Story
Is there a secular decline in disruptive patents? Correcting for measurement bias
Key takeaway
A study suggests the common perception of declining disruptive innovation may be due to measurement issues, not an actual slowdown. This is significant as it could shape innovation policy and public discourse.
Quick Explainer
This study examines recent claims of a decline in disruptive patents over time. It finds that the apparent decline is largely an artifact of measurement biases in the prior analysis. The key methodological improvements include accounting for truncation bias (by not excluding early citations) and exclusion bias (by considering patent applications). These corrections reveal that the number of highly disruptive patents has actually increased in recent years, contradicting the original conclusion of "remarkable stability." The findings suggest caution in using the prior results as a basis for policy changes, as the observed trends may not reflect true shifts in the nature of innovation.
Deep Dive
Technical Deep Dive: Secular Decline in Disruptive Patents
Overview
This study examines recent claims of a secular decline in disruptive patents over time. The 2023 Nature paper by Park et al. reported a substantial decrease in the disruptiveness of scientific papers and patents from 1945-2010, suggesting a fundamental shift in the nature of innovation. However, this analysis finds that the apparent decline is largely an artifact of measurement biases in the methodology used by Park et al.
Methodology
The study uses the same patent data as Park et al., focusing on granted USPTO utility patents from 1980-2016. It calculates a "Consolidating vs. Disruptive" (CD) index for each patent, which characterizes whether a patent is consolidating (building on prior work) or disruptive (pushing in new directions).
The key methodological differences are:
- Accounting for truncation bias: Park et al. excluded all backward citations to patents published before 1976. This was shown to artificially inflate the CD index for older patents.
- Accounting for exclusion bias: Park et al. did not consider citations to patent applications, which became increasingly common after a 1999 law change.
Results
- Correcting for truncation bias substantially decreases the apparent secular decline in patent disruptiveness reported by Park et al. The CD index shows only a marginal decrease from 1980-2010.
- Correcting for exclusion bias reveals an increase in the number of highly disruptive patents over time, particularly since 2008. This contradicts Park et al.'s claim of "remarkable stability" in the number of highly disruptive patents.
- The magnitude of truncation bias varies by technology field, with a stronger effect in Mechanical patents compared to IT patents.
Interpretation
The study's findings suggest caution in using the results from Park et al. as a basis for research policy changes or industry restructuring aimed at altering the innovation landscape. The apparent decline in disruptive patents is shown to be largely an artifact of measurement biases, rather than a true shift in the nature of innovation.
Limitations & Uncertainties
- The analysis is limited to USPTO patents and may not generalize to scientific publications or other innovation metrics.
- While the study corrects for two key biases, there may be other measurement issues that were not accounted for.
- The interpretation focuses on policy implications, but does not explore the underlying drivers of any real changes in disruptive innovation.
What Comes Next
Future research could:
- Investigate other datasets and innovation metrics beyond patents
- Explore the factors influencing trends in disruptive versus consolidating innovation
- Develop more robust methods for measuring disruptiveness that are resistant to biases
Sources:
