The Influence of Reverse Technology Spillover of Outward Foreign Direct Investment on Green Total Factor Productivity in China’s Manufacturing Industry

Abstract

Outward foreign direct investment (OFDI) is an important channel for China to obtain advanced technology spillover to promote green production upgrading. As a pillar of the national economy and also a large-scale pollution emission industry, can the manufacturing industry benefit from the reverse technology spillover of OFDI and improve its green total factor productivity (GTFP)? Based on the provincial data of China from 2005 to 2019, this paper analyzes the effect and moderating mechanism of reverse technology spillover of OFDI on domestic manufacturing GTFP theoretically and empirically. The results show that the reverse technology spillover of OFDI cannot significantly promote the growth of manufacturing GTFP in the sample period. The heterogeneity analysis further proves that the inhibition effect similar to that of the whole sample appears in the eastern region, but in the central and western regions, the reverse technology spillover of OFDI can significantly improve the manufacturing GTFP, and this positive effect in the central region is greater. Additionally, absorptive capacity characterized by human capital, economic development and financial development can actively adjust the impact of reverse technology spillover of OFDI on manufacturing GTFP, of which the positive moderating effect of financial development is the most significant.

1. Introduction

As an important pillar of China’s national economy, the manufacturing industry has played a decisive role in promoting the rapid economic development. However, its extensive development mode with high input, high energy consumption and high emission has led to serious resource waste and environmental pollution, which has become an important factor restricting high-quality economic development for a long time [1]. According to the 2020 Global Environmental Performance Index (EPI) Assessment Report [2], China ranks only 120th out of 180 countries, which illustrates the urgency of paying attention to environmental pollution. The report of the 19th National Congress of the Communist Party of China pointed out that “green development is an important means to transform the mode of economic development, and improving green total factor productivity (GTFP) is the key to achieving green development” [3]. GTFP can fully consider both energy input and environmental pollution in the process of economic development, which is more in line with the requirements of high-quality and sustainable economic development [4]. As the main source of energy consumption and environmental pollution emissions, upgrading the GTFP of the manufacturing industry and realizing its green transformation is the inevitable choice for China to achieve high-quality economic development.

The upgrading of the manufacturing GTFP cannot be separated from scientific and technological innovation. In the context of economic globalization, it is often difficult for a country, especially a developing country, to gain absolute technological advantage only by relying on its own strength. In this case, access to foreign technology spillovers through multiple channels can give play to the “late-mover advantage” of developing countries and achieve technology catch-up and upgrading [5]. There are three main forms of international technology spillovers, including trade, foreign direct investment (FDI) and outward foreign direct investment (OFDI) [6,7]. However, due to the protection of developed countries, it is often difficult for developing countries to obtain core technologies at the international frontier. Compared with the first two ways, the reverse technology spillover obtained through OFDI can be more targeted and proactive [8]. Nowadays, OFDI has become an important way for the international community, especially developing countries, to integrate into the global economy and acquire international advanced technology and other strategic assets [9,10,11]. In recent years, China has further expanded its opening up to the outside world, actively attracting foreign capital and encouraging local enterprises to “go global” [12,13]. Some other emerging countries have also boosted local economies through OFDI and upgraded their position in global value chains by promoting technological progress and improving the status of trade networks [14,15]. In 2020, China’s OFDI flow reached USD 153.71 billion, and the stock had risen to USD 2580.66 billion. Can OFDI of such a large scale help improve the GTFP of domestic manufacturing industry by obtaining advanced technology spillover from abroad? An accurate answer to this question is of crucial relevance and theoretical value for the green development of the manufacturing industry and for the choice of China’s future OFDI development direction.

Scholars have discussed this research topic from different perspectives and made various findings. The study of Kogut and Chang [9] first confirms that OFDI can indeed bring reverse technology spillovers to home countries through investment and learning from developed countries. Hong et al. [10] and Fahad et al. [11] further provide the similar conclusion that OFDI is actually an effective way to acquire advanced technologies. Based on this, some scholars directly affirmed the positive linear influence of OFDI on green production activities of the home country [16,17]. At the regional level, Pan et al. [18] proved that reverse technology spillover of China’s OFDI can significantly promote the growth of total factor carbon productivity (TFCP) according to provincial panel data and found that the reverse technology spillover of OFDI can increase the TFCP of neighboring provinces through the spatial spillover mechanism. At the firm level, Bai et al. [12] examined that OFDI to developed countries can transfer reverse green technology and promote the environmental innovation of manufacturing enterprises in emerging economies. Song et al. [13] found a similar significant positive impact of OFDI on GTFP using Chinese firm-level data and discussed the heterogeneity of the impact for different firm nature, different OFDI types and different regions. However, there are also some studies drawing the opposite conclusion that the promotion effect of reverse technology spillover of OFDI in the home country is not significant [19] or even negative [20,21]. Lee [22] empirically tested the relationship between reverse technology spillover of OFDI and home country productivity by taking OECD countries as research objects and concluded that the former does not play a significant positive role. Bitzer and Kerekes [23] believe that OFDI has a negative effect on the improvement of total factor productivity through analyzing the industrial level data of OECD countries, but this effect is significantly different among different countries. The research of Li et al. [24] found that reverse technology spillover of OFDI has a significant inhibitory effect on TFP growth in the home country which is more obvious when the market mechanism of the host country is not sound.

Different from the above studies, some scholars believe that the impact of reverse technology spillover of OFDI on green production in home countries is nonlinear. The research results of Zhu and Ye [25] show that only OFDI flowing to developed countries can bring significant technology spillover effect and promote China’s green technology progress, while this effect is insignificant for OFDI flows to transition countries or developing countries. He et al. [26] found that the impact of reverse technology spillover of OFDI on total factor energy efficiency is nonlinear, and that there is a single threshold effect at the regional level and a double threshold effect at the industry level. Some other scholars propose that the impact of OFDI reverse technology spillovers on green production in the home country is limited by other factors such as environmental regulation [27,28], institutional quality [29], national culture [30] and absorptive capacity [7]. Among these, absorptive capacity is one of the key capabilities of the home country in the process from acquiring foreign technologies to domestic effective application, which is manifested in financial, human and other aspects [31,32]. Tian [33] argues that OFDI can significantly improve GTFP and digital inclusive finance can strengthen the promoting effect. Zhou et al. [34] verify that whether China’s OFDI has green spillover effect on domestic economy depends on the heterogeneity of absorptive capacity and environmental supervision strength of provinces.

The above literature has highlighted the impact of OFDI on the green production of the home country from different perspectives, but the research strand on the impact of reverse technology spillover of OFDI on GTFP is relatively small. Therefore, this paper attempts to supplement some research gaps and makes several marginal contributions. First, many studies fail to fully consider the bidirectional impact of OFDI on GTFP in theory, so the estimation results may be biased. This paper considers the multiple effects in the whole process of a country obtaining foreign technology spillovers through OFDI and affecting the GTFP of the industry, providing a solid theoretical framework for the research. Second, many existing studies assess the impact of OFDI on green production in the home country from the perspective of OFDI flow or stock scale, which may cause deviation due to the selection of indicators. This paper directly calculates the reverse technology spillovers of OFDI and selects manufacturing industry as the research object, hoping to improve the accuracy and pertinence of the research conclusions. Third, many studies only select one variable as a reference when studying the moderating effect of absorptive capacity, which may not be comprehensive. In this paper, human capital, economic development and financial development are all included in the test of the moderating effect, which can not only make the regression results more convincing, but also draw more effective policy enlightenment through comparison.

Generally, the main objective of this paper is to analyze whether and how reverse technology spillover of OFDI can affect the domestic manufacturing GTFP theoretically and empirically. To achieve this, this paper (1) adopts the baseline regression to test the influence of reverse technology spillover of OFDI on domestic manufacturing GTFP; (2) constructs heterogeneity analysis to discuss the different performance of this influence in different regions; and (3) studies the moderating effect of absorptive capacity including human capital, economic development and financial development.

The rest of this work is organized as follows. Section 2 illustrates the mechanism function of reverse technology spillover of OFDI on GTFP. The design of the research scheme, including model introduction, indicator selection and data source, is shown in Section 3. The empirical results and analysis are presented in Section 4. Finally, Section 5 gives the conclusions and puts forward the policy implications.

2. Theoretical Analysis

2.1. Process of Reverse Technology Spillover of OFDI

The process of a country obtaining foreign technology spillovers through OFDI can be divided into three stages: acquisition stage, transfer stage and diffusion stage [35,36].

In the technology acquisition stage, multinational corporations can acquire advanced technologies in the host countries three ways. First is cross-border mergers and acquisitions (M&A). OFDI can be achieved through greenfield investment and M&A, which can bypass the protection barriers of developed countries to advanced technologies. In this way, the patent and production process, scientific and technological R&D and other tacit knowledge of the acquired company can be transferred to the subsidiary in the host country quickly and conveniently. Second is overseas R&D cooperation. By setting up R&D institutions in the host country and cooperating with local universities and research institutes, multinational corporations can make full use of local advanced technologies and resources, which is conducive to a more in-depth and intuitive understanding of the cutting-edge areas of technology development. Third is industrial agglomeration. After entering the technology-intensive industrial cluster area of the host country through OFDI, multinational companies can not only get in touch with the frontier development field of their own industries, but also acquire advanced technologies through communication and learning with other enterprises in the cluster.

In the technology transfer stage, the subsidiary in the host country needs to transfer the information resources, acquired technical knowledge and R&D achievements to the parent company, and the parent company subsequently digests and absorbs these technologies and finally puts them into production and operation. In order to realize this process, the government of the home country can attract the parent company, actively promoting the overseas subsidiary to transfer new technologies and skills back to home through policy incentives at the macro level. For example, governments can provide preferential policies for domestic multinational corporations investing in advanced technology industries in developed countries to save investment costs and increase profits. The micro transmission for enterprise level is mainly based on two-way personnel flow. On the one hand, the parent company can send domestic technicians overseas for training and learning, so that technicians can master overseas skills and then return to apply what they have learned to domestic production practice and product R&D. On the other hand, the parent company can also directly introduce talents from the overseas subsidiary to exchange and work domestically, which can not only shorten the cycle of reverse technology spillover, but also help the home country to absorb sophisticated technologies.

In the technology diffusion stage, the parent company should first identify and screen out the overseas technologies that are valuable to the home country and fit its situation. Subsequently, the parent company needs to make a localized interpretation of the technologies that meet the screening conditions, and combine the core and key technologies with each production link, to lay the foundation for technology transformation and application. Finally, these modified and innovative technologies will be applied to the production practice of the multinational corporations and even the whole industry. This way, reverse technology spillover completes the value embodiment in the internal optimization process of the home country.

2.2. Effect of Reverse Technology Spillover of OFDI on GTFP

In the process of reverse technology spillover of OFDI, the green production activities of the home country will be affected in many aspects [19,26,37], resulting in the change in GTFP. The influence effects mainly include:

(1) R&D feedback effect. On the one hand, OFDI activities can help enterprises in the home country gain direct access to advanced green production technologies in the host country, so as to help the parent company learn and imitate, and eventually apply these technologies to the whole industry. On the other hand, the transfer of foreign technologies to the home country will also stimulate local enterprises’ sense of competition, to maintain or increase their market share. Overall, the R&D feedback from reverse technology spillovers of OFDI will drive enterprises to learn and compete, thereby increasing the GTFP across the industry.

(2) Cost allocation effect. For manufacturing enterprises, it is essential to maintain technological advantages through increasing R&D expenses. However, excessive R&D costs have become invisible capital pressure on enterprises with relatively backward and unsophisticated production systems. OFDI activities can expand overseas markets for these enterprises, increase their income and reduce the pressure of R&D expenses. At the same time, when investing in a host country that actively welcomes overseas capital, it can also obtain tax and fiscal policy support from the host government which is conducive to share R&D costs. An example from India confirms this—to realize the “Made in India” plan, the Indian government has created some key factors to attract FDI investment, including establishing high-level government support and maintaining a large number of sustainable financial resources. According to the World Investment Report 2022, these measures have enabled India’s FDI inflows to reach USD 81.973 billion between 2020 and 2021, which created a record level and an increase of 10% over the previous fiscal year. With the fiscal support of the host government, the R&D cost pressure of the parent company could be reduced, which better helps improve the green production efficiency of the home country.

(3) Resource optimization effect. Through OFDI activities, the bi-directional flow of production factors at home and abroad will be promoted, especially the internalization of international resources. Since enterprises may tend to protect their core technologies, OFDI is somewhat easier than other routes, such as trade, to weaken technological barriers and help the home country gains access to advanced overseas production factors. At the same time, OFDI also facilitates the cross-border flow of human capital, and enterprises in the home country can hire high-level technical personnel from the host country to optimize local production resources and strengthen green production capacity.

(4) Capital crowding-out effect. OFDI activities are also essentially part of enterprises’ limited capital allocation. When large OFDI accelerates the outflow of domestic capital, resource allocation on other production links may change greatly. Crucially, funds which are originally available for independent R&D and innovation of enterprises may be partly occupied. Especially for enterprises with strong independent R&D capabilities, the negative impact of the crowding-out effect will be more obvious. In addition, due to the risks in the OFDI process, enterprises may suffer losses in overseas investment. This may also lead to insufficient supply of domestic R&D funds, which is not conducive to the development of enterprises’ independent innovation.

(5) External dependence effect. Although the reverse technology spillover of OFDI can help enterprises imitate innovation to a certain extent, it can also cause some excessive dependence on external technologies. For example, in order to save time and promote production, some enterprises tend to directly assign overseas technicians to take charge of the core production link, without digesting and absorbing the advanced technologies needed in this link. In the long run, the enthusiasm of enterprises for independent innovation will be reduced, and the independent innovation ability cannot be really improved. In other words, this mentality of dependence on external technologies will lead to local enterprises being locked in behind the advanced foreign technologies, thereby depressing the GTFP of the whole industry in the home country.

2.3. Moderating Effect of Absorptive Capacity on Reverse Technology Spillover of OFDI

Whether reverse technology spillovers of OFDI can improve the technology level of the parent company depends, to some extent, on whether the parent company can identify, absorb and transform the advanced technologies transferred back from its subsidiaries. In other words, absorptive capacity plays an important role in the process of reverse technology spillover of OFDI affecting GTFP in the home country. For different enterprises, even if the technical knowledge transferred back by their overseas subsidiaries is similar, the intensity of reverse technology spillover will be different due to different absorption capacities. Factors affecting the absorption capacity of spillover technology in the home country mainly include:

(1) Human capital. The level of human capital determines the ability of knowledge accumulation, and the knowledge system originally possessed by domestic human capital can effectively complement advanced foreign knowledge. High-quality human capital will help promote the progress that reverse spillover technology elements can be digested, absorbed and recreated after entering the home country [32]. Therefore, the improvement of human capital can make each absorption step of the home country gain more benefits, and further strengthen the effect of reverse technology spillover of OFDI.

(2) Economic development. First, the better the local economy develops, the more complete the construction of transportation, communication and other facilities, and the more conducive for enterprises to absorb foreign technologies. Second, economic development will improve the economic vitality of the region, thus enhancing the effectiveness of market competition. Under the pressure of competition, enterprises in the market will focus on R&D and innovation, which can actively improve the absorption and digestion of technological knowledge in the whole region. Third, economic development can drive regional market demand. This will encourage local enterprises to actively absorb foreign technologies in order to make them better localized and cater to local needs [34].

(3) Financial development. The improvement of the financial development level is conducive to relaxing the financing constraints of regions on enterprises. On the one hand, this will enable enterprises to increase their R&D investment, and gradually enhance their ability to absorb and digest foreign technologies while repeatedly learning and creating technologies [31]. On the other hand, the frequency of enterprises’ outward investment will be increased, which can help enterprises skillfully handle foreign technologies, so as to improve the overall technology absorption capacity of the region [33].

3. Research Design

3.1. Research Scheme

Many previous studies on this topic have laid the foundation for this paper, whose methods, variables, data and conclusions are shown in Appendix A. However, as mentioned in the introduction, although these literature works have studied relevant issues from multiple perspectives, there are still some research gaps that can be supplemented. Based on the above theoretical analysis, this paper proposes that the reverse technology spillover of OFDI promotes GTFP of the home country through R&D feedback effect, cost allocation effect and resource optimization effect, and inhibits through capital crowding-out effect and external dependence effect. It is worth noting that the performance of these effects may be different in different regions. At the same time, the impact of reverse technology spillover of OFDI on GTFP is affected by the heterogeneity of absorptive capacity. Basis on these, the complete mechanism function and empirical tests taken in this paper to verify the following three hypotheses are shown in Figure 1.

Hypothesis 1.

Reverse technology spillover of OFDI can significantly affect the manufacturing GTFP of the home country.

Hypothesis 2.

There is regional heterogeneity in the impact of reverse technology spillover of OFDI on the manufacturing GTFP.

Hypothesis 3.

Absorption capacity plays a moderating role in the impact of reverse technology spillover of OFDI on GTFP.

3.2. Model Specification

Coe and Helpmen [38] proposed the international R&D spillover model, believing that a country’s technological progress can be affected by not only by domestic but also foreign R&D level. On this basis, Lichtenberg and Potterie [39] introduced OFDI as an important channel of foreign R&D spillover and further constructed the L-P model, which provided research thinking for testing the impact of OFDI on technological progress. In this paper, the L-P model is referenced to investigate the impact of reverse technology spillover of OFDI on GTFP, which is described as:

$$\ln GTF{P}_{i,t}={\alpha }_0+{\alpha }_1\ln SOFD{I}_{i,t}+{\alpha }_{2}Control{s}_{i,t}+{\epsilon }_{i,t}$$

(1)

Among them, the subscripts $i$ and $t$ stand for provincial region and year, respectively. $\ln GTFP$ is the explained variable of this paper and denotes the logarithm value of GTFP. $\ln SOFDI$ is the core explanatory variable of this paper and denotes the logarithm value of reverse technology spillover of OFDI. $Controls$ represents a series of control variables. ${\epsilon }_{i,t}$ represents the random error term and ${\alpha }_0$ represents constant.

In addition, considering the continuity of the growth for GTFP, this paper adds $\ln GTF{P}_{i,t-1}$, the lag term of GTFP, to the empirical model, as shown in Equation (2). At the same time, there may be a bidirectional causality between the reverse technology spillover of OFDI and the GTFP, that is, the improvement of the GTFP in turn affects enterprises to “going out” and obtain foreign technologies. Therefore, endogeneity inevitably exists, and it is scientifically feasible to use the generalized method of moments (GMM) to overcome the problem when the lag term of the explained variable has been introduced in the model. Since the SYS-GMM is more suitable for solving the problem of weak instrumental variables than the DIF-GMM, this paper selects the former as the empirical estimation method to verify Hypothesis 1.

$$\ln GTF{P}_{i,t}={\beta }_0+{\beta }_1\ln SOFD{I}_{i,t}+{\beta }_2\ln GTF{P}_{i,t-1}+{\beta }_{3}Control{s}_{i,t}+{\epsilon }_{i,t}$$

(2)

Subsequently, to verify Hypothesis 3, this paper sets the following mechanism model to examine the “moderating effect” of absorptive capacity in the impact of reverse technology spillover of OFDI on GTFP, which is described as:

$$\ln GTF{P}_{i,t}={\gamma }_0+{\gamma }_1\ln SOFD{I}_{i,t}+{\gamma }_2\ln GTF{P}_{i,t-1}+{\gamma }_{3}INTE{R}_{i,t}+{\gamma }_{4}Control{s}_{i,t}+{\epsilon }_{i,t}$$

(3)

Among them, $INTE{R}_{i,t}$ denotes the interactive item between absorptive capacity and reverse technology spillover of OFDI. The variables representing absorption capacity are included in the control variables, which are described in the following sections.

3.3. Data Sources

Taking the data availability into consideration, this paper selects the manufacturing industry of 30 Chinese provincial regions from 2005 to 2019 as samples (except Tibet, Hong Kong, Macao and Taiwan). The original data come from the Statistical Bulletin of China’s Outward Foreign Direct Investment, the Organization for Economic Co-operation and Development Database (OECD), the World Bank Database, China Industry Statistical Yearbook, China Statistical Yearbook of Science and Technology, China Labor Statistical Yearbook, China Energy Statistical Yearbook, China Statistical Yearbook on Environment and statistical yearbooks of every provincial region.

3.4. Definition of Variables

3.4.1. Explained Variable

GTFP ($GTF{P}_{i,t})$ is employed as the explained variable in this paper. Referring to the methods of Tone [40] and Oh [41], the SBM-GML model considering unexpected output is adopted for its widely use in dealing with the relationship between multiple inputs and outputs to evaluate efficiency. The selection and detailed description of input and output indicators are shown below [19,26,33].

Input indicators: (1) labor input, which is expressed as the year-end number of employed people in the manufacturing industry of every provincial region; (2) capital input, which is expressed as additional fixed capital and calculated using the perpetual inventory method; (3) energy input, which is expressed as the industrial energy consumption converted to standard coal of every provincial region.

Output indicators: (1) expected output, which is expressed as the industrial sale value after deflating by the producer price index (PPI) of industrial products in each year; (2) unexpected output, which is expressed as comprehensive environmental pollution index synthesized by entropy method using industrial wastewater discharge, industrial smoke and dust discharge, industrial sulfur dioxide discharge and solid waste discharge.

Based on the above, this paper sets the GTFP of 2005 as 1, and the calculated change rate is multiplied to obtain the GTFP level from 2006 to 2019 of the whole China as well as eastern, central and western regions. To reduce the heteroscedasticity, logarithmic form is taken when the model is regressed.

3.4.2. Explanatory Variable

Reverse technology spillover of OFDI ($SOFD{I}_{i,t})$ is taken as the core explanatory variable in this paper. Referring to the common practices in the existing literature [11,26], it can be measured by calculating the international R&D capital stock obtained from OFDI, as shown below:

$$SOFD{I}_{p,t}={{\displaystyle \sum }}_{q}OFD{I}_{p,q,t}\times \frac{S_{q,t}}{GD{P}_{q,t}}$$

(4)

Among them, the subscripts $p$ and $q$ stand for home and host region, respectively. $OFD{I}_{p,q,t}$ stands for the investment stock from region $p$ to $q$ in period $t$; $GD{P}_{q,t}$ and $S_{q,t}$ stand for the gross domestic product (GDP) and R&D capital stock of the host region $q$, respectively. $S_{q,t}$ can be further expressed as:

$$S_{q,t}= S_{q,t-1}\times \left(1-\delta \right)+R{D}_{q,t}$$

(5)

Among them, $\delta$ denotes the depreciation rate, set as 5%; $R{D}_{q,t}$ stands for the R&D expenditure of region $q$ in period $t$. Considering that the sample period of this paper starts from 2005, $S_{q,2005}$ can be calculated as:

$$S_{q,2005}= \frac{R{D}_{q,2005}}{g+\delta }$$

(6)

Among them, $g$ denotes the average growth rate of R&D expenditure from 2005 to 2019. The R&D expenditure is obtained by multiplying the R&D intensity of country $q$ by its GDP denominated in USD. In order to eliminate the influence of price factors, the consumer price index of each country from 2005 to 2019 is used to convert the R&D expenditure with 2005 as the constant price. Based on data availability, and considering that technology seeking OFDI going to developed economies can better realize learning of foreign advanced technologies, this paper selects 11 countries and regions as the main sources for China to absorb international technology spillover: Hong Kong, Japan, South Korea, Singapore, Russia, Germany, The Netherlands, the United Kingdom, the United States, Canada and Australia. By the end of 2020, China’s OFDI stock in the aforementioned 11 countries and regions had totaled USD 1706.92 billion, accounting for 66.14% of China’s global OFDI stock. Using the R&D stock in 2005 of 11 countries and regions, $S_{q,t}$ from 2006 to 2019 can be obtained, which is further summed up to calculate international R&D capital stock from OFDI of the whole China.

Based on the above, reverse technology spillover of OFDI for every provincial region can be calculated using Equation (7), where $OFD{I}_{i,t}$ and $OFD{I}_{p,t}$ stand for non-financial OFDI stocks of provincial region $i$ and the whole China, respectively.

$$SOFD{I}_{i,t}= SOFD{I}_{p,t}\times \frac{OFD{I}_{i,t}}{OFD{I}_{p,t}}$$

(7)

3.4.3. Control Variables

Referring to existing literature [12,13,14,18,19], this paper includes the following control variables:

(1) Lag term of GTFP ($GTF{P}_{i,t-1})$. The growth of GTFP may be continuous, and the previous period of GTFP may lead the growth of the later. Logarithmic form is taken when the model is regressed to reduce the heteroscedasticity.

(2) Technology spillover of import ($SIM{P}_{i,t})$. Generally, the greater the actual import volume of a country, the stronger the reverse technology spillover. The calculation method of $SIM{P}_{i,t}$ is similar to that of reverse technology spillover of OFDI. Logarithmic form is taken when the model is regressed to reduce the heteroscedasticity.

(3) R&D stock of domestic manufacturing ($SR{D}_{i,t}$). Due to the relative lack of R&D data for provincial manufacturing, this paper uses the internal expenditure of R&D funds of industrial enterprises above the scale in every provincial region to replace the R&D funds of the manufacturing industry, and adopts the perpetual inventory method to convert it into stock data. Logarithmic form is taken when the model is regressed to reduce the heteroscedasticity.

(4) Human capital ($HU{M}_{i,t}$). The higher the level of human capital, the more conducive to the development of technological innovation. According to Barro and Lee [42], this paper employs the average years of education of the labor force to approximate the level of human capital. Logarithmic form is taken when the model is regressed to reduce the heteroscedasticity.

(5) Economic development ($GD{P}_{i,t})$. A higher level of economic development means a higher level of scientific and technological research and innovation in the region, and a stronger ability to absorb advanced technologies. The level of economic development is generally expressed by regional GDP. In order to eliminate the influence of population factors on GDP, this paper uses per capita GDP as the representation of economic development level.

(6) Financial development ($FI{N}_{i,t})$. The higher the level of financial development, the smaller the financing constraints of regional enterprises, and the more favorable it is for enterprises to carry out foreign investment activities. In this paper, the ratio of year-end credit to GDP of financial institutions is selected as the proxy index of financial development level.

Among the above control variables, human capital, financial development and economic development can reflect the local absorptive capacity for advanced technology spillover. Therefore, this paper selects these three variables as the moderating variables in the mechanism model, forming an interactive item with the core explanatory variable. That is, $INTE{R}_{i,t}$ in Equation (3) can be expressed as$HU{M}_{i,t}\times SOFD{I}_{i,t}$,$GD{P}_{i,t}\times SOFD{I}_{i,t}$ and $FI{N}_{i,t}\times SOFD{I}_{i,t}$. Descriptive statistics of each variable are shown in

Table 1. Descriptive statistics of variables.

Variables	Obs	Mean	Std	Min	Max
$\ln GTFP$	450	0.028	0.397	−1.337	1.278
$\ln SOFDI$	450	−0.634	2.1	−7.247	4.039
$\ln SIMP$	450	2.241	0.133	1.861	2.632
$\ln SRD$	450	3.12	1.63	−1.39	6.573
$\ln HUM$	450	3.858	1.643	−3.429	7.26
$GDP$	450	0.641	0.417	0.065	2.381
$FIN$	450	3.005	1.146	1.288	8.131

.

4. Empirical Results and Description

In order to avoid spurious regression of panel data in empirical analysis, this paper tests the stationarity of data before regression analysis. CADF test and CIPS test are used to improve the accuracy, and the null hypothesis of both tests is that there is a unit root. The results show that all variables have no unit root, that is, the variables are stationary.

4.1. Baseline Regression

The regression results of the baseline model are presented in

Table 2. Empirical results of baseline model using SYS-GMM model.

Variables	(1)	(2)	(3)	(4)	(5)	(6)	(7)
$\ln SOFDI$	−0.186 * (−1.83)	−0.020 *** (−3.07)	−0.040 *** (−3.43)	−0.040 ** (−2.41)	−0.070 *** (−3.03)	−0.928 ** (−2.58)	−0.810 *** (−3.25)
$\ln GTF{P}_{t-1}$		0.991 *** (16.22)	0.959 *** (25.65)	0.974 *** (28.39)	0.943 *** (25.36)	0.785 *** (26.24)	0.860 *** (26.37)
$\ln SIMP$			0.057 *** (3.76)	0.057 *** (3.76)	0.052 *** (3.04)	0.533 *** (3.40)	0.574 *** (3.78)
$\ln SRD$				0.026 (1.25)	0.014 (0.91)	0.345 (1.55)	0.288 ** (1.83)
$\ln HUM$					0.483 *** (3.05)	5.782 ** (2.43)	3.287 ** (2.38)
$GDP$						−1.197 ** (−2.22)	−1.560 *** (−2.88)
$FIN$							0.318 ** (2.14)
Constant	−0.090 (−0.83)	−0.019 (−1.82)	−0.280 *** (−3.84)	−1.930 *** (−2.79)	−1.420 *** (−3.12)	−8.667 * (−2.49)	−11.100 *** (−3.07)
AR(2)	1.59 (0.112)	0.43 (0.669)	0.89 (0.376)	0.79 (0.429)	1.01 (0.313)	0.79 (0.432)	0.87 (0.384)
Hansen	20.36 (0.119)	28.38 (0.291)	28.21 (0.453)	27.87 (0.472)	27.96 (0.467)	15.0 (0.307)	12.78 (0.465)
$N$	450	450	450	450	450	450	450

Note: (1) The numbers in brackets are the t statistics corresponding to the regression coefficient; ***, ** and * indicate that they have passed the significance test at the 1%, 5% and 10% confidence levels, respectively, with the same interpretations below. (2) The numbers in brackets of Hansen test are the P statistics to check the rationality of tool variables, the same as in the following table.

. Through stepwise regression, variables that may cause multicollinearity are eliminated. The P value of the Hansen test in column (7) is 0.465, which is smaller than 0.1, indicating that there is no over identification problem for the tool variables selected in the model, and the selection of tool variables is effective; the p value of the AR (2) test is 0.384, indicating that the random disturbance term of the residual does not have second-order sequence correlation, and the endogenous problem of the model has been well solved.

The regression results of columns (1)–(7) robustly show that the coefficient of reverse technology spillover of OFDI ($SOFDI)$ is always significant. That is to say, Hypothesis 1 raised in the theoretical analysis is acceptable. From the sign of coefficient, it can be observed that the impact of reverse technology spillover of OFDI on GTFP is negative. This is consistent with the research conclusions of Bitzer and Kerekes [23] and Li et al. [24]. Column (5) is the result of using the complete Equation (2), indicating that GTFP decreases by 0.81% for every 1% increase in reverse technology spillover of OFDI. This may be because the negative impact of China’s OFDI to developed countries, such as capital crowding-out effect and external dependence effect, is more obvious than the positive impact of R&D feedback effect, cost allocation effect and resource optimization effect.

The coefficient of lag term of GTFP ($GTF{P}_{t-1})$ is 0.860, which is significant at the level of 1%, indicating that the growth of GTFP of the manufacturing industry is not only affected by explanatory variables, but also by the GTFP of the previous year, and the GTFP of the last year plays a positive demonstration role for this year. For other control variables, the coefficient of technology spillover obtained through import ($SIMP$) is 0.574 and significant at the level of 1%, indicating that every 1% increase in technology spillover obtained through import will drive the GTFP of the manufacturing industry up by 0.574%. Import trade provides opportunities for domestic enterprises to learn, imitate and absorb foreign advanced technologies, thus improving the level of domestic manufacturing GTFP. R&D stock of domestic manufacturing ($SRD$) is significant at the level of 5%, with a coefficient of 0.288, indicating that the more domestic R&D investment accumulation to the purchase of advanced equipment and the introduction of technical personnel, the more it promotes the improvement of manufacturing GTFP. The coefficient of human capital level ($HUM$) is 3.287 and significantly positive, indicating that the higher the level of human capital, the better domestic enterprises can learn, identify and absorb foreign advanced technical knowledge. At the same time, the improvement of human capital will also promote the independent technological innovation of enterprises to a certain extent, thereby further driving the improvement of GTFP. The coefficient of economic development ($GDP$) equals −1.56 and significant at the level of 1%, indicating that the economic development level has a certain inhibiting effect on the manufacturing GTFP. The possible reason is that the improvement of China’s economic development level during the sample period is still based on large-scale energy consumption, and the economic development is not fully compatible with green development. The coefficient of financial development ($FIN$) is significantly positive, indicating that China’s financial development has actively promoted the absorption of reverse technology spillovers. Reverse technology spillover of OFDI is mainly concentrated in high-tech fields with high investment and risk. A sound financial system can facilitate enterprise financing, reduce financing costs and improve financing efficiency, thus promoting technology diffusion, transfer and absorption.

4.2. Robustness Test

Two methods are adopted for robustness testing in this section. First, this paper uses practice of Zhang and Li [43] for reference, removes 1% of the minimum and maximum values of GTFP and re-estimates Equation (2) to overcome the influence of the abnormal values and non-randomness of explained variable on the regression. Secondly, this paper adjusts the calculation method of core explanatory variables referring to the methods of Rong et al. [44], that is, replace $GD{P}_{q,t}$ with $K_{q,t}$ in the calculation formula of reverse technology spillover of OFDI. Based on this, Equation (4) is reformulated as below:

$$SOFD{I}_{p,t}={{\displaystyle \sum }}_{q}OFD{I}_{p,q,t}\times \frac{S_{q,t}}{K_{q,t}}$$

(8)

The robustness test results are shown in

Table 3. Empirical results of robustness test using SYS-GMM model.

Variables	(1)	(2)
$\ln SOFDI$	−0.785 ** (−2.64)	−0.793 *** (−3.28)
$\ln GTF{P}_{t-1}$	0.830 *** (6.45)	0.844 *** (6.48)
$\ln SIMP$	0.558 *** (2.92)	0.570 *** (3.87)
$\ln SRD$	0.268 (1.65)	0.287 * (1.77)
$\ln HUM$	2.957 * (2.01)	3.107 *** (2.35)
$GDP$	−1.466 ** (−2.20)	−1.540 *** (−2.84)
$FIN$	0.305 * (1.79)	0.326 *** (2.23)
Constant	−10.369 ** (−2.55)	−10.807 *** (−3.10)
AR(2)	−0.51 (0.608)	0.92 (0.358)
Hansen	13.54 (0.331)	12.41 (0.495)
$N$	441	450

Note: ***, ** and * indicate significance at the 1%, 5%, and 10% levels, respectively.

. It can be seen that the two methods both pass the Hansen test, and the values of AR (2) are greater than 0.1. The coefficients of the core explanatory variables are significantly negative at the levels of 5% and 1%, respectively, and all the other control variables pass the significance test except R&D stock of domestic manufacturing ($SRD$) in column (1), which is basically the same as the results in the baseline regression. The above results demonstrate that the conclusion of this paper is robust.

4.3. Heterogeneity Analysis

Different regions are quite different in resource endowment, economic development basis, policy support and technological innovation capacity, so the overall sample may be difficult to make a reasonable assessment of how reverse technology spillover of OFDI affect GTFP in different regions of China. It is of great practical significance to further analyze the regional heterogeneity. Thus, this paper divides 30 regions into the eastern region, the central region and the western region, and performs sub-sample regression to verify Hypothesis 2. On the basis of the baseline regression model, the heterogeneity analysis takes the eastern region as the reference system, introduces two dummy variables representing the central and western regions, and also employs the SYS-GMM method to estimate the dynamic panel. See

Table 4. Empirical results of regional heterogeneity test using SYS-GMM model.

Variables	Eastern	Central	Western
$\ln SOFDI$	−1.260 * (−2.08)	0.741 * (1.88)	0.678 * (1.86)
$\ln GTF{P}_{t-1}$	1.182 *** (5.25)
$\ln SIMP$	0.474 * (2.06)
$\ln SRD$	0.436 * (1.99)
$\ln HUM$	2.063 (1.66)
$GDP$	−1.463 * (−1.72)
$FIN$	0.450 * (1.79)
Constant	−8.698 ** (−2.20)
AR(2)	0.34 (0.735)
Hansen	14.36 (0.349)
$N$	450

Note: ***, ** and * indicate significance at the 1%, 5%, and 10% levels, respectively.

for the estimation results.

The regression results show that the impact of reverse technology spillover of OFDI on manufacturing GTFP varies significantly among regions. In the central and western regions, reverse technology spillover of OFDI play a positive role in promoting GTFP significantly with coefficients of 0.741 and 0.678, respectively, of which the effect in the central region is slightly greater than that in the western region. However, the coefficient of core explanatory variable in the eastern region is −1.260, indicating that for every 1% increase in reverse technology spillover of OFDI, the manufacturing GTFP in the eastern region will decline by 1.260%. This partly explains the results of full sample regression, as more than 80% of China’s technology seeking OFDI flowing to developed countries come from the eastern region. Although OFDI in the central and western regions has grown rapidly in recent years, their overall scales are still small.

Why does reverse technology spillover of OFDI in the eastern region inhibit domestic manufacturing GTFP? The core reason may be that enterprises in the eastern region have stronger independent R&D ability, a smaller technology gap with overseas and a larger scale of OFDI compared with the central and western regions. According to the conditional convergence theory, when the technology gap between the two countries is small, the technology learning space of OFDI activities will be limited, because the subsidiary in the host country can hardly acquire more advanced knowledge and cannot provide much help for the green production of the parent company in the home country. In this case, from the perspective of technology acquisition, a larger scale of OFDI from the eastern region will squeeze more investment in independent R&D of enterprises in the home country. Especially when there is “incomplete information” in OFDI activities, the R&D spillover investment of the host country will also have “repeated consumption” with the domestic R&D funds, resulting in resource waste. As a result, the crowding-out effect caused by reverse technology spillover of OFDI will be amplified in the eastern region, while the positive effects such as R&D feedback and resource optimization are relatively insignificant, which eventually lead to the suppression of manufacturing GTFP.

4.4. Moderating Analysis

The results of baseline regression show that reverse technology spillover of OFDI fails to promote the GTFP of the manufacturing industry in the sample period. So, does an increase in domestic absorptive capacity help ameliorate this negative effect? To answer this question, Hypothesis 3 is proposed in the theoretical analysis, and interactive items composed of human capital, economic development, financial development and reverse technology spillover of OFDI are added as auxiliary explanatory variables in the baseline model. The regression results using SYS-GMM are shown in

Table 5. Empirical results of moderating effect test using SYS-GMM model.

Variables	(1)	(2)	(3)
$\ln SOFDI$	−0.611 *** (−3.56)	−0.668 *** (−3.51)	−0.636 *** (−3.50)
$HUM\times SOFDI$	0.020 ** (2.13)
$GDP\times SOFDI$		0.020 ** (2.12)
$FIN\times SOFDI$			0.030 ** (1.85)
$\ln GTF{P}_{t-1}$	0.864 *** (8.17)	0.842 *** (7.30)	0.850 *** (7.59)
$\ln SIMP$	0.407 *** (3.82)	0.455 *** (3.84)	0.430 *** (3.65)
$\ln SRD$	0.204 * (1.91)	0.228 ** (1.89)	0.219 * (1.90)
$\ln HUM$	2.500 ** (2.46)	2.766 ** (2.48)	2.599 ** (2.43)
$GDP$	−1.080 *** (−2.92)	−1.260 *** (−2.96)	−1.143 ** (−2.72)
$FIN$	0.200 * (1.87)	0.229 * (1.90)	0.207 * (1.74)
Constant	−8.320 *** (−3.20)	−9.210 *** (−3.22)	−8.687 (−3.15)
AR(2)	0.13 (1.49)	0.15 (1.44)	0.14 (1.46)
Hansen	0.223 (16.51)	0.251 (15.97)	0.244 (16.09)
$N$	450	450	450

Note: ***, ** and * indicate significance at the 1%, 5%, and 10% levels, respectively.

.

It can be found that the coefficients of the three interactive items in the regression results are all significantly positive, and the coefficients of the core explanatory variables are still significantly negative, but their absolute values are reduced compared with those of baseline regression. On this basis, Hypothesis 3 is validated: the improvement of absorptive capacity can weaken the negative impact of reverse technology spillover of OFDI on manufacturing GTFP, and promote the improvement effect of the former on the latter. By contrast, the coefficient of the interactive item composed of financial development and core explanatory variable is equal to 0.030, which is significant at the 5% level and relatively higher than that of the other two groups representing absorptive capacity. This means that financial development can play a more active role in the process of reverse technology spillover of OFDI affecting the manufacturing GTFP. Through relaxing regional financial constraints, enterprises can gradually enhance the ability to absorb and digest foreign technologies acquired through OFDI, and amplify positive effects such as R&D feedback. At the same time, enterprises can also invest sufficient funds to improve their independent R&D capabilities, thereby weakening the crowding-out effect of reverse technology spillover of OFDI, and comprehensively promote the manufacturing GTFP. The coefficients of the other two interactive items are both equal to 0.020, significant at the 5% level. This indicates that the accumulation of human capital and the promotion of regional economic development can both effectively improve the absorptive capacity of local enterprises to advanced technologies, and then help enterprises to better use the reverse technology spillover of OFDI to improve the GTFP of the whole manufacturing industry.

5. Conclusions and Policy Implications

This paper clarifies the impact of reverse technology spillover of OFDI on the GTFP of China’s manufacturing industry by first conducting mechanism analysis theoretically. Then, the panel data of the manufacturing industry in 30 provinces of China and SYS-GMM regression model are adopted to quantitatively analyze the impact of reverse technology spillover of OFDI on GTFP and the moderating role of absorptive capacity in it. The results indicate that reverse technology spillover of OFDI has a negative effect on manufacturing GTFP on the whole. This conclusion is still valid after changing the sample size and the calculation method of the core explanatory variable. At a regional level, heterogeneity is reflected in that the reverse technology spillover of OFDI significantly inhibits manufacturing GTFP in the eastern region, while it has a positive impact both in the central and western regions, and the impact in the central region is greater than that in the western region. Moreover, the impact of reverse technology spillover of OFDI on manufacturing GTFP can be moderated by absorptive capacity. With the improvement of human capital, economic development and financial development, the inhibiting effect of reverse technology spillover of OFDI on manufacturing GTFP will be weakened, and the promoting effect will be enhanced. Among them, the moderating role of financial development as absorptive capacity is more positive than the other two indicators.

Based on the results of the empirical analysis, the following policy implications are put forward. First, it is necessary to implement regionally differentiated investment and improve investment quality. Due to the large differences in the scale of OFDI in different regions, reverse technology spillover of OFDI has different effects on manufacturing GTFP. Therefore, differentiated investment strategies should be implemented according to the actual region situation to maximize the positive effects of reverse technology spillover. For example, reverse technology spillover of OFDI in the central and western regions can significantly promote manufacturing GTFP, so regional governments should continue to provide various investment related policies and actively encourage enterprises to “go global”. For the eastern region, large-scale overseas investment cannot promote the upgrading of regional manufacturing GTFP. The underlying reason may be that despite the large scale of OFDI in the eastern region, the proportion of technology acquisition is relatively low, and domestic R&D investment may be squeezed at the same time. Therefore, enterprises in the eastern region should improve the quality of OFDI through targeted investment in projects whose technology spillovers are needed by local development. At the same time, the government should pay more attention to maintaining the ability of independent innovation and promote the green development of the manufacturing industry in a two-pronged approach. Second, the government should improve the regional absorptive capacity and actively transform foreign advanced technologies. Absorptive capacity, which can be reflected in the accumulation of human capital, economic development and financial development, can help to reduce the negative effect of reverse technology spillover of OFDI on GTFP and magnify the positive effect. Therefore, the local government should pay attention to the introduction and training of high-tech personnel through strengthening the cooperation between domestic and foreign universities and research institutes, promote local economic development to tap market demand matching advanced technology and promote technological competition, and build a more active financial support system and improve the level of regional financial development to reduce the financing constraints of enterprises in the process of “going global”.

To sum up, this paper takes China’s manufacturing industry as the research object to verify the effect and moderating mechanism of reverse technology spillover of OFDI on domestic manufacturing GTFP theoretically and empirically. The results show that the reverse technology spillover of OFDI cannot significantly promote the growth of manufacturing GTFP based on the provincial data of China from 2005 to 2019, which is contrary to the research conclusions of Pan et al. [18], Bai et al. [12] and Song et al. [13], but supports the findings of Zhang and Ren [21] and Li et al. [24] to some extent. On this basis, this paper also verifies the heterogeneity of the above relationship in the eastern, central and western regions, which can provide targeted OFDI development reference according to the actual region situation. This is one of the important application contributions of this study. Moreover, this paper further explores the moderating mechanism of absorptive capacity, finding that human capital, economic development and financial development can all play important roles in the impact of reverse technology spillover of OFDI on GTFP, of which the positive moderating effect of financial development is the most significant. This can help China’s manufacturing industry to select targeted forms and paths to amplify the positive promotion effect of OFDI on GTFP and encourage enterprises to carry out reverse technology spillover in the process of “going global”.

This study provides a theoretical basis and policy reference for China to promote reverse technology spillover of OFDI and improve GTFP in the manufacturing industry. Nonetheless, this study has the following limitations. First, because of the data limitations, the time span chosen in this study to examine the impact of reverse technology spillover of OFDI on manufacturing GTFP is from 2005 to 2019, and the latest relationship after the pandemic cannot be analyzed. Second, the enterprise is the main element of OFDI, but due to the restriction of database, we only conducted research through industry panel data, without exploring typical cases or conducting research. In future research, we will seek breakthroughs in these aspects.