Can FDI and ODI two-way flows improve the quality of economic growth? Empirical Evidence from China

ABSTRACT

Different from the existing literature which only studied the unilateral impact of the FDI or the ODI on economic growth, this paper took both the FDI and the ODI into the analysis framework of international capital flow on economic growth, and tried to introduce the mediating effect model to test the transmission mechanism and influence effect of international capital two-way flow on the economic growth. The results showed that both the FDI and the ODI can significantly improve the quality of economic growth, and the role of the FDI was stronger than that of the ODI. At the same time, the FDI and the ODI had obvious regional heterogeneity in promoting the quality of economic growth. International capital flow mainly requires three mediating effects to improve the quality of economic growth, which are the employment, the technology spillover as well as the output efficiency. The three effects of the FDI have a played significant role in promoting the quality of economic growth, while the ODI mainly plays a role through the improvement of output efficiency and technology reverse spillover. There are also significant regional differences in mediating effects.

KEYWORDS:

• FDI
• ODI
• capital flow
• quality of economic growth
• mediating effect

I. Introduction

Since the 2000’s, the scale of the foreign direct investment (shorted for the FDI) and outward direct investment (shorted for the ODI) has been increasing year by year, which has significantly promoted the economic development of countries in the world (Yao and Wei 2007; Mehic, Silajdzic, and Babic-Hodovic 2013; Nwaogu and Ryan 2015; Rehman 2016; Hayat 2018; Aust, Morais, and Pinto 2020). However, some studies showed that the positive impact of FDI on economic growth was conditional. FDI only started to play a role when the development of financial market exceeded the threshold level (Azman-Saini, Law, and Ahmad 2010). The effect of FDI on economic growth was not significant in the short-term, but it was obvious in the long-term (Lee and Chang 2009; Belloumi 2014). When a country had sufficient absorptive capacity and occurred in a more technologically advanced transition economy, the FDI will have a greater impact on economic growth (Silajdzic and Mehic 2016). The FDI promoted the improvement of economic performance through financial development, and its impact of financial development on economic growth was further enhanced by the inflow of foreign direct investment (Sirag, SidAhmed, and Ali 2018). Among the FDI inflow components, the reinvestment of equity and income had a significant impact on the change of GDP (Kosztowniak 2019). At the same time, the impact of the FDI on the economy had obvious heterogeneity. The research showed that the growth effect of the FDI varied greatly among different industries (Chakraborty and Nunnenkamp 2008). The FDI from manufacturing sector had a positive impact on the economic growth of host countries, while the FDI inflow into non-manufacturing sectors had no significant effect on promoting economic growth (Wang 2009). On the contrary, some studies had shown that the contribution of the FDI to economic growth was not clear (Alfaro et al. 2004; Zandile and Phiri 2019).

The impact of the FDI and the ODI on China’s economy has aroused extensive concern of many scholars. Research showed that the FDI and government expenditure were two important factors of the economic growth, and there was an inverted U-shaped relationship between the FDI and the growth of GDP (Wu et al. 2020). The FDI promoted economic growth by promoting the accumulation of physical capital and human resources capital (Liu et al. 2014). Privatization and the FDI inflow were two key factors to promote China’s economic growth (Zhao 2013). Both export and the FDI had strong positive effects on economic growth (Yao 2006). However, some scholars believed that the impact of the FDI and the ODI on China’s economic growth was not obvious (Gunby, Jin, and Robert Reed 2017). Besides that, the research on the mechanism and effect of the two-way flow of the FDI and the ODI on the quality of economic growth still needed to be further studied. Therefore, the purpose of this paper was to analyse and test the mechanism of the two-way flow of the FDI and the ODI on the quality of economic growth.

The innovation of this paper was mainly in three aspects: (1) this paper studied the high-quality growth of China’s economy from the perspective of international capital flow, and discussed the transmission mechanism and influence effect of two-way international capital flow on China’s economic growth by considering both the FDI and the ODI. (2) the paper improved and expanded the names and measurement methods of some specific indicators, while defining and measuring the quality of economic growth from the three dimensions of economic growth efficiency, economic growth stability and economic growth sustainability. (3) the paper attempted to introduce the mediating effect model into the analysis of the conduction effect of international capital flow on economic growth, and systematically studies the main effect of the FDI and the ODI on the quality of economic growth and the three mediating effects of employment, output efficiency and technology spillover.

The rest of the paper was arranged as follows: the second part was the theoretical analysis and research hypothesis; the third part was the measurement of the quality of economic growth; the fourth part was the model, variables and data; the fifth part was the empirical test and result analysis; the sixth part was the conclusion and enlightenment.

II. Theoretical analysis and hypothesis

Theoretical model

In this section, we refer to the research of Borensztein, De Gregorio, and Lee (1998) and Junhong and Xiaohong (2017). Following the previous studies related to the classical Solow model, we consider an economy where the production process conforms to the specific Cobb-Douglas form:

(1) $Y=A{H}^{\alpha }{K}^{1-\alpha }$(1)

where $Y$ is the production, $A$represents the technology, $H$ denotes the human capital and $K$ denotes the physical capital. Since the paper focuses on the impact of international capital flow on economic growth, we mainly analyse the behaviour of the enterprise sector and ignore the individual behaviour. This is a difference with the extant literature related to the Solow model. However, this simplification is necessary which does not affect our conclusion.

According to the above theoretical analysis, both FDI denoted by $F$ and ODI denote by the $O$ can make a difference to the quality of domestic capital, $\varphi \left(F,O\right)$. Since capital $K$ determines the amounts of output $Y$, the connotation of domestic capital’s quality includes employment effect, output efficiency, and technology spillover. Given this, $\varphi \left(F,O\right)$ should satisfy conditions $\frac{\mathrm{\partial }\varphi }{\mathrm{\partial }\mathrm{F}}>0$ and $\frac{\mathrm{\partial }\varphi }{\mathrm{\partial }\mathrm{O}}>0$, which means the quality of domestic capital will increase with FDI and ODI.

In this paper, at each instant in time, we assume that the stock of domestic capital is

(2) $K={\left\{\underset{0}{\overset{N}{\int }}\varphi {\left(j;F,O\right)}^{1-\alpha }M{\left(j\right)}^{1-\alpha }dj\right\}}^{\frac{1}{1-\alpha }}$(2)

where $M\left(j\right)$ is the amount of capital $j$, $N$ is the total amount of varieties of capital goods. According the regular practice of the Solow model, all the enterprises in this model would comply with the principle of profit maximization. Following the processes of Junhong and Xiaohong (2017), we can get the result:

(3) $\frac{\mathrm{\Delta }A}{A}=-\frac{1}{\varphi \left(F,O\right)}$(3)

Since $\frac{\mathrm{\partial }\varphi }{\mathrm{\partial }\mathrm{F}}>0$ and $\frac{\mathrm{\partial }\varphi }{\mathrm{\partial }\mathrm{O}}>0$, we deduce that both FDI and ODI could make a positive impact on the quality of economic growth.

Mechanism analysis and hypothesis

Foreign research on the quality of economic growth started earlier, which can be traced back to the 1950s. At the beginning, people focused on finding a variable to reflect the quality of economic growth. Solow (1956) put forward the concept of economic total factor productivity (the TFP), which became an important indicator for many scholars to study the quality of economic growth later and considered the quality of economic growth from multiple perspectives. High quality economic growth did not only mean the growth of economic quantity, the smooth circulation of technical equipment and the self-sufficiency of clothing, food, housing and transportation, but also included the pursuit of peace, freedom and happiness of mankind (Stefan 2012). Some scholars had also constructed a quality measurement system of the economic growth, including three comprehensive indicators, namely scale performance, economic structure and social coordination, to comprehensively evaluate the quality of economic development (Qi 2016). Some scholars believed that social welfare level, resource consumption and environmental pollution were three important factors affecting the quality of economic growth. As for the impact of international capital flow on the quality of economic growth, the mainstream academic view was that it can effectively promote the economic growth of host countries. However, the impact of the FDI and the ODI on the quality of economic growth was not directly affected, but mainly through the channels of investment, employment, output efficiency and technology spillover, in order to improve the quality of economic growth. Therefore, this paper analysed the mediating factors from three aspects of employment effect, output efficiency and technology spillover, and analysed whether there was mediating effect in each factor, and how they can improve the quality of economic growth. The mechanism of their impact on the quality of economic growth was shown in Figure 1:

Figure 1. Theoretical mechanism

It can be seen from Figure 1 that the FDI and the ODI of a country have an impact on the quality of its economic growth, and they were mainly transmitted through three mediating effects, which were employment effect, output efficiency and technology spillover. Therefore, the hypothesis 1 was proposed.

Hypothesis 1: The FDI and the ODI of a country can improve the quality of its economic growth, and it was mainly transmitted through three mechanisms by influencing employment, improving output efficiency, and realizing technology spillover.

Employment effect

From the perspective of employment effect of the FDI, this paper analysed the impact of the FDI of Italian on local employment growth of 12 manufacturing industries and 103 administrative provinces from 1996 to 2001. It showed that after controlling the local industrial structure and regional fixed effect, the relationship between the FDI and local employment growth was faster than the average level of domestic industry (Federico and Minerva 2008). Through the research on whether the FDI created employment opportunities for the manufacturing industry of Mexico’s non US-Mexico joint ventures, it was found that the FDI had a significant positive impact on the employment of Mexican manufacturing industry, and the employment promotion effect of the FDI was greater in the export-oriented industries (Waldkirch, Nunnenkamp, and Alatorre Bremont 2009). Based on the data of Chinese enterprises from 1998 to 2007, the different effects of the FDI on employment elasticity were studied. The analysis showed that when facing the impacts of income, capital, and output, the FDI would significantly affect the employment elasticity of enterprises. These effects varied greatly among industries with different factor strength and export status . Through the research on the impact of China’s FDI on Thailand, it was found that government expenditure played a leading role in increasing employment in Thailand’s industrial sector, while export played a leading role in reducing employment in Thailand’s industrial sector (Keorite and Mohamed 2016). The research on whether the FDI affected the transfer of employment in both host and home countries showed that the inward FDI and the outward FDI in developed countries led to the transfer of employment from industrial sector to service sector, while only inward FDI in developing countries led to employment transfer from service sector to industrial sector (Rozen-Bakher 2017). By analysing the impact of the FDI and labour market on China’s employment and its internal mechanism, the results showed that the employment effect of the FDI was positive, and labour market flexibility played a significant positive role of regulatory (Rong et al. 2020).

From the employment effect of the ODI, some scholars analysed the impact of foreign direct investment of Spanish enterprises on domestic employment by using the industry data from 1995 to 2011. The results showed that the ODI had a positive impact on total domestic employment, but the employment effect of investment in different regions and industries was significantly different (Bajo-Rubio and Diaz-Mora2015). By studying the employment development of Czech, a German-owned company during the great recession from 2008 to 2009, it was found that companies with German capital significantly promoted employment in the years before the crisis. Since 2008, German enterprises have played a stabilizing role in the Czech labour market during the recession (Schäffler, Hecht, and Moritz 2017). From 1996 to 2010, the ODI of Japanese MNCs in 59 countries showed that the ODI can enhance MNCs’ domestic employment mainly by seeking expansion of market size and scope, natural resources, or strategic assets (Hong, Lee, and Makino 2019). However, according to the data of Korean industry from 2007 to 2014, it was found that the significant increase in the proportion of temporary workers offset the promotion effect of ODI on employment and in that condition, it had no impact on the employment level (Kang and Whang 2018). Taking 787 Japanese multinational corporations operating in 60 countries from 1996 to 2010 as the sample, the study showed that unconventional core business activities reduced the domestic employment level of multinational corporations’ overseas direct investment (Lee, Hong, and Makino 2020).

In China, the FDI has a positive impact on China’s employment and the quality of economic growth due to its long experience. However, China’s ODI is still at the beginning stage and the employment effect of adverse spillover of the ODI was uncertain, which showed differences in various industries, periods and regions. Therefore, the hypothesis 2 was proposed.

Hypothesis 2: The FDI and the ODI have employment effect on the quality of economic growth of a country. The employment effect of the FDI was positive, while the employment effect of the ODI showed strong heterogeneity.

Output efficiency

The output efficiency mainly represents the output of capital per unit. If the host country was a developing country, most of the national economic development was still in an extensive state. The inflow of the FDI can greatly accelerate the pace of industrial structure upgrading of the countries for being invested and improve the efficiency and sustainability of its economic growth. The research on Lithuanian enterprises showed that the FDI can produce positive productivity spillover, which was related to domestic and foreign joint projects, but not to fully owned foreign capital (Javorcik 2004). The results showed that during the early stage of low output, capital inflow improved with the increase of output, but it was not the cause of output growth. When the output was high, the increase of output was related to the capital outflow, because the growth of local savings exceeded the local demand for capital (Aharonovitz and Miller 2010). By examining the long-term impact of the FDI on domestic output in developing countries, it was found that the FDI had a significant positive long-term effect on domestic output (Herzer 2011). This paper studied the spillover effect of the FDI in Yangtze River Delta and Pearl River Delta as well as its impact on urban productivity. The results showed that the FDI had a positive impact on the local urban productivity in the Pearl River Delta and Yangtze River Delta, but the spatial spillover effect was different (Zhonglin et al. 2004).

The FDI promoted productivity mainly through ownership management, which was not indirectly. As the ODI of a developed country, it can bring rich profits to developed countries. At the same time, the international capital flow was an important symbol of the integration of global industrial chain, which maximized its comparative advantage (Iwasaki and Tokunaga 2016). From the perspective of the output level, after analysing the data of Chinese industrial enterprises from 2003 to 2007, it was found that the ODI can alleviate the problem of productivity decline caused by factor price distortion, and boost the output level of enterprises (Cheng et al. 2020). Using the quantile regression method, this paper studied whether the FDI weakened or enlarged the output growth volatility of 11 countries in the Economic Community of West Africa (the ECOWAS) region. It was found that the inhibitory effect of the FDI on output growth volatility was not across the quintile, but only in countries with high volatility (Ajide and Osode 2017).As the two sides of international capital flow, they had concentrated on giving full play to the local industrial advantages, improving the output efficiency, and improving the quality of economic development. Therefore, the hypothesis 3 was proposed.

Hypothesis 3: a country’s FDI and ODI have an output efficiency spillover effect on its economic growth quality, therefore, both the FDI and the ODI can improve the country’s output efficiency, and then improve the quality of economic growth.

Technology spillover

Generally speaking, technology spillover of the FDI can be divided into positive spillover of the FDI and reverse spillover of the ODI. The results showed that the FDI will produce technology spillover effect on local companies in transition economies, and the spillover effect depended on the FDI and the characteristics of local enterprises (Sinani and Meyer 2004). Capital flow, the FDI and technology spillover effects had significant impacts on the technology development and total factor productivity (TFP) of Arab countries (Sadik and Bolbol 2001). Technology spillover played an important role in promoting the technological innovation of local enterprises in emerging countries. Based on the research on the technology spillover effect of Shanghai, the largest foreign-funded city in China, it was found that the increase of the FDI inhibited the technological growth of local enterprises (Zeng, Wan, and Tam 2009). In the decentralized innovation network of the multinational companies, the dynamic development of its subsidiary’s technology and strategic role had an impact on the spillover effect of the FDI (Ha and Giroud, 2015). The FDI did not only promote China’s rapid economic growth, but also affected China’s environmental quality through technology spillover effect (Song, Tao, and Wang 2015). Research on the mechanism of technology diffusion through the FDI showed that the location in the supply chain was especially important to obtain the FDI spillovers (Jude 2016). By studying the geographical scope and spatial diffusion of the technology spillover of the FDI in China, the results showed that the direct impact of the FDI on specific regions itself may be negative (Lin and Kwan 2016). Some studies had shown that there was a double threshold effect in the process of the FDI technology spillover in different regions of China, and the technology spillover effect of the FDI gradually increased (Wang et al. 2016). In addition, many scholars had carried out more detailed research and confirmed the technology spillover effect from the aspects of environmental regulation, industrial chain, human capital, region, innovation and ownership (Hamida 2013;Wang and Aiqi 2016).

With the rapid growth of the foreign direct investment (the ODI), the research on the spillover effect of the ODI were gradually increasing. Through the analysis of China’s provincial panel data, it was found that the ODI would have adverse impact on the local environment, but the technology reverse spillover effect brought by ODI improved the domestic technology level (Hao et al., 2020). Some scholars believed that the technology spillover of the ODI had an important impact on the industrial structure of emerging economies as well as their dependence on domestic absorptive capacity. The results showed that the impact of China’s FDI on industrial structure was generally positive, but there was regional heterogeneity in absorptive capacity (Jiang, Luo, and Zhou 2020).

In general, the technology spillover effect of the FDI mainly through vertical spillover and horizontal spillover to promote and its effect of promoting the quality of economic growth was obvious. However, the ODI mainly brought about the technology reverse spillover effect, which means that the foreign investment enterprises integrated the technology, talents, and resources of foreign enterprises, fully absorbed, and then fed back to the local development. As one of the most important motivations for the ODI, technology reverse spillover had been verified by many scholars. It had an impact on the quality of economic growth by promoting the technological progress of investment countries. Therefore, the hypothesis 4 was proposed.

Hypothesis 4: both the FDI and the ODI had technology spillover effect on the quality of economic growth of a country, and the ODI showed the reverse spillover effect.

III. Measurement of the quality of economic growth

This paper adopted the comprehensive index evaluation system constructed by Guangguang and Tinghua (2014) to measure the quality of economic growth, including three classification indexes, which were economic growth efficiency (Effi), economic growth stability (Stab) and economic growth sustainability (Sustain). At the same time, on the basis of defining and measuring the quality of economic growth from the three dimensions of economic growth efficiency, economic growth stability and economic growth sustainability, six of the 22 specific indicators were improved, adjusted, replaced and eliminated in this paper. Among them, the calculation methods of the three indicators of capital financing efficiency, labour accommodation efficiency and trade structure were improved and expanded, the industrial upgrading process indicators were adjusted, the secondary school enrolment rate was replaced, and the world economic freedom index was eliminated.

The specific improvements were as follows: (1) the financing efficiency, which took into account the number of institutions related to the size of the problem, the number of financial institutions was replaced by financial industry output value/GDP; (2) considering that the unemployment rate index was more official and unified, the number of employment agencies per 10,000 people was replaced by the unemployment rate; (3) In terms of trade structure, considering that there were obvious port differences in net export indicators, for example, Shanghai Yangshan Port imports far more than exports, which did not mean that Shanghai’s trade indicators were poor. Therefore, the net import and export/GDP was replaced with its trade openness (total import and export/GDP); (4) In the process of industrial upgrading, the proportion of the secondary industry and the proportion of the tertiary industry were not scientific in the indicators of economic growth stability. For example, the increase of the proportion of tertiary industry brought by industrial upgrading cannot explain the instability of economic growth. Therefore, the proportion of secondary and tertiary industries would be adjusted to the sustainability of economic growth in this paper; (5) For the ratio of students to teachers in compulsory education stage, considering that the enrolment rate of secondary schools in various provinces in China was very high and the difference was not large, the ratio of students to teachers in the junior high education stage was replaced by the ratio of students to teachers in compulsory education stage, which was used to indicate the richness of educational resources; (6) Considering that the provincial data did not involve the world economy, the world economic freedom index was excluded. The improved and updated measurement standards were shown in Table 1.

Table 1. Economic growth quality classification and measurement criteria

The index name			Calculating mode	Index attribute
				Positive	Negative	Medium
Efficiency of economic growth	Factor productivity	Capital productivity	GDP/capital stock	√
		Labour productivity	GDP/employees	√
	Production organization efficiency	Total factor productivity	TFP	√
	Market efficiency	Product circulation efficiency	Total retail sales/GDP	√
		Financing efficiency	Financial sector output/GDP	√
		Technology integration efficiency	Technology market turnover/GDP	√
		Efficiency of labour integration	The unemployment rate		√
Stability of economic growth	Stability of factor supply	Capital price stability	Growth rate of real interest			√
		Labour price stability	Real wage growth rate			√
	Output stability	Output fluctuations	The GDP growth rate			√
		Price fluctuations	Rate of inflation			√
	Structural stability	National economic structure	Output value/GDP of private enterprises	√
		The trade structure	Total imports and exports/GDP	√
Economic growth sustainability	Long–term growth drivers	Industrial upgrading process	Tertiary industry/GDP	√
		Knowledge increment	The number of new patents issued by 10,000 people	√
		Research and development strength	R&D funds/GDP	√
	Human resource dynamics	Nonaging degree	Proportion of the population under 65	√
		Student – teacher ratio in compulsory education	Number of pupils/teachers in primary and junior middle schools		√
	Ecological environmental cost	Degree of air pollution	Sulphur dioxide emissions per unit of output		√
		Sewage discharge	Discharge of sewage per unit of output		√
		Solid waste discharge	Emissions of solid waste per unit of output		√

The indicators that needed further explanation in Table 1 were as follows:

(1) Capital stock: capital stock was calculated by perpetual inventory method $L_{it}=L_{i,t-1}\ast \left(1-\mathrm{\partial }\right)+I_{it}.$ The value of $L_{i,t-1}$ was the capital stock of the region in the previous year, ∂ is the capital depreciation rate, which was defined as 5% by convention. The $I_{it}$ was the total investment in fixed assets in the year t. The calculation of fixed assets in 2003 was based on the sustainable growth model $L_{i2003}=I_{i2003}/\left(\mathrm{\partial }+g_i\right).$ Among them, the $g_i$ was the average growth rate of China’s fixed asset investment from 2003 to 2017.

(2) For the TFP: The Data Envelopment Analysis (DEA) was adopted in this paper and Malmquist productivity index were used to measure and decompose TFP. As the most widely used total factor productivity estimation method in recent years, this method had the characteristics of nonparametric estimation and less data constraints and can overcome all kinds of biases caused by artificial selection of data. At the same time, the TFP can also be decomposed into efficiency change TEC and technical progress TC. The formula was as follows:

(4) $\begin{array}{r}\mathrm{M}\left(y_{t+1},x_{t+1},y_t,x_t\right)=\frac{D^{t+1}\left(x_{t+1},y_{t+1}\right)}{D^t\left(x_t,y_t\right)}\\ \times [\frac{D^t\left(x_{t+1},y_{t+1}\right)}{D^{t+1}\left(x_{t+1},y_{t+1}\right)}\times \frac{D^t\left(x_t,y_t\right)}{D^{t+1}\left(x_t,y_t\right)}{]}^{1/2}\end{array}$(4)

In formula (4), TFP was on the left and the TEC was in the first half of the right, and the TC was in the second half.

(3) The ratio of students to teachers in compulsory education stage: the index was calculated by the following formula,

(5) $\mathrm{S}=\left(\frac{1}{3}\right)\ast S_M+\left(\frac{2}{3}\right)\ast S_P$(5)

In formula (5), $S_M$ and $S_P$ represented the ratio of students to teachers in junior high school education and that in primary education.

(4) Positive, reverse and moderate index treatment. The positive index in Table 1 remained unchanged, the reverse index was taken as the reciprocal, and the moderate index was the reciprocal of the deviation. Considering the problem of incoherence between different indexes, the dimensionless method of averaging was used to unify the dimensions of data.

(5) Synthesis of comprehensive indexes. The synthesis of comprehensive indexes generally included entropy weight method, relative index method and principal component analysis method. The principle of entropy weight method was to quantify and synthesize the information of each unit to be evaluated, and to weight each factor, but it was easy to ignore the direct correlation of each index. The relative index method turned each index into a comparable form and obtained the specific index through the weighted sum method, which was too subjective. The principal component analysis (PCA) focused on the comprehensive evaluation of information contribution influence, and weights according to the mathematical characteristics of the data, avoiding the problem of subjectivity. Therefore, the principal component analysis was adopted to measure the quality of economic growth in this paper.

Other indicators were obtained through calculation, which were limited to space and would not be described in detail here. We used the data of 31 provinces, municipalities, and autonomous regions of Chinese mainland from 2003 to 2017 to measure the quality of economic growth. The results were shown in Table 2:

Table 2. Measurement results of economic growth quality

Province	Economic growth quality		Growth efficiency		Growth stability		Growth sustainability
	mean value	standard deviation	mean value	standard deviation	mean value	standard deviation	mean value	standard deviation
Beijing	2.16	1.19	1.37	0.75	0.69	0.27	2.60	1.70
Shanghai	1.72	1.30	1.43	0.92	0.52	0.20	1.80	1.81
Guangdong	1.13	0.79	0.35	0.58	1.08	0.18	0.86	0.95
Tianjin	0.92	1.02	0.79	0.88	0.07	0.15	1.01	1.04
Zhejiang	0.79	0.87	0.46	0.62	0.34	0.25	0.77	0.97
Hainan	0.70	0.85	0.34	0.60	0.37	0.48	0.54	0.67
Jiangsu	0.56	0.83	0.15	0.71	0.47	0.25	0.44	0.84
Tibet	0.31	0.55	0.30	0.58	−0.12	0.39	0.43	0.33
Fujian	0.24	0.60	0.45	0.55	0.17	0.19	−0.07	0.59
Chongqing	0.04	0.71	0.26	0.60	−0.08	0.28	−0.04	0.66
Shandong	−0.05	0.51	−0.05	0.54	0.12	0.20	−0.17	0.44
Hunan	−0.07	0.45	0.40	0.38	−0.20	0.24	−0.19	0.49
Hubei	−0.11	0.77	0.19	0.76	−0.12	0.33	−0.20	0.47
Inner Mongolia	−0.28	0.38	−0.09	0.41	−0.33	0.14	−0.34	0.31
Heilongjiang	−0.28	0.59	0.07	0.77	−0.30	0.17	−0.16	0.47
Liaoning	−0.29	0.59	0.13	0.75	−0.22	0.19	−0.29	0.32
Guangxi	−0.29	0.43	0.07	0.31	−0.04	0.29	−0.60	0.30
Sichuan	−0.30	0.55	−0.06	0.55	−0.15	0.25	−0.33	0.49
Jiangxi	−0.31	0.39	−0.07	0.32	0.04	0.22	−0.64	0.31
Jilin	−0.33	0.52	0.05	0.66	−0.30	0.12	−0.21	0.34
Henan	−0.40	0.46	−0.30	0.33	−0.02	0.25	−0.58	0.42
Shaanxi	−0.40	0.62	−0.39	0.67	−0.22	0.29	−0.23	0.45
Anhui	−0.48	0.50	−0.39	0.44	−0.23	0.24	−0.39	0.38
Hebei	−0.49	0.36	−0.29	0.43	−0.13	0.22	−0.57	0.28
Xinjiang	−0.53	0.41	−0.71	0.28	−0.11	0.21	−0.44	0.24
Gansu	−0.57	0.58	−0.68	0.51	−0.05	0.36	−0.38	0.40
Yunnan	−0.58	0.42	−0.50	0.42	−0.22	0.20	−0.50	0.27
Guizhou	−0.61	0.46	−0.74	0.42	−0.18	0.27	−0.42	0.32
Shanxi	−0.63	0.37	−0.69	0.53	−0.27	0.19	−0.44	0.36
Qinghai	−0.77	0.50	−0.96	0.57	−0.22	0.27	−0.64	0.20
Ningxia	−0.82	0.33	−0.91	0.33	−0.36	0.20	−0.61	0.22

The mean value is the average value for the period 2003–2017.

According to the measurement results in Table 2, there was a big gap in the quality of economic growth among provinces. Beijing, Shanghai, Guangdong, Tianjin, and Zhejiang occupied the top five in the quality of economic growth, while Shanxi, Qinghai and Ningxia were relatively weak in the performance of economic growth quality. In terms of various classification indicators, Shanghai, Beijing and Tianjin ranked among the top three in terms of economic growth efficiency; Guangdong, Beijing and Shanghai ranked among the top three in terms of economic growth stability; and Beijing, Shanghai and Tianjin ranked among the top three in terms of sustainable growth. This was generally consistent with the general impression of economic development, which showed that the evaluation system had good persuasion.

In recent years, the FDI and ODI of China have increased largely. As shown in Figure 2, FDI has increased from 53.51 billion dollars in 2003 to 138.14 billion dollars in 2019; on the meanwhile, ODI has increased from 2.85 billion dollars to 136.91 billion dollars. The growth rates of FDI and ODI have far exceeded the growth of GDP during the same period. Besides that, the upward trend of the quality of economic growth is basically consistent with that of FDI and ODI. Intuitively, we tend to conjecture that both FDI and ODI have promoted the improvement of the quality of economic growth.

Figure 2. FDI, ODI and quality of economic growth

Data source: National Bureau of Statistics of China.

According to the measurement of the quality of economic growth, we get the Figure 3. As we can see, the comprehensive index of economic growth quality and the other three indicators have shown a similar upward trend. Specifically, both the indices of the efficiency and that of sustainability show a steady upward trend. While showing the same upward trend, the stability of economic growth has a greater fluctuation. Given that, all three indicators above would enhance the economic quality improvement from different perspectives. Moreover, FDI and ODI tend to play a more important role than the other one.

Figure 3. The quality of economic growth from 2003-2017

IV. Research design

Baseline model

The main objective of this paper was to investigate the impact path of the FDI and the ODI on high-quality economic growth. Referring to the mediating effect model of Wen and Ye (2014), the model with the optimal relative testing power was established in this paper. The test steps were as follows:

step 1: tested the significance of the main effect, as shown in models (6) and (7):

(6) $Valu{e}_{it}=C_f+c_f\ast FD{I}_{it}+{\beta }_f\ast Contro{l}_{it}+{\epsilon }_{\begin{array}{c}it\\ \hspace{0.17em}\end{array}}$(6)

(7) $Valu{e}_{it}=C_f+c_o\ast OD{I}_{it}+{\beta }_o\ast Contro{l}_{it}+{\epsilon }_{\begin{array}{c}it\\ \hspace{0.17em}\end{array}}$(7)

In formula (6) and formula (7), $Value$ was the quality of economic growth, the $FDI$ and $ODI$ were foreign direct investment and outward direct investment, $Control$ was the control variable, which were government revenue (gov), financial market activity (sec), urban-rural gap (ur) and opening degree (open). If the coefficient c was significant, the next step was based on the mediating effect.

step 2: tested the following two mediating effect models (8), (9), and found the significance of the observation coefficient:

(8) $M{V}_{it}=C_o+a\ast I{V}_{it}+\beta \ast Contro{l}_{it}+{\epsilon }_{\begin{array}{c}it\\ \hspace{0.17em}\end{array}}$(8)

(9) $Valu{e}_{it}=C_o+c{ }^{\mathrm{\prime }}\ast I{V}_{it}+b\ast M{V}_{it}+\beta \ast Contro{l}_{it}+{\epsilon }_{\begin{array}{c}it\\ \hspace{0.17em}\end{array}}$(9)

In formula (8) and formula (9), $MV$ was the mediating variable, and $IV$ was the independent variable (FDI, ODI). Equation (8)(8) $M{V}_{it}=C_o+a\ast I{V}_{it}+\beta \ast Contro{l}_{it}+{\epsilon }_{\begin{array}{c}it\\ \hspace{0.17em}\end{array}}$(8) showed the effect of international capital flow on the mediating variable, expressed by coefficient a. The formula (9) represented the regression equation of the main effect after adding mediating variables. The coefficient $c{ }^{\mathrm{\prime }}$was the direct effect of international capital flow on the quality of economic growth, and b was the effect of mediating variables on the quality of economic growth. Theoretically, the mediating effect was the difference between the total effect and the direct effect, c-c′. Therefore, if the coefficients a and b were significant, we can conclude that the indirect effect was significant and proceed to the fourth step; if at least one of the coefficients was not significant, then proceed to the step three to continue the test.

Step 3: for the hypothesis H: ab = 0, Bootstrap method was used to test. If the BC confidence interval did not contain 0, the indirect effect was significant; if the BC interval contained 0, there was no significant effect, and the test was stopped.

Step 4: test formula (9). If the coefficient $c{ }^{\mathrm{\prime }}$was significant, it was a partial mediating effect; if the coefficient $c{ }^{\mathrm{\prime }}$ was not significant, it was a complete mediating effect.

Step 5: calculate the values of ab and c, $c{ }^{\mathrm{\prime }}$, if the values of ab and c were the same, then it was the mediating effect, and the proportion of mediating effect in the total effect was (c-$c{ }^{\mathrm{\prime }}$)/c.

Data description

China Economic Information Network Database (CEI) and China Stock Market & Accounting Research Database (CSMAR). Based on the actual needs and data availability, the annual data of 31 provinces in Chinese mainland from 2003 to 2017 for fifteen years were used, which were 5115 samples. Among them, the data of foreign direct investment was from the annual Statistical Bulletin of China’s Foreign Direct Investment, the data related to high-tech industry was from the Statistical Yearbook of China’s High-Tech Industry, the data of employment and population was from the Statistical Yearbook of China’s Population and Employment, and the rest is from the statistical yearbook of various provinces and cities, China Economic Information Network Database (CEI) and China Stock Market & Accounting Research Database (CSMAR).

Variables

Variable selection and description were shown in Table 3:

Table 3. Variable table

Variable type	Variable name	Specific connotation
Explained variable	Quality of economic growth	Comprehensive index of economic growth quality
Explanatory variables	Foreign direct investment	Adjusted per capita FDI stock (unit: CNY100/per captia)
	Outward direct investment	ODI stock per capita after adjustment (unit: CNY100/per captia)
Mediating variable	Employment effect index	Incomes of employees after price adjustment (unit: CNY)
		rate of employment (unit: %)
	Output efficiency index	Adjusted per capita GDP (unit: CNY)
		Energy consumption per 10,000 Yuan GDP (unit: CNY)
	Technology spillover index	Proportion of high-tech industry output value (unit: %)
		Number of patents granted per capita
		Total factor productivity (technical progress TC) (unit: %)
Control variable	Government revenue	General budget revenue/GDP (unit: CNY10,000/per captia)
	Financial market activity	Price adjusted per capita stamp revenue (unit: CNY100/per captia)
	Urban rural gap	Per capita consumption expenditure in urban/rural areas (unit: %)
	Degree of opening up	total imports and exports/GDP (unit: %)

In terms of explanatory variables, considering the time lag between the FDI and the ODI, this paper used the stock method to set indicators. The stock of the FDI was like the perpetual inventory method of capital stock, with a depreciation rate of 6% and a price adjustment value of GDP deflator from 1996 to 2017. Taking 1996 as the base period, the stock of the FDI in the base period was as follows:

(10) $FD{I}_{1996}=\frac{fd{i}_{1996}}{g+\theta }$(10)

In formula (10), g was the average growth rate of GDP from 1996 to 2003, and θ was the depreciation rate. The perpetual inventory method was adopted:

(11) $FD{I}_n=\left(1-\theta \right)\ast FD{I}_{n-1}+fd{i}_n\ast \alpha$(11)

The FDI data of each year came from the item of ‘the amount of foreign capital actually utilized’ in the statistical yearbook of each province; the ODI data came from the Statistical Bulletin of China’s Foreign Direct Investment of each year, and the price was adjusted.

In terms of mediating variables, employment effect index, output efficiency index and technology spillover index were measured by multiple sub indicators, which can measure specific effects more completely and accurately. For example, in the employment effect, income and employment rate were two important aspects, and a single indicator cannot fully reflect the employment situation. In terms of output efficiency, if we only considered GDP per capita, we would ignore the cost factor of energy consumption. Therefore, energy consumption of 10000 Yuan GDP was added as a supplement. In terms of technology spillover, high-tech industry was the main carrier of China’s technological progress, and the output value can reflect the development of high-tech industry. As the main result of industrial innovation activities, patent was an important aspect of technological progress. With the TC variable of technological progress of the whole society, it can reflect the degree of technological progress more completely and accurately.

In the selection of control variables, four variables were selected in this paper: government income, financial market activity, urban-rural gap and the degree of opening up. As the main body of China’s macro-control, the government has an important impact on economic development and future planning. Government revenue can affect the quality of economic development through regional development planning and specific measures. To a certain extent, the activity of financial market represented the enthusiasm of people to participate in capital market investment, showed the vitality of wealth re-creation, and was the embodiment of economic and financial prosperity. The urban-rural gap was the ratio of per capita consumption expenditure between urban and rural areas in each province, which showed social problems such as the gap between the rich and the poor, wealth redistribution and so on, which directly affected the sustainability of economic development. As the most commonly used control variable, the degree of opening to the outside world reflected the depth of a region’s economic exchanges with foreign countries and had an impact on economic development. Therefore, the above four variables were selected as control variables to discuss.

In the aspect of data processing, firstly, the dimensionless processing of averaging was used to eliminate the dimensional difference, and then principal component analysis was adopted to extract the comprehensive index.

V. Results and discussion

Descriptive statistics

This paper was mainly divided into two parts: the first part was to test the regression and mediating effect of the whole sample; the second part was to test the mediating effect by adding regional heterogeneity into four regions.

The variables involved in the empirical analysis included: quality of economic growth (value), foreign direct investment (FDI), foreign direct investment (ODI), employment effect index (job), output efficiency index (output), technology spillover index (tech), government revenue (gov), financial market activity (sec), urban-rural gap (ur) and degree of opening up (open). The statistical results were shown in Table 4:

Table 4. Descriptive statistical results

Variable code	Variable name	Minimum value	maximum value	mean value	standard deviation
value	Quality of economic growth	−1.38	4.70	0.00	0.95
fdi	foreign direct investment	0.80	220.66	39.29	44.23
odi	Outward direct investment	0.00	181.43	5.45	15.74
job	Employment effect index	−2.63	2.43	0.00	0.72
output	Output efficiency index	−1.70	4.90	0.00	1.16
tech	Technology spillover index	−1.39	3.92	0.00	0.92
gov	Government revenue	0.03	1.62	0.27	0.29
sec	Financial market activity	0.01	2.6	0.31	0.39
ur	Urban rural gap	1.59	7.2	2.90	0.73
open	Degree of opening up	0.02	1.59	0.29	0.35

The results of descriptive statistics showed that the standard deviation of other variables was close to the mean value and was relatively stable, except for the indicators with negative value and foreign direct investment.

Full sample test

The results of Hausmann test showed that there were significant time fixed effect and individual fixed effect in all regression. Therefore, according to the test results, this paper used the two-way fixed effect model. At the same time, in order to solve the problem of autocorrelation of disturbance term, clustering robust standard error was used.

Full sample main effect test

The first step was to test the main effect c according to the models (3) and (4) established in 4.1. The main effect regression results were shown in Table 5.

Table 5. Main effect regression results of full sample

variable	value	variable	value
fdi	0.007189***	odi	0.011362***
	(3.09)		(6.84)
gov	0.7963763**	gov	0.5633327**
	(2.23)		(3.71)
sec	0.6417909**	sec	0.4379969**
	(2.55)		(3.19)
ur	0.0344234	ur	0.0155521
	(1.32)		(0.51)
open	0.0275619**	open	0.3509627**
	(0.12)		(1.57)
_cons	−1.077191***	_cons	−0.9032473***
	(−8.17)		(−8.01)
rho	0.76296501	rho	0.79729694
N	465	N	465
R^2	0.8462	R^2	0.8478

*, * *, and *** represented significance levels of 10%, 5% and 1% respectively. Regression coefficient was shown in the table, and T value was shown in brackets.

It can be seen from Table 5 that both the FDI and the ODI were positively significant at the level of 1%, indicating that both the FDI and the ODI had a more significant positive role in promoting the quality of economic growth. At the same time, the regression coefficient of the FDI was significantly higher than that of the ODI, indicating that the promotion effect of per FDI on the quality of economic development was stronger than that of the ODI.

In terms of control variables, government revenue, financial market activity and degree of opening to the outside world had positive significance. However, there was no negative correlation between the urban-rural gap and the quality of economic growth in theory. Through the observation of the data, it was found that the gap between urban and rural areas was not small in some provinces with high quality of economic growth. The most obvious example was Guangdong Province. The gap between the cities represented by the Pearl River Delta and the rural areas represented by the southwest of Guangdong Province was very large.

From the whole process of mediating effect test, the main effect regression of whole sample was the first step. From the test results, the strong significance enabled us to argue with the ‘mediating effect’, and then we can carry out the next test steps.

Full sample mediating effect test

According to the models (5) and (6) established in 4.1, we tested the mediating effect of the employment effect index, the output efficiency index and the technology spillover index. Combined with the test results of the models (3) and (4), we got the coefficients c, a, B and $c{ }^{\mathrm{\prime }}$. At the same time, combined with bootstrap test, the mediating effect test results were obtained. The whole test steps and results of mediating effect were shown in Table 6:

Table 6. Test results of mediating effect of whole sample

Core variable	FDI			ODI
Intermediary variable	Employment effect	Output efficiency	Technology spillover	Employment effect	Output efficiency	Technology spillover
Step1: main effect c	0.007189*** (3.09)			0.011362*** (6.84)
Step2: coefficient a	0.00867*** (3.42)	0.00699*** (3.6)	0.00706** (2.14)	−0.0007 (−0.30)	0.0047*** (2.23)	0.0083*** (3.67)
Step2: coefficient b	0.0837 (1.12)	0.4019*** (4.05)	0.4911*** (7.86)	0.1965*** (2.78)	0.4044*** (5.10)	0.4631*** (8.45)
Step2: coefficient c ‘	0.00646*** (2.84)	0.00438** (2.36)	0.00372** (3.24)	0.0115*** (7.74)	0.0094*** (5.10)	0.0075*** (6.03)
Step3: bootstrap test	[0.0274 0.0424]	/	/	[0.0068 0.0122]	/	/
Step4: Is there a mediating effect	Yes	Yes	Yes	Yes	Yes	Yes
Step5: Proportion of intermediary effect	10.14%	39.07%	48.25%	−1.21%	16.90%	35.40%

*, **, and *** represented significance levels of 10%, 5% and 1% respectively. The results of Bootstrap test were adjusted by PC.

The test results in Table 6 showed that not only the FDI and the ODI had a significant positive effect on the quality of economic growth, but also there were three mediating effects: employment effect, output efficiency and technology spillover. This showed that the FDI and the ODI can significantly improve the quality of economic growth, and mainly through the impact of employment, improve output efficiency and realize technology spillover. Further analysis showed that the sum of the three mediating effects was 97.46% in the process of the FDI, which showed that the FDI can improve the quality of economic growth mainly through indirect effects, and the promotion effect through direct effects was small. In the process of the ODI, the sum of the three mediating effects also reached 51.09%. The specific analysis of the mediating effects was as follows:

From the perspective of employment effect, the employment effect of the FDI accounts for 10.14% of the total effect, indicating that the promotion effect of the FDI on the quality of economic growth was partly reflected through the employment effect. It showed that foreign-funded enterprises had a certain contribution in increasing employment and improving income level, thus promoting high-quality economic growth. The mediating effect of the ODI on employment was– 1.21%, which indicated that the ODI outflow had a weak inhibitory effect on employment level, which may be related to the loss of a small number of jobs caused by the ODI outflow. Therefore, the employment effect was mainly reflected in the inflow of the FDI, which can be well explained by the high salary and job growth of foreign-funded enterprises, while the ODI had little impact on employment.

In terms of output efficiency, there was a significant mediating effect of the FDI on the quality of economic growth, accounting for 39.07%. And because the coefficients a and b were significant, it showed that the inflow of the FDI promoted the improvement of the whole output efficiency, and then improved the quality of economic growth. Among the effects of the ODI on the quality of economic growth, the mediating effect of output efficiency reached 16.9%, which was far lower than that of the FDI. This result basically verified the previous theoretical analysis that the production efficiency of foreign direct investment enterprises and foreign investment enterprises was higher than the average market efficiency of China’s enterprises, and then the output efficiency level and the quality of economic development can be improved through the driving role and demonstration role.

From the perspective of technology spillover, technology spillover had a mediating effect in the process of the FDI and the ODI’s impact on the quality of economic growth, and both had strong explanatory power. In the process of FDI, technology spillover explained 48.25% of the main effect, which showed that nearly half of the impact of the FDI on the quality of economic growth was realized through technology spillover. Similarly, the reverse technology spillover of the ODI can also explain 35.4% of the main effect. Technology spillover was the most powerful mediating factor in the process of international capital two-way flow.

Regional heterogeneity test

The empirical results of the whole sample showed that the main effect and mediating effect of international capital flow on the quality of economic growth were obvious. However, due to the large regional differences in China and the level of economic development, there were significant differences in the purpose, channel, and effect of the FDI in the eastern region and the western region. Therefore, it was necessary and practical to test the heterogeneity in different regions. Due to the great difference in the development speed between northeast China and other regions in the east and central China in recent 15 years, the industrial type and economic vitality of northeast China were quite different from those of the eastern coastal area. Therefore, this paper divided China into four regions: eastern region, western region, central region and northeast region. Each region included the following provinces: the eastern region: Beijing, Tianjin, Shanghai, Hebei, Shandong, Jiangsu, Zhejiang, Fujian, Guangdong, Guangxi and Hainan; the central region: Shanxi, Inner Mongolia, Anhui, Henan, Hubei, Hunan and Jiangxi provinces, cities and autonomous regions; the western region: Chongqing, Sichuan, Yunnan, Guizhou, Tibet, Shaanxi, Gansu, Qinghai, Ningxia, Xinjiang, 10 provinces, autonomous regions and municipalities directly under the central government; the northeast China: Liaoning, Jilin and Heilongjiang provinces. The results of mediating effect test in different regions were as follows:

Eastern region

The test results in Table 7 were analysed as follows:

Table 7. Test results of mediating effect in Eastern

Core variable	FDI			ODI
Mediating variable	Employment effect	Output efficiency	Technology spillover	Employment effect	Output efficiency	Technology spillover
Main effect C	0.0125702***			0.0103735***
	(4.96)			(8.56)
Step2: coefficient a	0.00377	0.00687*** (3.43)	0.01701***	0.001	0.0029	0.0076***
	(1.15)		(3.7)	(0.37)	(1.05)	(3.4)
Step2: coefficient b	0.3022***	0.3226	0.4072***	0.3533***	0.4465**	0.4325***
	(4.34)	(1.18)	(5.04)	(3.28)	(2.27)	(6.63)
Step2: coefficient c^’	0.01143*** (4.63)	0.01035*** (4.06)	0.00564** (2.74)	0.0100*** (7.33)	0.0090***	0.0071*** (7.33)
					(4.84)
Bootstrap test	[0.0142 0.0654]	[0.1240	/	[−0.003 0.0208]	[0.0111 0.0310]	/
		0.2594]
Mediating effect results	9.07%	17.70%	55.13%	No mediating effect results	13.20%	31.70%

*, * *, and *** represented significance levels of 10%, 5% and 1% respectively.

(1) From the perspective of mediating effect, the FDI and the ODI can significantly improve the quality of economic growth, and the role of the FDI was far higher than the national level. It showed that in the eastern region, per unit of the FDI had a higher pulling effect on the quality of economic growth, which may be due to the relatively mature introduction and supporting system of foreign investment in the eastern region, owning strong absorption capacity of foreign capital.

(2) Employment played a mediating role in the process of the FDI’s influence on the quality of economic growth, but there was no mediating effect in the process of the ODI. As the most densely populated area of human capital in China, the eastern region can undertake relatively high-end foreign capital. For example, Tesla recently built a factory in Shanghai. Such high-end manufacturing foreign investment can improve the wage level of manufacturing workers, improve the employment level, and then improve the quality of economic development.

(3) However, the proportion of output efficiency was not high in the whole sample. From the statistical and test results, the main reason was that although there was a positive correlation between the level of output efficiency and the quality of economic growth, it was far less than the whole sample. The results showed that the output efficiency played a smaller role than technological progress, employment improvement and environmental protection in the process of improving the quality of economic growth in the eastern region. For example, Shanghai’s per capita GDP, average wage and per capita new patents were increased by 2.32 times, 3.79 times and 4.2 times respectively from 2003 to 2017. Compared with employment and innovation, the output efficiency represented by GDP per capita had a smaller improvement, and its contribution to the quality of economic growth may be smaller than other factors. At the same time, as the eastern region with the highest output efficiency in China, the output efficiency of local enterprises may not be inferior to that of foreign enterprises. Therefore, the mediating effect of international capital output efficiency was not large.

(4) Technology spillover was the most significant indicator in the eastern region, which had a strong explanation for the main effect. The technology spillover effect of the FDI was 55.13%, and that of the ODI was 31.7%. As a research hotspot of international capital flow, technology spillover effect had been well reflected in the eastern region. The technology spillover effect in the eastern region was slightly higher than that in the whole sample, which basically constituted the main effect. The significant technology spillover effect in the eastern region can be attributed to the strong absorptive capacity. Therefore, as an advantageous area of human resources, infrastructure and scientific research funds, the eastern region had a higher degree of opening to the outside world and had a better ability to absorb technology spillovers.

Central region

The test results in Table 8 were analysed as follows:

Table 8. Test results of mediating effect in Central

Core variable	FDI			ODI
Mediating variable	Employment effect	Output efficiency	Technology spillover	Employment effect	Output efficiency	Technology spillover
Main effect c	−0.01141474			0.0953559**
	(−0.55)			−2.73
Second step coefficient a	/			−0.0077 (−0.30)	0.0489 (1.62)	0.0855* (2.14)
Second step coefficient b	/			0.0743 (0.63)	0.5041* (2.15)	0.5599*** (4.3)
Second step coefficient c^’	/			0.0959** (2.65)	0.0707** (2.73)	0.0475* (2.17)
Bootstrap test	/			[−0.0213 0.2582]	[0.0126 0.2568]	/
Mediating effect results	/			No mediating effect results	25.80%	50.20%

*, * *, and *** represented significance levels of 10%, 5% and 1% respectively.

(1) In terms of main effect, the ODI had a significant role in promoting the quality of economic growth in the central region, but the FDI had no significant correlation. After studying the data, it was found that there was no significant correlation between FDI. The main reason was that the Inner Mongolia Autonomous Region and Jiangxi Province did not reflect the corresponding quality of economic growth in the case of the high FDI. However, the ODI can promote the quality of economic growth in central China. At the same time, the ODI coefficient of 0.095 was much higher than that of the whole sample, which may be due to the small overall scale of the ODI in central China and the strong promotion effect of per unit of the ODI.

(2) The employment effect was not significant in the ODI process. It showed that the employment effect did not play a role in the role of the ODI on the quality of economic growth in central China, which was basically consistent with the results of the whole sample. Due to the strict requirements of the ODI to feed back the employment of domestic enterprises through foreign investment enterprises, the employment mediating effect in central China had not been reflected.

(3) In terms of output efficiency and technology spillover, both of them can show significant mediating effect in the ODI process, the former was 25.8%, the latter was as high as 50.2%. The level of production and technology in the central region was between the eastern and western regions. The indicators of infrastructure, human capital and scientific research level in the central region were only inferior to those in the eastern region. Therefore, foreign direct investment can feed the advanced technology and management concept learned by the investment enterprises back to local enterprises by means of reverse technology spillover, in order to improve the production efficiency and technical level of the central region, and then to improve the quality of economic development.

Western region

The test results in Table 9 were analysed as follows:

Table 9. Test results of mediating effect in Western

Core variable	FDI			ODI
Mediating variable	Employment effect	Output efficiency	Technology spillover	Employment effect	Output efficiency	Technology spillover
Main effect c	0.0083409*** (3.66)			−0.0046968 (−0.22)
Second step coefficient a	0.00895** (3.15)	0.01009** (2.32)	0.00044 (0.11)	/
Second step coefficient b	0.1412 (1.22)	0.2833** (2.63)	0.4237*** (4.88)	/
Second step coefficient c^’	0.00707** (2.68)	0.00548*** (3.29)	0.00815*** (5.61)	/
Bootstrap test	[0.0147 0.0901]	/	[0.0651 0.1345]	/
Mediating effect results	15.20%	34.30%	2.28%	/

*, * *, and *** represented significance levels of 10%, 5% and 1% respectively.

(1) In the western region, the FDI had a significant role in promoting the quality of economic development, while the ODI had no corresponding effect. As a relatively backward area of economic development, the overall level of employment and technology in the western region lagged the average level of the eastern region and foreign-funded enterprises. Therefore, foreign-funded enterprises, such as Intel, P & G and other multinational companies, had much higher production efficiency and technical level than local enterprises, and can effectively drive the local economic development level. Therefore, the FDI had a good role in promoting the quality of economic growth, and the coefficient reached 0.08, which was slightly higher than the average level of the whole sample. In terms of the ODI, the absolute stock of the ODI in the western region was not high, which was less than one tenth of that in the eastern region.

(2) In terms of employment effect, there was a significant mediating effect in the process of the FDI. It showed that foreign enterprises can improve the quality of economic growth by improving the level of employment. The employment level in the western region was poor, and the income level that foreign businessmen can provide was far higher than the local average level, thus forming a mediating effect. For example, in Chengdu, the southwest central city, the average wage of employees from foreign-funded enterprises and Hong Kong, Macao and Taiwan investment enterprises were RMB 78,973 Yuan in 2018, while that of private enterprises in urban areas was only RMB 45,108 Yuan, with a huge gap. The mediating effect accounted for about 15.2%, which was higher than the national average level, indicating that employment in the western region was a particularly important aspect to benefit from the FDI.

(3) As an mediating variable, output efficiency also had significant mediating effect in the process of the FDI. The output efficiency of western region was low, and the improvement of output efficiency brought by foreign direct investment can significantly promote the quality of economic growth. This mediating effect accounted for 34.3%, which showed that the output efficiency of foreign investment had been well reflected in the western region.

(4) In the process of the FDI, the mediating effect of technology spillover was not prominent, which was only 2.28%. This conclusion was in good agreement with the absorptive capacity perspective mentioned above. The infrastructure, human resources and industrial coordination in the western region were relatively backward. The operation of foreign-funded enterprises, especially high-tech enterprises, required highly educated researchers, developed communication networks and transportation infrastructure. The current development level of western region was difficult to reach the threshold of absorbing technology spillover. Therefore, the mediating effect of technology spillover in this region was not prominent.

Northeast China

The test results in Table 10 showed that northeast China was special, and the main effects of the FDI and the ODI were not significant. Therefore, the mediating effect test did not need to continue.

Table 10. Test results of mediating effect in Northeast

Core variable	FDI			ODI
Mediating variable	Employment effect	Output efficiency	Technology spillover	Employment effect	Output efficiency	Technology spillover
Main effect c	0.00280044			0.0220558
	(2.09)			(0.89)

*, * *, and *** represented significance levels of 10%, 5% and 1% respectively.

From the regression coefficient, the FDI and the ODI had a positive impact on the quality of economic growth, but the effect was not significant. In terms of data, northeast China, especially Liaoning Province, had the largest scale of foreign direct investment in China at the beginning of the 21st century, but a large number of foreign investments had not been transformed into the quality of economic development. Northeast China was the first developed region of the Republic with an old industrial base. However, due to historical and practical reasons, the economic vitality of the region had gradually declined. Foreign direct investment and foreign direct investment had no significant effect on the quality of economic growth, indicating that the quality of economic development in northeast China needed more other factors to boost, such as internal investment, talent introduction, industrial transformation and so on.

Comprehensive analysis of mediating effect

The test results of comprehensive mediating effect were shown in Table 11:

Table 11. Test results of comprehensive mediating effect

Region	FDI			ODI
	Employment effect	Output efficiency	Technology spillover	Employment effect	Output efficiency	Technology spillover
Full samples	0.007189*** (3.09)			0.011362*** (6.84)
	10.14%	39.07%	48.25%	−1.21%	16.9%	35.4%
East region	0.0125702*** (4.96)			0.0103735*** (8.56)
	9.07%	17.7%	55.13%		13.2%	31.7%
Central region	−0.0114147 (−0.55)			0.0953559** (2.73)
					25.8%	50.2%
West region	0.0083409*** (3.66)			−0.0046968 (−0.22)
	15.2%	34.3%	2.28%
Northeast region	0.0028004 (2.09)			0.0220558 (0.89)

*, * *, and *** represented significance levels of 10%, 5% and 1% respectively.

The results of comprehensive mediating effect test in Table 11 were as follows:

(1) The regional heterogeneity of the impact of international capital flow on the quality of economic growth was very obvious. Both the FDI and the ODI in the eastern region had a positive impact on the quality of economic growth, while only the ODI in the central region had an impact, and the FDI only had an impact in the western region.

(2) The employment effect generally existed in the process of the FDI, and the western region benefited the most. The FDI created jobs and raised salary more directly. Therefore, the employment effect was more significant. There were also some regional differences in this effect. The western region was more likely to benefit from the backward employment.

(3) Output efficiency was reflected in the process of the FDI and the ODI. As an important impact of international capital flow, the improvement of output efficiency had been well reflected in China and had improved the quality of economic development. However, the local enterprises in the eastern region had relatively high output efficiency, and the effect of international capital was not as good as that of other regions.

(4) Technology spillover was the most significant mediating effect in the two processes, especially in the central and eastern regions. Technology spillover and reverse spillover, as important motivations of international capital flow, had been fully verified in China. Especially in the eastern region, the contribution rate of technology spillover in the process of the FDI and the ODI reached 55.13% and 31.7% respectively. The mediating effect was the most obvious, which played an important role in improving the quality of economic growth.

Endogeneity and robustness

Discussion and test of endogeneity

There were two main reasons for endogenous problems: one was the problem of missing variables. For example, the quality of economic growth in this year was likely to be affected by the stock of the FDI and the ODI in the previous year, thus causing endogenous problems. The other was causality. For example, international capital flow had an impact on the quality of economic growth, while the quality of economic growth may also have an impact on international capital flow. In order to alleviate the impact of endogenous problems, this paper used the instrumental variable method to estimate:

(12) $Valu{e}_{it}=C_f+\alpha \ast L.I{V}_{it}+c_f\ast I{V}_{it}+{\beta }_f\ast Contro{l}_{it}+{\epsilon }_{\begin{array}{c}it\\ \hspace{0.17em}\end{array}}$(12)

(13) $Valu{e}_{it}=C_o+\alpha \ast L.I{V}_{it}+c^{\prime }\ast I{V}_{it}+b\ast M{V}_{it}+\beta \ast Contro{l}_{it}+{\epsilon }_{\begin{array}{c}it\\ \hspace{0.17em}\end{array}}$(13)

In formula (12) and equation (13)(13) $Valu{e}_{it}=C_o+\alpha \ast L.I{V}_{it}+c^{\prime }\ast I{V}_{it}+b\ast M{V}_{it}+\beta \ast Contro{l}_{it}+{\epsilon }_{\begin{array}{c}it\\ \hspace{0.17em}\end{array}}$(13) , $IV$ represented the core explanatory variables of the FDI and the ODI, which were potential endogenous variables, and $L.IV$ was the first-order lag term of the FDI and the ODI to alleviate the problem of missing variables. The second-order lag term of the explained variable was used as the tool variable, and then the panel instrumental variable method was used for regression. The results were shown in Table 12:

Table 12. Regression results of instrumental variable method

	FDI					ODI
IV equation		Full samples	Employment effect	Output efficiency	Technology spillover	Full samples	Employment effect	Output efficiency	Technology spillover
L.IV		−0.0147	−0.0119	−0.0118	−0.0153**	−0.086*	−0.057*	−0.055*	−0.073*
		(−1.6)	(−1.4)	(−1.5)	(−2.1)	(−1.9)	(−1.8)	(−1.9)	(−1.9)
fdi\		0.02820***	0.02023**	0.01782**	0.02280***	0.0899**	0.0645**	0.0586**	0.0731**
odi		(2.93)	(2.19)	(2.06)	(2.93)	(2.3)	(2.36)	(2.39)	(2.12)
job			0.4171**				0.6220**
			(7.51)				(9.51)
outpour				0.6045***				0.7684***
				(10.3)				(9.87)
tech					0.6527***				0.5478***
					(13.8)				(5.14)
gov		1.229***	0.9188***	0.2656	0.0832	0.2966	−0.184	−0.860*	−0.545
		(7.37)	(6.01)	(1.6)	(0.54)	(0.51)	(−0.4)	(−1.8)	(−1.3)
sec		0.6122***	0.5535	0.2080*	0.3171***	0.3904	0.2473	−0.12	0.2173
		(4.29)	(4.34)	(1.7)	(2.8)	(1.59)	(1.36)	(−0.7)	(1.09)
ur		−0.289***	−0.123***	−0.173***	−0.272***	−0.570***	−0.240***	−0.306***	−0.480***
		(−8.2)	(−3.07)	(−5.3)	(−9.7)	(−5.9)	(−3.6)	(−5.4)	(−5.4)
open		0.044	−0.282*	−0.437***	0.2612	1.2437***	0.6646**	0.3518	1.125***
		(0.25)	(−1.68)	(−2.7)	(1.86)	(2.69)	(2.06)	(1.23)	(2.84)
Fixed effect		Control	Control	Control	Control	Control	Control	Control	Control
Number of Provinces		31	31	31	31	31	31	31	31
N		403	403	403	403	403	403	403	403
R^2		0.8465	0.8743	0.8865	0.9023	0.5108	0.7404	0.7771	0.6594

*, * *, and *** represented significance levels of 10%, 5% and 1% respectively.

It can be seen from Table 12 that the first-order lag term with explanatory variables was not significant and the coefficient was negative, indicating that the lag term cannot have an impact on the current quality of economic growth. At the same time, after using the instrumental variable method regression, the FDI and the ODI were still more significant, which showed that after considering the endogenous problem, the explanatory variables still had a positive impact on the explained variables.

Robustness test

Two methods of variable substitution and Bootstrap sampling were used to test the robustness of the model. The former replaces government income with per capita disposable income, while the latter takes 1000 self-service samples. The test results were shown in Table 13:

Table 13. Robustness test

samples	Replace variable		Bootstrap sampling
	FDI	ODI	FDI	ODI
Full samples	0.0068128***	0.0117465***	0.0071890***	0.011362***
	(2.78)	(5.00)	(5.37)	(5.51)
Eastern region	0.0141749*** (4.53)	0.0118277***	0.0125702***	0.0103735***
		(12.1)	(6.06)	(4.27)
Central region	−0.0032052	0.1011927*** (3.97)	−0.0114147	0.0953559*** (4.22)
	(−0.19)		(−1.16)
Western region	0.0066461**	−0.0021438	0.0083409*** (3.06)	−0.0046968
	(2.28)	(−0.10)		(−0.33)
Northeast region	0.007485**	0.0119022	0.0028004	0.0220558
	(5.05)	(0.33)	(1.23)	(0.67)

*, * *, and *** represented significance levels of 10%, 5% and 1% respectively.

According to the results of robustness test, the regression coefficient and significance level of independent variables of the FDI and the ODI were basically the same as before, which showed that the original regression had good robustness and strong persuasion ability.

VI. Conclusion and implication

This paper combined the mediating effect model and the panel data model to analyse the mechanism of the two-way flow of foreign direct investment and outbound direct investment on the quality of economic growth in China, and tested the heterogeneity of four regions in China, which were the east, the middle, the west and the northeast of China. The conclusion is as follows.

Both the FDI and the ODI can significantly improve the quality of economic growth, and the effect of the FDI on the quality of economic growth was stronger than that of the ODI. At the same time, there were obvious differences between the FDI and regional economic growth. Both the FDI and the ODI in eastern China can promote the quality of economic growth obviously. The promotion of the ODI in the central region was more significant, while that in the western region was significant. However, both the FDI and the ODI in northeast China had no significant effect on the quality of economic growth.

Further research showed that the mechanism of the FDI and the ODI to improve the quality of economic growth was mainly through the influence of employment, technology spillover and output efficiency. The FDI promoted the quality of economic growth through employment, output efficiency and technology spillover, while the ODI promoted the quality of economic growth mainly through output efficiency and technology reverse spillover. Similarly, the mediating effect had significant regional heterogeneity. The FDI and the ODI in the eastern region mainly relied on technology spillover to play a mediating effect. The ODI in the central region had an mediating effect through output efficiency and reverse technology spillover, while the FDI in the western region had an mediating effect through employment and output efficiency. Therefore, the employment effect was mainly reflected in the western region, the output efficiency was more obvious in the central and western regions, and technology spillover constituted the main mediating effect in the middle and eastern regions.

Therefore, China should continue to adhere to the strategy of ‘bringing in’ and ‘going out’ and continue to play the role of the FDI and the ODI in improving the quality of economic growth. Overall, we should adhere to the principle of high technology and high efficiency in the choice of foreign investment and outbound direct investment. Therefore, in the process of attracting foreign capital and investment, high-tech and high-efficiency foreign-funded joined ventures and foreign-funded entities should be given the highest priority. In terms of regions, the eastern region should encourage the introduction of high-tech industries and investment in high-tech industries in developed countries. The central region should encourage foreign investment, and the standard can be relaxed compared with the eastern region. The western region should speed up the pace of attracting foreign investment and focus on the ability of foreign-funded enterprises to promote employment and improve efficiency. The importance of introducing foreign capital and foreign investment can be reduced.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by the National Natural Science Foundation of China [71573170;71573050,72003123];The National Social Science Fund of China [21ZDA094];Science and Technology Commission of Shanghai Municipality [20692191900].