An Analysis of the Impact of International R&D Spillovers and Technology Innovation in China

Abstract

To test the impact of international R&D spillovers on China’s technological innovation, we collect and use province-country-level data from 23 provinces from 2001 to 2020 to fill a measurement gap of international R&D spillovers, so that our measurement can avoid ‘aggregation bias’ unlike the calculation methods in previous studies. We find that imports act as an effective international R&D spillover channel for improving technological innovation. Meanwhile, international R&D spillover through inward FDI and imports positively promotes China’s innovation performance. It suggests to policymakers that continuing to open up the economy and attract high-quality inward FDI is still required. Moreover, our results confirm that the eastern region enjoys a more significant international R&D spillover effect because of a more effective innovation environment. Hence, we suggest that given the inherent short board in the central and western regions, preferential policies to make up for this short board should be proposed to improve the innovation environment in the inland regions so as to enjoy a more significant international R&D spillover effect. Finally, we also observe that different periods of economic growth and the development levels of source countries also result in heterogeneous innovation impacts of international R&D spillover effects.

1. Introduction

In the endogenous growth theory, technological progress is regarded as an endogenous process. Commercially oriented innovation efforts are potential factors in technological progress [1,2]. It is also a source of productivity improvement and sustainable economic growth. As is well known, innovation is based on the accumulated knowledge of R&D experience over time [3]. Compared to developed countries, developing countries are less endowed with accumulated knowledge, but the knowledge endowment gap could be caught up by the direct channel of purchasing efforts in the technology market or by the indirect channel of international R&D spillovers. Thus, international R&D spillover is essential for industrialization and catch-up in developing countries [4].

Coe and Helpman [5] demonstrate that in a world of international trade in goods and services, FDI, and knowledge diffusion and exchange, one country’s productivity is affected not only by domestic R&D activities but also by the R&D efforts of other trading partners. Then, Aitken and Harrison [6], Potterie and Lichtenberg [7], Wang et al. [8], and Branstetter [9] illustrate that inward FDI is another effective channel for international R&D spillover, except for imports.

After China’s access to the WTO, changes in the scale of international trade and inward FDI have been remarkable. Not only has it brought dramatic economic growth, but the accompanying knowledge through FDI and imports has also influenced many aspects of China’s economy. However, as China gradually lost its comparative advantage in labor-intensive production activities, the economy entered a “new normal” mode. Then, the government’s task has focused on technological innovation, industrial restructuring, and the cultivation of new industries. Especially after 2012, an innovation-oriented development strategy began to be widely implemented in China.

As a technological latecomer, China has made multi-aspect efforts to catch up with the technological gap. In particular, a leap forward has been made in R&D inputs. For example, the ratio of R&D expenditures to GDP has exceeded 2% and has maintained a continuous growth trend since 2013 (National Science and Technology Expenditure Statistics Bulletin). According to the report on R&D personnel development status in 2019, the R&D personnel number in China has been the world’s No. 1 since 2008. Of course, the technological performances brought about by the soaring R&D investment are also remarkable, especially in patents, SCI papers, and high-tech industries.

Currently, China’s purchasing behaviors in the technology market are restricted due to environmental factors and trade pressure. It is more challenging to purchase behaviors in the technology markets. Conversely, domestic R&D activities and international R&D spillovers are less restricted by environmental factors and trade pressure to some extent, and their importance is more significant.

Therefore, our study aims to ascertain whether domestic R&D activities and international R&D spillovers affect technological innovation or not in China. In other words, it tests whether China’s opening-up policy can effectively promote technological innovation. Then, it also examines which aspects of knowledge creation, corporation innovation, innovation environment, innovation performance, and knowledge acquisition are affected by international R&D spillovers. Furthermore, assuming that the international R&D spillover effect exists, what channels of international R&D spillovers may affect technological innovation in China? That is, do imports and inward FDI act as effective international R&D spillover channels for improving China’s technological innovation?

Regarding the previous studies, due to the complexity of measuring knowledge capital and data availability, the existing studies all used country-level data, and the share of each province was used as a weighting value [10,11,12,13,14]. This would lead to ‘aggregation bias’ to measure international R&D spillover effects. Thus, to avoid overestimating the technological innovation promotion effect of international R&D spillovers, we calculate it more accurately with reference to the measurement from Lichtenberg and Pottelsberghe de la Potterie [15] and use 23 province-level data between 2001 and 2020. The data set begins with China’s access to the WTO until the latest available data. Thus, the first contribution of our study is to fill a measurement gap of international R&D spillover, which our measurement can avoid ‘aggregation bias.’

Second, compared to the existing studies, we comprehensively observe multiple aspects of heterogeneous innovation impact of international R&D spillover and domestic R&D activities in our study, in terms of the source country’s development level (developing and developed countries), technological innovation measurements (knowledge creation; knowledge acquisition; corporation innovation; innovation performance; innovation environment), inland and east regions, and stages of economic development (before and after 2010).

2. Literature Reviews

Regarding the relationship between international R&D spillovers and technological innovation, most of the previous research has mainly been based on developed countries [16,17,18,19,20]. However, it is hard to directly apply developed countries’ theories to developing countries’ situations.

Developing countries generally lag behind developed countries in technology innovation and management [21]. Krammer [4] indicates that imports are the main spillover channel for transition countries to promote productivity, which is similar to the conclusion of Seck [22]. Similarly, Ang and Madsen [23] point out that imports are probably the most important factor in increasing Asian miracle economies’ productivity. The GVC for emerging economies can pressure that the inter-firm linkages afforded by being part of a chain are crucial for transferring knowledge [24]. Although FDI-related spillover is significantly weakened, it is an essential source of know-how for transition countries [4,25].

Unlike the studies mentioned before, Salim et al. [26] argue that FDI could not act as a spillover channel directly. Nevertheless, when foreign subsidiaries’ technological capabilities in Iran act as mediating factors, FDI can be positively affected through two spillover channels: the demonstration effect and the training effect. With reference to the previous studies, FDI and imports, as the primary channels of international R&D spillover, can affect technological innovation through some paths, summarized in Figure 1.

As a developing country with a unique economic system, China’s technological innovation has also become a hot topic that has attracted much attention during the period of its development strategy transformation. Liu and Buck [29] indicate that international R&D spillover and indigenous efforts jointly determine the innovation performance of Chinese high-tech sectors, which is similar to the study of Liu et al. [30]. Li [31] draws a similar conclusion when studying the innovation capabilities of state-owned enterprises in China’s high-tech sectors. Similarly, Qin and Du [32] illustrate that the internal and external technology spillover effects improve innovation efficiency in China.

Since the potential offered by globalization and a liberal trade regime, the positive innovation impacts of international technology spillover can parallel indigenous innovation efforts when the region can satisfy the requirements of absorptive capacity, governance structures, the role of regional industrial specialization and diversity [8,33,34,35]. However, Feng and Li [11] put forward that inward FDI has no significance in improving innovation efficiency, but imports and outward FDI can. Chen et al. [36] interpret that in terms of the structure and quality of China’s FDI, international R&D spillover is not apparent, especially through the manufacturing activities of FDI enterprises.

Conversely, knowledge spillovers from imported products exhibit a medium- to long-term facilitation effect [37]. Seck [22] emphasizes that imports appear more conducive to R&D spillover. Developing countries that enjoy more enormous benefits tend to exhibit a more extensive stock of human capital, more openness to trade and foreign activities, and more powerful institutions. Furthermore, Sun and Du [38] denote that technology transfer from foreign countries and the domestic technology market plays an insignificant role in industrial innovation. Only in-house R&D is the driving force of China’s industrial innovation, similar to the conclusions of Huang et al. [34].

In general, we have not reached a consensus on the impact of international R&D spillover on technological innovation or productivity in different contexts, sectors, spillover channels, and innovation measurement. The relationship between international R&D spillover and technological innovation is still inconclusive. It is also an open question as to what channels international R&D spillovers affect technological innovation in China. Especially in previous studies with China as the context, due to the availability of data, there are still limitations in calculating international R&D spillovers. Thus, it is still worthy of in-depth exploration.

3. Methodology and Variables

3.1. Model Specification

To analyze the impact of R&D and knowledge spillover on productivity, Griliches [39] derived the Knowledge Production Function, in which the basic assumption is that the output of R&D investment and R&D capital is technological innovation. Then, based on the Knowledge Production Function, with reference to the studies of Hong et al. [12], Zhou et al. [14], and Zhang [40], we extend and develop our model below:

$${\mathrm{lninv}}_{\mathrm{it}}={\mathsf{\beta }}_0+{\mathsf{\beta }}_1{\mathrm{lnSD}}_{\mathrm{it}}+{\mathsf{\beta }}_2{\mathrm{lnSD}}_{\mathrm{it}}^{\mathrm{fdi}}+{\mathsf{\beta }}_3{\mathrm{lnSD}}_{\mathrm{it}}^{\mathrm{imp}}+{\mathsf{\beta }\mathsf{\chi }}_{\mathrm{it}}+{\mathsf{\epsilon }}_{\mathrm{it}}$$

(1)

where ${\mathrm{inv}}_{\mathrm{it}}$ is the proxy variable for the technological innovation of province i in year t. ${\mathrm{SD}}_{\mathrm{it}}$ denotes domestic R&D capital stock of province i in year t. ${\mathrm{SD}}_{\mathrm{it}}^{\mathrm{fdi}}$ represents province i gains from the international R&D spillover through FDI inflows from country j in year t. ${\mathrm{SD}}_{\mathrm{it}}^{\mathrm{imp}}$ measures that province i gains from international R&D spillover through imports from country j in year t. ${\mathsf{\chi }}_{\mathrm{it}}$ is a collection of the control variables. $\mathsf{\epsilon }$ represents the error term. We take the logarithm of all variables.

Most previous studies used patent-related variables to measure technological innovation [14,41,42,43,44]. Although most widely used, it has certain disadvantages, especially in the Chinese context. Xiao and Xie [44] point out that the composition of Chinese patents is less distributed in invention and that most of them are applied by foreign affiliates. In addition, only the number of patents is distinguished, and it is impossible to distinguish the heterogeneity of individual importance. They put forward that new product sales are a more appropriate measure of technological innovation in China because they reflect market demand quickly, thereby conveying the value of innovations [8].

Of course, except for the proxy variables mentioned earlier, some studies use multiple variables to form a composite indicator [45,46]. Different methods are used to calculate the composite indicator, such as the weighted average, entropy weighting method, and principal component analysis. Each calculation method has its advantages and disadvantages.

To address these issues, the proxy variable used in our study for measuring technological innovation will be the comprehensive indicator of the regional innovation index issued by the Chinese Ministry of Science and Technology. The index includes all the variables above and serves as a calculated indicator to evaluate regional technological innovation capacity. It contains 20 secondary indicators below five primary indicators. Among them, the knowledge creation indicator evaluates a region’s ability to create new knowledge. The knowledge acquisition indicator reflects the ability to utilize globally helpful knowledge. The corporation innovation indicator evaluates regional corporations’ level of applying new knowledge or launching new products. The innovation environment reflects the ability to provide a responsive environment for knowledge formation, movement, and application, and the innovation performance demonstrates the region’s innovation output capability [47].

3.2. Measurement of International R&D Spillovers and Control Variables

We will use the perpetual inventory method to calculate foreign R&D capital stock. In the first step, because our data set is from 2001 to 2020, we should calculate the R&D capital stock of a foreign country in the base year 2001:

$${\mathrm{SD}}_{\mathrm{j}.2001}^{\mathrm{d}}=\frac{{\mathrm{SD}}_{\mathrm{j},2001}}{\mathsf{\delta }+\mathrm{g}}$$

(2)

where ${\mathrm{SD}}_{\mathrm{j},2001}$ denotes the R&D expenditure of country j in 2001, and ${\mathrm{SD}}_{\mathrm{j},2001}^d$ is the R&D capital stock of country j in 2001. Moreover, $\mathsf{\delta }$ is the depreciation rate; we take the value of 9.6% with reference to Huang et al. [34]; $\mathrm{g}$ is the average growth of R&D expenditure of country j from 2001 to 2020. Then,

$${\mathrm{SD}}_{\mathrm{jt}}^d=\left(1-\mathsf{\delta }\right){\mathrm{SD}}_{\mathrm{j},\mathrm{t}-1}^{\mathrm{d}}+{\mathrm{RD}}_{\mathrm{jt}}$$

(3)

where ${\mathrm{SD}}_{\mathrm{jt}}^{\mathrm{d}}$ is the R&D capital stock of country j in year t, ${\mathrm{RD}}_{\mathrm{jt}}$ is the R&D expenditure of country j in year t. The domestic R&D stock of each province in China is calculated as the same as the foreign R&D stock mentioned before. According to the studies of Chen et al. [10], Feng and Li [11], Hong et al. [12], Li et al. [13], and Zhou et al. [14], we find that due to data availability, they used country-level data and then the ratio of each province as a weighting value to measure the international R&D spillover effect of each province. However, with reference to Lichtenberg and Pottelsberghe de la Potterie [15], we use province-level data from provincial statistical yearbooks to calculate more accurately.

$${\mathrm{SD}}_{\mathrm{it}}^{\mathrm{fdi}}={\displaystyle \sum }_{\mathrm{j}}\frac{{\mathrm{fdi}}_{\mathrm{ijt}}}{{\mathrm{k}}_{\mathrm{jt}}}{\mathrm{SD}}_{\mathrm{jt}}^{\mathrm{d}}$$

(4)

$${\mathrm{SD}}_{\mathrm{it}}^{\mathrm{imp}}={\displaystyle \sum }_{\mathrm{j}}\frac{{\mathrm{imp}}_{\mathrm{ijt}}}{{\mathrm{y}}_{\mathrm{jt}}}{\mathrm{SD}}_{\mathrm{jt}}^{\mathrm{d}}$$

(5)

where ${\mathrm{SD}}_{\mathrm{it}}^{\mathrm{fdi}}$ represents province i gains from the international R&D spillover through FDI inflows from country j in year t. ${\mathrm{fdi}}_{\mathrm{ijt}}$ denotes FDI flows from country j to province i in year t, ${\mathrm{k}}_{\mathrm{jt}}$ is the gross capital formation of country j in year t. ${\mathrm{SD}}_{\mathrm{it}}^{\mathrm{imp}}$ measures that province i gains from the international R&D spillover through imports from country j in year t. ${\mathrm{imp}}_{\mathrm{ijt}}$ is the imports from country j to province i in year t, ${\mathrm{y}}_{\mathrm{jt}}$ is the gross domestic product of country j in year t.

Moreover, Ang and Madsen [23] stated that countries with similar economic growth experiences exhibit clustering characteristics geographically. Most of these clustered countries have similar cultures, a high frequency of international trade, and high levels of cross-border labor mobility. Therefore, the geographical distance between trading partners has been emphasized in studies of international trade and economic growth. The distance between trading partner countries not only reflects transportation costs but also has the potential to be an unfamiliarity or information barrier. Keller [27] pointed out that geographical proximity allows for non-public activities. Knowledge can be transferred through meetings, lectures, visits, and seminars; thus, geographical proximity positively correlates with knowledge spillover.

$${\mathrm{IRS}}_{\mathrm{t}}^{\mathrm{prox}}={\displaystyle \sum }_{\mathrm{j}}\sqrt{\left({\mathrm{TD}}_{\mathrm{i}}/{\mathrm{D}}_{\mathrm{ij}}\right)}{\mathrm{SD}}_{\mathrm{jt}}^{\mathrm{d}}$$

(6)

where ${\mathrm{TD}}_{\mathrm{i}}$ is the total geographical distance between Beijing city with all other foreign countries in kilometers; ${\mathrm{D}}_{\mathrm{ij}}$ denotes the geographical distance between Beijing city with the capital of country j in kilometers.

In addition, human capital is regarded as the essential input for technological innovation. We use the average years of schooling in each province to measure it. The grade levels are considered years of education. When entering a grade, all previous grades are recognized as completed. The coefficients are given below. Nineteen years for a master’s degree and above; 16 years for college; 12 years for high school; 9 years for junior high school students; 6 years for elementary school; and 0 years if no schooling. The population is the population aged six years and above. Note that all variables are constant prices in US dollars during the year 2003.

Considering that there are significant differences in innovation level and the scale of FDI, international trade between the eastern coastal and inland regions of China, we control for them using dummy variables. In other words, the eastern regions (Shanghai, Beijing, Shandong, Guangdong, Jiangsu, Hebei, Zhejiang, and Fujian) take a value of 1, and the rest of the regions take a value of 0. Then, the descriptive statistics are reported in

Table 1. Descriptive statistics.

Variable	Definition	Obs	Mean	Std. Dev.	Min	Max
lninv	Technological innovation	459	3.335	0.352	2.709	4.129
${\mathrm{lnSD}}^{\mathrm{fdi}}$	International R&D spillover through inward FDI	441	19.282	2.62	8.911	26.659
${\mathrm{lnSD}}^{\mathrm{imp}}$	International R&D spillover through imports	449	20.833	1.971	15.78	24.925
lnSD	Domestic R&D stock	460	13.276	1.613	8.806	16.564
lnedu	average years of schooling in each province	460	2.154	0.123	1.798	2.548
lngeo	International R&D spillover through geographical proximity	460	31.898	0.204	31.525	32.144
east	East regions:1; inland regions:0	460	0.346	0.476	0	1

below.

3.3. Econometric Issue

Technological innovation is interpreted as a process of knowledge accumulation based on the accumulated knowledge of R&D experience over time [3,48]. Hence, we introduce the lag variable of the dependent variable. It is important to consider the dependent variable’s internal factors and time series influence.

In general, there is a correlation between the lag variables of the dependent variables and the fixed effects included in the error term, which may lead to “dynamic panel bias” and inconsistency problems if OLS is used. In other words, there may be a potential endogeneity problem between the error term and the lag variables of the dependent variable [49]. Moreover, it cannot be confirmed that the other explanatory variables used are strictly exogenous. The system GMM method is proposed to be used in our study.

As Aitken and Harrison [6] pointed out, there exists an endogeneity problem in estimating the technological innovation effect of international R&D spillovers through FDI and imports due to the often-distributed cluster in some regions with high innovation output. We follow Lai et al. [50] and Zhang et al. [51] to lag one year to reduce the possible endogeneity problem. Then, the model is specified below:

$${\mathrm{lninv}}_{\mathrm{it}}={\mathsf{\beta }}_0+{\mathsf{\beta }}_1{\mathrm{lninv}}_{\mathrm{it}-1}+{\mathsf{\beta }}_2{\mathrm{lnSD}}_{\mathrm{it}-1}+{\mathsf{\beta }}_3{\mathrm{lnSD}}_{\mathrm{it}-1}^{\mathrm{fdi}}+{\mathsf{\beta }}_4{\mathrm{lnSD}}_{\mathrm{it}-1}^{\mathrm{imp}}+{\mathsf{\beta }\mathsf{\chi }}_{\mathrm{it}-1}+{\mathsf{\epsilon }}_{\mathrm{it}}$$

(7)

The system GMM method combines a level GMM equation and a difference GMM equation, which has several advantages in estimating dynamic panels: (1) It can control for country-specific effects that vary over time. (2) It can address the endogeneity problem caused by lag-dependent variables. (3) The other explanatory variables allow for some degree of endogeneity problems.

In the level GMM equation, the lag variable of the difference variable can be used as an instrument variable. In the difference GMM equation, the lag variable of the level variable can be used as an instrument variable. Therefore, it has the advantage of improving estimation efficiency and allows the estimation of time-invariant variables compared to the difference GMM because the estimation of the level GMM equations is also included in the system GMM.

In order to identify the efficiency of the system GMM estimators, two tests are required. The first is to test the autocorrelation of the error terms. The null hypothesis should be rejected in the first period and not in the second period. A prerequisite for obtaining consistent estimates in the difference GMM equations is that there should be no autocorrelation in the error terms. Thus, the first-level difference variables of the error term remain autocorrelated in the first period and should not be autocorrelated in the second or more periods.

Next is the over-identification test. Suppose the over-identification test rejects the null hypothesis that all instrument variables are valid. In that case, the added instrumental variables are correlated with the error term and cannot be considered valid instrument variables. Therefore, to determine the validity of the instrument variables, the Hansen or Sargan test should not reject the null hypothesis [49].

4. Empirical Results

4.1. Overall Level

Before the empirical analysis, a panel unit root test was performed on our dataset to avoid spurious regressions. Since it is an unbalanced panel, we obtained the result of no unit root for all variables using the Fisher-type test.

Table 2. Overall empirical results.

	(1)	(2)	(3)	(4)	(5)	2001– 2010	2011– 2020	(8)	(9)	(10)	(11)
	Panel OLS	FE	System GMM	System GMM	System GMM	System GMM	System GMM	System GMM	System GMM	System GMM	System GMM
L.lninv			0.570 ***	0.522 ***	0.918 ***	0.549 ***	0.520 ***	0.571 ***	0.527 ***	0.786 **	0.768 ***
			(0.11)	(0.08)	(0.04)	(0.09)	(0.12)	(0.08)	(0.08)	(0.38)	(0.08)
L.${\mathrm{lnSD}}^{\mathrm{imp}}$	0.072 ***	−0.006		0.046 **	0.014 **	0.069 ***	0.016	0.021 *		−0.003	−0.047
	(0.01)	(0.01)		(0.02)	(0.01)	(0.03)	(0.02)	(0.01)		(0.07)	(0.06)
L.${\mathrm{lnSD}}^{\mathrm{fdi}}$	−0.006	0.000	0.009		0.003	0.007	−0.006		−0.000	0.006	0.007
	(0.00)	(0.00)	(0.01)		(0.01)	(0.03)	(0.01)		(0.00)	(0.01)	(0.01)
L.lnSD	0.130 ***	0.196 ***	0.083 ***	0.046 *	−0.012	0.003	0.092 **	0.066 **	0.050 **	0.001	0.052 *
	(0.02)	(0.07)	(0.03)	(0.03)	(0.01)	(0.02)	(0.04)	(0.03)	(0.02)	(0.06)	(0.03)
L.lngeo	−1.023 ***	−1.133 ***	−0.378 **	−0.629 ***	−0.030 ***	0.265	−0.478 *	−0.441 ***	−0.459 ***	0.017	0.352
	(0.06)	(0.34)	(0.18)	(0.13)	(0.01)	(0.22)	(0.26)	(0.15)	(0.14)	(0.05)	(0.54)
L.lnedu	0.504 ***	0.965 ***	0.188	0.207	0.463 ***	0.208 *	0.244	0.154	0.321 **	0.062	−1.247
	(0.10)	(0.20)	(0.16)	(0.15)	(0.14)	(0.11)	(0.20)	(0.15)	(0.16)	(0.61)	(1.30)
L.${\mathrm{lnSD}}^{\mathrm{fdiing}}$								−0.001
								(0.00)
L.${\mathrm{lnSD}}^{\mathrm{fdied}}$								0.000
								(0.02)
L.${\mathrm{lnSD}}^{\mathrm{imping}}$									−0.011
									(0.01)
L.${\mathrm{lnSD}}^{\mathrm{imped}}$									0.039 **
									(0.02)
east#L.${\mathrm{lnSD}}^{\mathrm{imp}}$										0.005
										(0.01)
east#L.${\mathrm{lnSD}}^{\mathrm{fdi}}$											0.011
											(0.01)
_cons	31.782 ***	34.866 ***	11.804 **	19.661 ***	0.000	−8.900	14.832 *	13.871 ***	14.252 ***	0.000	−7.621
	(1.88)	(9.92)	(5.65)	(4.14)	(0.00)	(6.74)	(8.00)	(4.46)	(4.21)	(0.00)	(13.79)
N	421	421	389	396	395	173	205	395	386	399	399
Adj R-sq	0.79	0.33
AR(1)			0.000	0.000	0.001	0.002	0.001	0.000	0.000	0.061	0.001
AR(2)			0.478	0.837	0.992	0.147	0.739	0.809	0.787	0.931	0.715
Hansen			0.146	0.698	0.901	0.357	0.219	0.634	0.771	0.685	0.997
Sargan			0.253	0.275	0.742	0.482	0.343	0.639	0.485	0.447	0.257

Notes: 1. Robust standard errors are in parentheses, * p < 0.10, ** p < 0.05, *** p < 0.01. 2. ${\mathrm{lnSD}}^{\mathrm{fdied}}$: International R&D spillover through inward FDI from developed countries; ${\mathrm{lnSD}}^{\mathrm{fdiing}}$: International R&D spillover through inward FDI from developing countries; ${\mathrm{lnSD}}^{\mathrm{imped}}$: International R&D spillover through imports from developed countries; ${\mathrm{lnSD}}^{\mathrm{imping}}$: International R&D spillover through imports from developing countries; 3. The time fixed effect is controlled in system GMM. To reduce the number of instrument variables, we use ‘collapse’ demand in Stata.

reports the empirical results of the relationship between international R&D spillovers and technological innovation using data from 2001 to 2020 for 23 Chinese provinces.

Comprehensively observing the results of system GMM in columns (4) and (5) of Table 2, we find that international R&D spillover through imports statistically significantly facilitates technological innovation in China. That is, imports are an effective, positive channel of international R&D spillover to stimulate China’s technological innovation. The results are similar to Krammer [4], Feng and Li [11], Zhou et al. [14], Ang and Madsen [23], Liu and Buck [29].

The international division of the production process is an essential channel for technology transfer [42,43]. As mentioned before, imports can generate international R&D spillover through the following channels. The technology contained in imported intermediate goods can provide learning opportunities for importer countries, commonly referred to as the “learning-by-doing” effect. Its positive spillover effect will be more pronounced, especially when importing high-tech products or high-quality production facilities [13]. In addition, domestic firms can re-innovate based on the absorption of imported products. In addition, introducing imported products will pressure domestic firms to innovate to remain competitive.

China’s economic growth rate remained high until 2010. Except for the period before 2003, and 2009, the GDP growth rate was greater than 10% in the remaining periods. As the GDP growth rate has become slower since 2011, China has tried to change from an extensive growth pattern to an intensive one and has sought to improve productivity simultaneously through technological innovation. Therefore, we use 2010 as the time boundary to distinguish between China’s rapid and stable economic growth periods.

According to our analysis results in columns (6) and (7) of Table 2, international R&D spillover through imports positively affects technological innovation in China during the rapid growth period. However, only domestic R&D capital stock is statistically significant during the stable growth period. We interpret that the sharp increase in the scale of international trade and FDI inflows after China accessed the WTO is accompanied by foreign knowledge transfer. This would increase the knowledge endowment in China and contribute to spontaneous R&D activities in the future.

Moreover, with the technology gap with advanced countries gradually narrowing, China will not enjoy the ‘advantage of backwardness’ [44,52]. Meanwhile, the international economic environment for China is changing during this period. Therefore, the impact of international R&D spillover through imports on China’s technological innovation changed around 2010. This result is similar to that of Yang et al. [53], who consider the economic development level as the absorptive capacity of China’s labor-intensive manufacturing sector for international R&D spillover.

Unfortunately, the effects of international R&D spillover through inward FDI are unclear and statistically insignificant. The result is similar to Chen et al. [36]. There are several reasons for this. First, the purpose of MNCs setting up subsidiaries overseas is not to transfer their technological advantage to local firms, but to even have incentives to protect their technological advantage. Furthermore, MNCs may take protection measures for their technological advantage in order to maintain a comparative advantage in the market competition with local Chinese enterprises continually growing after the reform and opening up. Thus, international R&D spillover through inward FDI may have an insignificant impact on technological innovation in China. Note that international R & D spillover through inward FDI has no statistically significant impact on the composite indicator-regional innovation capacity indicator. This may affect other aspects of innovation.

Pietrucha and Zelelazny [54] argue that the direction of trade affects the extent of international R&D spillover. There may be heterogeneity in knowledge endowment in innovation systems among different development-level countries; thus, international R&D spillover from different development-level countries may lead to a heterogeneous impact on technological innovation. In our study, FDI and imports are classified into groups of developed and developing countries to observe the heterogeneous impacts, as shown in columns (8) and (9). The results illustrate that imports from developed countries are an essential channel of international R&D spillover for China.

As a latecomer in technological innovation, China still has gaps in various technology fields with advanced countries since it has achieved many incredible innovation performances in recent years. Consequently, many policy efforts have been made to catch up with these existing technological gaps in advanced countries. For instance, the meaning of the “market for technology” strategy is to open the domestic market for imports or attract FDI, inducing foreign enterprises to technology or knowledge transfer and diffusion, thereby strengthening domestic R&D capabilities by learning and applying advanced foreign technologies. The ultimate goal of the strategy is to improve China’s technological innovation level.

Meanwhile, according to the 2019 China Import Development Report, China is still in a seller’s international market for high-quality goods. Although some developed countries have also introduced restrictive policies on exports to China, the share of European and North American products in China’s imports has gradually increased. This underscores the urgency of domestic supply-side reforms. Thus, a virtuous cycle of introducing, learning, absorbing, applying, and re-innovating foreign technologies can be constituted by importing products from developed countries. The formation of a virtuous circle effectively improves effective domestic supply and ultimately realizes supply-side reform and sustainable development.

In addition, Chinese importers have access to more diverse and superior goods at lower prices than non-importers, thereby increasing their profits (economies of scale effect) and investing more in R&D activities. Additionally, pressure from competition from imported goods can be seen as a driving force to stimulate domestic R&D activities.

Finally, dummy variables were used to control for heterogeneity across regions within China [11]. The results reported in columns (10) and (11) confirm that the heterogeneous impact of international R&D spillovers through imports and FDI on technological innovation between eastern and inland regions is insignificant. For the other control variables, we find that the signs of control variables with statistical significance were consistent. In other words, an increase in human capital and a decrease in international R&D spillovers through geographic proximity promote technological innovation in China.

4.2. Regional Innovation Index Classification

4.2.1. Innovation Performance Results

Innovation performance is an index used to evaluate regional innovation output capacity. The analysis results of the overall level in

Table 3. Innovation performance results.

	(1)	(2)	(3)	(4)	(5)	2001– 2010	2011– 2020	(8)	(9)	(10)	(11)
	Panel OLS	FE	System GMM	System GMM	System GMM	System GMM	System GMM	System GMM	System GMM	System GMM	System GMM
L.lninvper			0.639 ***	0.342 ***	0.464 ***	0.512 ***	0.394 **	0.258 **	0.267 **	0.270 **	0.251 ***
			(0.15)	(0.10)	(0.08)	(0.11)	(0.16)	(0.11)	(0.11)	(0.11)	(0.08)
L.${\mathrm{lnSD}}^{\mathrm{fdi}}$	−0.006	0.010 *	0.029 **		0.007 *	−0.004	−0.013		0.001	−0.004	−0.002
	(0.01)	(0.01)	(0.01)		(0.00)	(0.01)	(0.04)		(0.01)	(0.04)	(0.00)
L.${\mathrm{lnSD}}^{\mathrm{imp}}$	0.103 ***	0.021		0.044 ***	0.061 **	0.086 **	0.059 *	0.061 ***		0.049 **	0.067 **
	(0.01)	(0.01)		(0.01)	(0.02)	(0.04)	(0.03)	(0.02)		(0.02)	(0.03)
L.lnedu	0.349 ***	0.309	−0.078	0.086	−0.168	0.125	−0.055	0.150	0.097	0.181	−0.408
	(0.13)	(0.33)	(0.16)	(0.20)	(0.42)	(0.17)	(0.14)	(0.25)	(0.32)	(0.24)	(1.10)
L.lnSD	0.054 **	0.285 ***	0.030	0.067 ***	0.011	−0.006	0.063	0.092	0.085 *	0.061	0.059
	(0.02)	(0.05)	(0.02)	(0.02)	(0.03)	(0.03)	(0.12)	(0.06)	(0.04)	(0.06)	(0.05)
L.lngeo	−0.187**	−0.623 **	0.015 *	−0.215 *	0.021	−0.792 ***	0.176	−0.453	−0.323	−0.282	−0.126
	(0.08)	(0.26)	(0.01)	(0.13)	(0.03)	(0.21)	(0.57)	(0.29)	(0.20)	(0.35)	(0.17)
L.${\mathrm{lnSD}}^{\mathrm{fdiing}}$								−0.000
								(0.00)
L.${\mathrm{lnSD}}^{\mathrm{fdied}}$								−0.015
								(0.04)
L.${\mathrm{lnSD}}^{\mathrm{imping}}$									0.006
									(0.02)
L.${\mathrm{lnSD}}^{\mathrm{imped}}$									0.035
									(0.03)
east#L.${\mathrm{lnSD}}^{\mathrm{imp}}$										0.003
										(0.00)
east#L.${\mathrm{lnSD}}^{\mathrm{fdi}}$											0.004
											(0.00)
_cons	5.964 **	18.172 **	0.000	7.165 *	0.000	24.962 ***	−5.247	14.541	10.680 *	9.388	5.324
	(2.52)	(7.54)	(0.00)	(3.85)	(0.00)	(6.64)	(18.28)	(9.19)	(6.03)	(10.58)	(4.61)
N	422	422	391	407	397	174	210	387	388	387	383
Adj R-sq	0.63	0.50
AR(1)			0.001	0.000	0.000	0.004	0.005	0.002	0.001	0.001	0.000
AR(2)			0.149	0.129	0.115	0.982	0.125	0.362	0.274	0.365	0.260
Hansen			0.280	0.824	0.698	0.832	0.489	0.147	0.101	0.338	0.964
Sargan			0.370	0.838	0.880	0.914	0.196	0.157	0.110	0.154	0.475

Notes: 1. Robust standard errors are in parentheses, * p < 0.10, ** p < 0.05, *** p < 0.01. 2. ${\mathrm{lnSD}}^{\mathrm{fdied}}$: International R&D spillover through inward FDI from developed countries; ${\mathrm{lnSD}}^{\mathrm{fdiing}}$: International R&D spillover through inward FDI from developing countries; ${\mathrm{lnSD}}^{\mathrm{imped}}$: International R&D spillover through imports from developed countries; ${\mathrm{lnSD}}^{\mathrm{imping}}$: International R&D spillover through imports from developing countries; 3. The time fixed effect is controlled in system GMM. To reduce the number of instrument variables, we use ‘collapse’ demand in Stata.

show a statistically robust innovation stimulation effect of international R&D spillover through imports and inward FDI. The results are in line with Zhou et al. [14], Liu and Buck [29], Ito et al. [55], Zhang [40], and Goel [56]. In interpreting how imports affect innovation performances, except for the economies of scale effect mentioned before, importing knowledge-intensive intermediate goods would induce the demand for skilled labor in China, thereby facilitating later participation probability in knowledge-intensive activities (innovation cluster effect).

Although the recent rise in China’s innovation performance has been rapid, core technologies are still dependent on developed countries. The most important part of China’s imports from developed countries is core technology and intellectual property rights. In addition, with the increasing barriers to technology imports, developed countries have put much trade pressure on exporting technology and M&A to China due to national security and other reasons. As a result, the difficulties and costs of importing core and cutting-edge technologies continue to increase. Nonetheless, taking into account the positive international R&D spillover through imports, China still needs to actively engage in technology imports.

The composition of patents in China is mainly based on design and utility models, and the proportion of inventions is relatively low. Among them, most of the applicants for invention patents are from foreign subsidiaries in China. Thus, it can be proven that FDI is an effective channel of international R&D spillovers for promoting innovation performance. Unfortunately, there are no robustness results for the remaining variables.

Then, we observe some differences between the rapid and stable growth periods reported in columns (6) and (7) of Table 3. The factor influencing the innovation performance of Chinese provinces was international R&D spillover through imports, but the coefficients have some change. We find that the positive effect of international R&D spillover through imports decreases. It satisfies the marginal diminishing effect of the spillover effect.

4.2.2. Innovation Environment Results

In

Table A3. Innovation environment results.

	(1)	(2)	(3)	(4)	(5)	2000– 2010	2011– 2020	(8)	(9)	(10)	(11)
	Panel OLS	FE	System GMM	System GMM	System GMM	System GMM	System GMM	System GMM	System GMM	System GMM	System GMM
L.lninver			0.237 **	0.196 *	0.700 ***	0.232 *	0.385 ***	0.256 ***	0.397 ***	0.172 **	0.201 ***
			(0.09)	(0.10)	(0.23)	(0.14)	(0.10)	(0.09)	(0.12)	(0.09)	(0.07)
L.${\mathrm{lnSD}}^{\mathrm{fdi}}$	−0.007	−0.006	−0.000		0.027	−0.008	−0.003		0.165***	−0.041	0.004
	(0.01)	(0.00)	(0.01)		(0.07)	(0.01)	(0.02)		(0.05)	(0.08)	(0.01)
L ${\mathrm{lnSD}}^{\mathrm{imp}}$	0.051 ***	−0.026 *		0.025	0.022 *	0.049 **	0.008	0.063		0.012	0.007
	(0.01)	(0.01)		(0.02)	(0.01)	(0.02)	(0.01)	(0.05)		(0.01)	(0.03)
L.lnSD	0.132 ***	0.094	0.136 ***	0.118 ***	−0.006	0.100 ***	0.111*	0.042	0.028	0.163	0.098 ***
	(0.02)	(0.09)	(0.03)	(0.03)	(0.06)	(0.03)	(0.06)	(0.06)	(0.06)	(0.11)	(0.03)
L.lnedu	0.384 ***	0.795 **	0.388	0.464 *	0.223 *	0.309	0.429	0.227	0.295	0.423 **	0.373 **
	(0.11)	(0.37)	(0.24)	(0.25)	(0.12)	(0.24)	(0.37)	(0.28)	(0.20)	(0.20)	(0.19)
L.lngeo	−0.995 ***	−0.557	−0.965 ***	−0.983 ***	−0.012	−1.039 ***	−1.057***	−0.549 ***	0.008	−1.069 **	−0.743 ***
	(0.07)	(0.47)	(0.20)	(0.14)	(0.01)	(0.22)	(0.36)	(0.21)	(0.01)	(0.42)	(0.18)
L ${\mathrm{lnSD}}^{\mathrm{fdiing}}$								0.000
								(0.00)
L ${\mathrm{lnSD}}^{\mathrm{fdied}}$								0.004
								(0.01)
L.${\mathrm{lnSD}}^{\mathrm{imping}}$									−0.039
									(0.03)
L ${\mathrm{lnSD}}^{\mathrm{imped}}$									−0.087
									(0.09)
east#cL ${\mathrm{lnSD}}^{\mathrm{imp}}$										0.006 **
										(0.00)
1.east#cL.${\mathrm{lnSD}}^{\mathrm{fdi}}$											0.007 **
											(0.00)
_cons	31.544 ***	18.712	30.621 ***	30.985 ***	0.000	32.879 ***	33.311 ***	17.498 ***	0.000	34.204 **	23.919 ***
	(2.20)	(13.44)	(6.16)	(4.38)	(0.00)	(6.83)	(11.03)	(6.47)	(0.00)	(13.66)	(5.60)
N	422	422	406	407	387	177	223	396	377	387	383
Adj R-sq	0.67	0.30
AR(1)			0.000	0.000	0.002	0.004	0.000	0.000	0.066	0.043	0.000
AR(2)			0.433	0.307	0.272	0.315	0.495	0.513	0.255	0.606	0.801
Hansen			0.434	0.238	0.135	0.100	0.265	0.724	0.495	0.367	0.707
Sargan			0.198	0.066	0.101	0.127	0.206	0.702	0.297	0.264	0.116

Notes: 1. Robust standard errors are in parentheses, * p < 0.10, ** p < 0.05, *** p < 0.01. 2. ${\mathrm{lnSD}}^{\mathrm{fdied}}$: International R&D spillover through inward FDI from developed countries; ${\mathrm{lnSD}}^{\mathrm{fdiing}}$: International R&D spillover through inward FDI from developing countries; ${\mathrm{lnSD}}^{\mathrm{imped}}$: International R&D spillover through imports from developed countries; ${\mathrm{lnSD}}^{\mathrm{imping}}$: International R&D spillover through imports from developing countries; 3. The time fixed effect is controlled in system GMM. To reduce the number of instrument variables, we use ‘collapse’ demand in Stata.

, we report the results of the analysis of the impact of international R&D spillovers on the innovation environment. The results show no robust, statistically significant results in testing the impact of international R&D spillovers on China’s innovation environment at the overall level in columns (3) to (5) of Table A3.

Although international R&D spillover through imports is not statistically significant at the overall level, it facilitated the innovation environment in China between 2001 and 2010. The increase in imports could improve the efficiency of customs clearance of goods and the facilitation of the cross-border movement of service trade, which is conducive to enhancing China’s domestic business environment. In addition, imports can constitute an open, fair, and competitive market order and promote the free flow of capital, labor, technology, and other production factors among the international community. An advanced management system can also provide opportunities for Chinese enterprises to adapt to international market norms [57].

In addition, we find that domestic R&D capital stock is positively associated with the improvement of China’s innovation environment during the rapid and stable growth period. This may be because domestic firms are more likely to access and disseminate the innovations generated from domestic R&D activities, thereby building an effective innovation environment and reward system for innovation performance.

Considering the heterogeneity across regions, our results confirm that international R&D spillovers more significantly affect the innovation environment in the eastern region than in the central and western regions. As an earlier region to open up, the east is more excellent than the west and central regions in terms of marketization level and business environment. Meanwhile, the incentive system for innovations in the east with a higher economic level has been in place for a long time, contributing to the active cooperation between industry and academia. Local governments, which have more financial resources, have also introduced many support policies for innovation development.

4.2.3. Corporation Innovation Results

Corporation innovation indicators evaluate the ability of regional corporations to apply new knowledge or launch new products, and the results are reported in

Table A4. Corporation innovation results.

	(1)	(2)	(3)	(4)	(5)	2000– 2010	2011– 2020	(8)	(9)	(10)	(11)
	Panel OLS	FE	System GMM	System GMM	System GMM	System GMM	System GMM	System GMM	System GMM	System GMM	System GMM
L.lninvcor			0.654 **	0.461 ***	0.553 ***	0.214	0.626 ***	0.609 ***	0.364 ***	0.435 ***	0.581 ***
			(0.26)	(0.06)	(0.10)	(0.16)	(0.14)	(0.20)	(0.08)	(0.10)	(0.07)
L.${\mathrm{lnSD}}^{\mathrm{fdi}}$	0.012	0.000	0.006		0.137 **	0.071 **	−0.036		0.012 **	0.051 ***	−0.006
	(0.01)	(0.01)	(0.01)		(0.05)	(0.03)	(0.06)		(0.00)	(0.02)	(0.01)
L.${\mathrm{lnSD}}^{\mathrm{imp}}$	0.067 ***	−0.021		0.056 *	0.037	0.069 **	−0.051	0.012		0.022	−0.047
	(0.02)	(0.03)		(0.03)	(0.11)	(0.03)	(0.09)	(0.07)		(0.02)	(0.06)
L.lnSD	0.178 ***	0.377 ***	0.095	0.078	−0.151	0.020	0.260 *	−0.047	0.042	0.009	−0.077
	(0.03)	(0.12)	(0.07)	(0.06)	(0.13)	(0.04)	(0.13)	(0.06)	(0.05)	(0.04)	(0.11)
L.lnedu	−0.218	1.450 ***	−0.195	−0.077	−0.137	−0.037	−0.148	−0.414	−0.207	−0.230	5.331
	(0.15)	(0.41)	(0.18)	(0.22)	(0.41)	(0.23)	(0.39)	(0.28)	(0.24)	(0.19)	(3.76)
L.lngeo	−1.494 ***	−2.754 ***	−0.765 *	−0.716 ***	0.011	−1.007 ***	−0.860	0.373	−0.575 **	−0.077	−1.075 *
	(0.10)	(0.60)	(0.41)	(0.25)	(0.03)	(0.31)	(0.70)	(0.52)	(0.28)	(0.29)	(0.56)
L.${\mathrm{lnSD}}^{\mathrm{fdiing}}$								−0.008
								(0.01)
L.${\mathrm{lnSD}}^{\mathrm{fdied}}$								0.101 **
								(0.05)
L.${\mathrm{lnSD}}^{\mathrm{imping}}$									−0.016
									(0.02)
L.${\mathrm{lnSD}}^{\mathrm{imped}}$									0.092 ***
									(0.03)
east#L.${\mathrm{lnSD}}^{\mathrm{imp}}$										0.007 **
										(0.00)
east#L.${\mathrm{lnSD}}^{\mathrm{fdi}}$											0.004
											(0.00)
_cons	47.539 ***	83.551 ***	24.593 *	22.617 ***	0.000	31.768 ***	27.266	−11.114	18.568 **	3.289	26.339 **
	(2.97)	(17.32)	(13.16)	(7.92)	(0.00)	(9.87)	(22.01)	(16.15)	(8.54)	(9.08)	(12.72)
N	422	422	406	402	397	176	215	383	391	387	397
Adj R-sq	0.68	0..43
AR(1)			0.019	0.000	0.001	0.006	0.038	0.011	0.000	0.000	0.000
AR(2)			0.969	0.744	0.426	0.818	0.708	0.105	0.184	0.144	0.541
Hansen			0.212	0.308	0.214	0.641	0.660	0.213	0.526	0.393	0.988
Sargan			0.136	0.165	0.198	0.673	0.168	0.148	0.409	0.105	0.156

Notes: 1. Robust standard errors are in parentheses, * p < 0.10, ** p < 0.05, *** p < 0.01. 2. ${\mathrm{lnSD}}^{\mathrm{fdied}}$: International R&D spillover through inward FDI from developed countries; ${\mathrm{lnSD}}^{\mathrm{fdiing}}$: International R&D spillover through inward FDI from developing countries; ${\mathrm{lnSD}}^{\mathrm{imped}}$: International R&D spillover through imports from developed countries; ${\mathrm{lnSD}}^{\mathrm{imping}}$: International R&D spillover through imports from developing countries; 3. The time fixed effect is controlled in system GMM. To reduce the number of instrument variables, we use ‘collapse’ demand in Stata.

. We observe that robust results were not obtained at the overall level from columns (3)–(5) of Table A4.

Based on the country-type analysis results, international R&D spillover through inward FDI and imports from developed countries positively influences corporation innovation. These results are similar to those of Wang et al. [8], Ito et al. [55], Wang and Wu [58].

Significant gaps exist in employment, assets, sales, new product sales, exports, and labor productivity between domestic firms and foreign subsidiaries. Among them, foreign subsidiaries outperformed domestic firms in almost all economic indicators [58]. The Chinese subsidiaries of MNCs can also bring capital from home countries, advanced management models, specialized experience, etc.

Therefore, foreign subsidiaries can show a good demonstration effect on domestic firms in China. Also, the training effect on the local labor force employed by foreign subsidiaries will appear, which will then spread to domestic companies in China as the labor force leaves. Thus, the increased market competition through FDI inflows may pressure domestic firms to innovate to maintain market share. Moreover, because technologically advanced foreign firms require their domestic partners to maintain high-quality products, domestic firms can learn technology from foreign firms through industrial linkages.

In addition, importing products from developed countries through introduction, learning, absorption, application, and re-innovation can improve the knowledge endowment of Chinese enterprises. It will also provide opportunities for Chinese companies to adapt to international market norms.

We also find that the drivers of corporation innovation change across periods of economic growth in columns (6) and (7) of Table A4. During periods of rapid economic growth, international R&D spillovers positively influence Chinese corporations’ innovation through imports and inward FDI. During the period of stable growth, only domestic R&D stock has a positive impact on corporation innovation. Lastly, our results also illustrate that international R&D spillover through imports is more significant in eastern regions than inland regions.

4.2.4. Knowledge Creation and Knowledge Acquisition Results

Finally, the results of knowledge creation and acquisition are reported in the Appendix A of

Table A1. Knowledge creation results.

	(1)	(2)	(3)	(4)	(5)	2001– 2010	2011– 2020	(8)	(9)	(10)	(11)
	Panel OLS	FE	System GMM	System GMM	System GMM	System GMM	System GMM	System GMM	System GMM	System GMM	System GMM
L.knowcreat			0.460	0.226 *	−0.020	0.076	0.955 ***	0.257 **	0.632 ***	0.210	0.467 *
			(0.85)	(0.13)	(0.29)	(0.09)	(0.22)	(0.13)	(0.16)	(0.33)	(0.26)
L.${\mathrm{lnSD}}^{\mathrm{fdi}}$	−0.037 ***	−0.003	−0.095		−0.031	0.012	0.163 *		−0.046	−0.001	−0.008
	(0.01)	(0.01)	(0.18)		(0.07)	(0.04)	(0.10)		(0.04)	(0.07)	(0.02)
L.${\mathrm{lnSD}}^{\mathrm{imp}}$	0.019	0.018		0.000	0.017	−0.020	−0.037			−0.004	−0.009
	(0.02)	(0.02)		(0.02)	(0.02)	(0.05)	(0.07)			(0.02)	(0.04)
L.lnSD	0.244 ***	0.292 ***	0.258	0.159 ***	0.219	0.21 6 **	−0.214	0.128	0.039	0.182	0.139 *
	(0.03)	(0.05)	(0.43)	(0.05)	(0.17)	(0.09)	(0.16)	(0.15)	(0.04)	(0.12)	(0.08)
L.lnedu	1.543 ***	0.903 **	0.886	1.193 ***	1.685 ***	1.347 *	−0.519	0.680	0.354	1.028 *	0.875
	(0.15)	(0.36)	(1.60)	(0.28)	(0.55)	(0.74)	(0.77)	(0.71)	(0.29)	(0.62)	(0.56)
L.lngeo	−1.286 ***	−1.150 ***	−1.331	−0.948 ***	−0.086 ***	−0.753 ***	1.621	−0.785	−0.011	−0.858	−0.866 **
	(0.10)	(0.28)	(1.94)	(0.22)	(0.03)	(0.27)	(1.02)	(0.63)	(0.01)	(0.54)	(0.35)
L.${\mathrm{lnSD}}^{\mathrm{fdiing}}$								−0.028
								(0.02)
L.${\mathrm{lnSD}}^{\mathrm{fdied}}$								0.057
								(0.10)
L.${\mathrm{lnSD}}^{\mathrm{imping}}$									−0.035 ***
									(0.01)
L.${\mathrm{lnSD}}^{\mathrm{impedi}}$									0.085 **
									(0.04)
east#${\mathrm{lnSD}}^{\mathrm{imp}}$										0.002
										(0.00)
east#L.${\mathrm{lnSD}}^{\mathrm{fdi}}$											0.001
											(0.00)
_cons	37.898 ***	33.524 ***	40.646	27.983 ***	0.000	21.402 **	−50.076	23.365	0.000	25.229	25.869 **
	(2.99)	(8.04)	(59.02)	(6.50)	(0.00)	(8.39)	(31.03)	(19.37)	(0.00)	(16.23)	(10.49)
N	422	422	391	402	401	177	215	391	378	395	396
Adj R-sq	0.70	0.36
AR(1)			0.050	0.004	0.026	0.025	0.077	0.046	0.046	0.092	0.012
AR(2)			0.201	0.193	0.427	0.737	0.163	0.523	0.156	0.424	0.136
Hansen			0.579	0.865	0.986	0.554	0.549	0.500	0.698	0.104	0.816
Sargan			0.676	0.912	0.971	0.463	0.119	0.113	0.736	0.098	0.807

Notes: 1. Robust standard errors are in parentheses, * p < 0.10, ** p < 0.05, *** p < 0.01. 2. ${\mathrm{lnSD}}^{\mathrm{fdied}}$: International R&D spillover through inward FDI from developed countries; ${\mathrm{lnSD}}^{\mathrm{fdiing}}$: International R&D spillover through inward FDI from developing countries; ${\mathrm{lnSD}}^{\mathrm{imped}}$: International R&D spillover through imports from developed countries; ${\mathrm{lnSD}}^{\mathrm{imping}}$: International R&D spillover through imports from developing countries. 3. The time fixed effect is controlled in system GMM. To reduce the number of instrument variables, we use ‘collapse’ demand in Stata.

and

Table A2. Knowledge acquisition results.

	(1)	(2)	(3)	(4)	(5)	2000– 2010	2011– 2020	(8)	(9)	(10)	(11)
	Panel OLS	FE	System GMM	System GMM	System GMM	System GMM	System GMM	System GMM	System GMM	System GMM	System GMM
L.lninvget			0.708 ***	0.417 ***	0.860 ***	0.240	0.189	0.427 ***	0.892 **	0.331	0.108
			(0.18)	(0.10)	(0.16)	(0.16)	(0.40)	(0.14)	(0.38)	(0.21)	(0.10)
L.${\mathrm{lnSD}}^{\mathrm{fdi}}$	−0.011	−0.010	0.081		0.076 **	0.029	−0.027		0.046	−0.334	0.154 *
	(0.01)	(0.01)	(0.05)		(0.04)	(0.03)	(0.07)		(0.16)	(0.22)	(0.09)
L.${\mathrm{lnSD}}^{\mathrm{imp}}$	0.128 ***	0.111***		−0.091	−0.027	0.123 *	−0.032	0.356 **		0.692	−0.393
	(0.02)	(0.02)		(0.09)	(0.02)	(0.07)	(0.07)	(0.18)		(0.64)	(0.28)
L.lnSD	0.068 **	0.092 ***	−0.066	0.292 *	−0.055	0.021	−0.002	0.097	−0.036	0.263	−0.002
	(0.03)	(0.03)	(0.04)	(0.17)	(0.04)	(0.14)	(0.17)	(0.14)	(0.11)	(0.18)	(0.09)
L.lnedu	1.255 ***	1.333 ***	0.327	−0.160	0.135	0.555	6.959	1.879 *	−0.036	1.326	1.288 **
	(0.19)	(0.18)	(0.38)	(1.06)	(0.34)	(0.37)	(4.46)	(1.05)	(1.12)	(1.22)	(0.65)
L.lngeo	−1.362 ***	−1.494 ***	−0.018	−1.041 ***	0.002	−0.531	−3.226 **	−2.029 **	0.015	−4.097	0.431
	(0.12)	(0.17)	(0.02)	(0.22)	(0.02)	(0.35)	(1.35)	(0.94)	(0.06)	(2.63)	(0.98)
L.${\mathrm{lnSD}}^{\mathrm{fdiing}}$								0.055
								(0.04)
L.${\mathrm{lnSD}}^{\mathrm{fdied}}$								−0.350
								(0.24)
L.${\mathrm{lnSD}}^{\mathrm{imping}}$									−0.004
									(0.05)
L.${\mathrm{lnSD}}^{\mathrm{imped}}$									−0.023
									(0.14)
east#L.${\mathrm{lnSD}}^{\mathrm{imp}}$										−0.086
										(0.12)
east#L.${\mathrm{lnSD}}^{\mathrm{fdi}}$											0.084*
											(0.05)
_cons	40.469 ***	44.345 ***	0.000	33.322 ***	0.000	14.795	91.739 **	59.885 **	0.000	118.921	−9.231
	(3.71)	(5.35)	(0.00)	(7.94)	(0.00)	(11.03)	(36.41)	(27.93)	(0.00)	(73.29)	(27.26)
N	422	422	391	398	387	196	223	383	382	387	374
Adj R-sq	0.60	0.37
AR(1)			0.006	0.000	0.006	0.032	0.022	0.106	0.027	0.067	0.104
AR(2)			0.167	0.290	0.136	0.163	0.782	0.142	0.165	0.235	0.250
Hansen			0.556	0.993	0.759	0.491	0.928	0.702	0.135	0.864	0.754
Sargan			0.658	0.173	0.826	0.339	0.856	0.299	0.779	0.497	0.207

Notes: 1. Robust standard errors are in parentheses, * p < 0.10, ** p < 0.05, *** p < 0.01. 2. ${\mathrm{lnSD}}^{\mathrm{fdied}}$: International R&D spillover through inward FDI from developed countries; ${\mathrm{lnSD}}^{\mathrm{fdiing}}$: International R&D spillover through inward FDI from developing countries; ${\mathrm{lnSD}}^{\mathrm{imped}}$: International R&D spillover through imports from developed countries; ${\mathrm{lnSD}}^{\mathrm{imping}}$: International R&D spillover through imports from developing countries. 3. The time fixed effect is controlled in system GMM. To reduce the number of instrument variables, we use ‘collapse’ demand in Stata.

. We find that the coefficients of international R&D spillovers are insignificant at the overall level. In terms of knowledge creation, the results are only statistically significant when the origin countries of the imports are distinguished. The results are similar to those of Sun and Du [38].

4.3. Robustness Check

Due to space issues, readers interested in the results can request them from the authors.

(1)

For the choice of depreciation rate, we use 9.6% in the main text. However, in previous studies, multiple values appeared, such as 5%, 15%, and 20%. Therefore, we replaced the depreciation rate and performed a robustness check. According to the results, the change in the depreciation rate did not change our results.

(2)

According to the studies of Chen et al. [10], Feng and Li [11], Hong et al. [12], Li et al. [13], and Zhou et al. [14], we find that they used country-level data and then the ratio of each province as a weighting value to measure the international R&D spillover effect of each province, which are reported in Equations (8) and (9). We recalculated the international R&D spillover using this method, and the results showed that the sign and significance did not change, except for a magnifying effect on the coefficients.

$${\mathrm{S}}_{\mathrm{t}}^{\mathrm{fdi}}={\displaystyle \sum }_{\mathrm{j}}\frac{{\mathrm{fdi}}_{\mathrm{jt}}}{{\mathrm{k}}_{\mathrm{jt}}}{\mathrm{SD}}_{\mathrm{jt}}^{\mathrm{d}}$$

(8)

$${\mathrm{S}}_{\mathrm{it}}^{\mathrm{fdi}}=\frac{{\mathrm{fdi}}_{\mathrm{it}}}{{{\displaystyle \sum }}_{\mathrm{i}}{\mathrm{fdi}}_{\mathrm{it}}}{\mathrm{S}}_{\mathrm{t}}^{\mathrm{fdi}}$$

(9)

where ${\mathrm{S}}_{\mathrm{t}}^{\mathrm{fdi}}$ denotes the aggregate international R&D spillover effect of China in t period, and ${\mathrm{fdi}}_{\mathrm{jt}}$ is inward FDI from j country to China in t period. ${\mathrm{k}}_{\mathrm{jt}}$ and ${\mathrm{SD}}_{\mathrm{jt}}^{\mathrm{d}}$ represents the gross capital formation and R&D capital stock of j country in t period. ${\mathrm{fdi}}_{\mathrm{it}}$ denotes inward FDI value for i province of China in the t period. ${\displaystyle \sum }_{\mathrm{i}}{\mathrm{fdi}}_{\mathrm{it}}$ is inward FDI value for China in t period. ${\mathrm{S}}_{\mathrm{it}}^{\mathrm{fdi}}$ measures the international R&D spillover effect of i province of China in t period.

5. Conclusions

To test the impact of international R&D spillovers on technological innovation in China’s provinces, we concentrate on inward FDI and import channels, as the previous studies did. However, unlike the widely used data on country-level inward FDI and imports, we fill a gap in measuring international R&D spillovers by using detailed data for each province-country level in China to decrease ‘aggregation bias.’ Of course, it also leads to the inability to obtain some data.

In addition, we consider the heterogeneous innovation effect of international R&D spillovers from multiple levels, in terms of the different development levels of the source countries; five primary technological innovation indicators; the east and inland regions; and different economic growth periods. The results are summarized as follows.

First, international R&D spillover stimulates China’s technological innovation at the overall level, and imports can be regarded as an effective channel of international R&D spillover. It illustrates the effectiveness of the opening-up policy in China and provides empirical evidence of the need to expand the openness policy. On the other hand, technological innovation in China during the rapid growth period is influenced by international R&D spillover through imports. However, during a stable growth period, the impact of international R&D spillover is insignificant.

Second, international R&D spillover can significantly promote innovation performance. In other words, it confirms the importance of foreign subsidiaries’ presence in increasing innovation performance. Moreover, we find that the impact of international R&D spillover through imports on innovation performance is gradually weakened, which is in line with the marginal diminishing effect of the spillover effect.

Third, in terms of the country-type analysis results, international R&D spillover through imports from developed countries positively affects technological innovation, knowledge creation, and corporation innovation. A virtuous cycle of introducing, learning, absorbing, applying, and re-innovating foreign technologies can be constituted by importing products from developed countries. This may lead to an increase in knowledge endowment, thereby contributing to many aspects of China’s technological innovation.

Additionally, international R&D spillover through inward FDI from developed countries significantly promotes corporation innovation. For domestic enterprises in China, we confirm that positive spillover effects, such as the demonstration effect and training effect of foreign subsidiaries from developed countries, still exist.

Fourth, considering the regional heterogeneity, the results confirm that the east is more affected by international R&D spillovers than the central and western regions in terms of the innovation environment. The eastern region of China is ahead of the central and western regions in terms of economic development, market environment, and human capital. This leads to differences between eastern and inland regions in the ability to absorb international R&D spillover effects.

Fifth, the impact of international R&D spillover differs between China’s rapid and stable growth periods. As China enters a stable growth period, international R&D spillover will become insignificant. During the rapid growth period, international R&D spillover is caused by imitating and importing to enjoy ‘advantage of backwardness,’ which promotes technological innovation in all regions of China. However, as catching up and the technology gap narrowed, it became difficult to absorb international R&D spillovers. However, the gradual accumulation of knowledge endowments can contribute to later spontaneous R&D activities.

Considering the recent anti-globalization process and trade pressure for China, we propose some policy implications based on our analysis results.

First, a strategy of continuing to expand opening-up should be persistent. Based on our empirical results, openness, primarily through inward FDI and imports, can promote many aspects of technological innovation in China. Expanding imports of high-quality goods and inward FDI flows could form a strong attraction to global factor resources to improve the allocation level of China’s production factors, ultimately achieving productivity improvement and technological innovation.

Second, attracting high-quality inward FDI would contribute to spontaneous domestic innovation. However, relying on complete dependence on foreign technology will lead to a loss of its R&D capacity. Only by adhering to independent innovation can we not fall into the passive position of technological innovation. The government should take advantage of the global R&D strategies of MNCs and try to attract knowledge-intensive FDI and guide the development of local R&D activities and cooperation with domestic firms’ R&D activities.

Third, we find that an excellent innovation environment has a positive impact on absorbing international R&D spillovers from the empirical results. The Chinese government still needs to devote its efforts to implementing policies, such as the “Rise of Central China” and “Western Development,” to promote innovation environment development in the inland regions. In view of the inherent short board in the central and western regions, preferential policies to make up for this short board are proposed to improve the innovation environment.

Finally, our study has the following limitations. First, although imports and inward FDI as international R&D spillover channels distinguish between country types, there are areas for improvement in distinguishing between intermediate materials, final products, and different purposes of FDI. Second, the measurement of international R&D spillover only measured the direct spillover effect. However, since indirect effects also exist, the follow-up study intends to include them in the analysis.

Endnote

The province sample includes Shanghai, Beijing, Guangxi, Hebei, Gansu, Shanxi (West), Yunnan, Ningxia, Xinjiang, Henan, Fujian, Qinghai, Anhui, Shanxi (Middle), Jiangsu, Zhejiang, Liaoning, Heilongjiang, Shandong, Guangdong, Jiangxi, Henan, and Chongqing.

The foreign countries include Argentina, Colombia, Cyprus, Israel, Malta, Poland, Sweden, Australia, Croatia, Germany, Italy, Mexico, Portugal, Tajikistan, Austria, Cuba, Greece, Japan, Moldova, Romania, Thailand, Belarus, Czech Republic, Hungary, Kazakhstan, Mongolia, Russia, Tunisia, Belgium, Denmark, Iceland, Korea, Netherlands, Singapore, Turkey, Brazil, Egypt, India, Kuwait, New Zealand, Slovak Republic, Ukraine, Bulgaria, Finland, Iran, Luxembourg, Norway, and South Africa. The classification of developing and developed countries follows the World Bank standards.