Do global innovation networks influence the status of global value chains? Based on a patent cooperation network perspective

The paper examines whether and how a country’s integration into global innovation networks (GINs) affects its position in global value chains (GVCs). Motivated by the parallel globalization of trade and innovation, and mixed evidence on innovation’s role in GVC upgrading, the study addresses two questions: (1) What is the relationship between GINs and GVCs, and how does participation in GINs influence a country’s GVC position and through what mechanisms? (2) Do GIN characteristics affect countries differently depending on their development stage, and why? The context is the increasing internationalization of R&D and the shift toward open, cross-border, interdisciplinary innovation that firms cannot achieve alone. The study argues that GINs may upgrade GVC positions via knowledge flows, industry linkages, and human capital flows, but effects likely differ by country capabilities and constraints. The paper contributes by adopting a global perspective, integrating social network analysis with econometric modeling, and providing network-level evidence relevant for policy on cross-border innovation cooperation.

The literature review covers: (1) Global innovation networks (GINs): introduced by Ernst (2009) and developed by others, GINs integrate dispersed R&D, product development, and engineering across borders. They are characterized by geographic breadth, network scope, and innovative outcomes, enabling knowledge transfer and technology diffusion beyond traditional procurement and trade. (2) Global value chains (GVCs): building on Porter’s value chain and Gereffi’s GVC concept, GVCs are cross-enterprise global networks spanning production, distribution, consumption, and recycling. UNIDO’s definition underscores the global coordination of value-added activities. (3) GVC upgrading: Schmitz and Humphrey’s four types (process, product, functional, and chain) are noted; innovation is central to upgrading, with studies highlighting capability constraints in emerging economies and the productivity/upgrading effects of innovation. (4) Relationship between GINs and GVCs: Theoretical channels include knowledge flows (MNC R&D and partnerships embedded in GINs), linkage effects coupling intangible and tangible chains, and human capital mobility generating spillovers. Evidence is mixed: some find positive effects of GINs on upgrading (e.g., Pietrobelli & Rabellotti; Sachwald), while others highlight unequal value capture and dependence shaped by absorptive capacity and governance (e.g., Dedrick et al.). Existing research is heavy on case studies and often neglects the globalization of innovation itself, motivating broader empirical analysis.

Design: The study constructs a global innovation cooperation network using international co-invention patent data and empirically evaluates how GIN characteristics influence countries’ GVC positions using panel econometric models with fixed effects. Data and sample: - Innovation cooperation: WIPO PCT international patent database. Co-invention between inventors from different countries proxies bilateral innovation cooperation. Sample includes 46 countries (covering ≥80% of global GDP and >70% of global R&D and PCT patents) across multiple regions. Period: 2000–2018 (limited by data availability). - GVC position: OECD TiVA database for value-added trade measures. - Controls: World Bank indicators. Network construction and measures: - For each year t, build adjacency (unweighted) matrix A_t where a_ij=1 if any co-invention occurs between countries i and j; otherwise 0. Build weighted matrix W_t where w_ij equals the count of co-inventions (rescaled to 0–1 by dividing by the max). - Overall network density computed with UCINET 6.0 shows an increasing trend from 2000 (≈7.01) to 2018 (≈17.43), indicating denser global innovation collaboration. - Node-level GIN indicators: • Degree centrality (deg): number of direct ties per country, indicating breadth of cooperation. • Eigenvector centrality (deg*): centrality weighted by the importance of neighbors; nodes connected to central nodes score higher. • Network connectivity (strength; str_i=Σ_j w_ij): sum of weighted ties, reflecting depth/intensity of cooperation. • Structural holes (constraint; Con_i=Σ_j (p_ij + Σ_q p_iq p_qj)^2): higher values imply more closure/redundancy and stronger constraints; lower values imply access to diverse, non-redundant partners. A global downward trend in constraint indicates liberalization of innovation cooperation over time. - Descriptive kernel density analyses illustrate evolving distributions of degree, eigenvector centrality, and strength. Outcome variable (explained): - GVC position index per country-year: GVC_Position_it = ln(1 + IV_it/E_it) − ln(1 + FV_it/E_it), where IV is domestic value added of indirect exports, FV is foreign value added in exports, and E is total exports. Higher values imply more upstream positions and less dependence on foreign inputs. Econometric specification: - Baseline fixed-effects panel model with country and year FE: GVC_Position_it = α + α1 deg_it + α2 str_it + α3 con_it + α4 pgdp_it + α5 open_it + α6 pop_it + α7 fcr_it + α8 rd_it + Country_i + Year_t + ε_it. - Controls: economic development (pgdp), trade openness (open), population (pop), physical capital (fcr: fixed capital share of GDP), and R&D intensity (rd: GERD share of GDP). - Standard errors clustered at the country level. Endogeneity and robustness: - Use one-year lags of endogenous regressors as instruments (per Wooldridge) and an alternative dependent variable—domestic value added (DVA) share—from OECD TiVA as an alternative measure of GVC position (Johnson, 2014). Results are compared for consistency. Heterogeneity analysis: - Split-sample regressions for developed, emerging, and developing countries to assess differential effects of GIN characteristics by development stage.

Baseline results (Table 2, fixed effects, N=874): - Network linkage strength (str) positively and significantly relates to GVC position (coef ≈ 0.162, SE 0.051, p<0.01 with controls), indicating that deeper innovation cooperation is associated with more upstream GVC positions. Authors interpret elasticities as: a 1% increase in link strength raises GVC position by about 16.2%. - Structural hole constraint (con) negatively and significantly relates to GVC position (coef ≈ −0.022, SE 0.006, p<0.01 with controls), meaning fewer constraints (more openness to diverse, non-redundant ties) are associated with higher GVC positions. Authors interpret that a 1% decrease in constraint increases GVC position by about 2.2%. - Degree centrality (deg) is insignificant without controls, but becomes significantly negative with controls (coef ≈ −0.009, SE 0.003, p<0.01), implying that broad but indiscriminate expansion of partnerships does not enhance, and may reduce, GVC position. Further analysis with eigenvector centrality (deg*, Table 3): - Eigenvector centrality is positively and significantly related to GVC position both without and with controls (coef ≈ 0.203, SE 0.044, p<0.01; with controls ≈ 0.161, SE 0.054, p<0.01). This supports Hypothesis 4: partnering broadly with highly innovative, central partners improves GVC position. Robustness (Table 4): - Using one-year lag instruments: deg*, str remain positive and significant; con remains negative and significant (R^2 ~0.51–0.53). - Using DVA-based outcome: results hold with strong significance (deg* and str positive; con negative; R^2 ~0.87+), confirming robustness across measures. Heterogeneity (Table 5 and discussion): - Developed countries: Results mirror baseline—eigenvector centrality significantly positive; simple degree centrality and strength often insignificant; structural hole constraint not significant within same-type groups, implying similar constraint structures among developed peers. - Emerging economies: International innovation cooperation shows weaker or even adverse effects; findings suggest autonomous innovation yields better upgrading outcomes than external cooperation, consistent with technological blocking by developed countries and latecomer dynamics. Reported examples: in 2018 China (2,894 collaborations; constraint 0.454) and India (1,123; constraint 0.532) had higher constraints than some developed peers (e.g., Denmark constraint 0.374; France 0.385). - Developing countries: Expanding international innovation partnerships tends to positively affect GVC position, as external knowledge access and information exchange compensate for weak domestic innovation ecosystems. Overall: Hypotheses 2 and 3 are supported (strength positive; constraint negative). Hypothesis 1 (degree centrality) is not supported; instead, a refined measure—eigenvector centrality—shows positive effects, validating Hypothesis 4.

The findings indicate that not all forms of integration into GINs equally promote GVC upgrading. Mere breadth of connections (degree centrality) may dilute resources and add low-value ties, potentially lowering GVC position. In contrast, depth (stronger ties) and high-quality central positioning (eigenvector centrality) enhance knowledge flows, trust, and access to advanced technologies, thereby facilitating movement upstream in GVCs. Lower structural constraints (fewer redundant, closed ties) allow countries to bridge structural holes, gather heterogeneous knowledge, and attract talent, further improving GVC status. These results align with theoretical channels linking GINs to GVC upgrading via knowledge, linkage, and human capital flows. The heterogeneity analysis clarifies that benefits from GINs depend on domestic innovation capacity and geopolitical/technological frictions: developed countries benefit from deeper, higher-quality integration; emerging economies may face technology blocking and insufficient absorptive capacity, making autonomous innovation more effective; developing countries typically gain from broader international cooperation due to weak local innovation systems. Together, the results directly address the research questions by identifying which GIN attributes matter, through what mechanisms, and for whom.

The study shows that GIN characteristics significantly shape countries’ positions in GVCs. Stronger network linkages and higher eigenvector centrality improve GVC status, while higher structural constraints reduce it. Simple degree centrality alone can be detrimental, but when partner quality and network influence are considered (eigenvector centrality), broader cooperation becomes beneficial. Heterogeneity analyses reveal that developed countries gain from deeper, high-quality network integration; emerging economies may benefit more from strengthening autonomous innovation due to technological blockages and constraints; and developing countries can improve their GVC positions by expanding international innovation exchanges and integrating into global R&D. Policy recommendations include: prioritize high-quality, innovation-intensive partnerships; reduce network closure to access diverse knowledge; for emerging economies, bolster indigenous R&D and reduce dependence on constrained, low-value partnerships; for developing countries, build innovation ecosystems and leverage international collaboration for capability building. The paper highlights the value of combining social network analysis with econometric modeling to inform strategies for GVC upgrading.

Data limitations include: (1) Patent cooperation data are available only through 2018 due to database constraints, limiting analysis of more recent dynamics. (2) The sample consists of 46 countries, which, while covering a large share of global GDP and R&D, excludes many economies. (3) Innovation cooperation is proxied by international co-invention patents, which may not capture all forms of cross-border innovation collaboration. (4) Despite fixed effects and instrumental variable strategies, potential residual endogeneity and measurement error cannot be fully ruled out.