Spillover and Re-Spillover in China’s Collaborative Innovation

Abstract

The spillover effect serves as the basis of regional collaborative innovation. Existing research on innovation spillover focuses on the overall impact of a region's innovation factors on local and other regions' innovation activities. However, re-spillover may occur since the flow of innovation factors between any two regions may influence the innovation in third-party regions. This study quantifies labor flow, capital flow, and institutional learning between regions in China using a gravity model and a social network analysis model, and then applies a spatial econometric model to investigate innovation spillover and re-spillover. The results show that re-spillover can better explain levels of regional innovation. Capital, government support, labour flow, capital flow, and institutional learning have a positive spillover effect on local innovation, while labour flow also has positive spillover effects to other regions.

Keywords

spillover effect re-spillover Spatial Econometric Model regional collaborative innovation China

Introduction

Innovation is the determinant of economic growth (Romer 1990) and an important engine of regional development (Wang et al. 2019). However, it is increasingly difficult for a single innovation subject to assume all innovation activities in the current complex economic environment characterized by burgeoning social systems. Interaction with productive individuals can produce better results (Buera and Oberfield 2020) because close ties can promote learning among those individuals (Howard et al. 2016). Therefore, interaction between innovation subjects has a profound impact on economic development (McCann and Simonen 2005). People hope to create synergy through competition, cooperation, and complementarity (An et al. 2014). Governments also seek collaborative innovation by directly subsidizing enterprises and indirectly supporting universities and research institutions (Ji and Miao 2020). Collaborative innovation is a development strategy to achieve synergy by optimizing organizational structures (Kodama 2015), which is not a new phenomenon (Ketchen, Ireland, and Snow 2007). For the innovation subject, collaborative innovation provides the opportunity to acquire knowledge and market information from external sources (Heil and Bornemann 2018), and the deeper and broader cooperation that may occur can further promote innovation when the collaboration effect is enhanced (Liu Ying and Wu 2017). Accordingly, if the innovation subject cannot obtain knowledge and market information externally, that subject may be eliminated (Rowley, Behrens, and Krackhardt 2000; Emden, Calantone, and Droge 2006). For the innovation system, collaboration can improve the frequency of knowledge collision and integration (Sternberg 1996), enlarge the innovation effect (Xu, Yan, and Xiong 2019), improve innovation efficiency (Fan, Lian, and Wang 2020), and create public value (Neumann et al. 2019). The challenge of studying innovation which existing research have to face is that the knowledge mobility and the complexity of technology make doing so difficult (Scuotto et al. 2020), and the influencing factors on innovation even involve third-party governance (Bai et al. 2020), culture (Barboza and Pede 2012), and reputation (Wang et al. 2020). However, the basic logic of collaborative innovation remains the spillover effect of innovation, whether it is knowledge spillover or technology spillover.

Spillover effect refers to a subject that obtains (Li et al. 2019) and uses (Amir 2000; Gama, Maret, and Masson 2018) innovation from another subject at low cost to reduce local innovation cost (Klette and Kortum 2004). Innovation is not evenly distributed in space, and it is concentrated in some places (Paci and Usai 2000). With the accumulation of innovation (Lööf and Nabavi 2015) and quality ladder structures (Acemoglu et al. 2013), innovation will produce spillover effect. The work of Scott (1988), Porter (1990), and Krugman (1991) drew attention to innovation spillover. Because of innovation spillover, the surrounding areas with strong innovation capability can obtain the space discount version of innovation (Desmet and Hansberg 2014). Sometimes, the innovation spillover obtained externally can replace the internal innovation efforts of an entity (Audretsch and Feldman 1996). Therefore, the innovation capacity of a region depends on the region itself, but also on other regions (Asheim, Smith, and Oughton 2011). According to the theory of spatial correlation and spatial spillover, innovation between two regions are interrelated and realized through spatial spillover. However, the existing research considers this type of spillover to be point to point, so does the ‘line’ of innovation connection between the two regions also have spillover? If there are only seven cities in a closed country (as shown in Figure 1(a)), the innovation spillover from city A to city B can be expressed as a solid blue line, that is, the spillover line. In the process of innovation, the blue solid line will pass through city C and may also affect other cities adjacent to city C, even those cities not adjacent to city C. Existing studies only focus on the spillover of cities (points) to cities (points) and do not consider that the spillover line between two cities (points) may also produce spillover.

Figure 1.

(a) The spillover line between cities. (b) Network innovation spillover between cities.

If the cities themselves produce innovation spillover, and the innovation links between cities produce innovation spillover also, the external innovation impact of a city originates directly from the innovation spillover of other cities and from the secondary innovation spillover of the spillover line in the innovation network. Taking cities A and D in Figure 1b as examples, the innovation spillover of city D is directly affected by city A and is expressed as AD. At the same time, city A overflows to other cities, such as city B, city C, city E, city F, and city G, which may further affect city D, as shown in the dotted line in Figure 1. Therefore, the innovation spillover of city A to city D includes two parts, direct innovation spillover and secondary innovation spillover of the spillover line through the innovation network. To verify the existence of linear innovation spillover (the solid line in Figure 1b) and the spillover line in the innovation network spillover (the dotted line in Figure 1b), this study conducts an empirical analysis.

Literature Review

Spillover From Points

Although spillover of innovation is closely related to spatial location, growth theory does not take space into account (Lucas 1989), and Jaffe (1989)’s pioneering achievements have attracted researcher attention to the spatial dimension of innovation externality. The spatial dimension of early innovation spillover mainly uses the knowledge production function and patent cooperation data (Autant-Bernard, Fadairo, and Massard 2013). Based on the first law of geography that all things are related to other things, but closer things are more related to farther things (Tobler 1970), spatial econometrics emerged in the 1970’s to study the treatment of spatial data. Since then, with the application of spatial econometric models in many fields (Kelejian and Piras 2017), these models have been widely applied in the research on innovation spillover (Autant-Bernard 2012). Spatial econometric models explain regional economic growth from two aspects, local innovation and external innovation (Furková 2019), which leads to the testing of a large number of innovation spillover characteristics and the analysis of the results using spatial econometric models. For example, for developed economies, Hammadou, Paty, and Savona (2014) and Furková (2019) used spatial econometric models to analyse innovation spillovers in the EU. Recognizing that regional spillovers are limited by national boundaries (Moreno-Serrano, Paci, and Usai 2005), some studies pay more attention to the domestic characteristics of innovation spillovers (Acs, Audretsch, and Feldman 1994). In Europe, spatial econometric models have explored innovation spillover in countries such as France (Autant-Bernard and Lesage 2011), Italy (Cardamone 2018; Benedetti, Palma, and Postiglione 2019), Spain (Cabrer-Borrás and Serrano-Domingo 2007), and Hungary (Varga 2007). In emerging countries, China’s innovation spillover studies at the provincial level (Wang et al. 2016b), the urban agglomeration level (Sheng et al. 2019), and the urban level (Ning, Wang, and Li 2016) also adopt spatial econometric models. At present, the research on innovation spillover focuses on two aspects: one is analysing the influencing factors of innovation spillover through spatial econometric models, including environmental regulation and foreign direct investment (FDI) (Feng et al. 2019). On the other hand, research analyses the explanation of innovation spillover according to other phenomena, including social development (Moralles, Silveria, and Rabelatto 2018), ecological environment (Liu 2018; Benedetti, Palma, and Postiglione 2019), and tourism (Romão and Nijkamp 2019). However, spatial econometric models abstract the enterprise or region as a point, explore the externality of this point to the whole system, and ignore the connection between points.

Connection Between Lines

Spatial econometric models focus on the analysis of the spillover effect of innovation subjects, but does not quantify the innovation relationship between the subjects. Spillover is not unintentional, but a large component is the result of the interaction between innovation subjects (Zucker, Darby, and Brewer 1998; Breschi and Lissoni 2001; Ott and Rondé 2019). With the development of technology, innovation has changed in spatial form (Cao, Zeng, and Ye 2018), from ‘local space’ to ‘flowing space’ (Castells 1996; Ai and Miao 2010). Because knowledge is invisible, and its movement is difficult to observe (Keller and Yeaple 2013), quantifying the innovative connection between subjects is complex. However, the gravity model offers a new perspective. The gravity model originated from Newton’s law of universal gravitation and has been gradually applied to the field of social science to explain various behaviours in spatial interaction (Chen and Huang 2018). In economic research, the gravitation model is widely used to analyse the flow of goods (Cantore and Cheng 2018), services (Shepherd 2019), and trade (Carrère 2006; Kabir, Salim, and Al-Mawali 2017) in international trade. International trade is also accompanied by the flow of innovation factors, so relevant studies (Márquez-Ramos and Martínez-Zarzoso 2010; Kandilov and Grennes 2012; Cipollina, Demaria, and Pietrovito 2016) include innovation in the gravity model to analyse international trade issues. In the study of innovation, interregional cooperation is affected by distance attenuation (Boschma 2005), and knowledge flow among subjects is limited by geographical factors (Scherngell and Hu 2011). Correspondingly, spatial proximity has a significant impact on the intensity of cooperation between subjects (Katz and Martin 1997). On the one hand, regional resource endowments and gravitational factors formed by endowments can explain regional innovation connections (Hjaltadóttir, Makkonen, and Mitze 2020). On the other hand, because the gravity model is not limited by the number of observations, it is more suitable for static comparative innovation cooperation (Hardeman et al. 2014). Therefore, the gravity model is gradually applied to innovation cooperation (Bröcker and Mitze 2019). The research on North America (Hardeman et al. 2014), the European Union (Hjaltadóttir et al. 2020), and China’s provincial (Scherngell and Hu 2011) and municipal levels (Wang Yang and Qin 2020) also reflect the general applicability of the gravity model to innovation connection.

Construction of the Area

Research on innovation has been criticized for neglecting the impact of innovation networks by focusing on the functional level exclusively (Capello and Lenzi 2018a, 2018b). The functional level has been the focus because the economic and social value of innovation depends on its internal benefits and the benefits to the network (Arieli et al. 2020). The cooperation of multiple agents is a dynamic complex network (Yström and Agogué 2020), and an innovation network is the carrier of innovation flow (Hua and Wang 2015). Regarding innovation networks as social networks (Min, Kim, and Swang 2020) can explain the regional differences in innovation capability (Lawson and Lorenz 1999; Quatraro 2009). For innovation subjects, a large number of partners with strong innovation ability provide a better innovation environment for themselves (Davis and Dingel 2019), and integrating into innovation networks can help innovation subjects obtain more resources (Sternberg 2000; Borgatti and Foster 2003). Innovation networks promote information exchange, technology transfer (Gardet and Mothe 2011) and knowledge creation (Nieves and Osorio 2013), and innovation is spread through networks (Folke et al. 2005; Aswegen and Retief 2020) and integrated by networks (Harrisson and Laberge 2002). Therefore, innovation networks direct future innovation research (Rothwell 1992). The main methods used to study innovation networks are simulation and social network analysis. Simulation is an agent-based system dynamics model (Hua and Wang 2015; Mao et al. 2020) that sets many rules to ensure that the results resemble the facts. Social network analysis is based on the dual characteristics of innovation space (Caccamo 2020). For innovation social network analysis, one type of research uses patent cooperation to describe the structural characteristics of the innovation network, whose objects cover the United States (Mukherji and Silberman 2013), China (Wang et al. 2016a), and the world (Liu et al. 2017). However, due to the protections of patent intellectual rights (Cao, Zeng, and Ye 2019), other informal network cooperation (Brown and Duguid 1991), and tacit knowledge (Howells 2002; Gertler 2003), patent cooperation cannot reflect the overall innovation network dynamic (Cohen and Kozak 2002). The other type of research is relational data constructed by gravity models (Wang Yang and Qin 2020) to analyse the network location of innovation subjects. The network location reflects the connection between one innovation subject and other innovation subjects (Fang et al. 2014). The higher the innovation subject’s central position in the innovation network, the more benefits the subject has (Leenders and Dolfsma 2016), and the more innovation it can generate (Powell, Koput, and Smith-Doerr 1996; Tsai 2001).

Summary

From the existing research, we find that the spatial characteristics of innovation have been analysed from the perspective of three dimensions: point, line, and area, and through the research of innovation spillover of spatial regression, innovation connection of gravity models, and social innovation network analysis. However, the focus of all types of analysis is not the same. The innovation spillover research focuses on the effect of a point on itself and the total effect on other points (as shown in Figure 2(a)) but ignores the linear influence between points. Innovation connection research focuses on quantifying the connection between these points (as shown in Figure 2(b)) although the connection has not actually occurred but has the potential to do so. The research on innovation networks only focuses on the network influence created by points and lines on another point (as shown in Figure 2(c)). However, connection between points occurs through the lines that join those points, and the network formed between lines will also act on those points. Therefore, all the spatial characteristics of innovation must be considered when analysing the spatial effect of innovation.

Figure 2.

Spatial characteristics of innovation.

Models and Data

The Spatial Econometric Model

The spatial econometric model is the main tool used to analyse spatial spillover. Spatial econometric models can include geographical location information of elements in the model by setting a spatial weight matrix. The basic spatial econometric models include spatial auto regression (SAR), the spatial errors model (SEM), and the spatial Durbin model (SDM) (Anselin, Le Gallo, Jayet 2008), as shown in formulas (1), (2), (3).

SAR: y = ρ W y + X β + ε

(1)

SEM: y = X β + μ, (μ = λ W μ + ε)

(2)

SDM: y = ρ W y + X β + W X γ + ε

(3)

Here, y is the dependent variable. In terms of measuring regional innovation, patents and innovation have a strong positive correlation (Guo 2015), which is an important factor of innovation diffusion (Keller 2004). Although patents have limitations (Forman, Goldfarb, and Greenstein 2016), they are still the best tool to measure intangible knowledge (Hall, Adam, and Manuel 2005), and many studies have chosen patents as an alternative indicator of innovation (Anselin, Varga, and Acs 1997; Bottazzi and Peri 2003). Therefore, the number of patents is chosen as the explanatory variable to reflect regional innovation ability.

X represents the independent variables and control variables, and β is the influence coefficient. There is no unified theory for organizational elements in collaborative innovation research (Ollila and Yström 2016), but considering that the flow of labour and capital provides convenient conditions for innovation flow around the world (Kaygalak and Reid 2016), this study selects capital and labour as core explanatory variables. Accordingly, labour flow promotes the creation of new knowledge and the spread of innovation (Almeida and Kogut 1999) while capital flow also promotes the spillover of knowledge and the transfer of technology (Blomström and Kokko 1998). Therefore, labour and capital flows are the core explanatory variables of this study. In addition, government intervention is considered key to innovation in developing countries (Dunning, Kim, and Park 2007), and institutional proximity has a positive impact on the flow of innovation (Capello 1999; Knoben and Oerlemans 2006). This study also selects institution and institutional learning (i.e. institutional flow) as core explanatory variables. Labour force, capital, and government support are respectively represented by the number of people engaged in scientific and technological innovation, the stock of fixed assets, and the scientific and technological expenditure in the government’s financial expenditure. Labour flow, capital flow, and institutional learning are represented by the flow of the number of people engaged in scientific and technological innovation, the stock of fixed assets, and the scientific and technological expenditure in the government’s financial expenditure. Talent education, industrial structure, infrastructure, and the level of opening-up are selected as control variables, which are respectively expressed by education expenditure in local general public budget expenditure, the proportion of secondary industry in GDP, the number of Internet broadband access users, and the total amount of FDI.

W is the spatial weight matrix, and ρ is the spatial autoregressive parameter, which measures the influence of Wy on y. Wxγ is the influence of adjacent variables, and γ is the corresponding coefficient. ε is the disturbance term, and μ is the spatial disturbance term while λ is its spatial influence coefficient.

To solve the difficult problem of explaining the coefficients of spatial econometric models, Lesage and Pace (2009) proposed direct effects, indirect effects, and total effects. For the spatial Durbin model, the model can be rewritten as

(I_{n} - ρ W) y = X β + W X γ + ε

(4)

I_n is the unit matrix. If $A^{\bar{d}}$ represents averaging the diagonal elements of matrix A, the direct effect is the average value of the diagonal elements of the above matrix, which can be expressed as:

D E = {{(I - ρ W)}^{- 1} \times {(β I + γ W)}^{\bar{d}}}

(5)

The indirect effect is the average of the non-diagonal elements of the matrix. If $A^{\bar{r s u m}}$ represents the row average of the non-diagonal elements of matrix A, the indirect effect can be expressed as:

I E = {{(I - ρ W)}^{- 1} \times (β I + γ W)}^{\bar{r s u m}}

(6)

The total effect is the sum of the direct effect and the indirect effect, which can be expressed as follows:

T E = D E + I E

(7)

The Gravity Model

Before using the spatial econometric model to analyse the spillover of innovation, we first quantify the spillover line between cities by the flow of the number of people engaged in scientific and technological innovation, the stock of fixed assets, and the scientific and technological expenditure in the government’s financial expenditure. The gravity model quantifies the strength of the connection between two points. Although the connection strength is not the actual spillover, it reflects the potential spillover level to a certain extent. Therefore, this study chooses the connection strength to roughly reflect the bandwidth of the spillover line.

Gravity model is used to construct the spillover line’s characteristics of talent, capital, and institution. The basic gravity model can be expressed as follows:

F_{i j} = k \frac{M_{i} m_{j}}{D_{i j}^{b}}

(8)

In formula (8), F_ij is the innovative connection strength, the bandwidth of the spillover line, between city i and city j; M_i and m_j, respectively, represent the variables selected by city i and city j according to different measurement dimensions; D_ij represents the distance between city i and city j; b is the distance attenuation coefficient, which generally takes 2; and k is the empirical constant, which generally takes 1.

The spillover of innovation includes many elements of innovation, among which labour, capital, and the government system are the input elements of scientific and technological innovation and the most representative variables of the dynamic flow of innovation elements between cities, which are most likely to have an impact on the innovation links between cities. Therefore, this study measures the bandwidth of the spillover line among cities for the three dimensions of labour flow, capital flow, and institutional learning, which correspond to the three input factors of scientific and technological innovation. The formula is as follows:

L F_{i j} = \frac{w_{i} p_{j}}{D_{i j}^{2}}

(9)

In formula (9), w_i is the wage level of city i, which is represented by the average wage of on-the-job employees. Chen et al. (2020) found that people with the highest incomes in China are more than four times more likely to innovate than others. p_j is the number of innovative talents of city j, which is represented by the number of employees engaged in science and technology; and LF_ij is the attraction intensity of city i’s wage level to city j’s labour flow. This formula expresses the flow of labour to the cities with higher wages; that is, the approximate bandwidth of the labour spillover line.

C F_{i j} = \frac{R_{i} I_{j}}{D_{i j}^{2}}

(10)

In formula (10), R_i is the market activity of city i, which is represented by the total retail sales of social consumer goods; I_j is the capital strength of city j, which is represented by the total fixed asset investment of the whole society; and CF_ij is the attraction intensity of city i’s market activity to city j’s capital flow. This formula expresses the flow of capital to the cities with an active market environment; that is, the approximate bandwidth of the capital spillover line.

R F_{i j} = \frac{E_{i} G_{j}}{D_{i j}^{2}}

(11)

In formula (11), E_i is the innovation expenditure of city i, which is represented by the science and technology expenditure in the public finance expenditure; G_j is the economic benefit of city j, which is represented by the gross regional product; and RF_ij is the institutional learning attraction intensity of city i to city j. This formula expresses that institutions provide learning opportunities to the cities with a higher economic level; that is, the approximate bandwidth of the institutional learning spillover line.

The interaction between innovation subjects is influenced by proximity, which is emphasized by economic geography and related disciplines (Hansen 2014). Relevant studies have been analysed from the perspective of geographical proximity, economic proximity (Conley and Ligon 2002), and technical proximity (Fischer and Lesage 2008). Geographical proximity can promote the flow of innovation (Ter wal and Anne 2014), which is the precondition of innovation spillover (Autant-Bernard and Massard 2009), and the best method of communication (Desrochers 2001). Adjacent subjects are more likely to have innovation cooperation (Hoekman Franken, and Tijssen 2010), so D in this paper reflects geographical distance.

Social Network Analysis

Although the gravity model can reflect the strength of the relationship between cities, the quantitative results are in the form of a matrix whereby each city corresponds to n-1 connections, which cannot be calculated by the spatial econometric model. If we sum the bandwidth of each city’s corresponding spillover line simply, the results will have strong endogeneity with the original input index. Because attribute data are used in urban research and relational data are used in social network analysis, attribute data can be transformed into relational data using a modified gravity model. Then, the spatial connection intensity data can be integrated through network analysis. Attribute data are used in urban research while relational data are used in social network analysis. Therefore, attribute data can be transformed into relational data with the help of a modified gravity model, and the data of the approximate bandwidth of the spillover line can be integrated through network analysis.

The urban innovation network is defined as the explicit or implicit scientific and technological knowledge network formed by formal or informal cooperation (government, research institutions, and companies) within or between cities, which represent crucial support to enhance regional innovation capability. A social network is a geometric shape composed of multiple points and connecting lines, which is consistent with the analysis of an innovation network. The social network analysis method was first proposed by Brown in 1930. The method uses quantitative indicators to describe the interactive structural relationship and development changes between individuals to reflect the overall characteristics of the network structure and reveals the status of individuals in the network structure. Social network analysis provides a series of indicators to explain the innovation network. This study selects centrality to study the spatial structure and evolution characteristics of the spillover line among urban innovation networks.

Reflecting the importance (Wang et al. 2015) and status (Ibarra 1993) of the subject in the network, centrality is a quantitative analysis of individual power, which is used to measure the position of a node in the network mainly including degree centrality, intermediate centrality, and proximity centrality. In this study, degree centrality is selected to measure the central position of each city in the innovation network to simplify the form of the spillover line. If a city has innovation spillover with multiple cities, this indicates that the city is in a central position within the innovation network. The formula is as follows:

D C_{i} = \frac{r_{i j}}{n - 1}

(12)

In the formula, DC_i is the relative degree centrality of city i, which refers to the ratio of the absolute degree centrality of city i to the maximum possible degree of points in the network. r_ij represents the number of spillover lines between city i and other cities; that is, the absolute degree centrality of city I, and n is the number of innovation cities.

Data and Samples

Macro level innovation is abstract, and micro level research focuses more on fitting the technology diffusion curve (Comin and Hobijn 2010) while cities lie between the macro and micro level. Cities are the main driving force of space (Baycan, Nijkamp, and Stough 2017), but the understanding of urban spillovers are a much less significant driver (Fujita and Krugman 2004). The research of Audretsch and Feldman (1996) showed that it is more appropriate to select cities to study the spillover effect of innovation. Therefore, city level data are more suitable for this study.

In 2008, Shenzhen was determined to be the first national innovative city in China. By 2020, 78 innovative cities were approved. These cities have absolute advantages in both the level of economic development and innovative resources. Due to the availability of data and the late establishment of some innovative cities, this study selects 75 innovative cities in mainland China (excluding Shihezi, Changji and Lhasa). From 2003 to 2017, there were 285 cities in mainland China with long-term and comprehensive statistical data while there were 210 non-innovative cities. Taking 2017 as an example, 75 innovative cities have higher levels of economic development and larger urban areas than 210 non-innovative cities, and the labour and financial expenditures of innovative input and patents for innovative output were far higher than those of non-innovative cities (as shown in Figure 3).

Figure 3.

Comparison between innovative city and non innovative city in 2017.

The original data of each index in this study mainly come from the China Urban Statistical Yearbook and relevant statistical yearbook and bulletins for each city. The missing data are Supplemental Material by the exponential smoothing method. In addition, the geographic longitude and latitude coordinate data in the spatial econometric regression model are from the longitude and latitude query network of urban areas, and the inter-city distance data in the innovative connection degree calculation are from the national urban mileage query network.

Empirical Analysis

Spatial Statistical Analysis

When the fixed-effect model for benchmark regression is used before the spatial econometric analysis, labour flows and capital flows are found to have no significant impact on innovation output. This may be due to the lack of consideration of spatial spillovers. If the identification problem is not resolved, biased estimates may be generated (Stephen and Overman 2012). Therefore, instrumental variables are used to ensure the robustness of the results and exclude the reverse causality between labour flows, capital flows, and innovation output. Here, the cross-term of the number of regional dialects and time trend is used as the instrumental variable. Regional dialects may affect the flow of labour among different regions (Liu, Jiao, and Xu 2020; Chiswick and Miller 2015). Meanwhile, as a cultural factor formed in history, regional dialects will not directly affect the current innovation output. The logarithm of the urban green area is selected as the instrumental variable of capital flow. The urban green area is a manifestation of liveability and more liveable cities may attract more foreign capital for investment. At the same time, the urban green area will not directly affect innovation output. The 2SLS (two-stage least square) regression is carried out using instrumental variables. The data of the two instrumental variables, i.e., the number of regional dialects and the urban green area, come from the Dictionary of Chinese Dialects and the China Urban Statistical Yearbook, respectively. It is found that labour flows and capital flows still have no significant impact on innovation output, which is the same as the benchmark regression. This implies that the endogenous problem is not apparent. Additionally, the results of benchmark regression also show that the results without considering spatial factors deviate from reality.

Spatial econometric regression is based on spatial statistics. Therefore, this study conducts spatial statistical analysis on the scientific and technological innovation input of the independent variables and dependent variable through the global Moran’I index to reveal the spatial correlation and distribution characteristics. The results are shown in Table 1. During the measurement period, Moran’I is greater than 0.2 except for the data for the number of people engaged in scientific and technological innovation in 2011. This result indicates that the data selected in this study have strong spatial relevance, which is suitable for spatial regression. In addition, we also tested the cross-sectional dependence and found that there is no cross-sectional dependence, so it does not lead to downward attenuation bias in the estimates.

Table 1.

Results of Spatial Statistical Analysis.

	2003	2004	2005	2006	2007	2008	2009	2010
The number of patents	0.2533	0.2692	0.2973	0.3730	0.3713	0.4640	0.5180	0.5900
The number of people engaged in scientific and technological innovation	0.2195	0.2096	0.2300	0.2222	0.2655	0.2092	0.2219	0.2605
The scientific and technological expenditure	0.2335	0.2374	0.2763	0.3736	0.4740	0.4244	0.3545	0.4174
The stock of fixed assets	0.6582	0.6622	0.6709	0.6810	0.6859	0.6915	0.6966	0.6951
	2011	2012	2013	2014	2015	2016	2017
The number of patents	0.7937	0.8735	0.8234	0.7040	0.6557	0.6342	0.5130
The number of people engaged in scientific and technological innovation	0.1329	0.2862	0.2739	0.2901	0.3100	0.2470	0.1876
The scientific and technological expenditure	0.5436	0.5517	0.5236	0.4956	0.3572	0.2982	0.2763
The stock of fixed assets	0.6904	0.6917	0.6939	0.6970	0.6925	0.6824	0.6762

Network Association Analysis

In this study, degree centrality is selected to describe the importance of different cities in the network to integrate all the spillover lines corresponding to each city. To render the innovation network centrality of different time points comparable, the degree centrality index selected in this study is relative degree centrality. The network/centrality function of UCINET software is used to calculate the centrality value of degree centrality for the three dimensions of labour flow, capital flow, and institutional learning of each city from 2003 to 2017 (as shown in Figure 4). These calculations will preliminarily reveal the basic law of the bandwidth variation of the innovative urban spillover line.

Figure 4.

Innovation network degree centrality.

In terms of the spillover line for labour flow in the innovation network, except Beijing, Shanghai, and Nanjing in 2003, the degree centrality for each city is generally low, the spatial difference is significant, and the degree centrality of cities in the western region is extremely low. Among them, the degree centrality of Nanning, Haikou, and Urumqi is zero, and the high-value areas are mainly concentrated in Beijing-Tianjin-Hebei, Yangtze River Delta, and some provincial capitals of other regions. These findings show that the core node cities of labour flow in 2003 have weak centrality, limited radiation effect and driving ability, and there is significant disparity in status between each node city in the network with significant polarization. During the measurement period, all cities show a trend of rising fluctuation. After Beijing, the regions of Shanghai, Tianjin, Xi’an, Chengdu, Wuhan, Guangzhou, Hangzhou, and other provincial capitals stand out and drive the growth of the surrounding cities. These findings show that the degree centrality of labour flow tends to spread from the eastern region to the central and western regions, the main core node cities have significant spillover effects, and the overall development of the network tends to be balanced. However, the highest value of Urumqi, Jilin, and Harbin is less than 10 in the whole measurement period, which represents an important crux of restricting the development of labour flow in innovation networks.

In terms of the spillover line of capital flows in the innovation network, in 2003, the starting point of degree centrality of all sample cities is generally low, only Shanghai, Nanjing, and Suzhou in the eastern region have a value of more than 20, of which Shanghai has a maximum value of 24.32. The value of some non-capital cities in Central China, such as Inner Mongolia, Shanxi, Henan, Jiangxi, and Hubei, and all sample cities in Western China, such as Guizhou, Yunnan, Shaanxi, Gansu, Qinghai, Ningxia, and Xinjiang, is zero. These findings show that in 2003, the centrality of cities in terms of capital flow of innovation networks was weak, and their status in the network was almost the same. During the measurement period, Shanghai, Beijing, Tianjin, and cities in Jiangsu, Zhejiang, Shandong, and other eastern regions form high-value areas, which creates a gap with other cities. In a later period, the growth rate of cities in high-value areas gradually flattens, and the centrality value tends to be concentrated between 75 to 100 while the values of other cities are scattered between 0 to 60, indicating that the centrality of core nodes in the network has increased but the polarization phenomenon is significant. In 2017, the growth rate of cities previously in high-value areas gradually flattened while some central and western cities, such as Hengyang, Baoji, and Guiyang, experienced significant growth and are expected to enter high-value areas in the coming period. These findings imply that the capital flow centre of the innovation network is in the process of spreading to the central and western regions, the polarization situation has eased, and the overall development of the network is more balanced. The capital flow degree centrality of Jilin and Harbin is at a relatively high level in the whole measurement period, which is different from the lagged state of talent flow degree centrality.

In terms of the spillover line for institutional learning flow in the innovation network, the degree centrality of each city in 2003 is low. Except for Shanghai (41.89) and Beijing (24.32), other cities with relatively high degree centrality are concentrated in Suzhou and Hangzhou, but the degree centrality is not more than 20, and approximately 90% of the cities have degree centrality less than 10, which indicates that the core nodes of the network in 2003 are Shanghai and Beijing, and the spatial difference between cities is not obvious. During the measurement period, the degree centrality of most cities increased rapidly except for some cities in the western region. The reason is that each city increased its expenditure on science and technology, which has vastly improved the degree centrality of institutional learning in the innovative network as a whole. At the same time, the network as a whole began to differentiate into high value areas (including Beijing, Shanghai, Tianjin, and eastern cities) and low value areas (including most cities in the western region), and the western region gradually separated from the central and eastern regions. In 2017, the growth of cities in high value areas slowed while the degree centrality of cities in low value areas was still lower than that in high value areas, but the growth rate was relatively high. This finding shows that cities in the western region are accelerating their innovation spillover and moving closer to the core nodes in the central and eastern regions. The main core nodes have significant urban spillover effect, which benefits the surrounding areas and develops in a group way. The overall development of the network is more mature than before.

As a whole, the degree centrality of each city improved to a certain extent from 2003 to 2017 and maintained an overall upward trend, indicating that the degree of innovation connection between cities is constantly improving. The average degree centrality also shows that the internal connection of innovation networks of innovative cities is constantly expanding. From a comparison of cities, the average degree centrality of the three dimensions of talent flow, capital flow, and institutional learning in Beijing, Shanghai, and Nanjing for the whole measurement period is at a high level, which shows that these three cities are always at the core of the innovation network. In the spatial dimension, the core status of the eastern region is better than that of the central and western regions.

Spatial Regression Analysis

Using the quantitative spillover line and spatial econometric model, we can analyse the spillover between cities and the secondary spillover of the spillover line among cities. In this study, we calculate three cases, which include only the spillover among cities (case 1), only the spillover of the spillover line among cities (case 2), and both of these (case 3). For these indicators, because the spillover lines of labour flow, capital flow, and institutional learning after integration are non-normal, they are treated with logarithm.

The Wald spatial lag test, LR spatial lag test, Wald spatial error test, and LR spatial error test all reject the original hypothesis, indicating that the spatial Durbin model should be used, which can neither be transformed into a spatial lag model nor a spatial error model. In addition, the Hausman test supports the fixed effect. The regression results are shown in Table 2.

Table 2.

Results of Spatial Regression Analysis.

	Case 1	Case 2	Case 3
Labour	620.506		567.158
	(0.195)		(0.243)
Capital	0.003***		0.002***
	(0.000)		(0.000)
Government support	−0.050*		−0.020
	(0.057)		(0.451)
Labour flow		11369.280**	10872.580***
		(0.012)	(0.008)
Capital flow		−5113.896**	−3729.453*
		(0.021)	(0.069)
Institutional learning		3572.777**	3478.849**
		(0.016)	(0.012)
Talent education	117.274	−650.268***	−17.863
	(0.518)	(0.004)	(0.932)
Industrial structure	−3195.145***	−4078.953***	−4360.620***
	(0.010)	(0.001)	(0.001)
Infrastructure	−13.740	−11.976	−40.859
	(0.660)	(0.732)	(0.199)
Opening-up level	624.466	−2885.836***	118.284
	(0.321)	(0.000)	(0.857)
Rho	−0.521*	−0.781***	−0.547*
	(0.074)	(0.008)	(0.074)
R ²	0.6829	0.5798	0.6918
Adjusted R²	0.6809	0.5772	0.6890

Note: ***, **, * indicates statistical significance at the 1%, 5%, 10% level, respectively. The values in parentheses are p statistics.

In case 1, capital and government support have a significant impact on innovation without considering the spillover of spillover lines among cities. The impact coefficient of capital is 0.003, and the 1% significance test shows that capital has a positive role in promoting innovation. The influence coefficient of government support is −0.050, and it has passed the 10% significance test. This indicates that government support has a restraining effect on innovation, which may be due to the crowding out effect of government support on enterprise innovation. Labour force did not pass the significance test. Among the control variables, only industrial structure has a significant negative impact on innovation.

In case 2, labour flow, capital flow, and institutional learning have a significant impact on innovation without considering spillover among cities. Labour flow and institutional learning can promote innovation, and they passed the 5% significance test. Capital flow has a restraining effect on innovation and passed the 5% significance test, which may be due to the siphon effect of capital in the process of capital flow resulting in capital flows to the most innovative cities, thus reducing the innovation level of the whole innovation network. Among the control variables, talent education, industrial structure, and the level of opening-up all have a significant negative impact on innovation.

In case 3, considering both the spillover between cities and the second spillover of the spillover line between cities, capital has played a role in promoting innovation and passed the 1% significance test. Labour and government support, however, have no significant impact on innovation. Labour flow, capital flow, and institutional learning have a significant impact on innovation. Labour flow and institutional learning promote innovation, and they passed the 1% and 5% significance test, respectively. Capital flow has an inhibitory effect on innovation and passed the 10% significance test. Among the control variables, only industrial structure has a significant negative impact on innovation.

In all cases, capital plays a significant role in promoting innovation. Similarly, the promotion of labour flow and institutional learning with respect to innovation and the restraint of capital flow on innovation are significant. At the same time, we find that when considering all factors, the influence coefficients of capital, labour flow, capital flow, and institutional learning are smaller than those of case 1 and case 2, indicating that previous studies may exaggerate the influence of some factors. The significant inhibitory effect of government support in case 1 no longer has a significant impact on innovation after labour flow, capital flow, and institutional learning are added, indicating that the dependence and mobility between factors have a more direct impact on innovation. The reason for this phenomenon may be that the crowding out effect of government support may inhibit innovation when factor flow is not considered. However, in the case of factor flow, the crowded out elements and resources can flow through innovation networks without restraining innovation.

From a comparison of the spillover among cities and the second spillover of the spillover line among cities in case 3, we found that the labour force in cities does not have a spatial impact on innovation, but labour flow can promote innovation. Similar to the labour force, government support has no spatial impact on innovation, but institutional learning between cities can promote innovation. Capital can promote innovation, but capital flow restrains innovation because of the siphon effect. From a comparison of case 3 with other cases, and a comparison of the element itself and the element flow in case 3 when analysing innovation spillover, we should consider the point-to-point spillover characteristics of the element itself and the spillover characteristics of the spillover line generated by element flow.

Based on case 3, this study further splits the influence of explanatory variables and control variables to analyse the influence of elements on the city, other cities, and the whole, as shown in Table 3.

Table 3.

The split of the Influence Coefficient of Innovation Factors.

	Direct	Indirect	Total
Labour	−12.404	415.116	402.712
	(0.650)	(0.258)	(0.278)
Capital	0.0002***	0.002***	0.002***
	(0.000)	(0.004)	(0.002)
Government support	0.022***	−0.023	−0.001
	(0.000)	(0.192)	(0.962)
Labour flow	532.282*	7693.272**	8225.553**
	(0.070)	(0.044)	(0.037)
Capital flow	288.617**	−2915.708	−2627.091
	(0.038)	(0.144)	(0.202)
Institutional learning	762.605***	2229.031	2991.636**
	(0.000)	(0.106)	(0.038)
Talent education	−51.091***	15.765	−35.326
	(0.000)	(0.920)	(0.824)
Industrial structure	248.330***	−3064.316**	−2815.986**
	(0.000)	(0.014)	(0.026)
Infrastructure	−4.700***	−30.734	−35.434
	(0.000)	(0.172)	(0.118)
Opening-up level	29.106*	70.769	99.875
	(0.078)	(0.885)	(0.838)

Note: ***, **, * indicates statistical significance at the 1%, 5%, 10% level, respectively. The values in parentheses are p statistics.

Labour force has no significant spatial impact on innovation, which may be because innovation is a long-term process leading to a lag in the labour force’s impact on innovation. Capital plays a significant role in promoting innovation in local and other cities, so the total effect of capital is also positive, indicating that capital plays an important role in innovation. Government support only promotes local innovation, but does not affect innovation in other cities, so its total effect is not significant. Therefore, for the elements themselves, only capital spillover can promote innovation while government support can only promote local innovation, and the conclusion is consistent with the actual situation.

Labour flow has a significant role in promoting innovation in local and other cities, so the total effect of labour flow is positive, indicating that labour flow plays an important role in innovation. Capital flow can promote local innovation, but its effect on other cities is not obvious, so the overall effect on innovation is not significant as well. Institutional learning can promote local innovation but has no significant impact on other cities, so its total effect is to promote innovation. Therefore, for the mobile elements, labour flow can produce a positive spillover effect and promote innovation in other cities. Capital flow and institutional learning only affects the local area, but does not disturb other cities, which is consistent with the actual situation.

For the control variables, talent education inhibits local innovation because education is a long-term project. In the short term, investing capital and labour in education rather than innovation will inhibit innovation. However, in the long term, education can promote innovation by cultivating talents. The proportion of the secondary industry promotes local innovation because innovation mainly occurs in the secondary industry. When the proportion of the secondary industry of a city increases, it creates a crowding out effect on other cities reducing the proportion of the secondary industry of other cities and thus inhibiting the innovation of other cities. Infrastructure inhibits local innovation, while the level of opening-up promote local innovation, but they do not have a spillover effect.

Robustness Test

In order to test the reliability of the results, the robustness test is carried out from the following three aspects:

(1) Addition of a new control variable. With the increasing pressure brought by the environment, the world pays more attention to environmental protection. China takes the construction of ecological civilization as one of the main directions of development, and cities have accordingly issued a series of policies and paid more attention to environmental protection. Therefore, this paper adds environmental regulation as a new control variable to the spatial regression which expressed by SO₂ emission. The spatial regression result with the new control variable is generally consistent with the original result, indicating that the original result is robust.

(2) Replacement of spatial matrix. The original result uses the geographic distance matrix. However, in addition to geographical distance matrix, economic distance matrix and technical distance matrix are often used to analyze spatial problems. The technology distance matrix is generally constructed by the different types of patents in different cities, and the explanatory variable in this paper is the number of patents, which may lead to the deviation of the results. Therefore, this paper uses the economic distance matrix to replace the geographical distance matrix for robustness test, and uses the absolute value of per capita GDP between cities to construct the economic distance matrix. The spatial regression result of the replacement spatial matrix are generally consistent with the original result, indicating that the original result is robust.

(3) Replacement of spatial regression model. The static spatial model (including SAR, SEM ,and SDM models) only contains the spatial lag term. In contrast, the dynamic SDM model takes the time lag term of the explained variables into account. Here, the dynamic SDM model is used for spatial regression, and the function form is as follows. Compared with the SDM model, the dynamic SDM model takes into account the time lag term y’ and space-time lag term Wy’ of the explained variables, τ and μ are the corresponding coefficients. Considering that the completion of the patent takes a long time, we set the lag to 2. The result of the dynamic SDM model is generally consistent with the original result, indicating that the original result is robust.

dynamic SDM: y = τ y' + μ W y' + ρ W y + X β + W X γ + ε

(13)

Discussion and Conclusions

Discussion

The spillover of innovation may not only represent the point-to-point relationship between cities but also secondary spillover on the spillover line among cities. The spillover between points is realized by a linear relationship while the spillover of the spillover line is realized by the innovation network and a spatial relationship. First, this study finds that the input and output of innovation have strong spatial correlation by conducting a spatial statistical analysis. Then, this study quantifies the linear relationship between cities through the gravity model and social network model. The results show that: (1) the centrality of labour flow shows a trend of rising volatility, the main core node city spillover effect is significant, and the overall development of the network tends to be balanced. (2) During the measurement period, the polarization trend of capital flow centrality has eased, and the overall development of the network has become more balanced. (3) The centrality of institutional learning has increased significantly, and the main core node cities can benefit the surrounding areas and develop together.

Finally, this study quantifies the linear relationship between cities using the gravity model and the social network model. The models take the labour force, capital, government support, and the corresponding labour flow, capital flow, and institutional learning as explanatory variables, the number of patents as explanatory variables, and analyses the spatial impact of the factors themselves and the factor flows using a spatial econometric model. The results show that (1) there are secondary spillovers in the spillover lines of labour flow, capital flow, and institutional learning. (2) Capital, government support, labour flow, capital flow, and institutional learning promote local innovation. (3) Labour flow promote innovation in other cities, while capital flow and institutional learning have no significant impact on other cities. (4) The total effect of capital, labour flow, and institutional learning is positive, while the total effect of capital flow is not significant.

Based on the empirical analysis, we believe that considering both the spillover among points and re-spillover among lines may better explain the spillover effect of innovation. Firstly, by comparing case 3 with other cases, we found that case 3’s explanation of innovation reaches 69.18%, which is higher than case 1 (68.29%) and case 2 (57.98%), indicating that case 3’s model fitting is more in line with the actual situation. Secondly, if more variables are added, R2 will be higher but consume more degrees of freedom, which is not necessarily a better choice. Therefore, we also report the adjusted R2 in Table 2, also proving that case 3 is the best and not due to the addition of more variables. Thirdly, in case 1, government support inhibits innovation, which is far from people’s general cognition, while the result of case 3 is more in line with people’s general cognition. Finally, the coefficients of each core variable in case 3 are less than both of cases 1 and 2, which may mean that the spillover effect of innovation is the sum of the spillover among points and the re-spillover among lines.

Managerial and Policy Implications

Labour flow among cities is conducive to the traditional allocation of innovative resources and the dissemination of knowledge. In view of the positive impact of labour flow on innovation, cities should focus on talent cultivation and talent exchange. The labour flow among the western region of China and other regions is slow, which is not conducive to the western region transforming its current low level of innovation. To increase the knowledge spillover and innovation interaction among cities, cities should improve the talent exchange mechanism among cities or deepen household registration system reform, develop a new mode of collaborative innovation among cities, and break the barriers of inter-city talent flow.

Capital plays a significant role in promoting local innovation. Cities should form their own capital accumulation, particularly in the western regions whose capital organic composition has not reached the saturation state. In a short period, capital accumulation can still be used to develop capital biased innovation. The siphon effect of capital flow will reduce the innovation level of the network, and aggressive capital competition should be avoided between cities. In particular, the eastern region should avoid plundering the capital of the central and western regions.

In view of the role of institutional learning in promoting urban innovation, cities should follow a factor endowment structure, find their own advantages, assume the best position in the national innovation and development network, learn from the advanced experience of other cities' innovation and development, further improve on the basis of existing systems, and form an effective innovation policy system and governance structure.

Theoretical Contributions

From the results of spatial regression in different cases, the following general conclusions can be drawn:

(1) Considering the factors and their secondary effects together may be possible and would fit the regression equation better and improve interpretation. In this study, when considering labour, capital, and government support, we also consider labour flow, capital flow, and institutional learning so that the degree of fit for the spatial econometric regression equation is higher.

(2) Some factors may have a significant spatial impact on the results without considering the secondary impact, but they may not have a significant spatial impact when considering the secondary impact. In this study, government support can have a significant negative impact on innovation because of the crowding out effect, but when considering the flow of factors, the crowding out effect is diluted in the innovation network, and there is no significant spatial impact on innovation from government support.

(3) If we only consider the spatial effect of the factors themselves, or only consider the secondary effect of the flow of factors, the result may exaggerate the spatial effect of the factors themselves or the flow of factors. In this study, the coefficients of capital, labour flow, capital flow, and institutional learning are smaller when considering the spatial effects of all factors.

(4) The factors themselves may not have a significant spatial impact on the results, but the flow of factors may have a significant impact on the results. In this study, labour has no significant impact on innovation, but labour mobility has a significant role in promoting innovation.

(5) The factors themselves have significant spatial impacts on the results, but the spatial impacts of the flow of factors on the results may be not significant. In this study, capital only has a significant role in promoting innovation in local.

Limitations and Future Research Directions

We believe that considering both the spillover among points and the re-spillover among lines simultaneously may better explain the spillover effect of innovation and list some reasons. But these reasons are slightly insufficient to prove our point of view. Firstly, various scholars have various opinions on whether the value of R2 or adjusted R2 can prove the rationality of the model. We also agree that both the values of R2 and adjusted R2 have defects, but we have no better proof method at present. Similarly, the significance and coefficient of variables are also difficult to provide sufficient support for the model’s applicability. Through several apparent phenomena, we can only infer that the spillover effect of innovation should include the spillover among points and the re-spillover among lines. But more scientific and accurate proof needs further discussion.

Because of the universal spatial relationship, spillover occurs in other fields. In addition, innovation is a macro concept, which involves labour flow, capital flow, institutional learning, and the flow of other factors, which may also generate spillover. Therefore, the conclusion of this study may be more general, and the study of spillover should pay more attention to second spillover caused by the flow of factors.

Supplemental Material

Supplemental Material - Spillover and Re-Spillover in China’s Collaborative Innovation

Supplementary Material for Spillover and Re-Spillover in China’s Collaborative Innovation by Song Wang, Jiexin Wang, Yixiao Wang and Xueli Wang in International Regional Science Review

Footnotes

Acknowledgments

This article has been corrected by Editage () in English.

Declaration of Conflicting Interest

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The authors are thankful for the financial support from the National Natural Science Foundation of China (42071154), China Postdoctoral Science Foundation (2020M680960), Project of the Fundamental Research Funds for the Central Universities (N2206011), Postdoctoral foundation of Northeastern University (20210201), Major Program of National Fund of Philosophy and Social Science of China (21ZDA011), Youth project of high-end science and technology innovation think tank of China Association for science and technology (2021ZZZLFZB1207138), Humanities and Social Science Fund of Ministry of Education of the People’s Republic of China (20YJA790010).

ORCID iD

Song Wang

Supplemental Material

Supplemental material for this article is available online.

References

Acemoglu

Akcigit

Alp

Bloom

Kerr

W. R.

. 2013. Innovation, Reallocation and Growth. Cambridge, MA: NBER. NBER Working Papers.

Acs

Audretsch

Feldman

. 1994. “R&D Spillovers and Recipient Firm Size.” Review of Economics and Statistics 76 (2): 336‐340.

S. W.

Miao

C. H.

. 2010. “Space of Place”, “Space of Flows” and “Space Of Actor - Networks” from the Perspective of ANT.” Human Geography 2: 43‐49.

Almeida

Kogut

. 1999. “Localization of Knowledge and the Mobility of Engineers in Regional Networks.” Management Science 45: 905‐916.

Amir

2000. “Modelling Imperfectly Appropriable R&D Via Spillovers.” International Journal of Industrial Organization 18 (7): 1013‐1032.

Deng

Chao

Bai

. 2014. “Knowledge Management in Support of Collaborative Innovation Community Capacity Building.” Journal of Knowledge Management 18 (3): 574‐590.

Anselin

Le Gallo

Jayet

. 2008. “Spatial Panel Econometrics.” In The Econometrics of Panel Data: Fundamentals and Recent Developments in Theory and Practice, edited by Mátyás

Sevestre

, Berlin: Springer - Verlag.

Anselin

Varga

Acs

Z. J.

. 1997. “Local Geographic Spillovers Between University Research and High Technology Innovations.” Journal of Urban Economics 42 (3): 422‐448.

Arieli

Babichenko

Peretz

Young

H. P.

. 2020. “The Speed of Innovation Diffusion in Social Networks.” Econometrica 88 (2): 569‐594.

10.

Asheim

B. T.

Smith

H. L.

Oughton

. 2011. “Regional Innovation Systems: Theory, Empirics and Policy.” Regional Studies 45 (7): 875‐891.

11.

Aswegen

Retief

F. P.

. 2020. “The Role of Innovation and Knowledge Networks as a Policy Mechanism Towards More Resilient Peripheral Regions.” Land Use Policy 90: 104259.

12.

Audretsch

D. B.

Feldman

M. P.

. 1996. “R&D Spillovers and the Geography of Innovation and Production.” American Economic Review 86: 630‐640.

13.

Autant-Bernard

2012. “Spatial econometrics of Innovation: Recent Contributions and Research Perspectives.” Spatial Economic Analysis 7 (4): 403‐419.

14.

Autant-Bernard

Fadairo

Massard

. 2013. “Knowledge Diffusion and Innovation Policies within the European Regions: Challenges Based on Recent Empirical Evidence.” Research Policy 42 (1): 196‐210.

15.

Autant-Bernard

Lesage

J. P.

. 2011. “Quantifying Knowledge Spillovers Using Spatial Econometric Models.” Journal of Regional Science 51 (3): 471‐496.

16.

Autant-Bernard

Massard

. 2009. “Report on the Underlying Mechanisms of knowledge diffusion.” IAREG Deliverable 4, 2b.

17.

Bai

Wei

Yang

Chen

. 2020. “Third-Party Relational Governance and Collaborative Innovation Performance: The Role of IPR Protection.” International Journal of Innovation Studies 4 (1): 1‐15.

18.

Barboza

G. A.

Pede

V. O.

. 2012. “A Spatial Econometric Analysis of Cultural Dimensions, Technology Knowledge and Economic Growth.” Review of Development Economics 16 (2): 262‐275.

19.

Baycan

Nijkamp

Stough

. 2017. “Spatial Spillovers Revisited: Innovation, Human Capital and Local Dynamics.” International Journal of Urban and Regional Research 41 (6): 962‐975.

20.

Benedetti

Palma

Postiglione

. 2019. “Modeling the Impact of Technological Innovation on Environmental Efficiency: A Spatial Panel Data Approach.” Geographical Analysis 52: 1‐23.

21.

Blomström

Kokko

. 1998. “Multinational Corporations and Spillovers.” Journal of Economic Surveys 12: 247‐277.

22.

Borgatti

S. P.

Foster

P. C.

. 2003. “The Network Paradigm in Organizational Research: A Review and Typology.” Journal of Management 29: 991‐1013.

23.

Boschma

2005. “Proximity and innovation: a critical assessment.” Regional Studies 39 (1): 61‐74.

24.

Bottazzi

Peri

. 2003. “Innovation and spillovers in regions: Evidence from European Patent Data.” European Economic Review 47 (4): 687‐710.

25.

Breschi

Lissoni

. 2001. “Knowledge Spillovers and Local Innovation System: A Critical Survey.” Industrial and Corporate Change 10, 975‐1005.

26.

Brown

J. S.

Duguid

. 1991. “Organizational Learning and Communities of Practice: Towards a Unified View of Working, Learning and Organization.” Organization Science 2: 40‐57.

27.

Bröcker

Mitze

. 2019. Factor Mobility and Migration Models. Handbook of Regional Science. Berlin: Springer, https://doi.org/10.1007/978-3-642-36203-3_42-1

28.

Buera

F. J.

Oberfield

. 2020. “The Global Diffusion of Ideas.” Econometrica 88 (1): 83‐114.

29.

Cabrer-Borrás

Serrano-Domingo

. 2007. “Innovation and R&D Spillover Effects in Spanish Regions: A Spatial Approach.” Research Policy 36 (9): 1357‐1371.

30.

Caccamo

2020. “Leveraging Innovation Spaces to Foster Collaborative Innovation.” Creativity and Innovation Management 29, 178‐191.

31.

Cantore

Cheng

C. F. C.

. 2018. “International Trade of Environmental Goods in Gravity Models.” Journal of Environmental Management 233, 1047‐1060.

32.

Cao

Zeng

Teng

. 2018. “The best Spatial Scale of Firm Innovation Networks: Evidence from Shanghai High‐Tech Firms.” Growth and Change 49, 696‐711.

33.

Cao

Zeng

. 2019. “The Structure and Proximity Mechanism of Formal Innovation Networks: Evidence from Shanghai High‐Tech ITISAs.” Growth and Change 50: 569‐586.

34.

Capello

1999. “Spatial transfer of Knowledge in High Technology Milieux: Learning Versus Collective Learning Processes.” Regional Studies 33, 353‐365.

35.

Capello

Lenzi

. 2018a. “The Dynamics of Regional Learning Paradigms and Trajectories.” Journal of Evolutionary Economics 28 (4): 727‐748.

36.

Capello

Lenzi

. 2018b. “Regional Innovation Patterns from an Evolutionary Perspective.” Regional Studies 52 (2): 159‐171.

37.

Cardamone

2018. “Firm innovation and Spillovers in Italy: Does Geographical Proximity Matter?” Letters in Spatial and Resource Sciences 11, 1‐16.

38.

Carrère

2006. “Revisiting the Effects of Regional Trade Agreements on Trade Flows with Proper Specification of the Gravity Model.” European Economic Review 50 (2): 223‐247.

39.

Castells

1996. The Rise of the Network Society. Cambridge, MA: Blackwell.

40.

Chen

Y. S.

de Jong

J. P. J.

Hippel

. 2020. “Household Sector Innovation in China: Impacts of Income and Motivation.” Research Policy 49 (4): 103931.

41.

Chen

Huang

. 2018. “A scaling Approach to Evaluating the Distance Exponent of the Urban Gravity Model.” Chaos, Solitons & Fractals 109: 303‐313.

42.

Chiswick

B. R.

Miller

P. W.

. 2015. International Migration and the Economics of Language. Handbook of the Economics of International Migration, 1, 211‐269. Amsterdam, The Netherlands: Elsevier.

43.

Cipollina

Demaria

Pietrovito

. 2016. “Determinants of Trade: The Role of Innovation in Presence of Quality Standards.” Journal of Industry, Competition and Trade 16: 455‐475.

44.

Cohen

D. H.

Kozak

R. A.

. 2002. “Research and Technology: Market - Driven Innovation in the Twenty‐First Century.” Forestry Chronicle 78 (1): 108‐111.

45.

Comin

Hobijn

. 2010. “An Exploration of Technology Diffusion.” American Economic Review 100, 2031‐2059.

46.

Conley

T. G.

Ligon

E. A.

. 2002. “Economic Distance, Spillovers, and Cross Country Comparisons.” Journal of Economic Growth 7, 157‐187.

47.

Davis

D. R.

Dingel

J. I.

. 2019. “A Spatial Knowledge Economy.” American Economic Review 109 (1): 153‐170.

48.

Desmet

Hansberg

E. R.

. 2014. “Spatial Development.” American Economic Review 104 (4): 1211‐1243.

49.

Desrochers

2001. “Geographical Proximity and the Transmission of Tacit Knowledge.” Review of Austrian Economics 14 (1): 25‐46.

50.

Dunning

J. H.

Kim

Park

. 2007. Old Wine in New Bottles: A Comparison of Emerging-Market TNCs Today and Developed-Country TNCs Thirty Years Ago. Camberley: Edward Elgar.

51.

Emden

Calantone

R. J.

Droge

. 2006. “Collaborating for New Product Development: Selecting the Partner with Maximum Potential to Create Value.” Journal of Product Innovation Management 23: 330‐341.

52.

Fan

Lian

Wang

. 2020. “Can Regional Collaborative Innovation Improve Innovation Efficiency? An Empirical Study of Chinese Cities.” Growth & Change 51: 440‐463.

53.

Fang

X. Y.

Ren

Zhou

. 2014. “Antecedents of Network Capability and Their Effects on Innovation Performance: An Empirical Test of Hi - tech Firms in China.” Creativity and Innovation Management 23 (4): 436‐452.

54.

Feng

Wang

. 2019. “Effects of Environmental Regulation and FDI on Urban Innovation in China: A spatial Durbin Econometric Analysis.” Journal of Cleaner Production 235: 210‐224.

55.

Fischer

M. M.

LeSage

J. P.

. 2008. “Spatial Growth Regressions: Model Specification, Estimation and Interpretation.” Spatial Economic Analysis 3: 275‐304.

56.

Folke

Hahn

Olsso

Norberg

. 2005. “Adaptive Governance of Social-Ecological Knowledge.” Annual Review of Environment and Resources 30: 441‐473.

57.

Forman

Goldfarb

Greenstein

. 2016. “Agglomeration of Invention in the Bay Area: Not Just ICT.” American Economic Review: Papers & Proceedings 106 (5): 146‐151.

58.

Fujita

Krugman

. 2004. “The New Economic Geography: Past, Present and the Future.” Papers in Regional Science 83: 139‐164.

59.

Furková

2019. “Spatial Spillovers and European Union Regional Innovation Activities.” Central European Journal of Operations Research 27: 815‐834.

60.

Gama

Maret

Masson

. 2018. “Endogenous Heterogeneity in Duopoly with Deterministic One‐Way Spillovers.” Annals of Finance 15: 103-123.

61.

Gardet

Mothe

. 2011. “The Dynamics of Coordination in Innovation Networks.” European Management Review 8: 213‐229.

62.

Gertler

2003. “Tacit Knowledge And The Economic Geography of Context or the Undefinable Tacitness of Being (There).” Journal of Economic Geography 3 (1): 75‐99.

63.

Guo

2015. “On the Spillovers Between Patents and Innovation in Japan.” Journal of Economics and Finance 39 (3): 590‐605.

64.

Hall

B. H.

Adam

Manuel

. 2005. “Market Value and Patent Citations.” RAND Journal of Economics 36 (1): 16‐38.

65.

Hammadou

Paty

Savona

. 2014. “Strategic Interactions in Public R&D Across European Countries: A Spatial Econometric Analysis.” Research Policy 43 (7): 1217‐1226.

66.

Hansen

2014. “Juggling with Proximity and Distance: Collaborative Innovation Projects in the Danish Cleantech Industry.” Economic Geography 90 (4): 375‐402.

67.

Hardeman

Frenken

Nomaler

Ter Wal

A. L. J.

. 2014. “Characterizing and Comparing Innovation Systems by Different \\"Modes\\" of Knowledge Production: A Proximity Approach.” Science and Public Policy 12, 530-548.

68.

Harrisson

Laberge

. 2002. “Innovation, Identities and Resistance: The Social Construction of an Innovation Network.” Journal of Management Studies 39 (4): 497‐521.

69.

Heil

Bornemann

. 2018. “Creating shareholder Value Via Collaborative Innovation: The Role of Industry and Resource Alignment in Knowledge Exploration.” R&D Management 48 (4): 394‐409.

70.

Hjaltadóttir

R. E.

Makkonen

Mitze

. 2020. “Inter-Regional Innovation Cooperation and Structural Heterogeneity: Does Being a Rural, or Border Region, or Both, Make a Difference?” Journal of Rural Studies 74: 257‐270.

71.

Hoekman

Frenken

Tijssen

R. J. W.

. 2010. “Research collaboration at a distance: Changing Spatial Patterns of Scientific Collaboration within Europe.” Research Policy 39 (5): 662‐673.

72.

Howard

Steensma

H. K.

Lyles

Dhanaraj

. 2016. “Learning to Collaborate Through Collaboration: How Allying with Expert Firms Influences Collaborative Innovation within Novice Firms.” Strategic Management Journal 37: 2092‐2103.

73.

Howells

J. R.

2002. “Tacit Knowledge, Innovation and Economic Geography.” Urban Studies 39 (5-6): 871‐884.

74.

Hua

Wang

. 2015. “The Impact Of Network Structure on Innovation Efficiency: An Agent‐Based Study in the Context of Innovation Networks.” Complexity 21: 111‐122.

75.

Ibarra

1993. “Network Centrality, Power And Innovation Involvement: Determinants of Technical And Administrative Roles.” Academy of Management Journal 36 (3): 471501.

76.

Jaffe

A. B

. 1989. “Real Effects of Academic Research.” American Economic Review 79: 957‐970.

77.

Miao

. 2020. “Corporate Social Responsibility and Collaborative Innovation: The Role Of Government Support.” Journal of Cleaner Production 260: 121028.

78.

Kabir

Salim

Al-Mawali

. 2017. “The Gravity Model and Trade Flows: Recent Developments In Econometric Modeling and Empirical Evidence.” Economic Analysis and Policy 56: 60‐71.

79.

Kandilov

I. T.

Grennes

. 2012. “The Determinants of Service Offshoring: Does Distance Matter?” Japan and the World Economy 24 (1): 36‐43.

80.

Katz

J. S.

Martin

B. R.

. 1997. “What is research collaboration?” Research Policy 26: 1‐18.

81.

Kaygalak

Reid

. 2016. “Innovation and Knowledge Spillovers in Turkey: The Role of Geographic and Organizational Proximity.” Regional Science Policy & Practice 8 (1-2): 45‐60.

82.

Kelejian

Piras

. 2017. Spatial Econometrics. London: Academic Press, Elsevier.

83.

Keller

2004. “International technology diffusion.” Journal of Economic Literature 42: 752‐782.

84.

Keller

Yeaple

S. R.

. 2013. “The Gravity of Knowledge.” American Economic Review 103 (4): 1414‐1444.

85.

Ketchen

D. J.

Ireland

R. D.

Snow

C. C.

. 2007. “Strategic Entrepreneurship, Collaborative Innovation, and Wealth Creation.” Strategic Entrepreneurship Journal 1: 371‐385.

86.

Klette

T., J.

Kortum

. 2004. “Innovating Firms and Aggregate Innovation.” Journal of Political Economy 112 (5): 986‐1018.

87.

Knoben

Oerlemans

L. A. G.

. 2006. “Proximity and Inter‐Organizational Collaboration: A Literature Review.” International Journal of Management Reviews 8: 1‐89.

88.

Kodama

2015. Collaborative Innovation: Developing Health Support Ecosystems. New York, NY: Routledge.

89.

Krugman

1991. Geography and Trade. Cambridge, MA: MIT Press.

90.

Lawson

Lorenz

. 1999. “Collective Learning, Tacit Knowledge and Regional Innovative Capacity.” Regional Study 33 (4): 305‐317.

91.

Leenders

R. T. A. J.

Dolfsma

W. A.

. 2016. “Social Networks for Innovation and New Product Development.” Journal of Product Innovation Management 33 (2): 123‐131.

92.

LeSage

J. P.

Pace

R. K.

. 2009. Introduction to Spatial Econometrics. Boca Raton, FL: Taylor and Francis.

93.

Zhou

. 2019. “Managerial Overconfidence, Overinvestment, and R&D Spillover.” Managerial & Decision Economics 40: 858‐861.

94.

Liu

Gui

Duan

. 2017. “Structural Heterogeneity and Proximity Mechanism of Global Scientific Collaboration Network Based on Co‐Authored Papers.” Acta Geographica Sinica 72 (4): 737‐752.

95.

Liu

2018. “Dynamic Evolution, Spatial Spillover Effect of Technological Innovation and Haze Pollution in China.” Energy & Environment 29 (6): 968‐988.

96.

Liu

Jiao

. 2020. “Promoting or Preventing Labor Migration? Revisiting the Role of Language.” China Economic Review 60: 101407.

97.

Liu

Ying

. 2017. “Catch‐up through Collaborative Innovation: Evidence from China.” Thunderbird International Business Review 59 (4): 533‐545.

98.

Lööf

Nabavi

. 2015. “The Joint Impact of Innovation and Knowledge Spillovers on Productivity and Growth for Exporting Firms.” The World Economy 38 (4): 730‐750.

99.

Lucas

R. E

. 1989. “On the Mechanics of Economic Development.” Journal of Monetary Economics 22 (1): 3‐42.

100.

Mao

Zhou

Harms

Fang

. 2020. “Knowledge Growth in University-Industry Innovation Networks - Results from A Simulation Study.” Technological Forecasting and Social Change 151: 119746.

101.

Márquez-Ramos

Martínez-Zarzoso

. 2010. “The Effect of Technological Innovation on International Trade.” Economics the Open Access Open Assessment E Journal 4: 1‐37.

102.

McCann

Simonen

. 2005. “Innovation, Knowledge Spillovers and Local Labour Markets.” Papers in Regional Science 84 (3): 465‐485.

103.

Min

Kim

Sawng

Y. W.

. 2020. “The Effect of Innovation Network Size and Public R&D Investment on Regional Innovation Efficiency.” Technological Forecasting and Social Change 155: 119998.

104.

Moralles

H. F.

Silveira

N. J. C.

Rebelatto

D. A. D. N.

. 2018. Spatial Spillovers of Innovation and Transport in Brazil, Area Development & Policy 7 (1): 1‐17.

105.

Moreno-Serrano

Paci

Usai

. 2005. “Spatial Spillovers and Innovation Activity in European Regions.” Environment & Planning A 37 (1): 1793‐1812.

106.

Mukherji

Silberman

. 2013. “Absorptive Capacity, Knowledge Flows, and Innovation in U.S. Metropolitan Areas.” Journal of Regional Science 53 (3): 392‐417.

107.

Neumann

Matt

Hitz-Gamper

B. S.

Schmidthuber

Stürmer

. 2019. “Joining Forces for Public Value Creation? Exploring Collaborative Innovation in Smart City Initiatives.” Government Information Quarterly 36 (4): 101411.

108.

Nieves

Osorio

. 2013. “The role of Social Networks in Knowledge Creation.” Knowledge Management Research & Practice 11 (1): 62‐77.

109.

Ning

Wang

. 2016. “Urban Innovation, Regional Externalities of Foreign Direct Investment and Industrial Agglomeration: Evidence from Chinese Cities.” Research Policy 45 (4): 830‐843.

110.

Ollila

Yström

. 2016. “Exploring Design Principles of Organizing for Collaborative Innovation: The Case of an Open Innovation Initiative.” Creativity and Innovation Management 25 (3): 363‐377.

111.

Ott

Rondé

. 2019. “Inside the Regional Innovation System Black Box: Evidence from French Data.” Papers in Regional Science 98: 1993‐2026.

112.

Paci

Usai

. 2000. “Technological Enclaves and Industrial Districts: An Analysis of the Regional Distribution of Innovative Activity in Europe.” Regional Studies 34 (2): 97‐114.

113.

Porter

M. E.

1990. The Competitive Advantage of Nations. London: Macmillan.

114.

Powell

W. W.

Koput

K. W.

Smith-Doerr

. 1996. “Interorganizational Collaboration and the Locus of Innovation: Networks of Learning in Biotechnology.” Administrative Science Quarterly 41: 116‐145.

115.

Quatraro

2009. “Diffusion of Regional Innovation Capabilities: Evidence from Italian Patent Data.” Regional Study 43 (10): 1333‐1348.

116.

Romão

Nijkamp

. 2019. “Impacts of Innovation, Productivity and Specialization on Tourism Competitiveness - A Spatial Econometric Analysis on European Regions.” Current Issues in Tourism 22 (10): 1150‐1169.

117.

Romer

1990. “Endogenous Technological Change.” Journal of Political Economy 98 (II): 71‐102.

118.

Rothwell

1992. “Successful Industrial Innovation: Critical Factors for the 1990s.” R&D Management 22, 3: 221‐239.

119.

Rowley

Behrens

D. D.

Krackhardt

. 2000. “Redundant Governance Structures: An Analysis of Structural and Relational Embeddedness in the Steel and Semiconductor Industries.” Strategic Management Journal 21: 369‐386.

120.

Scherngell

. 2011. “Collaborative Knowledge Production in China: Regional Evidence from a Gravity Model Approach.” Regional Studies 45 (6): 755‐772.

121.

Scott

A. J.

1988. New Industrial Spaces: Flexible Production Organization and Regional Development in North American and Western Europe. London: Pion.

122.

Scuotto

Beatrice

Valentina

Nicotra

Di Gioia

Briamonte

M. F.

. 2020. “Uncovering the Micro-Foundations of Knowledge Sharing in Open Innovation Partnerships: An Intention-Based Perspective of Technology Transfer.” Technological Forecasting and Social Change 152: 119906.

123.

Sheng

Zhao

Zhang

Song

Miao

. 2019. “Innovation efficiency and spatial spillover in urban agglomerations: A case of the Beijing‐Tianjin‐Hebei, the Yangtze River Delta, and the Pearl River Delta.” Growth and Change 50: 1280‐1310.

124.

Shepherd

2019. “Mega-Regional Trade Agreements and Asia: An Application of Structural Gravity to Goods, Services, and Value Chains.” Journal of the Japanese and International Economies 51: 32‐42.

125.

Stephen

Overman

H. G.

. 2012. “Mostly pointless spatial econometrics?” Journal of Regional Science 52 (2): 172‐191.

126.

Sternberg

2000. “Innovation Networks and Regional Development—Evidence from the European Regional Innovation Survey (ERIS).” European Planning Studies 8: 389‐407.

127.

Sternberg

R. J.

1996. Successful intelligence: How practical and creative intelligence determine success in life. New York, NY: Simon & Schuster.

128.

Ter Wal

Anne

L. J.

. 2014. “The dynamics of the inventor network in German biotechnology: Geographic Proximity Versus Triadic Closure.” Journal of Economic Geography 14 (3): 589‐620.

129.

Tobler

W. R.

1970. “A Computer Movie Simulating Urban Growth in the Detroit Region.” Economic Geography 46: 234‐240.

130.

Tsai

2001. “Knowledge Transfer in Intraorganizational Networks: Effects of Network Position and Absorptive Capacity on Business Unit Innovation and Performance.” Academy of Management Journal 44: 996‐1004.

131.

Varga

2007. “Localised knowledge inputs and innovation: The Role of Spatially Mediated Knowledge Spillovers in Hungary.” Acta Oeconomica 57 (1): 1‐20.

132.

Wang

Yang

Qin

. 2020. “City Centrality, Migrants and Green Inovation Efficiency: Evidence from 106 Cities in the Yangtze River Economic Belt of China.” International Journal of Environmental Research and Public Health 17: 652.

133.

Wang

Zhao

. 2015. “Network Centrality, Organizational Innovation, and Performance: A Meta‐Analysis.” Canadian Journal of Administrative Sciences 32 (3): 146‐159.

134.

Wang

Lin

Fan

. 2020. “Influence of Reputation on Cooperative Behavior in Crowd Innovation Network.” Social Science Quarterly 101 (1): 172‐182.

135.

Wang

Zeng

Lyu

. 2016a. “Structural Evolution of Innovation Networks of China's Equipment Manufacturing Industry.” Acta Geographica Sinica 71 (2): 251‐264.

136.

Wang

Cheng

Wei

Y. H. D.

. 2016b. “Analyzing the Space–Time Dynamics of Innovation in China: ESDA and Spatial Panel Approaches.” Growth and Change 47 (1): 111‐129.

137.

Wang

Zhang

Fan

Yang

. 2019. “The Symbiosis of Scientific And Technological Innovation Efficiency and Economic Efficiency in China - An Analysis Based on Data Envelopment Analysis and Logistic Model.” Technology Analysis & Strategic Management 31 (1): 67‐80.

138.

Yan

Xiong

. 2019. “Network characteristics and organizational innovation capability: A Study of the Inter‐Organizational Collaboration Network of New Drug Development in Shanghai, China.” Strategic Change 28: 499‐506.

139.

Yström

Agogué

. 2020. “Exploring practices in Collaborative Innovation: Unpacking Dynamics, Relations, and Enactment in in‐Between Spaces.” Creativity and Innovation Management 29: 141‐145.

140.

Zucker

L. G.

Darby

M. R.

Brewer

M. B.

. 1998. “Intellectual Human Capital and the Birth of US biotechnology Enterprises.” American Economic Review 88 (1): 290‐306.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.03 MB