How Can Innovation Screening Be Improved? A Machine Learning Analysis With Economic Consequences for Firm Performance

51 Pages Posted: 14 May 2021 Last revised: 19 Jan 2024

See all articles by Xiang Zheng

Xiang Zheng

University of Connecticut - Department of Finance

Date Written: January 18, 2024

Abstract

In this study, I utilize data from the United States Patent and Trademark Office (USPTO) patent applications to explore potential improvements in the U.S. patent examination process through machine learning algorithms. I find that combining human examiners' expertise with machine learning predictions would yield a 13.2% to 15.5% gain in patent generality and a 29.6% to 35.6% gain in patent citations for granted patents. Using machine learning predictions as a benchmark, I calculate the false acceptance rate of each patent examiner and study its impact on firms. I show that falsely accepted patents are more likely to expire early. Furthermore, such errors adversely affect the operating performance of public companies and diminish the likelihood of successful exits (IPOs or M&A) for private firms. Overall, this study documents the achievable social and economic gain of integrating machine learning as a robo-advisor in the current patent screening system.

Keywords: Machine Learning, Patent Screening, Firm Performance

JEL Classification: C55, G32, O31

Suggested Citation

Zheng, Xiang, How Can Innovation Screening Be Improved? A Machine Learning Analysis With Economic Consequences for Firm Performance (January 18, 2024). Available at SSRN: https://ssrn.com/abstract=3845638 or http://dx.doi.org/10.2139/ssrn.3845638

Xiang Zheng (Contact Author)

University of Connecticut - Department of Finance ( email )

School of Business
2100 Hillside Road
Storrs, CT 06269
United States

HOME PAGE: http://www.xiangzheng.info/

Do you have negative results from your research you’d like to share?

Paper statistics

Downloads
341
Abstract Views
1,663
Rank
159,063
PlumX Metrics