Mapping the landscape of university technology flows in China using patent assignment data

Universities are key sources of knowledge for regional economic growth, generating revenue and skills for enterprises through technology flows that drive innovation. Many countries have emulated the Bayh-Dole Act to boost regional competitiveness, but university technology flows (UTFs) underperform outside the U.S. In China, despite incentives and major investments, the commercialization rate of academic patents remains below 5% versus over 50% in the U.S. This study addresses gaps in understanding within-country UTF networks (UTFN), the central technology fields, roles of different university types, and regional policy implementation. It examines UTFs via patent assignment data to identify popular technologies, key organizations, and regional patterns, thereby informing the commercialization of intellectual property in Chinese universities and contributing to literature on universities’ third mission and regional innovation systems. Research questions include: What is the structure of the UTFN? Who are the key actors? What are the most popular technological fields? What roles do different regions play?

Universities produce human capital and innovative knowledge that supports business innovation and regional competitiveness. In China, universities historically focused on human resource development, later adopting research as a second mission and regional development as a third mission in the mid-1990s, prompting policies to foster technology transfer. Challenges remain: underinvestment in basic research, regional disparities, and reliance on foreign technology—issues closely tied to UTFs that can reduce dependence on foreign technologies. UTFs are often geographically localized, yet universities have emphasized quantity over quality in innovation outputs. Research on UTFs has used patents as indicators due to their rich technological information. Patent citations reveal explicit knowledge spillovers but poorly capture economic value and tacit flows; joint patents capture collaboration but not directionality or economic value. Patent assignments, by contrast, reflect both economic value and the direction of technology and associated tacit knowledge flows. Social network analysis (SNA) is suitable for mapping relational structures like UTFs, using metrics such as centrality and density to identify key actors and trends. Existing work in China lacks comprehensive, quantitative, spatiotemporal analyses of UTFN; this study bridges these gaps using SNA across technological, organizational, and regional lenses.

Data sources: The study focuses on invention patent assignments, the primary university patent transaction channel in China. Patent data (2001–2021) were obtained from the incoPat platform, including assignor/assignee information, legal status, and IPC classes. Assignor type was set to universities (regular universities, junior colleges, adult colleges) and assignee type to enterprises to extract university-to-enterprise transfers. University names were disambiguated using the Ministry of Education’s 2021 list; enterprise names via Qichacha. Manual checks ensured accuracy. Final dataset: 65,055 patents transferred from 882 universities to 24,869 companies in mainland China.

Network construction and indicators: The UTFN was built with nodes as universities or firms and links as assignment relationships, segmented into three stages: 2001–2007, 2008–2014, 2015–2021. Key indicators: degree centrality, weighted degree centrality (sum of link weights), network density, and centralization (in- and out-). Gephi was used to compute network metrics and visualize networks; ArcGIS mapped city-level spatial distributions and overlaid network visualizations.

Technology classification: IPC section and subclass levels were used to analyze technology domains and temporal trends, including Sankey analysis of top 10 subclasses in three-year intervals.

Overall trends and policy stages:

• 65,055 patents were transferred from 882 universities to 24,869 enterprises (2001–2021), with an average annual growth rate near 40%.
• Three stages aligned with policy shifts: 2001–2007 (low, volatile transfers; universities lacked disposal rights), 2008–2014 (rights delegated but revenues retained by central government; transfers remained below ~3000/year), 2015–2021 (2015 law allowed universities/researchers to retain all income; transfers surged, peaking at 9,092 in 2019; further spikes in 2020–2021 possibly due to COVID-19-induced substitution toward domestic knowledge).

Technology fields (IPC):

• Section-level structure stabilized after 2008; Categories C (chemistry, metallurgy) and G (physics) together ~50% of assignments. C declined gradually; G increased. Categories A (human necessities), B (performing operations, transporting), and H (electricity) together ~40%; A and H declined, B grew; D, E, F together ~10%.
• Subclass trends: G01N (investigating/analyzing materials) took top rank after 2012; G06F (electric digital data processing) ranked second after 2015. A61K (preparations for medical/dental/toiletry) was consistently active but declined recently. H04L (digital information transmission) rose to 4th in 2016–2018, then 9th in 2019–2021. C02F (water/effluent treatment) rose to 2nd in 2013–2015, then declined. Other persistent C subclasses included C07C, C07D, C08L. B01J (chemical/physical processes) rose from 10th (2013–2015) to 3rd (2019–2021). No popular fields in D, E, F.

Spatial distribution and mismatch:

• Supply (university-origin patents) and demand (enterprise-acquired patents) concentrated in eastern coastal regions and provincial capitals (Beijing, Shanghai, Nanjing, Hangzhou, Guangzhou; Wuhan, Hefei, etc.). Northeast capitals (Harbin, Changchun, Shenyang) are strong suppliers but face limited local absorption; Pearl River Delta shows high demand exceeding local supply, sourcing externally. Mismatches observed between local university technology supply and regional industrial demand (e.g., Tianjin’s supply vs demand by IPC subclasses).

Network topology (organization level):

• Nodes and links increased rapidly while density decreased: nodes 362→4144→22,348; links 282→4178→24,271; density 0.00216→0.00024→0.00005 (2001–2007→2008–2014→2015–2021).
• Centralization trends: in-centralization 0.009→0.017→0.002; out-centralization 0.056→0.050→0.026, indicating decentralization in outflows over time.
• Average outdegree and indegree (and weighted variants) rose; weighted outdegree far exceeds outdegree, implying universities often transact multiple patents with many firms. Enterprises tend to connect with a single university (small indegree vs weighted indegree differences). Universities dominate outflows.
• Heterogeneity increased: coefficients of variation of weighted outdegree/indegree rose over time; differences among central universities narrowed between 2008–2014 and 2015–2021.

Key actors:

• Highly skewed participation: ~67% of assignments from top 10% of universities; 10% of enterprises purchased ~51% of patents.
• 985/211 project universities occupy central positions (higher outdegree/weighted outdegree), though their share of total weighted outdegree declined (83.1%→68.1%→41.7%), reflecting network expansion.
• Science, engineering, and comprehensive universities became increasingly central; average weighted outdegree for these types: 5.816, 27.252, 85.065 across the three periods vs 5.921, 22.895, 51.880 for others.
• Rapidly rising universities in the Yangtze River Delta (e.g., Changzhou University, Nantong University, Zhejiang Sci-Tech University) show strong recent growth, indicating regional demand pull.
• Top universities by weighted outdegree include Tsinghua, Zhejiang University of Technology, Jiangnan University, Harbin Institute of Technology, Shanghai Jiao Tong University, Changzhou University, Jiangsu University, Xi'an Jiaotong University, Zhejiang University, Beijing University of Technology.
• Top enterprises by weighted indegree include IP service firms (e.g., Guangdong Gaohang IP Operations; Zhejiang Pinchuang IP Service) and university-run intermediaries (e.g., Liyang Changda Technology Zhuanyi Center; Jiangyin Zhichanghui IP Operation) and technology firms (HIT Robot Group). State Grid is also a major acquirer.

Geography of flows (distance and scales):

• Nearly 45% of assignments occur within 100 km. Sharp drop from 100–400 km; relatively flat from 500–1100 km with a modest bump at 900–1100 km due to inter-major-city flows; <1% beyond 2000 km.
• Average transfer distance rose over time, peaking at 561 km in 2020. Intra-city transfers dominated early (peaked at 74% in 2003) but declined; inter-city across provinces increased rapidly (by 2004) and stabilized around 43%; inter-city within provinces increased to 2013 and then fluctuated between 14–18%. Overall pattern indicates gradual delocalization.

Inter- and intra-regional evolution:

• Intra-regional transfers expanded from 32 cities (2001–2007) to 188 (2015–2021). Beijing, Shanghai, Hangzhou led in 2015–2021 (2287, 1472, 1366, respectively), with strong growth in Wuhan, Guangzhou, Xi’an, Harbin, and Yangtze River Delta cities (Changzhou, Suzhou, Zhenjiang, Wuxi).
• Inter-regional network evolved from sparse, hierarchical diffusion among capitals (2001–2007) to dense, trapezoidal concentration in five megalopolises (2015–2021): Beijing-Tianjin, Yangtze River Delta, Pearl River Delta, Chengdu-Chongqing, and Harbin-Changchun-Shenyang. Knowledge mainly flows west→east and north→south; Yangtze River Delta and Beijing act as national centers; Pearl River Delta is a strong importer.
• City roles via normalized in/outdegree: Beijing is a national hub throughout; Nantong a persistent importer; Shenzhen, Suzhou, Guangzhou moved from periphery to importers; Shanghai shifted from exporter (2008–2014) to importer (2015–2021); Xi’an and Nanjing remained exporters; Hangzhou moved from periphery to importer. Potential future hubs/exporters include Shanghai, Nanjing, Wuhan, Chengdu, Chongqing; Huzhou and Jiaxing may become importers.

The study answers the research questions by mapping the UTFN’s structure, identifying central actors, revealing popular technologies, and classifying city roles. Patent assignments effectively capture economically meaningful, directed technology flows including tacit components, overcoming limitations of citation and co-patent analyses. Findings show that policy reforms (2007, 2015) are tightly linked to transfer volumes, underscoring the importance of incentive design. Technologically, enterprise demand has concentrated in measurement/analysis (G01N), computing (G06F), digital transmission (H04L), and selected chemical domains (C07D, C02F), aligning with areas where domestic firms seek to strengthen capabilities. Organizationally, UTFs are dominated by prestigious, science- and engineering-focused universities, with growing participation from regional universities in dynamic economic regions; IP intermediaries and university-run entities are pivotal in brokering and commercialization. Spatially, while UTFs remain localized, a clear delocalization trend emerges due to mismatches between local university supply and regional industrial demand and differing technological trajectories. The network’s decentralization in outflows reflects broader participation and diversification of channels and actors. These insights refine regional innovation system perspectives by evidencing multi-scalar roles of universities (local, regional, national) and challenge assumptions of strictly localized spillovers, indicating that policy should be tailored to network positions and regional absorptive capacities. Policy implications include prioritizing R&D in high-demand technologies, focusing support on central universities for commercialization capacity, and designing differentiated regional strategies to bridge supply-demand gaps and build inter-regional transfer networks.

This study provides a comprehensive, multi-level analysis of university technology flows in China using patent assignment data (2001–2021). It demonstrates strong policy sensitivity of transfer volumes, technological concentration in certain IPC domains (notably chemistry and physics at section level; G01N, G06F, H04L, C07D at subclass level), and organizational heterogeneity with central roles for prestigious science/engineering universities and IP service intermediaries, including university-run entities. Spatial analysis reveals localization alongside progressive delocalization driven by regional supply–demand mismatches, with major megalopolises structuring inter-regional flows and distinct city roles (hubs, importers, exporters). The study contributes to theories of the university’s third mission and regional innovation by empirically detailing UTFN structures and dynamics.

Future research should compare patent assignments with other formal and informal channels (e.g., licensing, contract research, satellite institutes), investigate technology-domain-specific flow mechanisms, and further examine the spatial mechanisms underlying delocalization and network evolution to inform targeted policy and management interventions.

The analysis focuses on invention patent assignments, which, although reflective of economic value and directionality (including tacit transfer), do not capture all UTF channels (e.g., licensing, consulting, contract research, joint labs, satellite institutes). Results may vary across technology domains, suggesting the need for domain-specific analyses. Data rely on disambiguation and publicly available platforms; despite manual checks, residual identification errors may exist. Network measures capture observed assignments within 2001–2021 and may underrepresent very recent transactions pending database updates.