Understanding formation mechanisms of regional collaborative innovation networks is crucial for effective innovation policy design, yet existing research lacks comprehensive empirical examination of how multiple mechanisms operate simultaneously across network evolution phases.
This study conducts the first longitudinal multi-mechanism analysis by examining 57,846 collaborative patent applications among 3,582 organizations in the Guangdong-Hong Kong-Macao Greater Bay Area (2009–2023). Using phase-segmented exponential random graph models across three developmental stages, we assess five key mechanisms: preferential attachment, triadic closure, brokerage potential, organizational homophily, and experience accumulation.
Our findings reveal a paradoxical “elite-driven community closure” trajectory. Despite network expansion, declining density indicates heightened partner selectivity. Preferential attachment effects persist throughout all phases, sustaining elite hub dominance, while experience accumulation effects diminish over time. Strengthening triadic closure alongside persistently negative brokerage effects creates closure-dominated topologies limiting cross-boundary knowledge flows. Most critically, persistent organizational homophily constrains industry-university-research integration despite policy incentives, demonstrating that structural inertia resists policy intervention more than previously assumed.
The analysis focuses on collaboration quantity rather than quality outcomes and applies specifically to patent-based collaboration networks.
Three targeted policy adjustments are recommended: (1) reducing cross-type collaboration transaction costs through institutional innovation, (2) counteracting preferential attachment concentration via peripheral engagement strategies, and (3) creating reputation-sharing infrastructure to extend trust beyond homophily clusters.
Our mechanism-based approach advances network evolution theory by revealing how micro-level collaboration decisions aggregate into macro-level structural patterns.
