding the selection of instrumental variables, we considered that the current dependent variable is not influenced by the lagged independent variable, and there exists a strong correlation between the current and lagged independent variables, which meets the criteria for instrumental variable selection. Thus, the lagged one-period environmental regulation was used as an instrument, and two-stage least squares (2SLS) estimation was conducted.

Table 7. Result of 2SLS.

Variables	$\mathit{G}\mathit{T}\mathit{F}\mathit{P}$	$\mathit{F}\mathit{D}\mathit{I}$	$\mathit{G}\mathit{T}\mathit{F}\mathit{P}$
	(1)	(2)	(3)
$ER$	0.0062 ***	0.0011 ***	0.0059 ***
	(0.0011)	(0.0002)	(0.0011)
$FDI$			0.2253 ***
			(0.0822)
First-stage regression (IV)	0.7243 ***	0.7243 ***	0.7243 ***
	(0.0098)	(0.0098)	(0.0098)
Cragg–Donald Wald F	5438.93	5438.93	5438.93
Kleibergen–Paaprk Wald F	40.13	40.13	40.13
Year FE	Yes	Yes	Yes
City FE	Yes	Yes	Yes
Controls	Yes	Yes	Yes
Observations	3456	3456	3456
R-squared	0.7454	0.6673	0.7459
City	288	288	288

Standard errors in parentheses, *** p < 0.01.

demonstrates the regression results.

The first-stage regression findings from Table 7 indicate that all of the instrumental variable coefficients are statistically positive at the 1% level. Additionally, all Cragg–Donald Wald F statistics and Kleibergen–Paaprk Wald F statistics are greater than the critical values at the 10% level, indicating the absence of weak instrument problems and confirming the effectiveness of the selected instruments. After estimating using 2SLS, both the coefficients for environmental regulation and FDI were positive and passed the 10% significance test. This suggests that environmental regulation and FDI considerably enhance GTFP, implying that the regression results are relatively solid.

4.3.2. Difference-in-Difference Approach

China’s low-carbon city pilot program is an initiative to explore green development methods and encourage cities to develop low-carbon economies. Low-carbon pilot cities are required to comprehensively incorporate climate change and other work into urban development planning. Therefore, these cities have far higher levels of environmental control than non-pilot cities, serving as an alternative indicator of environmental regulation. An efficient quasi-natural experiment was provided to investigate the effects of low-carbon city pilot policies on GTFP by the National Development and Reform Commission of China, which, from 2010 to 2018, released three batches of lists of low-carbon city pilot projects totaling 81 cities. To alleviate potential endogeneity issues, this study established a multi-period difference-in-differences model and used policy variables as alternative indicators of environmental regulation to re-examine the empirical evidence. The regression results are displayed in

Table 8. Results of difference-in-difference approach.

Variables	$\mathit{G}\mathit{T}\mathit{F}\mathit{P}$	$\mathit{F}\mathit{D}\mathit{I}$	$\mathit{G}\mathit{T}\mathit{F}\mathit{P}$
	(1)	(2)	(3)
$Time\ast Treat$	0.0468 ***	0.0012	0.0465 ***
	(0.0078)	(0.0016)	(0.0078)
$FDI$			0.2798 ***
			(0.0843)
Year FE	Yes	Yes	Yes
City FE	Yes	Yes	Yes
Controls	Yes	Yes	Yes
Observations	3744	3744	3744
R-squared	0.6746	0.0477	0.6756
City	288	288	288

Standard errors in parentheses, *** p < 0.01.

.

In Table 8, the $Time$ variable represents the low-carbon pilot city’s issue time, $Treat$ indicates whether the city is within the scope of the low-carbon city program, and $Time\ast Treat$ represents the policy variable. The low-carbon city program and FDI have positive coefficients, and all of them pass the 1% significance test with the exception of column (2), according to the regression results. This also indicates that environmental regulation and FDI have significant promoting effects on GTFP.

5. Heterogeneity Analysis

5.1. Level of Economic Development

In the previous theoretical analysis, FDI has a dual impact on GTFP. When cities have strong independent innovation capability, the technology spillover effect of FDI may be weakened. When the technology spillover effect is lower than the pollution transfer effect, it may hinder GTFP. Generally, regional independent innovation capability is positively correlated with economic development level. To verify Hypothesis 5, this study ranked urban GDP per capita from 2008 to 2020. The top 25% of regions were classified as having high levels of economic development, and the bottom 75% as having medium to low levels. There were 223 cities with medium to low levels of economic development and 65 cities with high levels once the sample data were grouped.

Table 9. Heterogeneity tests based on economic development level.

	Developed Cities			Less Developed Cities
Variables	$\mathit{G}\mathit{T}\mathit{F}\mathit{P}$	$\mathit{F}\mathit{D}\mathit{I}$	$\mathit{G}\mathit{T}\mathit{F}\mathit{P}$	$\mathit{G}\mathit{T}\mathit{F}\mathit{P}$	$\mathit{F}\mathit{D}\mathit{I}$	$\mathit{G}\mathit{T}\mathit{F}\mathit{P}$
	(1)	(2)	(3)	(4)	(5)	(6)
$ER$	0.0264 ***	0.0037 ***	0.0257 ***	0.0055 ***	0.0007 ***	0.0053 ***
	(0.0058)	(0.0009)	(0.0058)	(0.0006)	(0.0002)	(0.0006)
$FDI$			0.1963			0.2993 ***
			(0.2339)			(0.0721)
Year FE	Yes	Yes	Yes	Yes	Yes	Yes
City FE	Yes	Yes	Yes	Yes	Yes	Yes
Controls	Yes	Yes	Yes	Yes	Yes	Yes
Observations	845	845	845	2899	2899	2899
R-squared	0.3795	0.1093	0.3800	0.8309	0.0645	0.8320
City	65	65	65	223	223	223

Standard errors in parentheses, *** p < 0.01.

displays the regression findings.

The coefficients of environmental regulation in Table 9’s columns (1) to (3) are all statistically positive at the 1% level, suggesting that in areas with high levels of economic growth, environmental regulation promotes GTFP and FDI. In column (3), the FDI coefficient is positive but not statistically significant, indicating that FDI has no discernible effect on GTFP in areas with high levels of economic growth. This implicitly supports Hypothesis 5. Cities with high levels of economic growth have strong independent innovation abilities and lower dependence on foreign investment in green technology. The entry of low-quality foreign investment not only squeezes out the innovation resources of the city but also causes spatial transfer of pollution, which may have an inhibition effect on GTFP to some extent. The coefficients of FDI and environmental regulation in columns (4) to (6) are both significantly positive, suggesting that environmental regulation both increases FDI and boosts GTFP in places with medium to low levels of economic growth. When the coefficients are compared, it can be shown that environmental regulation in high economic development level cities is greater than that in medium to low-level cities, which indicates that environmental regulation has a greater promoting effect on GTFP in regions with high economic development levels. This is closely related to the long-term growth model of China, which emphasizes high input, high energy consumption, and high emissions. The higher the economic development level of an area, the more severe its ecological environmental problems and the greater the influence of environmental regulation.

5.2. City Scale

This research divided the sample data based on the 2014 State Council “Notice of the State Council on Adjusting the Standards for Categorizing City Sizes” to investigate the connection between environmental regulation, FDI, and GTFP concerning the scale of cities. Large-sized cities were specifically defined as those having a permanent population of one million or more, while medium-sized and small-sized cities were defined as those with a population of less than one million. After dividing the sample according to the city size, there were a total of 213 medium-sized and small-sized cities and 75 large-sized cities.

Table 10. Heterogeneity tests based on city size.

	Large-Sized Cities			Medium and Small-Sized Cities
Variables	$\mathit{G}\mathit{T}\mathit{F}\mathit{P}$	$\mathit{F}\mathit{D}\mathit{I}$	$\mathit{G}\mathit{T}\mathit{F}\mathit{P}$	$\mathit{G}\mathit{T}\mathit{F}\mathit{P}$	$\mathit{F}\mathit{D}\mathit{I}$	$\mathit{G}\mathit{T}\mathit{F}\mathit{P}$
	(1)	(2)	(3)	(4)	(5)	(6)
$ER$	0.0201 ***	0.0038 ***	0.0191 ***	0.0057 ***	0.0006 ***	0.0055 ***
	(0.0029)	(0.0007)	(0.0030)	(0.0008)	(0.0002)	(0.0008)
$FDI$			0.2644 *			0.3374 ***
			(0.1495)			(0.0989)
Year FE	Yes	Yes	Yes	Yes	Yes	Yes
City FE	Yes	Yes	Yes	Yes	Yes	Yes
Controls	Yes	Yes	Yes	Yes	Yes	Yes
Observations	975	975	975	2769	2769	2769
R-squared	0.6021	0.1047	0.6035	0.7271	0.0647	0.7284
City	75	75	75	213	213	213

Standard errors in parentheses, * p < 0.1, *** p < 0.01.

exhibits the regression outcomes.

The coefficients of FDI and environmental regulation were both positively significant at the 10% level, as Table 10 illustrates. This implies that environmental regulations have a direct stimulating influence on GTFP in medium-sized and small cities alike, in addition to an indirect stimulating effect through FDI. Through coefficient comparison, it can be observed that big cities are more affected by environmental regulation on GTFP than are medium and small cities and that medium and small cities are more affected by FDI on GTFP. This may be attributed to the relatively weak urban innovation foundation and the independent innovation capability of medium and small cities. The inflow of FDI not only provides financial support but also facilitates technology spillovers through training, demonstration and learning, and industry linkages, thereby improving green technology ability and significantly boosting GTFP in receiving areas.

5.3. Geographic Locations

The degree of foreign capital usage and the strength of environmental regulations vary significantly across eastern, central, and western China. To examine whether the geographical location of cities affects the impact of environmental regulation and FDI on GTFP, this paper divided the sample data into 100 cities in the eastern region and 188 cities in the central and western regions according to the geographical location of cities, grouped for empirical test.

Table 11. Heterogeneity tests based on geographic location.

	Eastern Regions			Central and Western Regions
Variables	$\mathit{G}\mathit{T}\mathit{F}\mathit{P}$	$\mathit{F}\mathit{D}\mathit{I}$	$\mathit{G}\mathit{T}\mathit{F}\mathit{P}$	$\mathit{G}\mathit{T}\mathit{F}\mathit{P}$	$\mathit{F}\mathit{D}\mathit{I}$	$\mathit{G}\mathit{T}\mathit{F}\mathit{P}$
	(1)	(2)	(3)	(4)	(5)	(6)
$ER$	0.0049 ***	0.0003	0.0049 ***	0.0245 ***	0.0014 ***	0.0238 ***
	(0.0009)	(0.0002)	(0.0009)	(0.0019)	(0.0004)	(0.0019)
$FDI$			0.1646			0.5233 ***
			(0.1342)			(0.1050)
Year FE	Yes	Yes	Yes	Yes	Yes	Yes
City FE	Yes	Yes	Yes	Yes	Yes	Yes
Controls	Yes	Yes	Yes	Yes	Yes	Yes
Observations	1300	1300	1300	2444	2444	2444
R-squared	0.6678	0.0953	0.6682	0.7284	0.1310	0.7314
City	100	100	100	188	188	188

Standard errors in parentheses, *** p < 0.01.

displays the regression outcomes.

Table 11 displays the regression results for the eastern part of China. All of the environmental regulation coefficients are positive, with the exception of column (2), which passes the 1% significance test. The FDI coefficient, on the other hand, is negative but not statistically significant. This suggests that while environmental regulations have the potential to boost GTFP in the eastern region, they have little effect on FDI. Due to the eastern region’s relatively high autonomous innovation ability and innovation efficiency, as well as the relatively small technology gap with foreign-funded enterprises, the effect of FDI on GTFP is not immediately apparent. As a result, there is only a slight marginal improvement effect of FDI on green technology innovation. Low-quality foreign investment may even prevent innovation efficiency from improving, which once again confirms research Hypothesis 5. In the regression outcomes for the central and western regions, the coefficients of environmental regulation and FDI are both strongly positive. This indicates that environmental regulation can not only promote GTFP in these regions but also further enhance GTFP by increasing FDI.

6. Conclusions and Implications

Environmental regulation and FDI have been controversial in academia regarding their impact on GTFP. This paper examines the relationship between environmental regulation, FDI, and GTFP using panel data from Chinese cities spanning from 2008 to 2020, employing various methods, such as the mediation effect model, 2SLS method, and difference-in-differences method. The research findings are the following: (1) GTFP greatly benefits from environmental regulation. (2) On the whole, FDI contributes to the improvement of GTFP, and environmental regulations have an indirect impact on GTFP through FDI. (3) The impact of environmental regulation and FDI on GTFP exhibits regional heterogeneity. In regions with higher levels of economic development and larger urban sizes, as well as in the central and western regions, environmental regulation significantly improves GTFP. Small and medium-sized cities are more affected by FDI’s promotion of GTFP than large cities, whereas the eastern areas with higher levels of economic growth are not greatly impacted.

According to the research findings, the following policy implications can be drawn: First, it is recommended to moderately enhance the intensity of environmental regulation. Initially, the government needs to bolster oversight of polluting sectors and promote industrial restructuring. Based on the current status of industrial development and the effectiveness of environmental governance, dynamic adjustments should be made to the negative list, clearly specifying categories of projects prohibited for investment and construction. Industries with large emissions of pollutants, severe overcapacity, and prominent environmental issues should be subject to key control measures, gradually phasing out backward and overcapacity sectors and promoting industrial restructuring. Then, the coordination of multiple environmental regulation tools should be promoted, and a diversified environmental regulation system should be established. Led by command-and-control regulations, the implementation of market-incentive regulations should be gradually advanced while using voluntary environmental regulations as a complement. This will facilitate the coordination of different types of environmental regulation tools, fully leveraging their synergistic effects to induce green technological innovation and enhance corporate initiative. Moreover, while increasing the intensity of environmental regulations, it also increases funding for corporate green technology R&D and incentivizes enterprises to carry out green technological innovation.

Second, it is imperative to intensify efforts to attract FDI. Firstly, government support should be enhanced, the catalog of encouraged foreign investment industries and the list of foreign-funded projects should be expanded, tax support policies should be implemented, and financial support should be increased to provide high-quality financial services and financing support for eligible foreign investment projects. Secondly, the foreign investment environment should be optimized. A good investment environment can enhance the attractiveness of foreign enterprises, establish a fair and open market environment aligned with international high-standard trade and economic rules, and create a more free and convenient trade and investment environment, fully implementing national treatment for foreign enterprises. Thirdly, it is crucial to implement differentiated policies to attract investment. For regions with high innovation capacity and efficiency, it is vital to suitably raise the barriers to foreign investment entry, eliminate foreign enterprises that use a lot of energy and produce a lot of pollution, and promote the upgrading of foreign investment structure. For regions with weak innovation capacity and insufficient R&D investment, the government should appropriately relax foreign investment access standards in some industries, give full play to cost and resource advantages, strengthen cooperation and exchanges with foreign enterprises, and promote green technology innovation. In addition, a favorable investment environment can also increase the demand for human capital, which is broadly shown to be a key factor for modern economic growth [70,71].

The following are the limitations of this study. Due to the data availability, the index selection of environmental regulation is based on the perspective of economics, without considering the number of local laws and regulations related to environmental protection, the number of administrative cases accepted and other command and control regulation methods. In the selection of the index of the undesirable output of GTFP, the industrial “three wastes” are selected for measurement, which does not cover the environmental pollution of other production sectors. In the future, we will try to build a more comprehensive index system. Moreover, micro-level validation is insufficient. This paper is based on the city-level data. Enterprises are the main drivers of green technology innovation, as well as the main objects of environmental regulation. In the next step, we can examine the influence of environmental regulation on the efficiency of enterprise green technology innovation from a micro perspective.