Dynamics of knowledge production: A relational-event analysis of patent citation hazards in hydrogen technologies

Research output: Contribution to journalArticlepeer-review

1 Downloads (Pure)

Abstract

Technological improvement rates are shaped by the evolving structure of complex patent-citation networks. We model this dynamic link using a relational-event survival framework that estimates citation hazards for individual patents and subdomains within hydrogen technologies. The hazard, i.e. the probability that a patent receives a new citation given time since its last, represents the instantaneous rate of knowledge production. Applying this approach to 777,000 patents and 1.3 million citation events (1841–2023) reveals systematic subdomain differences. Hydrogen distribution shows the slowest citation hazard, implying a persistent knowledge-production bottleneck that, under standard learning assumptions, translates into long-run cost persistence. Katz centrality strongly predicts citation hazards and absorbs most subdomain effects, indicating that structural position in the knowledge network matters more than categorical labels. The weaker centrality effect in distribution reflects organizational and market frictions that slow inventive diffusion. The results demonstrate that proportional citation hazards capture stable improvement rates and that deviations identify structural bottlenecks. This provides a tractable diagnostic for prioritizing R&D and infrastructure policies that accelerate system-level decarbonization.
Original languageEnglish
Article number101460
JournalSustainable Futures
Volume10
Early online date15 Nov 2025
DOIs
Publication statusPublished - Dec 2025

Keywords

  • Dynamic networks
  • Citation rates
  • Relational event model
  • Technology domains
  • Technological improvement rate

Fingerprint

Dive into the research topics of 'Dynamics of knowledge production: A relational-event analysis of patent citation hazards in hydrogen technologies'. Together they form a unique fingerprint.

Cite this