Do Foreign Direct Investment Inflows in the Producer Service Sector Promote Green Total Factor Productivity? Evidence from China

Abstract

By exploring feasible pathways for coordinating the ecological environment and economic development, this study investigated the impact of FDI in the producer service sector (SFDI) on green total factor productivity (GTFP) across 20 provinces from 2006 to 2019 in China. We employed a panel data regression model and found that SFDI significantly promotes China’s GTFP, verifying the existence of the “pollution halo” effects of SFDI in China, where GTFP is estimated by the global Malmquist–Luenberger productivity index based on the slack-based measure and directional distance function. We also employed mediating and moderating models to test the mechanism and found that SFDI can affect GTFP through competition, green innovation, and resource allocation mechanisms. Notably, the impact of SFDI on GTFP exhibits regional heterogeneity, with the strongest impact observed in the eastern region, followed by the western region, and the weakest in the central region. Further analysis reveals that the enhancement of environmental regulations and the level of factor marketization can amplify the influence of SFDI. Finally, we offer specific recommendations encompassing the enhancement of openness, improvement of factor markets, and strengthening of environmental regulations.

1. Introduction

The concept of producer services was first proposed by Greenfield [1], referring to the industries that provide intermediate products or services for production processes. According to the International Monetary Fund, foreign direct investment (FDI) refers to the cross-border investment in which an investor who is resident in one economy has control or a significant degree of influence on the management of an enterprise that is resident in another economy. In terms of FDI in the producer service sector, it demonstrates notable characteristics of knowledge intensity, capital intensity, and industrial synergy [2]. Moreover, compared to the domestic counterparts, multinational corporations in this sector generally possess comparative advantages in knowledge and technology [3], making them pivotal conduits for international knowledge and technology spillovers.

With China’s deepening integration into the global economy, FDI inflows into the country have experienced a remarkable upsurge over the past two decades. According to the World Investment Report 2022, China has consistently ranked as the world’s second-largest recipient of FDI for five consecutive years. While the development of the export-oriented economy that leverages its labor force and natural resource advantages has propelled China’s industrial progress and economic growth, it has also led to intensified resource and environmental constraints. China is currently falling on the left-hand side of the environmental Kuznets Curve [4], demonstrating a challenge in balancing environmental pollution and economic growth. Hence, it is crucial to adopt a transformative development approach and facilitate a transition from an extensive to an intensive mode of attracting and utilizing foreign investment.

Total factor productivity (TFP) refers to the ratio of total output to the aggregate inputs of all factors of production. Green total factor productivity (GTFP) encompasses the inclusion of energy consumption and environmental pollution within the TFP measurement framework, which evaluates the level of development that integrates both economic and environmental benefits. Previous studies have contributed valuable insights to the relationship between the two variables. The relevant research closely related to these issues is summarized in the following section.

As cross-border investment increasingly shifts to the service sector, scholars are devoting greater attention to the impact of FDI in the service sector, particularly in the producer service sector within the host country. Yu et al. [5], based on the Diamond model, found that SFDI promotes China’s manufacturing productivity. Analogous findings have been discovered in the case of India, Ukraine, and Asia Pacific [6,7,8]. Employing data from the Centre for Monitoring Indian Economy (CMIE), BAS [9] demonstrates that service sector liberalization significantly fosters the export growth for local manufacturing, particularly for high-productivity manufacturing enterprises, and this effect is achieved through channels of technological innovation and production cost reduction. Bas and Strauss-Kahn [10] utilized China’s 2000–2006 transaction data to establish a quasi-natural experiment, revealing that by liberalizing upstream input sectors, the export product quality of downstream industries undergoes significant improvement.

From the perspective of subdivided industries within the producer service sector, Beverelli et al. [11] identified that the relaxation of foreign investment constraints in the financial industry can alleviate financing constraints for enterprises, thereby facilitating improvements in manufacturing production efficiency. However, based on bank-level data from Thailand, Lu and Mieno [12] found that the effect of foreign entry into the banking sector is limited. Rehman et al. [13] constructed the GINF index and found that the inflow of FDI into the transport, telecommunication, energy, and financial sectors facilitated the development of relevant sectors’ infrastructure. Orlic et al. [14] found that compared to FDI in the manufacturing and labor-intensive service sectors, FDI in the knowledge-intensive service sector is more effective in promoting manufacturing efficiency. Employing the GMM and GWR models, Li and Wang [15] found that the two-way FDI of the logistics industry promotes its TFP and that the positive impact of the logistics industry’s inward FDI is stronger.

Incorporating environmental considerations into the assessment of the impacts of FDI, many scholars focus on its influence on GTFP in the host country. However, there is currently a controversial debate ongoing regarding the effect of these impacts, mainly encompassing three viewpoints: positive effects, negative effects, and nonlinear effects. Regarding the positive perspective, Wang et al. [16] employed the improved weighted Russell directional distance function to measure GTFP and found that FDI generated a halo effect in China. Li et al. [17] discovered that the inflow of FDI promoted the improvement of GTFP in China’s equipment manufacturing industry, and this effect is influenced by the level of human capital, technological innovation, and environmental regulation. From the negative perspective, Qiu et al. [18] found, based on micro-level enterprise data in China, that FDI inflows hinder the improvement of GTFP in the central and western regions of China, but the enhancement of economic benefits mitigates the negative impact of FDI on GTFP. Chai et al. [19], utilizing the SBM model and Malmquist–Luenberger index to measure GTFP, revealed that FDI at the aggregate level significantly suppresses the improvement of GTFP, but the degree of institutional development plays a moderating role. You and Xiao [20] conducted a threshold regression analysis and discovered that as marketization levels increase, the environmental benefits of FDI initially decrease and then increase. Furthermore, maintaining a certain range of environmental regulations can enhance the positive impact of FDI on GTFP.

Based on the above literature review, it is clear that a substantial body of research has examined the relationship between FDI and GTFP. However, due to variations in research objects, timeframes, regulatory intensity, and approaches, inconsistencies in conclusions arise regarding the influence of FDI. Moreover, given the high knowledge and capital intensity associated with the producer service sector, there may exist environmental effects that are heterogeneous to those of the industrial and manufacturing sectors. The existing research on the environmental performance of the service sector, specifically the producer service sector, remains limited. Therefore, investigating its impact on China’s GTFP has become a topic of utmost theoretical and practical significance.

Relative to the existing literatures, the marginal contributions of this paper are as follows: First, in terms of the scope of the research, we incorporate environmental factors into our analysis framework, extending the research scope associated with the effect of SFDI. Second, from the perspective of estimation, we employ the global Malmquist–Luenberger productivity index based on the slack-based measure and directional distance function, improving the precision of the estimation. Third, regarding the research content, we investigate the mechanisms of SFDI, including competition, green innovation, and resource allocation, revealing the influence of SFDI on GTFP from multiple dimensions. In addition, by incorporating the degree of factor marketization and environmental regulation into our analytical framework, our study provides valuable insights into the operation of the domestic and international dual circulation. The remainder of this paper is organized as follows. Section 2 elucidates the pertinent theoretical mechanism. Section 3 presents the research design employed in this study. Section 4 describes the conduction of the empirical specification and robustness tests. Finally, Section 5 draws conclusions and offers policy suggestions.

2. Theoretical Framework

This section will discuss the three potential mechanisms of SFDI on GTFP, including the productivity spillover effect, environmental effect, and resource allocation effect. In addition, these mechanisms will be examined in Section 4.2.

2.1. Productivity Spillover Effect

First, SFDI has the potential to generate a horizontal spillover effect on the host country’s producer service sector. Multinational corporations operating in the producer service sector possess inherent competitive advantages in terms of management services, production efficiency, and product quality [21], enabling them to overcome entry barriers and facilitate the creation of horizontal spillovers [22,23]. The entry of multinational corporations intensifies market competition in related industries [24], simultaneously providing a learning paradigm for local enterprises. Local enterprises, in response to market dynamics and the need for efficiency improvement, engage in the process of learning and imitating the high-efficiency, high-profit production management methods demonstrated by multinational corporations. This “demonstration and imitation effect” [16] contributes to the enhancement of local enterprise efficiency. It is worth noting, however, that scholars have cautioned about the existence of a “crowding-out effect” [25,26] when there is a significant technological gap between domestic and foreign-funded enterprises. Intensified market competition, in such cases, may negatively impact local enterprises and hinder the overall development of the industry. Moreover, multinational enterprises, motivated by the imperative of competition in host nations, commonly safeguard their proprietary knowledge to prevent it from leaking to its competitors [27]. Therefore, the efficacy of foreign direct investment in the producer service sector leading to international technology diffusion encounters impediments.

Second, SFDI has the potential to generate vertical spillover effects, encompassing forward spillovers to downstream industries and backward spillovers to upstream industries. In the case of forward spillovers, SFDI is embedded in the production chain as advanced inputs, providing downstream manufacturing with a broader array and superior quality of services that local suppliers are unable to match. These inputs subsequently enhance the technological level and production efficiency of the downstream manufacturing sector. In the case of backward spillovers, multinational corporations facilitate the transfer of advanced management practices, production technology, and other knowledge to their upstream suppliers through mechanisms such as employee training, quality control, and inventory management. Additionally, their market entry expands the demand for related inputs by upstream enterprises, stimulating production scale expansion and potentially yielding economies of scale. Furthermore, the impact of SFDI is influenced by external technological disparities, human capital, and institutional levels [28,29]. Generally, smaller technological gaps, a relaxed market environment, and favorable human capital conditions are conducive to harnessing its positive externalities.

2.2. Environmental Effect

Regarding the environmental effect of SFDI, there are two opposing viewpoints, namely the “pollution haven” hypothesis and the “pollution halo” hypothesis [17,18]. The former postulates that there are lax environmental regulations in host countries which provide multinational corporations with a “refuge” for pollution [30,31,32,33], while the latter argues that FDI facilitates the diffusion of environmentally friendly production techniques and management methods in host countries [34,35]. In particular, from the positive perspective, SFDI serves as a crucial channel for transferring green technologies and sustainable management practices, where multinational corporations in the producer service sector exhibit superior environmental efficiency compared to domestic enterprises [36,37], thereby fostering spillover effects in the realms of green technology, processes, and management practices [38,39] as well as leading to technology upgrades through knowledge externalities and the introduction of new technologies. Moreover, multinational corporations in the producer service sector may actively publicize their environmental advantages in order to make a social and environmental impact, leading to increased environmental awareness among local residents. This heightened awareness, in turn, exerts pressure on the government to enact more stringent environmental policies [40,41].

Conversely, from the negative perspective, the environmental pollution resulting from SFDI in the host country follows a similar trajectory as that observed in the manufacturing sector. Influenced by environmental regulations and cost considerations, multinational corporations within the producer service sector may relocate more polluting segments to developing countries with comparatively lax environmental regulations [42]. As the developing countries are situated at the lower end of the global value chain, the host country may be inclined to trade off some environmental benefits to attract these multinational corporations in pursuit of economic growth. Subsequently, the entry of FDI can amplify pollutant emissions and engender negative externalities on the host country’s environment [43,44].

2.3. Resource Allocation Effect

Heterogeneity in the marginal returns of production factors across industries leads to factor flows. As multinational corporations enter the host country, they exhibit comparative advantages in terms of wages, benefits, and other aspects, resulting in a “wage effect” that attracts high-quality labor from less efficient sectors, consequently enhancing labor efficiency [45,46]. Simultaneously, producer service industries exhibit higher marginal returns on investment, making them more appealing to investors. The inflows of FDI in these related industries induce additional investment in the producer service sector, as well as its upstream and downstream sectors. Moreover, the entry of foreign financial institutions expands regional credit supply, broadens the eligible borrowers for local enterprises, and reduces their financing costs by introducing competitive mechanisms, which will help alleviate the problem of enterprise financing constraints caused by capital distortion [47,48]. When financial constraints are eased and liquidity risk is reduced, enterprises are more willing to invest in projects with longer return cycles, such as technological innovation and green production [49,50], which in turn promotes the improvement of the GTFP. The reallocation of labor and capital factors facilitated by SFDI among industries promotes the optimization and upgrading of the regional industrial structure, enhancing the efficiency of factor allocation among primary, secondary, and tertiary sectors. As a result, the regional GTFP is improved.

Based on the above analysis, we put forward the following hypotheses:

Hypothesis 1 (H1).

SFDI engenders heightened industry competition, leading to spillover effects alongside potential crowding-out effects. The ultimate impact is contingent upon the net outcome of the two factors.

Hypothesis 2 (H2).

The environmental implications of SFDI are influenced by the stringency level of environmental regulations in the host country.

Hypothesis 3 (H3).

SFDI can affect GTFP by competition within the producer service sector, allocation of resources across industries, and green innovation.

3. Data and Methodology

3.1. Empirical Model and Variables

This paper identifies the causal effect between SFDI and GTFP in several steps.

3.1.1. Base Model

The first step is the base model. We construct a panel data regression model where all the regressions are estimated, including random disturbance, to neutralize the risk of unobserved confounding factors. Following the approach of Gulen and Ion [51], we construct the base empirical model as follows:

$$GTF{P}_{it}={\beta }_0+{\beta }_{1}SFD{I}_{it}+{\delta }_{it}+{\epsilon }_{it}$$

(1)

where subscript $i$ denotes each province and $t$ is the time index. $GTF{P}_{it}$ is the explained variable, representing the green total factor productivity of province $i$ in year $t$. $SFD{I}_{it}$ is an explanatory variable, representing the FDI in the producer service sector of province $i$ in year $t$. $X_{it}$ is a vector of control variables and ${\epsilon }_{it}$ is the random disturbance.

The explained variable of the research is the GTFP. Scholars often employ data envelopment analysis to measure productivity. Nonetheless, traditional data envelopment analysis fails to incorporate unexpected outputs, leading to measurement errors. To address this, Chuang et al. [52] extended their approaches by introducing a new directional distance function which allows the measurement of productivity inclusive of unexpected outputs using the ML index to take place. However, the ML index has limitations, including infeasibility in linear programming and the inability to facilitate multi-period comparisons. Additionally, radial and angular approaches struggle to overcome the bias introduced by the relaxation variables. Consequently, to overcome the measurement errors resulting from the slack variables, Fukuyama et al. [53] proposed a generalized slack-based measure and directional distance function that is non-radial and non-angular. To improve the problem of the LM index, Oh [54] constructed a global production possibility set that enables multi-period comparisons. Considering the influence of slack variables and the comparability of indicators across years, we employ the slack-based measure and directional distance function and estimate productivity using the global Malmquist–Luenberger productivity index.

Firstly, for the production possibility set (PPS), we assume that there are $K$ decision-making units ($DM{U}_k$), namely each province, which transform $N$ input factors $x=\left\{x_1,\dots ,x_n\right\}\in R_n^{+}$ into $M$ expected output $y=\left\{y_1,\dots y_m\right\}\in R_M^{+}$ and I undesirable output $b=\left\{b_1,\dots ,b_i\right\}\in R_M^{+}$, for each time period $t=1,\dots ,T$. The input and output of the province $k$ in the period $t$ is represented by $\left(x^{k,t},y^{k,t},b^{k,t}\right)$. The PPS is denoted as follows:

$$P^t\left(x\right)=\left\{\begin{array}{l}\left(y^t,b^t\right):{\displaystyle \sum _{k=1}^K{\lambda }_k^t{y}_{km}^t,\forall m;{\displaystyle \sum _{k=1}^K{\lambda }_k^t{b}_{ki}^t=b_{ki}^t,\forall i;}}\\ {\displaystyle \sum _{k=1}^K{\lambda }_k^t{x}_{kn}^t\le x_{kn}^t,\forall n;{\displaystyle \sum _{k=1}^K{\lambda }_k^t=1,{\lambda }_k^t=0,\forall k}}\end{array}\right\}$$

(2)

where ${\lambda }_k^t$ is the weight of each production frontier and ${\displaystyle \sum _{k=1}^K{\lambda }_k^t}=1,{\lambda }_k^t\ge 0,$ represents variable returns to scale (VRS). Meanwhile, drawing on Lv et al.’s [50] practice, the global PPS is denoted as follows:

$$P^G\left(x\right)=\left\{\begin{array}{c}\left(y^t,b^t\right):{\displaystyle \sum _{t=1}^T{\displaystyle \sum _{k=1}^K{\lambda }_k^t{y}_{km}^t}}\ge y_{km}^t,\forall m;{\displaystyle \sum _{t=1}^T{\displaystyle \sum _{k=1}^K{\lambda }_k^t{b}_{ki}^t}}=b_{ki}^t,\forall i;\\ {\displaystyle \sum _{t=1}^T{\displaystyle \sum _{k=1}^K{\lambda }_k^t{x}_{kn}^t}}\le x_{kn}^t,\forall n;{\displaystyle \sum _{t=1}^T{\displaystyle \sum _{k=1}^K{\lambda }_k^t=1,{\lambda }_k^t\ge 0,\forall n}}\end{array}\right\}$$

(3)

Secondly, the slack-based directional distance function is defined following the approach of Ren et al. [55] as follows:

$$\begin{array}{l}\overrightarrow{S_V^t}\left(x^{t,k^{\prime }},y^{t,k^{\prime }},b^{t,k^{\prime }},g^x,g^y,g^b\right)=\underset{s^x,s^y,s^b}{\max }\frac{1}{2}\left[\frac{1}{N}{\displaystyle \sum _{n=1}^N\frac{S_n^x}{g_n^x}}+\frac{1}{M+I}({\displaystyle \sum _{m=1}^M\frac{S_m^y}{g_m^y}+{\displaystyle \sum _{i=1}^I\frac{S_i^b}{g_i^b})}}\right]\\ \mathrm{s}.\mathrm{t}. {\displaystyle \sum _{k=1}^K{\lambda }_k^t{x}_{kn}^t+s_n^x=x_{k^{\prime }n}^t},\forall n;{\displaystyle \sum _{k=1}^K{\lambda }_k^t{y}_{km}^t}-s_n^y=y_{k^{\prime }n}^t,\forall m;{\displaystyle \sum _{k=1}^K{\lambda }_k^t{b}_{ki}^t}+s_i^b=b_{k^{\prime }i}^t,\forall i;\\ {\displaystyle \hspace{1em}\hspace{1em}\sum _{k=1}^K{\lambda }_k^t=1},{\lambda }_k^t\ge 0,\forall k_i;s_m^y\ge 0,\forall m_i;s_i^b\ge 0,\forall i\end{array}$$

(4)

where the directional vector $\left(g^x,g^y,g^b\right)$ refers to the input reduction, expected output increase, and undesirable output decrease, respectively, and $\left(S_n^x,S_m^y,S_i^b\right)$ is the slack vector, representing the amounts of input redundancy, insufficient expected output, and excessive undesirable output. Similarly, the global slack-based directional distance function is expressed as follows:

$$\begin{array}{l}\overrightarrow{S_V^G}\left(x^{t,k^{\prime }},y^{t,k^{\prime }},b^{t,k_{\prime }},g^x,g^y,g^b\right)=\underset{s^x,s^y,s^b}{\max }\frac{1}{2}\left[\frac{1}{N}{\displaystyle \sum _{n=1}^N\frac{S_n^x}{g_n^x}}+\frac{1}{M+I}({\displaystyle \sum _{m=1}^M\frac{S_m^y}{g_m^y}+{\displaystyle \sum _{i=1}^I\frac{S_i^b}{g_i^b})}}\right]\\ \mathrm{s}.\mathrm{t}. {\displaystyle \sum _{t=1}^T{\displaystyle \sum _{k=1}^K{\lambda }_k^t{x}_{kn}^t}}+s_n^x=x_{k_{\prime }n}^t,\forall n;{\displaystyle \sum _{t=1}^T{\displaystyle \sum _{k=1}^K{\lambda }_k^t{y}_{km}^t}}-s_n^y=y_{k^{\prime }n}^t,\forall m;{\displaystyle \sum _{t=1}^T{\displaystyle \sum _{k=1}^K{\lambda }_k^t{b}_{ki}^t}}+s_i^b\\ \hspace{1em}\hspace{1em}=b_{k^{\prime }i}^t,\forall i;{\displaystyle \sum _{k=1}^K{\lambda }_k^t=1},{\lambda }_k^t\ge 0,\forall k_i{s}_m^y\ge 0,\forall m_i{s}_i^b\ge 0,\forall i\end{array}$$

(5)

Third, based on the directional distance function, the global Malmquist–Luenberger productivity index is described in Equation (4), where $GM{L}_t^{t+1}$ denotes the growth rate of GTFP from period $t$ to period $t+1$.

$$GM{L}_t^{t+1}=\frac{1+S_V^G\left(x^t,y^t,b^t;g\right)}{1+S_V^G\left(x^{t+1},y^{t+1},b^{t+1};g\right)}$$

(6)

Finally, the cumulative value is computed for each subsequent year to derive the green total factor productivity index, as expressed by Equation (5), where the base year is set as 2004, with a value of 1.

$$GTF{P}_{t+1}=GM{L}_t^{t+1}\times GTF{P}_t$$

(7)

In the computation of GTFP, the input indicators chosen encompass labor input, capital input, and energy input. Specifically, labor input is quantified by the number of employed individuals residing in urban areas at the conclusion of the year. Capital input is gauged by the aggregate fixed asset investment across the entire society. Energy input is determined by aggregate energy consumption. In terms of output, expected output is ascertained through regional GDP, while undesirable output consists of three distinct components: industrial sulfur dioxide emissions, industrial smoke (dust) emissions, and industrial wastewater emissions.

The explanatory variable investigated in this study is SFDI. In line with the Industrial Classification for National Economic Activities (GB/T 4754-2017), this study delimits the boundaries of the producer service sector to include five specific categories: communications and transportation, storage, and postal industries; information transmission, software, and technology services; the financial industry; tenancy and business service industries; and scientific research, technological service, and the geological prospecting industry. Considering this variable’s enduring impact, we measure it by employing the foreign stock capital in the producer service sector. Specifically, we first gather data pertaining to the actual utilized FDI in the producer service sector for each province. Then, we convert the amount of actual utilized FDI into CNY pricing based on the annual average exchange rate of CNY against the US dollar. Finally, we employ the perpetual inventory method to estimate the capital stock, calculated as $SFD{I}_{it}=\left(1-\alpha \right)SFD{I}_{i(t-1)}+SFD{I}_{it}$, where $\alpha$ represents the depreciation rate set at 6%, following the approach introduced by Hall and Jones [56].

In accordance with prior research on green total factor productivity (GTFP), this study incorporated a set of control variables, including (1) degree of openness (OU), measured by the share of total import and export trade value of the regional GDP; (2) the industrial structure level (IS), measured by the share of value added for the secondary industry of regional GDP; (3) the human resource level (HC), measured by the average years of education of an employed individual; (4) the technological input level (TEC), measured by the share of scientific and technological expenditure of regional GDP; (5) the environmental regulation level (ETAX), measured by the logarithm of sewage charges; and (6) the factor marketization level (FL), measured by the factor marketization index.

3.1.2. Mediating Model

The second step is the mediating model. Scholars commonly employ the mediation effect model to examine the significance of causal mechanisms, as demonstrated by previous studies such as the one conducted by Wang and Ye [57]. Therefore, this study adopts a similar approach to investigate the impact mechanism of foreign direct investment in the producer service sector on green total factor productivity. Considering the possible mechanisms of SFDI on GTFP, this paper employs green innovation capability (GRIN), industry competitiveness (HHI), and resource allocation efficiency (ALO) as mediating variables and sets up the mechanism testing model as follows:

$$HH{I}_t={\beta }_0+{\beta }_{1}SFD{I}_{it}+v{X}_{it}+{\delta }_{it}+{\epsilon }_{it}$$

(8)

$$GTF{P}_{it}={\beta }_0+{\beta }_{1}SFD{I}_{it}+{\beta }_{2}HH{I}_t+v{X}_{it}+{\delta }_{it}+{\epsilon }_{it}$$

(9)

$$GRI{N}_{it}={\beta }_0+{\beta }_{1}SFD{I}_{it}+v{X}_{it}+{\delta }_{it}+{\epsilon }_{it}$$

(10)

$$GTF{P}_{it}={\beta }_0+{\beta }_{1}SFD{I}_{it}+{\beta }_{2}GRI{N}_{it}+v{X}_{it}+{\delta }_{it}+{\epsilon }_{it}$$

(11)

$$AL{O}_{it}={\beta }_0+{\beta }_{1}SFD{I}_{it}+v{X}_{it}+{\delta }_{it}+{\epsilon }_{it}$$

(12)

$$GTF{P}_{it}={\beta }_0+{\beta }_{1}SFD{I}_{it}+{\beta }_{2}AL{O}_{it}+v{X}_{it}+{\delta }_{it}+{\epsilon }_{it}$$

(13)

where $HH{I}_t$ refers to the competitiveness of the producer service sector in year $t$. $GRI{N}_{it}$ and $AL{O}_{it}$ represent the green innovation capability and resource allocation efficiency of province $i$ in year t, respectively. The remaining variables are defined as in the initial step.

Regarding the mediating variables, the first is the competitive effect, which is estimated using the Herfindahl–Hirschman Index. To facilitate regression analysis, the negative value is adopted and calculated as follows:

$$HHI=-{{\displaystyle \sum _1^n\left(\frac{X_i}{X}\right)}}^2$$

(14)

In Equation (12), $X_i$ represents the total assets of the firm $i$ within the specific industry, while $X$ denotes the sum of total assets across the industry. The HHI index serves as a measure of industry concentration, with a higher HHI value indicating a lower level of concentration and a correspondingly greater degree of competition.

Second, the green innovation effect is examined by the natural logarithm of the number of green patent applications. Third, the resource allocation effect is defined in accordance with the method proposed by Li et al. [58] as follows:

$$AL{O}_{i,t}={\displaystyle \sum _{k=1}^3\frac{Y_{i,t,k}}{Y_{i,t}}}\times \frac{Y_{i,t,k}}{L_{i,t,k}}$$

(15)

where $Y_{i,t,k}$ represents the total production value of industry $k$ in province $i$ during period $t$. $Y_{i,t}$ denotes the regional gross domestic product of province $i$ during period $t$. $L_{i,t,k}$ refers to the employment level of industry $k$ in province $i$ during period $t$.

3.1.3. Other Empirical Specifications

The third step is the robustness test, including both the causal effect and mechanism. For the causal effect, several methods are employed, including modifying the core explanatory variable and altering the regression model, as discussed in Section 4.3. Regarding the impact mechanism, we introduce mediating variables as well as the interaction terms between the core explanatory variables and mediating variables to test. To mitigate the issue of multicollinearity, we also centralize the core explanatory variables, mediating variables, and their interaction terms. The model is specified as follows:

$$GTF{P}_{it}={\beta }_0+{\beta }_{1}C\text{\_}HHI\times SFD{I}_{it}+{\beta }_{2}C\text{\_}SFD{I}_{it}+{\beta }_{3}C\text{\_}HH{I}_t+v{X}_{it}+{\delta }_{it}+{\epsilon }_{it}$$

(16)

$$GTF{P}_{it}={\beta }_0+{\beta }_{1}C\text{\_}GIN\times SFD{I}_{it}+{\beta }_{2}C\text{\_}SFD{I}_{it}+{\beta }_{3}C\text{\_}GI{N}_{it}+v{X}_{it}+{\delta }_{it}+{\epsilon }_{it}$$

(17)

$$GTF{P}_{it}={\beta }_0+{\beta }_{1}C\text{\_}ALO\times SFD{I}_{it}+{\beta }_{2}C\text{\_}SFD{I}_{it}+{\beta }_{3}C\text{\_}AL{O}_{it}+v{X}_{it}+{\delta }_{it}+{\epsilon }_{it}$$

(18)

where $C\text{\_}HHI$, $C\text{\_}GIN$, and $C\text{\_}ALO$ are the mediating variables, and $C\text{\_}HHI\times SFDI$, $C\text{\_}GIN\times SFDI$, and $C\text{\_}ALO\times SFDI$ are the interaction terms between the mediating variables and the core explanatory variable.

The final stage entails conducting additional analyses to deepen understanding. First, regional heterogeneity is examined by the analysis of group regression. Second, the moderating effect is tested. The “pollution haven” hypothesis posits that the adverse environmental impact resulting from FDI in the host country is primarily attributed to insufficient environmental regulations in the host country. In addition, varying levels of factor market development in the host country may also shape the effects of FDI [59,60]. Hance, we will examine the potential moderating effects of the two variables. The moderating effect model is donated as follows:

$$GTF{P}_{it}={\beta }_0+{\beta }_{1}C\text{\_}ETAX\times SFD{I}_{it}+{\beta }_{2}C\text{\_}SFD{I}_{it}+v{X}_{it}+{\delta }_{it}+{\epsilon }_{it}$$

(19)

$$GTF{P}_{it}={\beta }_0+{\beta }_{1}C\text{\_}FL\times SFD{I}_{it}+{\beta }_{2}C\text{\_}SFD{I}_{it}+v{X}_{it}+{\delta }_{it}+{\epsilon }_{it}$$

(20)

where $C\text{\_}FL$ and $C\text{\_}ETAX$ are the moderating variables and $C\text{\_}ETAX\times SFDI$, $C\text{\_}FL\times SFDI$ are the interaction term between the moderating variables and core explanatory variable.

3.2. Data Description

The data pertaining to the input and output indicators of the dependent variable, along with the core explanatory variable, were obtained from the Provincial Statistical Yearbook across numerous years. The control variables, including human resource level, industrial structure level, technological input level, and degree of openness, derive from the China Statistical Yearbook and Provincial Statistical Yearbook. The level of environmental regulation is assessed using sewage charges. Specifically, the data from 2005 to 2008 and 2010 to 2017 were sourced from the Wind Database and measured in terms of the discharge fees paid into the warehouse account. For the year 2009, data were obtained from the China Environmental Yearbook and measured by the total revenue of the sewage charges. On 1 January 2018, the Environmental Protection Tax Law of the People’s Republic of China was implemented, leading to the substitution of sewage charges for the environmental protection tax. Consequently, data for the years 2018–2019 were updated to incorporate environmental protection tax data. The data regarding the level of factor marketization derive from the China Provincial Marketization Index Report. The Herfindahl index and its associated indicators were sourced from the CSMAR database, while data on green patents were acquired from the China National Intellectual Property Office. To ensure comparability, all variables related to GDP underwent adjustment using the GDP deflator for each province. The descriptive statistics of all variables are shown in

Table 1. Descriptive statistics of variables.

Variable Type	Variable	Mean	SD	Min	Max
Explained Variable	GTFP	1.567	0.597	0.608	4.979
Explanatory Variable	SFDI	13.514	2.197	3.673	17.384
Control Variable	OU	0.283	0.336	0.013	1.619
	IS	0.447	0.081	0.160	0.620
	HC	9.503	1.212	6.544	13.901
	TEC	0.020	0.014	0.004	0.068
	ETAX	10.857	0.972	7.495	12.791
	FL	5.578	2.424	1.210	15.870
Mediating Variable	HHI	−0.388	0.051	−0.481	−0.307
	GRIN	7.168	1.636	2.303	10.620
	ALO	17.112	8.192	4.736	56.485

Notes: SD refers to the standard deviation of variables, ETAX is logarithmic. The number of the sample is 280. The panel data are balanced. Other abbreviations are included in the Appendix A.

.

Table 1 shows the description of all the variables, including the explained variable, the explanatory variable, control variables, and mediating variables. The standard deviation values show that the employed data are fairly clustered around their means. Furthermore, to test potential issues of multicollinearity, we conducted the variance inflation factor (VIF) test. The results indicate that the VIF values for SFDI, OU, IS, HC, ETEC, ETAX, and FL are 2.56, 2.17, 2.22, 3.14, 2.79, 1.96, and 2.64, respectively. All the values are below the threshold of 10, confirming that there is no multicollinearity problem.

4. Results and Discussion

4.1. Baseline Regression Analysis

Before the baseline regression, the F-test, LM test, and Hausman test were conducted. The obtained p-value from the F-test is recorded as 0.0000, leading to the rejection of the original hypothesis that favors the adoption of a mixed OLS model and the subsequent selection of the fixed effects model. Similarly, the LM test invalidates the initial hypothesis in favor of the mixed OLS model, instead opting for the random effects model. The results of the Hausman test are presented in the final row of

Table 2. Results of the baseline regression.

Variable	(1)	(2)	(3)	(4)	(5)	(6)	(7)
SFDI	0.114 ***	0.217 ***	0.198 ***	0.129 ***	0.125 ***	0.103 ***	0.110 ***
	(0.031)	(0.034)	(0.033)	(0.035)	(0.032)	(0.024)	(0.026)
OU		−0.833 ***	−0.636 **	−0.405	−0.328	0.087	−0.006
		(0.134)	(0.248)	(0.300)	(0.356)	(0.208)	(0.213)
IS			−2.650 ***	−1.743 **	−1.733 **	−2.605 ***	−2.271 ***
			(0.676)	(0.675)	(0.670)	(0.823)	(0.789)
HC				0.196 **	0.187 **	0.170 **	0.139 **
				(0.077)	(0.074)	(0.064)	(0.059)
TEC					6.840	10.508 *	7.666
					(4.952)	(5.547)	(5.937)
ETAX						0.270 **	0.227 **
						(0.106)	(0.105)
FL							0.038 **
							(0.015)
_cons	0.033	−1.128 **	0.250	−1.152 *	−1.180 *	−3.442 ***	−3.058 **
	(0.473)	(0.520)	(0.616)	(0.646)	(0.635)	(1.184)	(1.144)
N	280	280	280	280	280	280	280
R2	0.174	0.501	0.547	0.573	0.579	0.616	0.625
Hausman	28.83 ***	33.89 ***	8.30 **	7.46	52.91 ***	5.18	22.94 ***

Notes: robust standard errors clustered at the province level in parentheses; ***, **, *, indicate significance at the 1%, 5%, and 10% levels.

. Following the inclusion of all control variables, the Hausman test reveals that the fixed effects model exhibits a superior estimation performance. Columns (1)–(7) present the regression results with the stepwise inclusion of control variables, where clustered robust standard errors at the province level are indicated in parentheses.

The baseline regression results reveal that the coefficient of the core explanatory variable SFDI is significantly positive at a 1% confidence level. With regard to the control variables, the results exhibit that the increase in added value for the secondary industry impedes the improvement of GTFP. Conversely, higher levels of human capital, environmental regulation, and factor marketization exert a significant positive influence on GTFP growth. However, the impact of the technological input level and the degree of openness is statistically insignificant.

4.2. Analysis of Mechanisms

To investigate the competition, green innovation, and resource allocation mechanisms, this study adopted a mediation effects model. The results of the baseline regression are reported in column (1), while columns (2)–(3), (4)–(5), and (6)–(7) display the regression results pertaining to the competition mechanism, green innovation mechanism, and resource allocation mechanism, respectively, as presented in

Table 3. Results of the mechanisms tested.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)
Variable	GTFP	HHI	GTFP	GRIN	GTFP	ALO	GTFP
SFDI	0.110 ***	0.009 **	0.085 ***	0.186 ***	0.141 ***	0.833 ***	0.068 ***
	(0.026)	(0.004)	(0.020)	(0.049)	(0.037)	(0.286)	(0.020)
HHI			2.892 ***
			(0.413)
GRIN					–0.167 *
					(0.085)
ALO							0.050 ***
							(0.017)
OU	–0.006	–0.022	0.058	–0.517	–0.092	–8.816 ***	0.435
	(0.213)	(0.024)	(0.183)	(0.363)	(0.203)	(1.902)	(0.289)
IS	–2.271 ***	–0.353 ***	–1.250 *	–0.600	–2.371 ***	–3.640	–2.089 ***
	(0.789)	(0.043)	(0.693)	(0.841)	(0.738)	(9.789)	(0.632)
HC	0.139 **	0.017 **	0.089 *	0.724 ***	0.260 ***	3.905 ***	–0.056
	(0.059)	(0.006)	(0.048)	(0.065)	(0.081)	(0.530)	(0.086)
TEC	7.666	1.436 ***	3.512	17.539 **	10.591	50.751	5.128
	(5.937)	(0.400)	(5.559)	(6.560)	(6.331)	(53.804)	(5.448)
ETAX	0.227 **	0.006	0.211 **	0.046	0.235 **	0.912	0.182 *
	(0.105)	(0.007)	(0.090)	(0.133)	(0.095)	(0.658)	(0.089)
FL	0.038 **	–0.002	0.045 ***	–0.036	0.032 *	1.264 ***	–0.025
	(0.015)	(0.002)	(0.014)	(0.021)	(0.016)	(0.374)	(0.018)
_cons	–3.058 **	–0.583 ***	–1.370	–2.460 *	–3.468 ***	–45.106 ***	–0.802
	(1.144)	(0.074)	(1.057)	(1.316)	(1.153)	(7.794)	(1.313)
N	280	280	280	280	280	280	280
R2	0.625	0.514	0.672	0.732	0.654	0.859	0.686

Notes: robust standard errors clustered at the province level in parentheses; ***, **, *, indicate significance at the 1%, 5%, and 10% levels.

.

Columns (1)–(3) indicate a significant intensification of competition in the host country’s producer service sector due to the entry of SFDI. Upon introducing the competition mechanism, the coefficient of SFDI decreases while remaining statistically significant, suggesting a partial mediating effect on GTFP.

The columns (1), (4), and (5) show that the increase in SFDI stimulates green innovation in China. Upon incorporating the green innovation mechanism, the coefficient of SFDI becomes larger. Due to the opposite signs of the coefficients between green innovation and SFDI, a “masking effect” exists, which still verifies the existence of an indirect channel. In addition, it is noteworthy that the increase in green innovation hinders the improvement of GTFP, which contradicts Hypothesis 2.

The results from columns (1), (6), and (7) provide evidence of the partial mediating effect of the resource allocation mechanism.

4.3. Robustness Test

4.3.1. Robustness Test of Causal Effect

This study employs several approaches to test the robustness of the dual impact of SFDI on GTFP in

Table 4. Robustness test of the causal effect.

	(1)	(2)	(3)	(4)	(5)	(6)
SFDI2	13.602 ***
	(3.458)
SFDI		0.257 ***	0.069 ***	0.096 ***	0.075 *	0.115 ***
		(0.078)	(0.012)	(0.022)	(0.043)	(0.028)
OU	0.545 ***	0.382	0.011	–0.040	–0.588 **	–0.389 ***
	(0.147)	(0.604)	(0.169)	(0.220)	(0.234)	(0.135)
IS	–1.882 *	–1.471 *	0.432	–2.324 **	–2.255 *	–2.056 ***
	(1.073)	(0.775)	(0.541)	(0.888)	(1.122)	(0.708)
HC	0.291 ***	0.400 **	0.157 ***	0.143 **	0.116	0.121 *
	(0.058)	(0.146)	(0.046)	(0.054)	(0.101)	(0.063)
TEC	7.802	0.924	6.009 *	7.727	1.112	3.201
	(5.898)	(6.955)	(3.219)	(5.870)	(6.692)	(4.799)
ETAX	0.248 **	0.242	0.173 *	0.222 **	0.072	0.158 *
	(0.101)	(0.164)	(0.088)	(0.106)	(0.116)	(0.085)
FL	0.031	0.039	–0.017	0.033 **	0.005	0.040 ***
	(0.018)	(0.023)	(0.010)	(0.015)	(0.027)	(0.014)
_cons	–3.691 ***	–8.230 ***	–3.103 ***	–2.786 **	–0.198	–2.117 **
	(1.130)	(2.669)	(1.023)	(1.107)	(0.924)	(0.917)
N	280	180	180	300	280	280
R2	0.607	0.491	0.722	0.642	0.348	0.619

Notes: robust standard errors clustered at the province level in parentheses; ***, **, *, indicate significance at the 1%, 5%, and 10% levels.

. First, the core explanatory variable is replaced with the proportion of actual utilized FDI in the producer service sector of the region (SFDI2) (1). Second, the time window is adjusted by shortening it to five years before and after the reference period, and the regression results for the periods 2011–2019 and 2006–2015 are analyzed (2–3). Third, the time window is extended by including sample data from 2005 (4). Last, the estimation model is modified by employing mixed effects and random effects models (5–6). The results consistently reveal positive coefficients for SFDI that are statistically significant at the 1% level, thereby demonstrating the robustness and reliability of the causal effect.

4.3.2. Robustness Test of Mechanisms

This study employed a mediation effects analysis to validate the influence of SFDI on GTFP through three distinct pathways: the competition effect, the resource allocation effect, and the green innovation effect. In this section, the robustness of the mechanism tests is further examined by including interaction terms, and the results are presented in

Table 5. Robustness test of the mechanisms.

Variables	(1)	(2)	(3)
C_HHI_SFDI	0.848 ***
	(0.191)
C_HHI	2.839 ***
	(0.383)
C_GRIN		–0.177 **
		(0.084)
C_GRIN_SFDI		0.012 *
		(0.007)
C_ALO_FDI			0.007 ***
			(0.002)
C_ALO			0.034 **
			(0.014)
C_SFDI	0.117 ***	0.174 ***	0.008
	(0.016)	(0.034)	(0.020)
OU	–0.170*	–0.075	0.629 *
	(0.100)	(0.218)	(0.351)
IS	–0.761 *	–2.221 ***	–1.737 ***
	(0.445)	(0.666)	(0.479)
HC	0.095 **	0.258 ***	0.024
	(0.039)	(0.076)	(0.071)
TEC	–2.826	9.296	4.787
	(3.290)	(6.291)	(4.623)
ETAX	0.129 *	0.227 **	0.137
	(0.067)	(0.097)	(0.082)
FL	0.037 ***	0.029 *	–0.033 **
	(0.009)	(0.014)	(0.014)
_cons	–0.526	–2.712 *	0.474
	(0.750)	(1.458)	(1.273)
N	280	280	280
R2	0.705	0.659	0.717

Notes: robust standard errors clustered at the province level in parentheses; ***, **, *, indicate significance at the 1%, 5%, and 10% levels.

. The results indicate that the coefficients of the interaction terms are significant and align with the direction established by the mechanism tests, thus reinforcing the robustness of the causal mechanisms under investigation.

4.4. Further Analysis

4.4.1. Regional Heterogeneity Analysis

Given the substantial variations in industrial foundation and resource endowments among the eastern, central, and western regions of China, this section aims to explore the heterogeneous impact of SFDI on GTFP across these regions. The detailed results are shown in

Table 6. Results of the regional heterogeneity test.

	Nationwide	Eastern	Center	Western
SFDI	0.110 ***	0.091 *	0.070 *	0.090 ***
	(0.026)	(0.043)	(0.028)	(0.019)
OU	–0.006	0.263	1.038	0.009
	(0.213)	(0.509)	(1.004)	(0.448)
IS	–2.271 ***	–6.985 **	–1.422 ***	–1.489
	(0.789)	(2.012)	(0.337)	(0.941)
HC	0.139 **	0.013	0.178 ***	0.176 *
	(0.059)	(0.064)	(0.032)	(0.083)
TEC	7.666	6.645	14.125 **	36.759 **
	(5.937)	(12.091)	(5.318)	(11.687)
ETAX	0.227 **	0.329 *	0.083	0.180
	(0.105)	(0.167)	(0.078)	(0.252)
FL	0.038 **	0.018	0.012	0.056 *
	(0.015)	(0.028)	(0.018)	(0.025)
_cons	–3.058 **	–1.027	–1.994 *	–2.772
	(1.144)	(2.473)	(0.801)	(2.306)
N	280	98	84	98
R2	0.625	0.548	0.760	0.774

Notes: robust standard errors clustered at the province level in parentheses; ***, **, *, indicate significance at the 1%, 5%, and 10% levels.

.

The results indicate that there is a significant positive effect on the enhancement of GTFP across the three regions. However, the magnitude of this impact exhibits regional heterogeneity, where the eastern region experiences the highest level of impact, followed by the western region, and the central region shows the weakest impact.

4.4.2. Moderating Effect Test

Considering the potential influence of environmental regulations’ stringency and factor market development on the impact of SFDI, this section investigates the moderating influence of environmental regulations and factor marketization. The results are presented in

Table 7. Results of the moderating effect.

Variable	(1)	(2)
C_ETAX×SFDI	0.027 **
	(0.012)
C_ETAX	0.184 *
	(0.099)
C_FL×SFDI		0.025 ***
		(0.007)
C_FL		0.013
		(0.019)
C_SFDI	0.129 ***	0.158 ***
	(0.024)	0.022
OU	0.039	0.169
	(0.228)	(0.290)
IS	–1.944 **	–2.193 ***
	(0.708)	(0.612)
HC	0.144 **	0.119 *
	(0.053)	(0.059)
TEC	7.345	6.490
	(5.770)	(5.370)
FL	0.035 **
	(0.015)
ETAX		0.178 *
		(0.097)
_cons	0.685	–0.784
	(0.612)	(1.271)
N	280	280
R2	0.632	0.656

Notes: robust standard errors clustered at the province level in parentheses; ***, **, *, indicate significance at the 1%, 5%, and 10% levels.

.

In column (1), the regression results exhibit the moderating effect of the environmental regulation level, with significantly positive coefficients observed for both SFDI and the interaction term. This suggests that stricter environmental regulations enhance the promoting effect of SFDI, thereby validating Hypothesis 3 proposed in the mechanism analysis. Moving to column (2), a moderating effect of factor marketization is reported, where the coefficients of SFDI and the interaction term are also significantly positive. This confirms that the improvement of the factor marketization can facilitate the enhancement of the impact of SFDI.

4.5. Discussion

This research investigated the dual effect of SFDI on GTFP, extending the research scope associated with the effect of SFDI, which mainly focuses on its influence on economic efficiency rather than environmental benefits [6,8,9]. According to Section 4.1, an increase in SFDI significantly promotes China’s GTFP, affirming the presence of a “halo effect” associated with SFDI in China. This result aligns with several studies [16,17] regarding the impact of SFDI on China’s GTFP, but contradicts the results of Chai et al. [19], who found a negative effect of FDI on GTFP. The reason behind this inconsistency may lie in the fact that the producer service sector exhibits stronger industry interconnections compared to the service sector, thereby facilitating the occurrence of green efficiency spillovers both within and across industries.

For the mechanisms tested in Section 4.2, it is found that the entry of SFDI positively impacts GTFP through industry competition, highlighting the positive spillover effects from multinational corporations in the producer service sector. This result aligns with the research of Nuruzzaman et al. [24] and Orlic et al. [14], which can be attributed to two possible reasons. First, the service sector has relatively low imitation barriers, making it more prone to generate spillovers in terms of management techniques and process design. Second, China’s producer service sector has developed at a relatively slow pace and has not fully satisfied the demands of the manufacturing sector, where SFDI entry does not result in severe crowding-out effects. Notably, the increase in green innovation [34,35] hinders the improvement of GTFP, which contradicts Hypothesis 2. The possible explanation for this phenomenon is that, in practice, while firms’ green innovation may generate environmental benefits during production, it does not necessarily lead to a reduction in marginal costs or translate innovation capability into competitive advantages. Instead, it may increase input costs and hamper firms’ research and development activities in other areas, consequently impeding the enhancement of production efficiency. In addition, the resource allocation mechanism is significant, where the increase in SFDI contributes to the optimization of resource allocation across primary, secondary, and tertiary industries. Consequently, this facilitates the transfer of high-quality resources to sectors exhibiting higher efficiency levels, thereby fostering the sustainable development of industries.

Qiu et al. [18] tested the regional heterogeneity of the influence of FDI on the service sector and found that it has a negative impact. However, our research shows different results, where the effects of SFDI are all positive and significant among the three regions: the eastern region exhibits the strongest effect, followed by the western region, and the central region exhibits the weakest effect. Several factors contribute to these divergent outcomes. First, the eastern region possesses higher levels of human capital and technology, narrowing the gap between multinational corporations and facilitating the assimilation of technology spillovers. Second, the eastern region primarily attracts high-tech industries characterized by low pollution levels and high added value, whereas industries with relatively higher pollution levels, such as the transportation industry, tend to relocate more to the central and western regions. Further analysis shows that the enhancement of the environmental regulatory degree and marketization level can promote the effect of SFDI on GTFP, which is consistent with the findings of Tong [59] and Wang [60].

5. Conclusions and Suggestions

5.1. Conclusions

Using panel data at the provincial level from 2006 to 2019, this paper focuses on the dual impact of SFDI on China’s GTFP and uses the global Malmquist–Luenberger productivity index based on the slack-based measure and directional distance function to estimate GTFP. The following research conclusions are drawn: First, from a holistic perspective, the entry of SFDI significantly improves GTFP in China, which has been tested for robustness in various ways. Second, for the impact mechanisms, the competition effect, green innovation effect, and resource allocation effect are all significant. Among them, the competition and resource allocation mechanisms have partial mediating effects, while the green innovation mechanism has a masking effect; the mechanisms are also robust. Third, the impact of SFDI exhibits regional heterogeneity: the eastern region exhibits the strongest effect, followed by the western region, and the central region exhibits the weakest effect. Finally, by incorporating the moderating effects of environmental regulatory degree and marketization level into the analytical framework, their improvement significantly enhances the effect of SFDI on GTFP.

5.2. Suggestions

At the national level, it is imperative to integrate environmental considerations into the assessment of FDI attraction. This necessitates a shift from the conventional “extensive” investment approach towards an “intensive” approach that prioritizes the attraction of high-added-value and environmentally sustainable industries. Moreover, to attract FDI capable of fostering the upgrading of local industry, the government should adopt targeted investment policies employing administrative measures and tax incentives to reduce barriers to entry in specific sectors. In terms of regional variations, the eastern region should capitalize on its comparative advantage in location, human capital, and business environment to continue attracting high-quality FDI. In the central and western regions, it is essential for local governments to carefully assess the environmental implications of FDI and endeavor to foster a conducive business environment. Furthermore, they should expedite the development of skilled professionals in environmental fields to enhance human capital accumulation and facilitate better coordination with multinational corporations.

The enhancement of factor markets plays a pivotal role in amplifying the impact of SFDI. Therefore, it is crucial to establish an efficient capital exit mechanism to eliminate obsolete and surplus capacity through methods such as bankruptcy liquidation and corporate mergers and acquisitions (M&A). Secondly, there is a need to enhance property rights protection and incentive systems to stimulate internal motivation for enterprise learning and innovation. In terms of market competition, it is evident that the influx of foreign investment intensifies market competition. While the current mechanism test demonstrates a positive effect of SFDI on GTFP through intensified competition, it is still crucial for the government to establish regulatory measures for FDI to ensure healthy market competition and prevent unfair competition from the multinational corporations that may crowd out local enterprises and lead to imbalanced industrial development.

For enterprises, the escalation of costs associated with green innovation may impede their propensity to engage in innovative activities. Thus, it is imperative for the government to allocate ample timeframes and furnish targeted tax incentives, subsidies, and other support mechanisms for green transformation. Simultaneously, the government should continue to promote the reform of “streamlining administration and delegating power, improving regulation and optimizing services”, granting certain autonomy to provinces and cities to enforce reforms based on their own circumstances rather than using the “one-size-fits-all” method. Moreover, strengthening environmental regulations holds paramount importance, necessitating the integration of environmental performance into officials’ evaluation systems. This strategic measure is indispensable for eliminating the prevalent “GDP-only” mindset and steering the economy towards a harmonious and synchronized coalescence within the ecological landscape.

Our findings also provide important implications for developing countries striving to achieve equitable economic and environmental benefits within the global economic system. First, in order to attract high-quality foreign direct investment more effectively, developing countries should focus on creating a favorable business environment and promoting the free flow of capital, technology, and talent. In addition, when selecting foreign direct investment, they should pay greater attention to the economic and environmental effects of FDI and attract foreign direct investment, particularly in the producer service sector, representing eco-friendly and highly industrialized sectors. At the same time, there may exist differences in industrial efficiency and production experience between developing countries and developed countries. Therefore, the government of the host country should exercise caution in selecting the industries to open up and the degree of openness, in order to prevent potential monopolistic practices and crowding-out effects that may be caused by multinational corporations from developed countries.

6. Limitation and Future Research Direction

First, this study conducted an empirical examination of the relationship between foreign direct investment in the producer service sector and green total factor productivity based on China’s experience. However, due to the influence of political, economic, and cultural factors, the relationship between the two variables may differ in other developing countries and developed countries. Nonetheless, given the significant variations in industrial-related statistical methodologies and the limited availability of data, this study did not differentiate the analysis by country, and relative research is essential.

Second, there is still room for improvement in measuring green total factor productivity, particularly in the selection of non-expected output indicators. Carbon emission measurements or a comprehensive life-cycle assessment of non-expected outputs using carbon footprints could be considered.

Third, it is worth noting that this study examined competition mechanisms, green innovation mechanisms, and resource allocation mechanisms. However, due to the high industry synergies associated with the producer service sector, the entry of multinational corporations into host countries may result in agglomeration effects with local firms, thereby generating externalities. This aspect also merits further investigation in future studies.