Patent Secrecy and Defense Innovation: Theoretical Foundations and Cross-National Evidence from Declassified Data

The use of invention secrecy for reasons of national security remains a relatively underexplored topic in the economic literature, particularly from the perspective of innovation studies. Nonetheless, the limited data available provide critical insights into the structure and strategic orientation of national Defence Technological and Industrial Bases (DTIBs). The intensity with which states apply patent secrecy measures can also be interpreted as an indicator of their perceived international threat environment. This article analyzes two datasets of declassified patents, 2 346 U.S. patents (1960–2020) and 1,152 Chinese patents (2000–2020), alongside aggregate statistics from the UK Intellectual Property Office. Our analysis highlights substantial variation in secrecy practices across these countries. The United States and the United Kingdom apply secrecy orders across a broad array of private sector actors, whereas China enforces secrecy more extensively, focusing on civilian-relevant technologies within entities that are entirely state-controlled. This divergence echoes Cold War-era distinctions between Eastern and Western bloc practices (Martens, 2021). More generally, patents held under secrecy by state institutions typically originate from closely affiliated private partners, if not from the institutions themselves. The restricted knowledge does not reflect absorptive processes from the civilian sector; rather, it serves as a mechanism of self-protection for national DTIBs. Our findings reveal a significant correlation between a nation's perceived threat level and the intensity of its use of patent secrecy, suggesting a likely global increase in the classification of technologies in the coming decades. Despite its importance, patent secrecy does not appear to constitute a fundamental limitation for studying defence-related innovation through patent data. The practice remains proportionally limited, with the average duration of secrecy being less than four years, thereby allowing for meaningful empirical analysis of innovation dynamics within DTIBs.