Knowledge Flows Within Chinese Administrative Provinces: A Patent Citation Analysis

doi:10.21203/rs.3.rs-1991012/v1

Download PDF

Research Article

Knowledge Flows Within Chinese Administrative Provinces: A Patent Citation Analysis

https://doi.org/10.21203/rs.3.rs-1991012/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

This paper explores the determinants of knowledge flows, proxied by patent citations, within 31 Chinese administrative provinces. In order to explain the number of citations from a Province to another, in addition to geographic and technological proximity, we use factor analysis to capture the effect of regional research structure. In particular, we measure the private versus public research intensity of each Chinese administrative province. Our econometric results show that, as expected, geographical and technological distance between Provinces are negatively correlated with patent citations. In addition, we find that Chinese administrative provinces that show bigger intensity with regard to public and to private research both receive more patent citations and cite more other patents in other Provinces. This effect is particularly significant when the two regions are specialized in private research. Finally, and at odds with prior research, we find that controlling for social proximity between Provinces does not impact the effect of geographical and technological proximity. These results, if confirmed by further studies, might have significant policy implications.

Other Economics

Knowledge spillovers

Patent citation

Geographical distance

Regional research structure

Proximity

Understanding the determinants of knowledge flows within the economy and the factors that might accelerate or hamper them is a critical task for at least three reasons. First, according to new growth theory, knowledge flows are central in order to generate endogenous long-run growth (Romer, 1986; Acs et al., 2012)[1]. Second, the local versus global dimension of knowledge flows is considered an important explanation of the process of convergence or divergence between countries and regions and to therefore significantly affect the possibility of catching-up for developing countries (Abramovitz, 1986). Third, and linked to the former point, localized knowledge spillovers are viewed as a critical driver of economic geographic concentration (Feldman, 1994; Audretsch and Feldman, 1996; Audretsch and Feldman, 2004).

The objective of this paper is therefore to analyze the determinants of knowledge flows between Chinese administrative Provinces. This work is one of the first attempts to apply economic analysis of R&D knowledge spillovers in developing countries, and it explores empirically the determinants of patent citation in China at regional level. Most empirical research about knowledge spillovers concentrate on developed countries and the number of studies on China, although growing, is still modest. In addition, most studies that look at knowledge flows in China focus on the firm level (Xiang et al., 2013; Hansen and Hansen, 2020) or try to estimate the impact of knowledge flows from abroad (Qiu et al., 2017). To our knowledge, very little study so far has attempted to understand the determinants of inter-regional knowledge spillovers in China. Those who did found that intraregional and international collaborations are the main channels of knowledge exchange in China, while inter-regional knowledge exchange is relatively weak (Gao et al., 2011). It might therefore be important to understand the determinants of inter-Provinces knowledge flows in China in order to develop policies that could unlock the existing impediments.

In recent decades, China’s innovation capability has improved significantly, and the trend does not show any sign of deceleration. In 2017, Chinese R&D intensity was equal to 2.12% of GDP and R&D total expenditure was equal to 280 billion US dollars, accounting for 20% of total world R&D expenditure. The total number of R&D personnel amounted to 3.759 million, accounting for 31.1% of the total world R&D personnel. China is also the country with the most patent applications in the world. The trend is similar with scientific research: China is now ranking second for the number of scientific publications authored by researchers working in China (Ministry of Science and Technology, 2017).

However, economic infrastructure and economic development in China are unevenly distributed across Provinces. Due to historical factors that persist or reinforce through time and to differentiated regional development policy, significant differences can be observed in the distribution of regional innovation capabilities across Chinese Provinces. For instance, Eastern regions are more developed than the Central and Western regions. Since 2006, and even though, innovation activities increased in all the administrative units, eastern coastal regions still occupy an overwhelming position in innovation output (Chen et al., 2017).

Focusing on knowledge flows between Chinese Provinces enables to examine standard research questions such as the effect of the geographical distance or technological proximity between two Provinces. But, and more originally, it also enables to explore to which extent the characteristics of the Provinces affect knowledge flows. In particular, in this paper, we focus on the effect of the research structure of each Chinese Province. We use factor analysis in order to measure the private versus public research intensity of each Chinese administrative province, thus introducing a distinction between Provinces that are more oriented towards public versus private research. This is in line with research that go beyond the geographic or technological effect and look at the institutional dimension in order to explain knowledge flows (Gittelman, 2006; Rodríguez-Pose, 2013).

We rely on patent citations to measure and capture knowledge spillovers (Jaffe and de Rassenfosse, 2017). For several decades, economic studies have used patent statistics to measure innovative activity[2]. Since the 1990s, patent citations have also been widely used to measure knowledge spillovers (Jaffe et al., 1993)[3]. Patent citations, however, remain an imperfect measure of knowledge externalities. First, because they only reflect knowledge codified in the patent document and neglect knowledge contained in other forms. Second, because patent examiners often add citations during the examination process, thus inducing bias and noise in the measure of knowledge flows (Jaffe et al., 2000; Alcacer and Gittelman, 2006; Alcacer et al., 2009; Lampe, 2012; Roach and Cohen, 2013; Moser et al., 2018; Corsino et al., 2019). Despite this important caveat, for Jaffe and de Rassenfosse (2019) there is today a broad consensus on the fact that patent citations remain an essential tool to measure inventions’ characteristics (their impact, value, generality, originality, etc.) and knowledge flows between economic actors.

In line with this literature, we use data on the patent inventors’ location in order to identify the geographic location of the cited and citing patents. Each citation between two patents is therefore associated to two of the 31 Chinese administrative Provinces (the citing and the cited Province). Relying on patent data from the US Patent and Trademark Office during the period 1995-2019, we are able to identify 27 118 patent citation pairs.

Our econometric results show that, as expected, geographical and technological distance between Provinces are negatively correlated with patent citations. In addition, we find that Chinese administrative provinces that show bigger intensity with regard to public and to private research both receive more patent citations and cite more other patents in other Provinces. This effect is particularly significant when the two regions are specialized in private research, thus suggesting that two Chinese Province strongly oriented towards private research are more likely to generate knowledge inflows and outflows one with each other. On the other hand, this effect is not significant when the two Provinces are specialized in public research. Finally, and at odds with prior research (Zucker et al., 1998; Breschi and Lissoni, 2003; Morrison, 2008; Frenken et al., 2010; Ott and Rondé, 2019), we find that controlling for social proximity between Provinces does not impact the effect of geographical and technological proximity. These results, if confirmed by further studies, might have significant policy implications as to the influence of Chinese Provinces research structure on knowledge circulation within China.

The remainder of the paper is structured as follows. Section 2 surveys the literature related to the determinants of knowledge spillovers. Section 3 details the empirical design of the research and provides information as to the data we use. Section 4 reports our main empirical findings. Section 5 concludes.

[1] One century ago, Marshall already put forward the link between knowledge flows and endogenous growth: “The mysteries of the trade become no mysteries; but are as if it were in the air, and children learn many of them unconsciously. Good work is rightly appreciated, inventions and improvements in machinery, in processes and the general organization of the business have their merits promptly discussed: If one man starts a new idea, it is taken up by others and combined with suggestions of their own; and thus it becomes the source of further new ideas” (Marshall, 1920, p. 225)

[2] As acknowledged by Griliches: “In spite of all the difficulties, patents statistics remain a unique resource for the analysis of the process of technical change. Nothing else even comes close in the quantity of available data, accessibility, and the potential industrial, organizational, and technological detail” (Griliches, 1990, p. 1702)

^{^[3]} The use of patent citation is expected to offer a solution to the challenge put forward by Paul Krugman: “knowledge flows are invisible ; they leave no paper trail by which they may be measured and tracked and there is nothing to prevent the theorist from assuming anything about them that she likes” (Krugman, 1991, p. 53-54).

Since the seminal work of Griliches (1957) and Jaffe (1986), the economic literature has highlighted a number of factors determining the intensity of knowledge spillovers. These factors can be grouped into four categories: geographic and national proximity (linked to cultural and linguistic proximity), technological proximity, temporal proximity, and social proximity (social network).

The geographic dimension is the one that has received the most attention. Extending the work of Jaffe (1989), the paper of Jaffe, Henderson, and Trajtenberg (1993) is generally considered as pioneer in the analysis of the geographic dimension of externalities using patent citations. Starting from a set of patents filed by American universities and firms and citations made to these patents, and comparing this sample of citing patents to a control group, they show that geographic proximity does indeed play a significant role in the probability to cite a patent. This is true at the national level but also at the regional level (citations have a higher probability of being made by a patent filed in the same country, in the same state, and even in the same metropolitan area).

If the methodology used by Jaffe et al. (1993) was criticized by Thompson and Fox-Kean (2005), who obtained less clear-cut results (although without completely questioning the importance of the geographical dimension), the vast majority of studies that have followed highlighted a significant geographic effect of knowledge spillovers. For example, Maurseth and Verspagen (2002), on European data, find that “Geographical distance has a negative and substantial impact on knowledge flows”. Likewise, Belenzon and Schankerman (2013) find that “Citations to patents decline sharply with distance and are strongly constrained by state borders”. Finally, Murata et al. (2014), revisits “the debate between Thompson and Fox-Kean and Henderson, Jaffe, and Trajtenberg on the existence of localized knowledge spillovers and find solid evidence supporting localization”. Interestingly, we also have evidence that progress in ICTs does not seem to lower the importance of geographic proximity for knowledge flows, on the contrary (Sonn and Storper, 2008).

Geographical distance reduces the flow of knowledge because it limits the circulation of tacit knowledge. Geographic proximity enhances the possibility of economic agents exchanging ideas and, particularly, tacit knowledge. Codified knowledge might easily circulate despite distance but tacit knowledge requires human interactions that, despite progress of ICTs, remain largely localized. This is why, as famously stated by Feldman (1994): “knowledge crosses more easily corridors and streets than oceans and continents”.

In relation to geographic proximity, most studies also highlight a country effect for knowledge spillovers. For example, Jaffe and Trajtenberg (1999) analyze patent citations between US, UK, France, Germany, and Japan and find that: “patents whose inventors reside in the same country are typically 30 to 80% more likely to cite each other than inventors from other countries, and these citations come sooner”. This result is confirmed by Branstetter (2001) who shows with data on the US and Japanese firms that knowledge spillovers are primarily intra-national and by Maurseth and Verspagen (2002) who find that the country effect remains even if regions share the same language. In the case of China, Qiu et al. (2017) find that only companies in the most advanced Chinese regions benefit from international externalities and that domestic collaboration has shown a larger positive impact on corporate innovation than international collaboration in recent years. These results suggest an effect of national institutions that is different from the simple effect of geographic proximity.

Following Jaffe (1986), the literature has also highlighted the importance of technological proximity to explain the flow of knowledge between economic actors. For example, Autant-Bernard (2001) shows, using French data, that technological proximity seems to play a more important role than geographical proximity. She also highlights the importance of human mobility in the circulation of knowledge, a factor to which we will come back. Maurseth and Verspagen (2002, p. 531) also show that while patent citations in Europe are affected by geographic proximity: “patent citations are industry specific and occur most often between regions that are specialized in industrial sectors with specific technological linkages between them”. Orlando (2004) obtains similar results on company data.

The importance of technological proximity is primarily explained by the fact that companies working on the same research projects have more to learn than companies working on very different things. The potential of knowledge transfer is thus much greater when the companies are technologically close. This had already been underlined by Griliches (1991): “The photographic equipment industry and the scientific instruments industry may not buy much from each other but may be, in a sense, working on similar things and hence benefiting much from each other’s research”. In addition, companies working in technologically related fields also find it easier to absorb knowledge (Cohen and Levinthal, 1989). It is indeed easier to understand and reuse knowledge developed by its technological neighbors.

The time dimension is also an important determinant of knowledge flows. Caballero and Jaffe (1993) and Jaffe and Trajtenberg (1999) have developed a model for the diffusion of knowledge over time based on two opposite effects: first, the more time passes, the more knowledge is likely to diffuse, which thus mechanically increases its potential for future dissemination. Second, the more time passes, the more knowledge becomes obsolete, which reduces its potential for future dissemination. These two opposing forces thus make it possible to envisage an optimum of diffusion over time, that is to say, the moment when a piece of knowledge has reached its peak of maximum diffusion. This diffusion peak obviously varies according to the country, sector, or type of organization. Jaffe and Trajtenberg show, for example, that the diffusion peak is reached more quickly in electronics than in chemistry or mechanics (i.e., technologies are obsolete more quickly in electronics). More recently, Mehta et al. (2010) using a modified method find that the diffusion peak arrives earlier than expected by Jaffe and Trajtenberg, but do not significantly question their result. Finally, Bacchiocchi and Montobbio (2009) use the same methodology to study the impact of knowledge-producing organizations on the dissemination of knowledge. They show that the technologies developed by universities diffuse more quickly than those developed by firms.

Finally, several studies have affirmed the importance of social network, that is to say of the previous existence of links between economic actors, to explain knowledge flows. This literature directly questions Marshall’s famous quote that “knowledge flows in the air”. For Breschi and Lissoni (2003, po. 6) “this view provides a naive portrait of the channels along which knowledge flow”. Of course, knowledge does not circulate in the air but between individuals. Consequently, social proximity between individuals is an essential determinant in explaining the flow of knowledge. For instance, Almeida and Kogut (1999) show that engineers’ inter-firm mobility influences the local transfer of knowledge. Also, Agrawal et al. (2006) use data on inventor mobility to highlight the importance of social proximity to explain knowledge flows. They find that the flow of knowledge between an inventor’s departure place and arrival place is 50% higher than if the inventor had not lived in the departure place. The overall evidence is therefore consistent with a view that social relationships, not just physical proximity, are important in determining observed patterns of knowledge diffusion.

Some studies even suggest that the effect of social proximity is so strong that the geographic effect disappears when properly controlled for social proximity. For instance, Breschi and Lissoni (2003) find that: “localization effects tend to vanish where citing and cited patents are not linked to each other by any network relationship. On the contrary, knowledge flows, as evidenced by patent citations, are strongly localized to the extent that labor mobility and network ties also are. We interpret these results as evidence that geography is not a sufficient condition for accessing a local pool of knowledge, but it requires active participation in a network of knowledge exchanges”. Rondé and Hussler (2005) are in the same line when demonstrating that “the impact on regional innovation of unintended knowledge flows decrease when voluntary actions (deliberate interactive competences building) are introduced in the model” and Singh (2005) goes also in the same direction: “The existence of a network tie is found to be associated with a greater probability of knowledge flow. Furthermore, the effect of regional or firm boundaries on knowledge flow decreases once interpersonal ties have been accounted for. In fact, being in the same region or firm is found to have little additional effect on the probability of knowledge flow among inventors who already have close network ties”.

In summary, a very large number of studies have explored the determinants of knowledge flows between economic actors. These flows are influenced by geographic proximity (possibly through social proximity which is constrained by geographic proximity), technological and temporal proximity. Also, the observation of a significant national effect seems to indicate that institutional proximity plays a role (Boschma, 2004). In the rest of this article, we focus on the determinants of knowledge flows between Chinese Provinces. Besides the geographical and technological effect, we are interested in the effect played by the research structure of the Provinces. As a matter of fact, when considering knowledge flows between economic actors, the magnitude of these flows depends not only on different forms of proximity between actors, but also on their capacities to emit and absorb knowledge. Since the work of Cohen and Levinthal (1989), we know that these capacities depend on the research investment of the different actors. Moreover, from a more macroeconomic perspective, investments in local institutional research and educational structures in order to higher the quantity and the quality of human capital furthermore determine the absorption of knowledge and innovation generated elsewhere (Barro, 1996; Rodriguez-Pose and Crescenzi, 2008). Applied to our Chinese regions, this means that we must consider the research structures of the different Provinces. Indeed, Rondé and Hussler (2005) have shown the positive effects of the scientific density of territories (Nuts 3 level) on the innovation capacities of firms via the development of relational capacities in order to generate and absorb knowledge flows. In an interesting study focusing on the biotechnology industry, Guittelman (2006) shows that research institutions have differentiated effect on knowledge flows according to their organizational structure which suggests that specific combinations of individuals’ knowledge with organizational capabilities matter to explain innovation outcome. “Technological performance is explained by the heterogeneity of organizations and individual relationships within countries, rather than an overall institutional effect that operates similarly upon all organizations in the countries” (Guittelman, 2006, p1067). In order to take into account the role of regional organizational research structures, we contrast the Provinces specializing in public research with those specializing in private research and analyze the consequences of this specialization on the flow of knowledge. To our knowledge, this effect has not been measured in the literature yet. However, it is likely that it reflects a form of institutional and/or cognitive proximity (Boshma, 2005; Boschma and Frenken, 2010) between the actors and therefore ultimately has a significant impact.

3.1 Data collection

We conduct our analysis of knowledge flows at the level of Chinese Provinces. We have therefore collected detailed information on 31 out of the 34 Chinese administrative Provinces. Taiwan, Hong Kong and Macao have been excluded from our analysis due both to their very specific situation and to missing information. The period considered is 1995-2019.

The data used in this study comes from two main sources: First, we use data provided by the China Statistical Yearbook, prepared by the National Bureau of Statistics of China. This yearbook displays every year detailed information as to the economic activity and performance of the Chinese Provinces. Second, information about patents and patent citations has been provided by the US Patent and Trademark Office (USPTO). We used PatentsView[1] to collect information about patent documents and, in particular, patent citation data. In detail, we downloaded the patent and patent citation data on the website of PatentsView by searching with the field "Inventor Country equals CN", and the information on patents granted during January 1995–December 2019 was able to be retrieved. The USPTO lists the names of all inventors on patents. The addresses (city and county) are listed for all inventors on patents. We use the inventor city to assign individual patents to their corresponding Provinces. Following the first-named inventor assignment principle (Acs et al., 2002; Bottazzi and Peri, 2003; Hu and Jaffe, 2003; Singh, 2008; Stolpe, 2002; Trajtenberg, 2001), we manually distinguish which Province the patents belong to by the address of the first inventor. After the process of cleaning the data, we further combine the citation pairs at the provincial unit level. We use USPTO citations to build a set of matrices that map citations between any two Provinces. Each cell of the matrix is the number of citations in patents with the first inventor resident in a Province to patents with the first inventor resident in another Province.

We focus on all the utility patents granted by the USPTO to the first inventors residing in China over the 1995–2019 period. Furthermore, we use patent citations to identify all the citations made between patents with the first inventors residing in China. In the considered period, we have a total of 27118 patent citation pairs that we can then assign to Chinese Provinces, based on the first inventors’ locations of citing and cited Provinces.

We focus on Chinese patents registered in the US patent office for three reasons. First, there is no patent citation information available in any of the Chinese patent databases, which makes it impossible for us to get patent citation data of the patents granted by the China National Intellectual Property Administration (CNIPA) and undertake an in-depth patent citation analysis. At the same time, the US patent database offers detailed information on patent citations, and knowledge spillover can be analyzed by patent citation data. Second, since the USPTO is the world’s highest-level patent organization and since the US is China’s most important destination country for patent applications overseas, focusing on USPTO data allows us to keep enough patents in our database. Third, there are certain studies that have used the USPTO database to assess knowledge spillovers in China (Yu and Wu, 2014; Xiang et al., 2013; Gao et al., 2011; Wu and Mathews, 2012), which supports our choice of the USPTO to handle our case in China.

Considering newer patents are referenced less frequently and there were fewer USPTO applications in the early years, we decided to take a cross-section of the entire period rather than breaking it up into separate periods (Maurseth and Verspagen, 2002). As a result, we are unable to evaluate the trends in data.

3.2 The public versus private research structure of Chinese Provinces

Since we seek to understand the effect of Provinces’ research structures on knowledge flows, we use the factor analysis method in order to characterize the research structure of each Chinese Provinces. Factor analysis is a multivariate statistical analysis method that converts multiple variables into a small number of factors that summarize the explanatory capacity of the original variables. The main idea is to reduce the dimensions and simplify the data in order to focus on a small number of common factors (Buesa et al, 2010; Jellema and Roland, 2011). In order to take into account the quality and quantity of institutional research structures, we chose an initial number of 20 variables extracted from the Statistical Yearbook of China in 2016. The factor analysis led to three significant common factors that are indicated in Table 1: an indicator of public research intensity (F1), an indicator of private research intensity (F2), and an indicator of transportation infrastructure quality (F3). The cumulative variance contribution rate reaches 90.64%, thus exceeding the general requirement of 85%.

Each province can then be characterized according to its performance related to F1, F2, and F3. Figure 1 indicates how the 31 Provinces perform along the F1 and F2 dimensions. We see that some Provinces are quite intensive in public research but much less in private research (for instance Beijing). Some Provinces, conversely, are quite intensive in private research but much less in public research (for instance, Zhejiang). Some are good at both public and private research (for instance, Guangdong). The bulk of the Provinces remain below zero both for public and private research.

Using each Province scores on F1, F2, F3, and the co-variance contribution rate of each common factor, we can calculate the comprehensive innovation scores of 31 Provinces, as shown in Appendix I. We can see that Beijing, Jiangsu, Guangdong, Sichuan, Shanghai, Shandong, and Zhejiang are province-level administrative regions with the strongest regional innovation capability. Most of these Provinces are located in the eastern region, which has a large number of famous firms and universities and a better innovation environment than the central region and western region. We see also that a majority of Provinces have a negative score, indicating innovation in China is concentrated in a minority of Provinces. Using the factor analysis method, we have therefore a preliminary understanding of China’s regional innovation capability. In the next part, we will introduce the scores of F1 and F2 of each province-level administrative region as the independent variables into the estimation model of knowledge flows[2].

Table 1 Factor analysis.
Component	F1	F2	F3
Papers published by the scientific research institutions	0.978
R&D personnel full-time equivalent in the scientific research institution	0.962
Publishing scientific works in the scientific research institutions	0.961
R&D expentures in the scientific research institutions with regard to GDP(%)	0.940
Number of R&D projects in scientific research institutions	0.948
Number of Scientific research institutions	0.853
Scientific Books Published in the University	0.780
University R&D expenture with regard to GDP (%)	0.786
Number of R&D projects in the University	0.670
Published scientific papers of the University	0.667
R&D personnel full-time equivalent in the university	0.649
R&D staff full-time equivalent in industrial enterprises above designated size		0.983
Sales revenue of new products of industrial enterprises above designated size		0.975
Number of new product projects of industrial enterprises above designated size		0.970
Number of R&D projects of industrial enterprises above designated size		0.968
Funds for developing new products of industrial enterprises above designated size		0.956
Firms R&D expenture with regard to GDP(%)		0.765
Railway business density			0.854
Long-distance optical cable line density			0.806
Highway density			0.772
Extraction Method: Principal Component Analysis. Rotation Method: Varimax with Kaiser Normalization

3.3 Econometric treatment

Since we are interested in the determinants of knowledge flows between Chinese Provinces we rely, as dependent variable, on the number of citations C_ij between Province i and Province j. Please note that we pay attention to the direction of the citations, which means that C_ij, which is the number of citations that Province i made to Province j’s patents, is different from C_ji, which is the number of citations that Province j made to Province i’s patents. In other words in C_ij, i is the citing Province and j the cited one. Therefore, as we consider 31 Provinces in our database, our econometric model is based on 31 * 30 = 930 observations.

In order to explain the number of citations between two Provinces, we rely on a set of independent variables that are summarized in Table 2 and Table 3.

First, and in line with the literature on the determinants of knowledge spillovers, we focus on geographical and technological distance. The geographical distance, denoted by D_ij, is great circle distance (in kilometers) between the regions of the citing and cited inventors, based on coordinates for the regions or their geographical centers.

Technological proximity, noted S_ij, is calculated by using the international patent classifications of patents in region i and region j. Following Jaffe (1986), we use the distribution of the regions’ patents over 8 patent classes to calculate the technological proximity between two regions. Jaffe’s formula was originally used to measure the proximity of the technological structure between firms. But it can also be used to capture the extent of technological overlap between province-level administrative units pairs.

Second, we also introduce as an independent variable the public versus private research intensity of each Province. This is done by introducing the F1 and F2 scores calculated in section 3.2. F1 indicates the score of public-sector R&D, and F2 is the score of the private-sector R&D. These scores capture differences into regional research structure. Some regions are more intensive in public research and others more intensive in private research. Moreover, as figure 1 delimitates four zones (F1 > 0 and F2 > 0, F1 > 0 and F2 <0, F1< 0 and F2 > 0, F1< 0 and F2 > 0), two of them are particularly remarkable, namely, the regions intensive in private research and the one intensive in public research. Therefore, it could be interesting to know whether the transmitters and receivers of knowledge comes from the same categories of regions (public vs private R&D intensive) or not. To check for this point, we introduce three dummies variables D1, D2 and D3, as detailed in Table 2.

As control variables, we include geography dummies for each cited and citing region. We also use the World Bank’s standards to measure the degree of economic development. According to this criteria, the 31 provincial administrative regions are divided into six economic regions: Northeast, Bohai, Southeast, Central, West, and North[3]. Finally, we use co-inventors measure in order to capture the social proximity (social network effect) between regions. In doing so, we want to contribute to the debate “unintended spillovers vs deliberate interactions” debate which, to our knowledge, has only been informed by empirical studies from American or European data. The number of co-inventors is this defined as the frequency of collaboration between inventors in region i and inventors in region j.

We express the independent variables of geographical distance, number of patents as natural logs in order to lessen the impact of outliers, and to reduce heteroscedasticity. Since our dependent variable is a discrete and positive variable with overdispersion, we use count data models such that Negative binomial regression and Tobit regression (Hausman, et al. 1984).

Table 2 Description of variables included in our model.
Variable	Definition
Dependent Variable
C_ij	The number of patent citations between provincial-level administrative unit i and j.
Independent Variable
Co-patenting_ij	The number of patent cooperation between provincial-level administrative unit i and j.
P_i	The total number of patents in provincial-level administrative unit i.
P_j	The total number of patents in provincial-level administrative unit j.
D_ij	Geography distance, the great circle distance (in kilometers) between the provincial-level administrative units of the citing and cited inventors, based on coordinates for the regions or their geographical centers.
S_ij	Technological proximity between provincial-level administrative unit i and j.
Traffic_i	The score of the traffic of region i.
Traffic_j	The score of the traffic of region j.
F1_i	The score of public-sector R&D performance of region i.
F2_i	The score of the private-sector R&D performance of region i.
F1_j	The score of public-sector R&D performance of region j.
F2_j	The score of the private-sector R&D performance of region j.
F1_i ×F1_j	Interaction of the score of public-sector R&D performance between region i and j.
F2_i×F2_j	Interaction of the score of private-sector R&D performance between region i and j.
D1	Dummy variable taking value 1 if the score of the public-sector R&D performance of region i and j both are above 0.
D2	Dummy variable taking value 1 if the score of the private-sector R&D performance of region i and j both are above 0.
D3	Dummy variable taking value 1 if the score of the public-sector R&D performance of region i (j) is above 0, and the score of the private-sector R&D performance of region j (i) is above 0.
Geography dummy variables	Geography dummy variables for each cited and citing region, include 6 regions of Northeast, Bohai, Southeast, Central, West, and North. 5 dummy variables for citing regions, and 5 dummy variable for cited regions.

Table 3 Descriptive statistics.
Variable	Obs	Mean	Std. Dev.	Min	Max
C_ij	930	7.113	44.64	0	762
Co-patenting_ij	930	18.73	99.36	0	1,328
S_ij	930	0.780	0.179	0	0.997
lnP_ij	930	11.23	3.211	1.099	20.41
lndis	930	14.02	0.597	11.54	15.17
F1_i	930	0	0.984	-0.699	5.062
F2_i	930	0	0.984	-0.876	2.958
F1_j	930	0	0.984	-0.699	5.062
F2_j	930	0	0.984	-0.876	2.958
Traffic_i	930	0	0.984	-1.345	3.550
Traffic_j	930	0	0.984	-1.345	3.550
D1	930	0.0774	0.267	0	1
D2	930	0.0968	0.296	0	1
D3	930	0.172	0.378	0	1

[1] Patents View is a patent data visualization and analysis platform intended to increase the value, utility, and transparency of US patent data. The initiative is supported by the Office of the Chief Economist in the USPTO, with additional support from the U.S. Department of Agriculture (USDA). (www.patentsview.org)

[2] The factor analysis has been conducted for different years, at the beginning, the middle and the end of our sample. The time does not affect the outcome of the factor analysis. For this reason, in our econometric treatment we retain the factor analysis performed at year 2016 and displayed in this section.

[3] The Northeast includes Heilongjiang, Jilin and Liaoning; The Bohai includes Beijing, Tianjin, Hebei and Shandong; The Southeast includes Shanghai, Jiangsu, Zhejiang, Fujian and Guangdong; The Central includes Henan, Hubei, Hunan, Anhui and Jiangxi; The Southwest includes Chongqing, Sichuan, Yunnan, Hainan, Guizhou and Guangxi; The Northwest includes Shanxi, Shaanxi, Gansu, Ningxia, Inner Mongolia, Xinjiang, Qinghai and Tibet. There are 5 fixed-effect dummies for citing regions and for cited regions, respectively.

All the estimation results of regression models are shown in Table 4 and Table 5. In Table 4, the Tobit model uses the logarithm of the citation count as the dependent variable, Model 1 and Model 2 reports the result of Tobit regression. Model 3 and Model 4 show results of the negative binomial estimates. Results of the generalized negative binomial model are reported in Model 5 and Model 6. Table 4 reports the results with the F1 and F2 variables and with interaction terms whereas Table 5 reports results with the dummy variables D1, D2 and D3. The overall results are generally consistent with our hypotheses, suggesting that geographical distance, technological proximity and the research structure of Provinces have significant impact on knowledge spillovers in China.

From Table 4, we observe that the coefficient of the geographical distance variable is negatively and highly significant in all the specifications, and the absolute value of the coefficient is relatively large in all the different regression models, ranging from 0.32 to 0.49. That clearly indicates a strong and significant effect of the geographical distance, even if we control for technological proximity. In other words, knowledge spillovers in China are strongly localized.

In the same line, the coefficient of the technological proximity variable is positive and significant at 1% level in all the specifications. That indicates that, as expected, technological proximity has a positive impact on knowledge flows between Chinese Provinces. When two Provinces share the same technological structure, when they specialize in the same technology, the likelihood that they cite each other is more important.

Table 4 Regression results (1).
	(1)		(2)		(3)		(4)	(5)
	Tobit		Tobit		Negative binomial		Negative binomial	Generalised negative
	ln(C_ij)		ln(C_ij)		C_ij		C_ij	C_ij
Co-patenting	0.001		0.001***		0.0004		0.0005*	0.0002
	(1.27)		(2.59)		(1.46)		(1.69)	(1.20)
ln(P_iP_j)*	0.691***		1.955***		0.698***		0.723***	0.701***
	(6.66)		(4.26)		(10.10)		(12.14)	(12.13)
S_ij	2.470***		0.608***		3.137***		2.885***	3.109***
	(3.86)		(7.99)		(6.75)		(6.25)	(7.09)
ln(d_ij)	-0.490***		-0.435***		-0.388***		-0.386***	-0.316***
	(-3.19)		(-3.79)		(-3.01)		(-3.26)	(-3.72)
F1_i	0.219**				0.189**			0.199***
	(2.07)				(2.16)			(3.15)
F2_i	0.135				0.158			0.239***
	(0.98)				(1.57)			(2.99)
F1_j	0.245**				0.239***			0.209***
	(2.30)				(3.61)			(3.61)
F2_j	0.229				0.237***			0.177**
	(1.63)				(2.61)			(2.07)
F1_i × F1_j			0.176***				0.150***
			(3.37)				(3.28)
F2_i× F2_j			0.130***				0.104***
			(3.58)				(3.79)
Traffic_i	0.130		0.117**		0.110		0.094*	0.142**
	(1.16)		(1.99)		(1.29)		(1.84)	(2.34)
Traffic_j	0.198*		0.130**		0.232***		0.162***	0.153**
	(1.68)		(2.16)		(3.00)		(2.75)	(2.29)
Constant	-4.965*		-4.073**		-6.805***		-7.123***	-7.775***
	(-1.94)		(-2.16)		(-3.67)		(-4.05)	(-5.56)
Observations	930		930		930		930	930
Geographical position	Control		Control		Control		Control	Control
lambda	1.300***		0.961***
	(4.11)		( 4.81)
ln(P_iP_j)*								-0.581***
								(-9.40)
S_ij								2.880**
								(2.22)
ln(d_ij)								0.516**
								(2.17)
Robust z-statistics in parentheses * p<0.01, p<0.05, * p<0.1
Table 5 Regression results (2).
		(1)		(2)		(3)
		Tobit		Negative binomial		Generalised negative binomial
		ln(C_ij)		C_ij		C_ij
Co-patenting		0.001		0.001**		0.0003*
		(1.48)		(1.96)		(1.66)
ln(P_iP_j)*		0.666***		0.700***		0.688***
		(5.39)		(10.14)		(11.31)
S_ij		2.305***		2.981***		2.958***
		(3.33)		(5.88)		(6.25)
ln(d_ij)		-0.431***		-0.309**		-0.257***
		(-2.64)		(-2.39)		(-2.83)
F1_i		0.264**		0.246***		0.261***
		(2.27)		(2.79)		(3.97)
F2_i		0.106		0.152		0.235***
		(0.71)		(1.49)		(2.97)
F1_j		0.295**		0.295***		0.268***
		(2.50)		(4.36)		(4.97)
F2_j		0.211		0.215**		0.177**
		(1.40)		(2.43)		(2.43)
D1		-0.077		-0.178		-0.044
		(-0.27)		(-1.01)		(-0.29)
D2		0.800***		0.813***		0.648***
		(2.71)		(4.30)		(3.56)
D3		0.079		-0.083		-0.107
		(0.36)		(-0.54)		(-0.68)
Traffic_i		0.205		0.214**		0.234***
		(1.62)		(2.50)		(3.29)
Traffic_j		0.283**		0.336***		0.252***
		(2.09)		(4.24)		(3.62)
Constant		-5.583*		-7.886***		-8.412***
		(-1.91)		(-4.15)		(-5.70)
Observations		930		930		930
Geographical position		Control		Control		Control
lambda		1.374***
		(4.11)
ln(P_iP_j)*						-0.588***
						(-8.44)
S_ij						3.554**
						(2.32)
ln(d_ij)						0.448*
						(1.89)
Robust z-statistics in parentheses * p<0.01, p<0.05, * p<0.1

More originally, we can see in Table 4 that the research structure of Chinese Provinces do impact the intensity of knowledge flows. In particular, F1 is positive and significant both for the citing and cited side, thus suggesting that Provinces that experience a good performance with regard to public research are more likely to both cite and be cited by other Provinces. The same result, although in a less clear-cut way (F2 is only significant in the case of the Generalized Negative Binomial), can be observed for private research.

Furthermore, an interesting result is given by the interaction terms between F1 i and j and F2 i and j, which are both positive and significant. This suggests that when two regions share the same research infrastructure, for instance when they experience both good performance with regard to public or private research, they are more likely to exchange knowledge. This result can be completed by results from Table 5 where we can see that coefficient of variable D2 is positive and strongly significant while coefficient of D1 is negative but not significant. This suggests that knowledge flows are particularly strong between Chinese Provinces that are good at conducting private research but not necessarily between Provinces that are good at conducting public research. This surprising result (the interaction terms between F1 i and j is positive while D1 is non significant) could be explained by the fact that D1 is a dichotomic variable that contains less information than the interaction term. This shows that citations between 'public' provinces only work when the infrastructure is of a high level (high score of F1 i and j) and that it is not enough that both provinces have a minimum of public infrastructure (positive D1). This result is consistent with the fact that patent and patent citations focus mostly on industrial innovation and less on scientific research. Maybe a different result would be observed by analyzing scientific publications citations. Another possible interpretation is that as public research is more science based than private one (Pavitt, 1984), it is possible that public spillovers go beyond Chinese borders. In other words, Chinese patents cite technological advances from Chinese Provinces and scientific advances from around the world. Another interesting result concerns the absence of a significant effect of the variable D3, which suggests that the regional combination of a high level of public and private research infrastructures does not guarantee a significant knowledge exchange flow. This surprising result is however consistent with previous findings by Guittelman (2006) suggesting that patent citations are associated with specific combinations of individuals’ knowledge and organizational structures. A high level of public and private research institutions in some region is not a sufficient condition of high knowledge flows. Further research is needed to identify the specific organizational conditions (especially between the private and public spheres) that enable a high level of knowledge exchange.

A last result deals with the effect of social proximity. The coefficient of the co-patenting variable is very low thus suggesting a weak effect of prior social links. Furthermore, in contrast with former studies, the introduction of social proximity in the analysis does not affect and diminish the importance of geographical and technological proximity. The coefficients of these two variables are not affected when we introduce social proximity in the regression (see appendix III, where the results of the regression without social proximity are displayed). This is a very surprising result compared to previous works, and, in any case, a Chinese specificity. For the moment, we have no interpretation of this phenomenon and further work, both theoretical and empirical, will undoubtedly be needed in order to understand this Chinese originality.

This paper investigated the determinants of knowledge spillovers between 31 Chinese administrative Provinces (Hong-Kong, Macao and Taiwan excluded). By looking at USPTO patent citations between years 1975–2019, we have been able to put forward, at the level of Chinese Provinces, strong geographical and technological effects. In addition, the importance of social proximity is not validated by our study, which stands in sharp contrast with other studies. Last, an important and original result of our research is that, in order to understand knowledge flows between regions, not only the technological, geographical and/or sectoral dimension matters, but also the type of research that is conducted in the region. Indeed, by conducting a factor analysis, we have been able to develop an indicator of the public versus private research specialization of each Chinese Province. By doing so, we have been able to show that the Provinces’ research structure (more or less strong with regard to public and or private research) strongly and significantly impact knowledge flows between Provinces. In particular, knowledge flows seem to be stronger when the two Provinces rank good with regard to private research. This result, which to our knowledge has never been put forward in the literature, might suggest the importance of the nature of the research that is conducted at the regional level in order to explain knowledge flows between regions.

Even if this research suffers from many limitations (for instance we do not control for citations added by USPTO examiners, we focus only on first inventors in the patent, or we remain at the province level) it contributes to feed discussions on the uneven economic development of China. Indeed, it is well known that localized knowledge spillovers are a major source of economic concentration and therefore of divergence between regions. This effect seems to be strong in China. Clearly, knowledge is not equally distributed in space, and not easily accessible at every point in space. As a consequence, the difference between Province that are leading and Provinces that are lagging behind is still increasing. In the less developed regions of China, governments should consider setting up incentives and mechanisms for fostering local and extra-local knowledge transfer networks to create learning regions. In order to fight against these divergences and to ensure an even smooth development of all Chinese Provinces, Chinese authorities might attempt to significantly improve the knowledge flows from most advanced to less developed regions in order to help the latter breaking out their “locked-in” development trajectories.

In the same vein, the very small effect of social proximity in our analysis might suggest that interpersonal mobility remains too weak between Chinese Provinces. Geographical constraints on knowledge flows can be overcome by fostering interpersonal links across regional boundaries. However, this is possible only if there is a minimum level of social interactions between Provinces. Therefore, Chinese authorities could think about feeding and nurturing inter-regional inter-personal networks through encouraging mobility and interaction of people outside their own regions, for instance through subsidizing joint R&D projects, joint conferences and encouraging local firms to tap into external collaborative networks.

Last, and even if more research needs to be conducted in order to confirm this effect, it seems that the research specialization of Chinese Provinces does impact knowledge flows which seem to be more important between Provinces good at private research. This might suggest the existence of a locking-in of some Province that remain specialized in public research and others specialized in private research. Again, appropriate economic policies could aim at unlocking the situation and favoring interactions and knowledge flows between Provinces specialized in public research and the ones specialized in private research.

Abramovitz, M. (1986). Catching up, forging ahead, and falling behind. The journal of economic history, 46(2), 385-406.

Acs, Z. J., Anselin, L., & Varga, A. (2002). Patents and innovation counts as measures of regional production of new knowledge. Research policy, 31(7), 1069-1085.

Acs, Z. J., Audretsch, D. B., Braunerhjelm, P., & Carlsson, B. (2012). Growth and entrepreneurship. Small Business Economics, 39(2), 289-300.

Agrawal, A., Cockburn, I., & McHale, J. (2006). Gone but not forgotten: knowledge flows, labor mobility, and enduring social relationships. Journal of Economic Geography, 6(5), 571-591.

Alcacer, J., & Gittelman, M. (2006). Patent citations as a measure of knowledge flows: The influence of examiner citations. The review of economics and statistics, 88(4), 774-779.

Alcácer, J., Gittelman, M., & Sampat, B. (2009). Applicant and examiner citations in US patents: An overview and analysis. Research Policy, 38(2), 415-427.

Almeida, P., & Kogut, B. (1999). Localization of knowledge and the mobility of engineers in regional networks. Management science, 45(7), 905-917.

Audretsch, D. B., & Feldman, M. P. (1996). R&D spillovers and the geography of innovation and production. The American economic review, 86(3), 630-640.

Audretsch, D. B., & Feldman, M. P. (2004). Knowledge spillovers and the geography of innovation. In Handbook of regional and urban economics (Vol. 4, pp. 2713-2739). Elsevier.

Autant-Bernard, C. (2001). The geography of knowledge spillovers and technological proximity. Economics of Innovation and New Technology, 10(4), 237-254.

Bacchiocchi, E., & Montobbio, F. (2009). Knowledge diffusion from university and public research. A comparison between US, Japan and Europe using patent citations. The Journal of Technology Transfer, 34(2), 169-181.

Barro, R. J. (1996). Determinants of economic growth: A cross-country empirical study.

Belenzon, S., & Schankerman, M. 2013. Spreading the word: Geography, policy, and knowledge spillovers. Review of Economics and Statistics, 95(3), 884-903.

Boschma, R. (2004). Proximité et innovation. Économie rurale, 280(1), 8-24.

Boschma, R., & Frenken, K. (2010). The spatial evolution of innovation networks: a proximity perspective. In The handbook of evolutionary economic geography. Edward Elgar Publishing.

Bottazzi, L., & Peri, G. (2003). Innovation and spillovers in regions: Evidence from European patent data. European economic review, 47(4), 687-710.

Branstetter, L. G. (2001). Are knowledge spillovers international or intranational in scope?: Microeconometric evidence from the US and Japan. Journal of international Economics, 53(1), 53-79.

Breschi, S., & Lissoni, F. (2003). Mobility and social networks: Localised knowledge spillovers revisited. Università commerciale Luigi Bocconi.

Buesa, M., Heijs, J., & Baumert, T. (2010). The determinants of regional innovation in Europe: A combined factorial and regression knowledge production function approach. Research policy, 39(6), 722-735.

Caballero, R. J., & Jaffe, A. B. (1993). How high are the giants' shoulders: An empirical assessment of knowledge spillovers and creative destruction in a model of economic growth. NBER macroeconomics annual, 8, 15-74.

Chen, X., Liu, Z., & Ma, C. (2017). Chinese innovation-driving factors: regional structure, innovation effect, and economic development—empirical research based on panel data. The annals of regional science, 59(1), 43-68.

Cohen, W. M., & Levinthal, D. A. (1989). Innovation and learning: the two faces of R & D. The economic journal, 99(397), 569-596.

Corsino, M., Mariani, M., & Torrisi, S. (2019). Firm strategic behavior and the measurement of knowledge flows with patent citations. Strategic Management Journal, 40(7), 1040-1069.

Feldman, M. P. (1994). The geography of innovation (Vol. 2). Springer Science & Business Media.

Frenken, K., Ponds, R., & Van Oort, F. (2010). The citation impact of research collaboration in science‐based industries: A spatial‐institutional analysis. Papers in regional science, 89(2), 351-271.

Gao, X., Guan, J., & Rousseau, R. (2011). Mapping collaborative knowledge production in China using patent co-inventorships. Scientometrics, 88(2), 343-362.

Gittelman, M. (2006). National institutions, public–private knowledge flows, and innovation performance: A comparative study of the biotechnology industry in the US and France. Research Policy, 35(7), 1052-1068.

Griliches, Z. (1957). Hybrid corn: An exploration in the economics of technological change. Econometrica, Journal of the Econometric Society, 501-522.

Griliches, Z. (1991). The search for R&D spillovers. National Bureau of Economic Research Working Paper Series, (w3768).

Hansen, T., & Hansen, U. E. (2020). How many firms benefit from a window of opportunity? Knowledge spillovers, industry characteristics, and catching up in the Chinese biomass power plant industry. Industrial and Corporate Change, 29(5), 1211-1232.

Hausman, J. A., Hall, B. H., & Griliches, Z. (1984). Econometric models for count data with an application to the patents-R&D relationship.

Hu, A. G., & Jaffe, A. B. (2003). Patent citations and international knowledge flow: the cases of Korea and Taiwan. International journal of industrial organization, 21(6), 849-880.

Jaffe, A. B. (1986). Technological opportunity and spillovers of R&D: evidence from firms' patents, profits and market value.

Jaffe, A. B. (1989). Real effects of academic research. The American economic review, 957-970.

Jaffe, A. B., & De Rassenfosse, G. (2019). Patent citation data in social science research: Overview and best practices. Research handbook on the economics of intellectual property law.

Jaffe, A. B., & Trajtenberg, M. (1999). International knowledge flows: Evidence from patent citations. Economics of innovation and new technology, 8(1-2), 105-136.

Jaffe, A. B., Trajtenberg, M., & Fogarty, M. S. (2000). Knowledge spillovers and patent citations: Evidence from a survey of inventors. American Economic Review, 90(2), 215-218.

Jaffe, A. B., Trajtenberg, M., & Henderson, R. (1993). Geographic localization of knowledge spillovers as evidenced by patent citations. the Quarterly journal of Economics, 108(3), 577-598.

Jellema, J., & Roland, G. (2011). Institutional clusters and economic performance. Journal of Economic Behavior & Organization, 79(1-2), 108-132.

Lampe, R. (2012). Strategic citation. Review of Economics and Statistics, 94(1), 320-333.

Maurseth, P. B., & Verspagen, B. (2002). Knowledge spillovers in Europe: a patent citations analysis. Scandinavian Journal of Economics, 104(4), 531-545.

Mehta, A., Rysman, M., & Simcoe, T. (2010). Identifying the age profile of patent citations: New estimates of knowledge diffusion. Journal of Applied Econometrics, 25(7), 1179-1204.

Morrison, A. (2008). Gatekeepers of knowledge within industrial districts: who they are, how they interact. Regional Studies, 42(6), 817-835.

Moser, P., Ohmstedt, J., & Rhode, P. W. (2018). Patent citations—an analysis of quality differences and citing practices in hybrid corn. Management Science, 64(4), 1926-1940.

Murata, Y., Nakajima, R., Okamoto, R., & Tamura, R. (2014). Localized knowledge spillovers and patent citations: A distance-based approach. Review of Economics and Statistics, 96(5), 967-985.

Orlando, M. J. (2004). Measuring spillovers from industrial R&D: on the importance of geographic and technological proximity. RAND Journal of Economics, 777-786.

Ott, H., & Rondé, P. (2019). Inside the regional innovation system black box: Evidence from French data. Papers in Regional Science, 98(5), 1993-2026.

Pavitt, K. (1984). Sectoral patterns of technical change: towards a taxonomy and a theory. Research policy, 13(6), 343-373.

Qiu, S., Liu, X., & Gao, T. (2017). Do emerging countries prefer local knowledge or distant knowledge? Spillover effect of university collaborations on local firms. Research Policy, 46(7), 1299-1311.

Roach, M., & Cohen, W. M. (2013). Lens or prism? Patent citations as a measure of knowledge flows from public research. Management Science, 59(2), 504-525.

Rodríguez-Pose, A., & Crescenzi, R. (2008). Research and development, spillovers, innovation systems, and the genesis of regional growth in Europe. Regional studies, 42(1), 51-67.

Romer, P. M. (1986). Increasing returns and long-run growth. Journal of political economy, 94(5), 1002-1037.

Ronde, P., & Hussler, C. (2005). Innovation in regions: What does really matter?. Research policy, 34(8), 1150-1172.

Singh, J. (2005). Collaborative networks as determinants of knowledge diffusion patterns. Management science, 51(5), 756-770.

Singh, J. (2008). Distributed R&D, cross-regional knowledge integration and quality of innovative output. Research Policy, 37(1), 77-96.

Sonn, J. W., & Storper, M. (2008). The increasing importance of geographical proximity in knowledge production: an analysis of US patent citations, 1975–1997. Environment and Planning a, 40(5), 1020-1039.

Stolpe, M. (2002). Determinants of knowledge diffusion as evidenced in patent data: the case of liquid crystal display technology. Research Policy, 31(7), 1181-1198.

Thompson, P., & Fox-Kean, M. (2005). Patent citations and the geography of knowledge spillovers: A reassessment. American Economic Review, 95(1), 450-460.

Trajtenberg, M. (2001). Innovation in Israel 1968–1997: a comparative analysis using patent data. Research Policy, 30(3), 363-389.

Wu, C. Y., & Mathews, J. A. (2012). Knowledge flows in the solar photovoltaic industry: Insights from patenting by Taiwan, Korea, and China. Research Policy, 41(3), 524-540.

Xiang, X. Y., Cai, H., Lam, S., & Pei, Y. L. (2013). International knowledge spillover through co-inventors: An empirical study using Chinese assignees' patent data. Technological Forecasting and Social Change, 80(1), 161-174.

Yu, F., & Wu, Y. (2014). Patent citations and knowledge spillovers: an analysis of Chinese patents registered in the USA. Asian Journal of Technology Innovation, 22(1), 86-99.

Zucker, L. G., Darby, M. R., & Armstrong, J. (1998). Geographically localized knowledge: spillovers or markets?. Economic Inquiry, 36(1), 65-86.

Competing interests: The authors declare no competing interests.

Appendix.docx

Download PDF

Version 1

posted

You are reading this latest preprint version

Knowledge Flows Within Chinese Administrative Provinces: A Patent Citation Analysis

Status:

Version 1

Abstract

Figures

1. Introduction

2. Related Literature On The Determinants Of Knowledge Flows

3. Empirical Design

4. Empirical Results And Discussion

5. Conclusion

References

Declarations

Supplementary Files

Status:

Version 1