How Does Outward Foreign Direct Investment Affect Green Total Factor Productivity? Evidence from Increment and Quality Improvement

Abstract

Based on the provincial data of China from 2011 to 2020, this paper analyzes the effect and influence mechanism of Outward Foreign Direct Investment (OFDI) on Green Total Factor Productivity (GTFP) theoretically and empirically. The results show that OFDI can significantly improve GTFP, and this conclusion is still valid after measurement replacement, instrumental variable method and systematic GMM treatment. The mechanism analysis demonstrates that digital inclusive finance can strengthen the promoting effect of OFDI on GTFP, and OFDI can promote invention innovation and utility model innovation, thus enhancing GTFP. The heterogeneity analysis shows that OFDI can play a stronger role in promoting GTFP in the eastern region, western region, and the regions with a low marketization level. Therefore, actively promoting OFDI, optimizing the accuracy of OFDI, speeding up the development of digital inclusive finance, and formulating policies according to local conditions, play a vital role in promoting GTFP and the transformation of economic structure.

1. Introduction

China’s economy has developed rapidly for more than 40 years since reform and opening up. Its extensive development mode has brought about the improvement of total economic output, but it has also triggered the problems of energy depletion and environmental pollution. The report of the 19th National Congress of the Communist Party of China puts forward suggestions persisting in saving resources, protecting the environment, and taking the path of green development. In the new stage of development, it is necessary to excavate and cultivate the driving force of green growth so as to maintain green development. Green Total Factor Productivity (GTFP) meanwhile, is a crucial aspect of green development capability. Therefore, the issue of how to steadily improve the growth of China’s GTFP is a significant topic that the government and scholars are concerned about. In recent years, China has continued to accelerate the pace of opening up. In 2016, global OFDI exceeded 10%. From industry’s perspective, the industry’s direct investment in 2016 was USD 36.57 billion, accounting for 27.15%.

Since the Chinese government implemented the “go global” strategy, the level of OFDI has increased substantially. According to the Statistical Bulletin of China’s OFDI in 2020, the flow of China’s OFDI reached USD 153.71 billion in 2020, ranking the highest in the world. A large number of reviews have studied the reverse technology backflow effect of OFDI [1,2,3,4,5], such as the research and innovation of manufacturing enterprises in Taiwan Province [6] and the research of GTFP in countries along “the Belt and Road Initiative” [7]. For example, it has been found from research in China that the impact of OFDI on GTFP has significant regional heterogeneity; there are also nonlinear studies on the absorptive capacity of human capital [5], factor endowment, and the marketization process in the home country [8].

According to previous studies, there have been many studies on the relationship between OFDI and GTFP, but there are few studies on the “increment” of OFDI, especially the role of digital inclusive finance. Therefore, the marginal contribution of this paper may be as follows. First, it is theoretically and empirically verified that digital inclusive finance can enhance the reverse green technology backflow effect of OFDI. Second, it is found that only high-level invention innovation, and utility model innovation, play an important mechanism role among the innovation channels that OFDI affects GTFP. Third, the compensation effect of OFDI on weak regions is found from the perspective of the marketization level. Based on provincial data of China from 2011 to 2020, this paper analyzes the effect and influence mechanism of OFDI on GTFP, theoretically and empirically.

The structure of the paper is as follows: the second chapter is a literature review and research hypothesis; the third chapter covers research design; the fourth is an empirical analysis; and the fifth chapter covers the research conclusion.

2. Literature Review and Research Hypothesis

GTFP is a productivity index that considers energy input and environmental pollution on the basis of TFP. The existing studies on OFDI and GTFP are mainly carried out from a linear and nonlinear perspective. There are many case studies which have found that OFDI can promote GTFP. For example, the study of using Japan’s direct investment in R&D resource-intensive industries in the United States of America [2] was the first to propose that OFDI may lead to the reverse technology backflow effect. Some studies carried out at the enterprise level [1] have found that OFDI can produce significant knowledge and technology spillovers for the investing company, by constructing the framework of international knowledge spillovers measured at the enterprise level. Jie et al. [4] believe that the reverse technology backflow effect will be greater, and will make it easier to promote green development, when the R&D and the development of the home country is concentrated in other countries. Some studies are carried out at the regional level. For example, Hu et al. [9] argue that OFDI can significantly promote the growth of GTFP, but its impact has short-term and long-term differences. In the short term, it has different promoting effects in different regions, and has the strongest performance in the eastern region. In the long term, it shows a stable promoting effect, and its short-term effect has been compensated. Zhang et al. [10] find that there are obvious regional differences in GTFP and its decomposition terms; OFDI has a significant spatial spillover effect on GTFP, forming a good regional demonstration role. Xie et al. [7] studied 56 countries along the Belt and Road Initiative and point out that China’s OFDI significantly promotes the growth of GTFP in countries within the initiative; however, its promoting effect on the home country was not stable. Other studies also think that OFDI can reduce GTFP. For example, Wang et al. [11] use the GML index to calculate GTFP, analyze the effects of OFDI, and the structural transformation of GTFP, proposing that OFDI can significantly reduce GTFP in China. OFDI has obvious regional differences on the impact of GTFP. OFDI plays a significant role in promoting GTFP in the eastern region, but it has no significant impact on GTFP in the central and western regions.

Some studies find that the influence of OFDI on GTFP is affected by the capability of the home country. For example, Zheng [12] studied the threshold impact of OFDI on the TFP of the home country, from the perspective of financial development, and proposes that there are two thresholds in the role of financial development. Some scholars have also applied the panel smooth transformation model to study the role of absorptive capacity, and found that the influence of OFDI on GTFP is nonlinear, due to factors such as technology R&D intensity and the technology gap [13]. Yang et al. [8] put forward that the promoting the effect of OFDI on GTFP is related to factors such as factor endowment and the marketization process. Kong Xi et al. [5] concluded that the human capital of the home country plays a role in the influence of OFDI on GTFP, and finds that only when human capital reaches a specific appropriate value can OFDI promote GTFP. However, enterprises with large financing constraints often encounter various development difficulties when carrying out OFDI, and it is more difficult for them to enter the international market.

Guangdong, Shandong, Jiangsu, Zhejiang, Shanghai, and Liaoning Province, are the main provinces and cities in which China conducts foreign direct investment. Obviously, the main force of China’s foreign direct investment is mainly concentrated in the southeast coastal areas. Overall, the eastern region has become the main area of China’s foreign direct investment. Secondly, compared with the western region, the central region is stronger than the western land. Finally, for the western region, only Sichuan, Chongqing, and Yunnan provinces have a large direct investment. In the eastern region, the top three provinces of the Green Full Productivity Index are Beijing, Shandong and Shanghai, respectively. In the central region, the average productivity of the Green Full Element is Jiangxi, Henan, and Hubei. In the western region, the top three provinces of the Green Full Factors are Shaanxi, Inner Mongolia, and Guizhou.

Financial resources can alleviate the financing constraints of enterprises, which has been demonstrated in many studies [14,15,16]. Financial development can significantly alleviate the cash flow of listed companies, while the investment behavior of enterprises is highly correlated with cash flow [17]. As a financial innovation combining big data, artificial intelligence, blockchain, and other technologies with traditional finance, digital inclusive finance can improve the distribution efficiency of financial elements, so it can also alleviate the financing constraints in development [18,19]. Therefore, this paper holds that the development of digital inclusive finance can “increment” OFDI, and then stimulate the promoting effect of OFDI on GTFP. Thus, the following hypothesis on “increment” is put forward:

H1. 

Digital inclusive finance improves the promoting effect of OFDI on GTFP.

OFDI has a reverse technology backflow effect [3]. Companies that carry out OFDI can learn from the technology and management experience of foreign advanced companies, or directly merge enterprises to acquire advanced technology, thus improving the innovation capability of their home countries. Yang et al. [6] employ the PSM method and found that OFDI could significantly improve the technical efficiency of manufacturing enterprises in Taiwan Province, based on the data of manufacturing enterprises in Taiwan Province, China, from 1987 to 2000. Mao and Xu [20] concluded that there is a significant causal relationship between OFDI and enterprise innovation, and the promoting effect is persistent, which also varies according to the classification of OFDI.

Li and Xiao [21] put forward that the main driving force for private enterprises to explore international opportunities for ODFI under the background of institutional constraints, is to stimulate the innovation ability of enterprises. Innovation is an important pivot in promoting GTFP. Ge et al. [22] state that both basic innovation and application innovation can significantly promote GTFP by using the panel data of countries along “the Belt and Road Initiative”. To sum up, the following hypothesis on “quality improvement” is proposed:

H2. 

OFDI promotes GTFP by stimulating innovation.

3. Research Design

3.1. Sample Data

The data used in this paper are provincial-level data of 30 provinces, municipalities, and autonomous regions in China, from 2011 to 2020. The data is from the CSMAR database and EPS database of the China Statistical Yearbook, the China Environmental Statistical Yearbook, and the China Energy Statistical Yearbook. Some missing values are filled by interpolation and by looking up local yearbooks. Tibet is a province in China; because the Tibet economy is underdeveloped, there is serious missing data in Tibet, therefore, the sample in Tibet is deleted. This article uses the STATA15.1 software to process data.

3.2. Definition of Variables

3.2.1. The Explained Variable

For GTFP (gtfp), this paper refers to common practices in the existing literature; takes energy utilization and environmental pollution into the calculation formula of TFP; adopts the SBM-GML model considering unexpected output; and further calculates GTFP at the provincial level in China. Its input-output indicators are as follows. Input indicators: (1) capital input is expressed as physical capital stock (CNY 100 million). Referring to the perpetual inventory method of Zhang [23], this paper sets the depreciation rate at 9.6%, takes 2003 as the base period, and uses the GDP deflator for processing; (2) labor input is the year-end number of employed people in each province (10,000 people); and (3) energy consumption is energy consumption converted from standard coal, including oil, coal and natural gas, and other energy consumption indicators. Output indicators: (1) expected output is gross domestic product (CNY 100 million), with 2003 as the base period and the GDP deflator for processing; and (2) unexpected output uses the industrial sulfur dioxide (10,000 tons) emitted in various provinces and cities. The calculated GTFP change rate is multiplied to obtain the GTFP level of China from 2011 to 2020. To reduce the heteroscedasticity, it is logarithmized when the model is regressed.

3.2.2. Explanatory Variable

For OFDI (ofdi), this paper refers to the practices of Bai and Liu [24]. It is measured by the proportion of OFDI in GDP. Since OFDI is settled in US dollars, the amount of OFDI is converted into RMB according to the annual exchange rate [24].

3.2.3. Mechanism Variables

For digital inclusive finance (index), this paper refers to the research of Guo et al. [25]. It uses the China Digital Inclusive Finance Index, jointly developed by the Institute of Digital Finance Peking University and Ant Group, as a proxy variable. For regional innovation (patent), this paper uses 10,000 invention patents (ipatent), 10,000 utility model patents (npatent), and 10,000 design patents (dpatent), to represent different aspects of regional innovation since patent data can better reflect the quality of innovation than the data of innovation investment, etc. In addition, the data of regional innovation came from the EPS database.

3.2.4. Control Variables

Referring to the research of Bai and Liu [24], this paper includes the following control variables: the level of economic development (pgdp), which is measured by the logarithmic value of per capita GDP; education level (edu), which takes the proportion of education expenditure in local fiscal expenditure as a proxy variable; science and technology investment (science), which takes the proportion of science and technology expenditure in local fiscal expenditure as a proxy variable; infrastructure level (jj), which is measured by logarithmic value of per capita road area; foreign direct investment (fdi), which is measured by the logarithmic value of foreign direct investment adjusted by the exchange rate; and the marketization level (market), which takes the marketability index measured by China’s sub-provincial Marketability Index Report as the original data, and divides it into two groups according to its median. If it is greater than the median, the market is assigned to 1, otherwise, it is 0. Descriptive statistics of main variables are shown in Table 1, and their values are similar to those of previous studies.

3.3. Empirical Model

Referring to the research of Bai and Liu [24], this paper establishes the following econometric regression Model (1):

(1)$lngtf{p}_{i,t}={\alpha }_0+{\alpha }_{1}ofd{i}_{i,t}+{\alpha }_{2}control{s}_{i,t}+{\vartheta }_i+{\phi }_t+{\epsilon }_{i,t}$

In this Model, subscripts $i$ and $t$ represent the area identification code and year identification code, respectively; $lngtfp$ is the logarithmic value of GTFP, and it is the explained variable of this paper; $ofdi$ is the OFDI level in this region, and it is the core explanatory variable in this paper; $controls$ represents a series of control variables at the regional level; ${\vartheta }_i$ is the fixed effect of region; ${\phi }_t$ is the fixed effect of time; ${\epsilon }_{i,t}$ is the random disturbance term according to time and individual changes. ${\alpha }_1$ is the core coefficient this paper focuses on, and it is expected that its coefficient is positive.

To verify Hypothesis 1, this paper sets the following Model (2) to examine the “increment” effect of OFDI on GTFP under the adjustment of digital inclusive finance:

(2)$$\begin{gathered} lngtf{p}_{it}={\beta }_0+{\beta }_{1}ofd{i}_{it}*I\left(X_{it}\le D_1\right)+{\beta }_{2}ofd{i}_{it}*I(X_{it}>D_1)+{\beta }_{2n-1}ofd{i}_{it}*I\left(X_{it} \\ \le D_{n-1}\right)+{\beta }_{2n}ofd{i}_{it}*I(X_{it}>D_{n-1})+{\beta }_{2n+1}control{s}_{it}+{\vartheta }_i+{\phi }_t+{\epsilon }_{i,t} \end{gathered}$$

In this Model, $X$ is the threshold variable, OFDI ($ofdi$), and digital inclusive finance ($index$). $D$ is the value corresponding to the threshold variable. The definitions of other variables are the same as above. It is expected that ${\beta }_1$-${\beta }_{2n}$ coefficient is positive.

To verify the previous Hypothesis 2, this paper sets the following mechanism model to examine the “quality improvement” effect of OFDI on GTFP.

(3)$paten{t}_{i,t}={\gamma }_0+{\gamma }_{1}ofd{i}_{i,t}+{\gamma }_{2}control{s}_{i,t}+{\vartheta }_i+{\phi }_t+{\epsilon }_{i,t}$

(4)$lngtf{p}_{i,t}={\delta }_0+{\delta }_{1}ofd{i}_{i,t}+{\delta }_{2}paten{t}_{i,t}+{\delta }_{3}control{s}_{i,t}+{\vartheta }_i+{\phi }_t+{\epsilon }_{i,t}$

In this Model, $patent$ is the regional innovation level, which includes 10,000 invention patents ($ipatent$), 10,000 utility model patents ($npatent$), and 10,000 design patents ($dpatent$). The definitions of other variables are the same as above. It is expected that ${\gamma }_1$ coefficient is significantly positive, and ${\delta }_2$ is significantly positive.

4. Empirical Analysis

4.1. Baseline Regression Analysis

Before the empirical analysis, it is necessary to ensure that there is no multicollinearity among the variables. Table 2 reports the results of the VIF test. The VIF test is the most commonly used and can represent multiple common linear inspections that can represent variables. It shows that the average value of VIF is 2.7, and the VIF value of each variable is less than 10, indicating that there is no serious multicollinearity problem [10].

Table 3 presents the results of baseline regression. Columns (1)–(5) are the results of gradually increasing control variables and fixed effects. It can be observed that the coefficient of OFDI (ofdi) is always significantly positive, indicating that this result is more robust. Among them, Column (5) is the result of using the complete Model (1). It can be seen that the coefficient of OFDI (ofdi) is 2.572, and t value is 3.77, which is significant at the level of 1%. This suggests that OFDI can significantly improve GTFP. Its economic implication is that GTFP increases by 2.572% for every 1% increase in OFDI. China’s OFDI can acquire foreign advanced technology, knowledge, and management experience, etc., exert reverse backflow effect, and then improve its GTFP. For the control variables, the coefficient of infrastructure level (jj) is −0.305, which is significant at the level of 1%. This indicates that the increase in per capita road area reduces GTFP, which might be due to the fact that the construction of roads can bring more environmental pollution problems. The coefficient of foreign direct investment (fdi) is −0.063, which is significant at the level of 1%. This implies that foreign direct investment is not conducive to the improvement of GTFP.

4.2. Robustness Test

1. Change the measurement method: Referring to the practice of Yang et al. [8], the logarithmic value of the total OFDI multiplied by the exchange rate is used as the proxy variable of OFDI level in each region. The results are reported in Column (1) of Table 4. It shows that the coefficient of OFDI (lnofdi) is 0.016, and t value is 2.07, which is significant at the level of 5%. 2. Control macro factors: The interactive item of provincial dummy variable and time dummy variable is included in the control variables to control the different influences of macro factors on different individuals. The results are displayed in Column (2) of Table 4. It can be seen that the coefficient of OFDI (ofdi) to GTFP (gtfp) is 2.567 after controlling the macro factors, and its t value is 4.10, which is significant at the level of 1%. The conclusion of this paper is relatively stable.

Because the higher GTFP may also lead to OFDI, and there may be some factors such as measurement error, it is faced with an endogenous problem. This paper uses the instrumental variable method and systematic GMM to solve this problem. 3. Instrumental variable method: As the previous-year OFDI is directly related to the development of the current-year OFDI, and the data of the previous-year OFDI has nothing to do with the current-year random disturbance, the endogenous problem can be better solved by using the previous-year OFDI. Columns (3) and (4) report the results of the instrumental variable method. Column (4) shows that the coefficient of OFDI (ofdi) is 4.168, which is significant at the level of 5%. In addition, the LM value and Wald-F value of Kleibergen-Paap tested by the instrumental variable method are 12.110 and 25.370, respectively, both of which pass the significance test. This verifies that the conclusion is still valid after being processed by the instrumental variable method. 4. Systematic GMM: There are some sequence-related problems in the development of GTFP; therefore, the systematic GMM is employed to solve this problem. The coefficient of OFDI in Column (4) is 1.408, which is significant at the level of 5%. The model also passes the AR(2) and the Hansen test. The above results demonstrate that the conclusion of this paper is robust.

4.3. Mechanism Analysis

4.3.1. The “Increment” Adjustment Function of Digital Inclusive Finance

It has been proved that OFDI can promote GTFP, so what is its influence mechanism? Theoretical analysis shows that digital inclusive finance can regulate the promotion effect of OFDI on GTFP; so can digital inclusive finance promote this effect empirically? When can it have the promotion effect? In this paper, the threshold model is applied for analysis. The xthreg developed by Wang [26] is used for panel threshold model demonstration. Meanwhile, OFDI and digital inclusive finance are used as threshold variables to test their threshold values and significance. It can be seen from Table 5 that OFDI passes the single threshold test when it is a threshold variable, and digital inclusive finance also passes the single threshold test.

Table 6 shows the mechanism test results of the threshold model. Column (1) is the estimation result when OFDI is the threshold variable. It can be seen that OFDI has a significant self-threshold effect on GTFP. When OFDI (ofdi) is less than 0.0147, the coefficient of OFDI (ofdi) to GTFP (lngtfp) is −1.990, which is not significant, indicating that OFDI cannot improve GTFP at this time. However, when OFDI (ofdi) is greater than 0.0147, the estimation coefficient of OFDI (ofdi) is 1.933, which is significant at the level of 5%, indicating that only when OFDI reaches a certain level can it promote GTFP.

Before testing Hypothesis 1, i.e., the “increment” effect of digital inclusive finance, this paper firsts test whether digital inclusive finance can improve OFDI level. Column (2) of Table 6 reports the corresponding estimation results. It can be seen that the coefficient of digital inclusive finance (index) is 0.0002, and its t value is 2.70, which is significant at the level of 1%. This suggests that the development of digital inclusive finance can significantly improve China’s OFDI level. In addition, the coefficient of foreign direct investment (fdi) to OFDI (ofdi) is 0.004, which is significant at the level of 5%. This indicates that FDI can significantly improve OFDI level, which is consistent with the research of Li et al. [27]. Column (3) reports the results when digital inclusive finance is the threshold variable. The coefficient of OFDI (ofdi) to GTFP (lngtfp) is 1.922 when the digital inclusive finance (index) is less than 282.77, and it is significant at the level of 1%. It shows that OFDI can still significantly improve GTFP when the development level of digital inclusive finance is not high. However, when digital inclusive finance passes the threshold value of 310.02, the coefficient of OFDI (ofdi) becomes 5.894, and the t value is larger. This indicates that the higher the development level of digital inclusive finance, the greater the promotion effect of OFDI on GTFP. The above results show that Hypothesis 1 is verified.

4.3.2. The Role of Regional Innovation in “Quality Improvement”

To further analyze the mechanism path of OFDI’s influence on GTFP, and verify Hypothesis 2, this paper uses Model (3) and Model (4) for analysis, and the results are shown in Table 7. Columns (1), (3), and (5), are the results of using Model (3). Columns (2), (4), and (6), are the results of using Model (4). Column (7) is the result of including all types of innovations. In Column (1), the coefficient of OFDI (ofdi) to invention patent (ipatent) is 36.501, which is significant at the level of 5%. This indicates that OFDI can significantly improve invention innovation. In Column (2), the coefficient of invention patent (ipatent) is 0.018, which is significant at the level of 1%, indicating that OFDI can further promote GTFP by stimulating invention innovation. In Column (2), the coefficient of OFDI (ofdi) to utility model patent (npatent) is 71.555, which is significant at the level of 5%, indicating that OFDI can significantly improve utility model innovation. In Column (4), the coefficient of utility model patent (npatent) is 0.01, which is significant at the level of 1%, indicating that OFDI can promote GTFP by stimulating utility model innovation. In Column (5), the coefficient of OFDI (ofdi) to design patent (dpatent) is 10.866, but it is not significant, which shows that OFDI has no significant effect on design innovation. In Column (6), the coefficient of design patent (dpatent) is −0.005, which is also not significant. The design patent is not the mechanism in the promoting effect of OFDI on GTFP. To further strengthen the credibility of the conclusion, this paper puts all innovation variables into the same model for testing. The results are shown in Column (7), and there is no obvious difference in the results. Thus, Hypothesis 2 is verified.

4.4. Heterogeneity Analysis

The effect of OFDI on GTFP is affected by the local economic development level, resource endowment, and other conditions. This effect may not be applicable to all regions. Therefore, it is of great practical significance to further analyze the heterogeneity of this effect.

Regional heterogeneity: China has different economic policies, a different emphasis on OFDI and environmental protection, and different levels of economic growth brought by its own capabilities. Therefore, this paper divides 30 regions into the eastern region, the western region, and the central region, and makes sub-sample regression. Table 8 shows the results of regional heterogeneity. Column (1) is the regression result of the eastern region. It shows that the coefficient of OFDI (ofdi) to GTFP (lngtfp) is 1.665, and its t value is 1.76, which is significant at the level of 10%. This indicates that OFDI in the eastern region has only marginal significant promotion effect on GTFP. In Column (2), the coefficient of OFDI (ofdi) is −3.661, but it is not significant, indicating that OFDI in the central region cannot significantly improve GTFP. In Column (3), the coefficient of OFDI (ofdi) is 4.066, and its t value is 2.89, which is significant at the level of 1%. This indicates that the development of OFDI in the western region can significantly promote GTFP. The possible reason for this is that the eastern region itself has a high level of GTFP and a high level of technological innovation. The reverse technology backflow effect brought by OFDI is relatively small, while the western region is relatively backward in terms of technological level and experience management. Therefore, OFDI activities can significantly improve its backward GTFP.

Heterogeneity of marketization level: The marketization level can reflect the resource allocation efficiency of regions to some extent, while the production efficiency of the regions with low resource allocation efficiency is often low. The reverse technology backflow effect brought by OFDI may bring more significant green growth. To test this analysis, this paper divides it into two groups, according to the marketization level. Column (1) in Table 9 shows the regions with a high marketization level. The coefficient of OFDI (ofdi) to GTFP (lngtfp) is 0.151, and its t value is 0.21, which is not significant. This suggests that OFDI cannot play a significant role in promoting GTFP when the marketization level is high. Column (2) shows the regions with a low marketization level. The coefficient of OFDI (ofdi) to GTFP (lngtfp) is 2.220, and its t value is 2.91, which is significant at the level of 1%. This indicates that OFDI can effectively improve GTFP in regions with a low marketization level.

5. Research Conclusions

Based on the data of 30 provinces in China, this paper analyzes the role, influence mechanism, and heterogeneity, of OFDI on GTFP, theoretically and empirically. This research finds that OFDI can significantly improve China’s GTFP. This conclusion is still valid after changing the measurement method, controlling macro factors, and using the instrumental variable method and systematic GMM. The influence of OFDI on GTFP is strongest in the western region, slightly weaker in the eastern region, and weakest in the central region. OFDI can play a better role in promoting GTFP in regions with a low marketization level. The results demonstrate that digital inclusive finance has an “increment” effect, which can improve the role of OFDI in promoting GTFP. Meanwhile, OFDI has a “quality improvement” effect, which can promote GTFP by stimulating invention innovation and utility model innovation.

The reverse green technology backflow effect of OFDI is vital to China’s economic growth. The policy implications in this paper are as follows: First, it is necessary to increase OFDI of enterprises and improve its accuracy, optimize the goal and structure of OFDI, and encourage enterprises to invest in foreign companies with innovative green technologies, especially those with comparative advantages in invention patents and utility model patents. Through mergers and acquisitions, enterprises can set up scientific research institutions, learn from experience, and carry out direct investment, etc., to ensure that green technologies can effectively flow back and promote the growth of domestic GTFP. Second, local governments should build appropriate development strategies according to local conditions. The eastern and western regions should strengthen OFDI, actively exert the reverse green technology backflow effect, and accelerate green development. The central region should build a reasonable and effective technological innovation mechanism and diffusion capacity, based on the business characteristics and resource endowment conditions of enterprises, thus facilitating the role of OFDI.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

Not applicable.

Conflicts of Interest

The authors declare no conflict of interest.

Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Tables

Table 1

Descriptive statistics.

Variable	Sample Size	Average Value	Standard Deviation	MinimumValue	Maximum Value
lngtfp	300	0.09	0.13	−0.16	0.78
ofdi	300	0.01	0.01	0	0.06
index	300	217.25	96.97	18.33	431.93
ipatent	300	1.97	3.48	0.12	28.90
npatent	300	7.06	7.99	0.26	47.88
dpatent	300	2.61	3.85	0.12	22.19
pgdp	300	10.76	0.44	9.63	12.07
edu	300	0.16	0.03	0.10	0.22
science	300	0.02	0.02	0.004	0.07
jj	300	3.47	0.62	1.42	4.96
fdi	300	11.26	1.38	7.95	14.5
market	300	0.5	0.50	0	1

Table 2

Multicollinearity test.

Variable	fdi	pgdp	Science	jj	Market	Edu	ofdi	Mean VIF
VIF	4.61	3.02	2.97	2.85	2.47	1.61	1.38	2.7

Table 3

Baseline regression results.

	OLS	OLS	RE	FE	FE
	(1)	(2)	(3)	(4)	(5)
	lngtfp	lngtfp	lngtfp	lngtfp	lngtfp
ofdi	4.540 ***	0.999 **	1.694 ***	2.354 ***	2.572 ***
	(4.84)	(2.20)	(2.83)	(2.99)	(3.77)
pgdp		0.065 ***	0.127***	0.287 ***	0.161 **
		(3.08)	(2.98)	(3.64)	(2.17)
edu		−1.158 ***	−1.013 ***	−1.073 ***	−0.259
		(−4.31)	(−2.96)	(−3.19)	(−0.51)
science		−0.936	−0.620	0.417	−0.348
		(−1.37)	(−0.84)	(0.32)	(−0.43)
jj		−0.021	−0.012	−0.306 **	−0.305 ***
		(−1.42)	(−0.52)	(-2.32)	(−3.29)
fdi		0.030 ***	0.025 *	0.001	−0.063 ***
		(3.58)	(1.78)	(0.06)	(−2.72)
market		0.059 ***	0.032	0.012	0.026 *
		(3.31)	(1.44)	(0.50)	(1.74)
_cons	0.060 ***	−0.706 ***	−1.372 ***	−1.803 ***	0.150
	(7.09)	(−2.68)	(−3.08)	(−3.35)	(0.23)
Year FE	NO	NO	NO	NO	YES
Prov FE	NO	NO	NO	YES	YES
N	300	300	300	300	300
R 2	0.1077	0.513		0.395	0.769

Note: *, **, and *** indicate that the estimation coefficient is significant at the level of 10%, 5%, and 1%, respectively; t value is in brackets; same as below.

Table 4

Robustness test results.

	Change the Measurement	Control Macro Factors	Instrumental Variable Method		GMM
	(1)	(2)	(3)	(4)	(5)
	lngtfp	lngtfp	ofdi	lngtfp	lngtfp
lnofdi	0.016 **	2.567
	(2.07)	(4.10)
iv			0.215 ***
			(3.25)
ofdi				4.168 **	1.408 **
				(2.30)	(2.51)
l.lngtfp					0.561 ***
					(2.80)
_cons	0.406				0.143
	(0.63)				(0.07)
Year FE	YES	YES	YES	YES	YES
Prov FE	YES	YES	YES	YES	YES
AR(1)					0.006
AR(2)					0.221
Hansen					0.798
N	300	300	270		270
R 2	0.762	0.810	0.080

Table 5

Threshold value test.

Threshold Variable	Threshold Sequence	Threshold Value	p Value	95% Confidence Interval	BS Times	Seed Value
ofdi	Single Threshold	0.0147	0.0467	[0.0113, 0.0152]	500	100
	Double Threshold	0.0303	0.1600	[0.0198, 0.0349]	500	100
index	Single Threshold	282.77	0.0400	[274.98, 285.28]	500	100
	Double Threshold	310.02	0.1867	[292.85, 316.88]	500	100

Table 6

Hypothesis 1: result analysis of “Increment”.

	(1)	(2)	(3)
	lngtfp	ofdi	lngtfp
pgdp	0.290 ***	−0.016 ***	0.292 ***
	(3.66)	(−3.13)	(3.85)
edu	−1.035 ***	−0.111 **	−0.828 **
	(−2.89)	(−2.38)	(−2.43)
science	0.806	−0.211 **	−0.351
	(0.63)	(−2.04)	(−0.28)
jj	−0.294 **	0.017 *	−0.283 **
	(−2.29)	(1.82)	(−2.12)
fdi	−0.001	0.004 **	−0.011
	(−0.04)	(2.09)	(−0.56)
market	0.012		0.023
	(0.53)		(1.04)
index		0.0002 ***
		(2.70)
ofdi*I_1	−1.990		1.922 ***
	(−1.01)		(2.98)
ofdi*I_2	1.933 **		5.894 ***
	(2.61)		(3.90)
_cons	−1.848 ***	0.055	−1.828 ***
	(−3.39)	(1.06)	(−3.61)
Year FE	YES	YES	YES
Prov FE	YES	YES	YES
N	300	300	300
R 2	0.407	0.650	0.433

Table 7

Hypothesis 2: result analysis of “Quality Improvement”.

	Invention Patent		Utility Model Patent		Design Patent		Comprehensive
	(1)	(2)	(3)	(4)	(5)	(6)	(7)
	ipatent	lngtfp	npatent	lngtfp	dpatent	lngtfp	lngtfp
ofdi	36.501 **	1.909 ***	71.555 **	1.799 ***	10.886	2.621 ***	1.512 ***
	(2.58)	(3.05)	(2.55)	(3.06)	(0.74)	(3.74)	(2.74)
ipatent		0.018***					0.014 ***
		(4.40)					(3.53)
npatent				0.011 ***			0.009 **
				(3.19)			(2.54)
dpatent						−0.005	−0.009 *
						(−0.73)	(−1.84)
_cons	18.121	−0.179	112.002 ***	−1.060	0.075	0.150	−1.100
	(1.64)	(−0.25)	(3.07)	(−1.45)	(0.01)	(0.23)	(−1.47)
Year FE	YES	YES	YES	YES	YES	YES	YES
Prov FE	YES	YES	YES	YES	YES	YES	YES
N	300	300	300	300	300	300	300
R 2	0.880	0.796	0.897	0.812	0.855	0.771	0.830

Table 8

Regional heterogeneity results.

	Eastern Region	Central Region	Western Region
	(1)	(2)	(3)
	lngtfp	lngtfp	lngtfp
ofdi	1.665 *	−3.661	4.066 ***
	(1.76)	(−1.02)	(2.89)
_cons	0.115	−1.693	−1.896
	(0.11)	(−1.12)	(−0.75)
Year FE	YES	YES	YES
Prov FE	YES	YES	YES
N	120	100	80
R2	0.777	0.730	0.857

Table 9

Results of heterogeneity of marketization level.

	High Marketization Level	Low Marketization Level
	(1)	(2)
	lngtfp	lngtfp
ofdi	0.151	2.220 ***
	(0.21)	(2.91)
_cons	−0.481	−0.856
	(−0.67)	(−0.36)
Year FE	YES	YES
Prov FE	YES	YES
N	150	150
R 2	0.780	0.622

References

1. Branstetter, L. Is OFDI a channel of knowledge spillovers? Evidence from Japanߣs FDI in the United States. J. Int. Econ.; 2006; 68, pp. 325-344. [DOI: https://dx.doi.org/10.1016/j.jinteco.2005.06.006]

2. Kogut, B.; Chang, S.J. Technological capabilities and Japanese OFDI in the United States. Rev. Econ. Stat.; 1991; 73, pp. 401-413. [DOI: https://dx.doi.org/10.2307/2109564]

3. Lall, S. The rise of multinationals from the third world. Third World Q.; 1983; 5, pp. 618-626. [DOI: https://dx.doi.org/10.1080/01436598308419716]

4. Jie, J.; Ji, S.B.; Wen, X.H. Reverse technology spillover of OFDI and Chinaߣs technological progress-Research trends and prospects. J. Int. Trade; 2013; 8, pp. 161-169.

5. Kong, Q.X.; Peng, D.; Wang, X.Y. Research on the regional differences of Chinaߣs ODI reverse technology spillover effect in an open economy-An explanation based on the absorptive capacity of human capital. Forum World Econ. Politics; 2019; 4, pp. 113-132.

6. Yang, S.F.; Chen, K.M.; Huang, T.H. Outward OFDI and technical efficiency: Evidence from Taiwanߣs manufacturing firms. J. Asian Econ.; 2013; 27, pp. 7-17. [DOI: https://dx.doi.org/10.1016/j.asieco.2013.04.007]

7. Xie, T.Z.G.; Xue, X.; Ge, P.F. The impact of Chinaߣs OFDI on the GTFP of countries along the "the Belt and Road Initiative". J. Shanghai Univ. Financ. Econ.; 2019; 21, pp. 96-110.

8. Yang, S.D.; Han, X.F.; Song, W.F. Does OFDI affect Chinaߣs GTFP. J. Shanxi Univ. Financ. Econ.; 2017; 39, pp. 14-26.

9. Hu, Y.X.; Qu, X.E.; Li, Y.Y. Chinaߣs green economic growth effect on OFDI in countries along the "the Belt and Road Initiative". Econ. Manag.; 2019; 41, pp. 5-21.

10. Zhang, W.K.; Ge, Y. The spatial effect of OFDI on GTFP. Chin. J. Manag. Sci.; 2021; 29, pp. 26-35.

11. Wang, K.L.; Xu, R.Y.; Zhao, B. OFDI, Structural Transformation and GTFP. J. Nanjing Univ. Financ. Econ.; 2020; 6, pp. 74-85.

12. Zheng, Q. Does OFDI promote the growth of total factor productivity in the home country-An empirical test based on the threshold model of financial development. J. Int. Trade; 2017; 7, pp. 131-141.

13. Wu, S.S.; Li, B. Research on nonlinear effects of OFDI reverse technology spillovers in China—An empirical analysis based on panel smooth transition model. World Econ. Res.; 2015; 9, pp. 74-85.

14. Rajan, R.; Zingales, L. Financial development and growth. Am. Econ. Rev.; 1998; 88, pp. 559-586.

15. Xie, W.M.; Fang, H.X. Financial development, financing constraints and enterprise R&D investment. J. Financ. Res.; 2011; 5, pp. 171-183.

16. Zhong, T.; Wang, C.Y. Financial development and enterprise innovation output-Based on the comparative perspective of different financing models. J. Financ. Res.; 2017; 12, pp. 127-142.

17. Shen, H.B.; Kou, H.; Zhang, C. Empirical study on financial development, financing constraints and enterprise investment. China Ind. Econ.; 2010; 6, pp. 55-64.

18. Tang, S.; Wu, X.C.; Zhu, J. Digital finance and technological innovation of enterprises-Structural characteristics, mechanism identification and effect differences under financial supervision. Manag. World; 2020; 36, pp. 52–66+9.

19. Wan, J.Y.; Zhou, Q.; Xiao, Y. Digital finance, financing constraints and enterprise innovation. Econ. Rev.; 2020; 1, pp. 71-83.

20. Mao, Q.L.; Xu, J.Y. Does the OFDI of Chinese enterprises promote enterprise innovation. World Econ.; 2014; 37, pp. 98-125.

21. Li, X.C.; Xiao, X. System Escape or Innovation Drive?-Institutional constraints and OFDI of private enterprises. Manag. World; 2017; 10, pp. 99–112+129+188.

22. Ge, P.F.; Huang, X.L.; Han, X.F. Innovation-driven and GTFP promotion of "the belt and road initiative"—Heterogeneous innovation analysis based on the new economic growth model. Econ. Sci.; 2018; 1, pp. 37-51.

23. Zhang, J.; Wu, G.Y.; Zhang, J.P. Chinaߣs provincial physical capital stock estimation: 1952–2000. Econ. Res. J.; 2004; 10, pp. 35-44.

24. Bai, J.H.; Liu, Y.Y. Can OFDI improve Chinaߣs resource mismatch. China Ind. Econ.; 2018; 42, pp. 60-78.

25. Guo, F.; Wang, J.Y.; Wang, F.; Kong, T.; Zhang, X.; Cheng, Z.Y. Measuring the development of digital inclusive finance in China: Index compilation and spatial characteristics. Econ. Q.; 2020; 19, pp. 1401-1418.

26. Wang, Q. Fixed-effect panel threshold model using Stata. Stata J.; 2015; 15, pp. 121-134. [DOI: https://dx.doi.org/10.1177/1536867X1501500108]

27. Li, L.; Xian, G.M.; Bao, Q. Does "bringing in" promote "going aboard"? -The impact of foreign investment on OFDI of Chinese enterprises. Econ. Res. J.; 2018; 53, pp. 142-156.