Patent Similarity Data and Innovation Metrics

• Published date: 01 September 2020
• Author: Leslie DeChurch,Noshir Contractor,Alina Lungeanu,Ryan Whalen
• DOI: http://doi.org/10.1111/jels.12261
• Date: 01 September 2020

Journal of Empirical Legal Studies

Volume 17, Issue 3, 615–639, September 2020

Patent Similarity Data and Innovation

Metrics

Ryan Whalen,*Alina Lungeanu, Leslie DeChurch, and Noshir Contractor

We introduce and describe the Patent Similarity Dataset, comprising vector space model-

based similarity scores for U.S. utility patents. The dataset provides approximately 640 mil-

lion pre-calculated similarity scores, as well as the code and computed vectors required to

calculate further pairwise similarities. In addition to the raw data, we introduce measures

that leverage patent similarity to provide insight into innovation and intellectual property

law issues of interest to both scholars and policymakers. Code is provided in accompanying

scripts to assist researchers in obtaining the dataset, joining it with other available patent

data, and using it in their research.

I. Introduction and Overview

For decades, empirical research on patent law and innovation has benefited from access

to increasingly high-quality patent datasets. Scholars have used these datasets to study

innovation in a wide variety of contexts at the national level,

1

firm level,

2

team level,

3

and

*Address correspondence to Ryan Whalen, NUS Faculty of Law, 469G Bukit Timah Road 259776, Singapore;

email: ryanwhalen@nus.edu.sg. Whalen is Assistant Professor of Law at the National University of Singapore, Singa-

pore.; Lungeanu is Research Assistant Professor at Northwestern University School of Communication, Evanston.;

DeChurch is Professor at Northwestern University School of Communication, Evanston.; Contractor is the Jane

S. & William J. White Professor of Behavioral Sciences in the School of Engineering, School of Communication

and the Kellogg School of Management, Evanston.

This research was supported by NSF award number 1856090.

1

See, e.g., Raffaele Paci, Antonio Sassu & Stefano Usai, International Patenting and National Technological Spe-

cialization, 17 Technovation 25 (1997).

2

See, e.g., Michele Grimaldi, Livio Cricelli, Martina Di Giovanni & Francesco Rogo, The Patent Portfolio Value

Analysis: A New Framework to Leverage Patent Information for Strategic Technology Planning, 94 Technological

Forecasting & Soc. Change 286 (2015).

3

See, e.g., Margherita Balconi, Stefano Breschi & Francesco Lissoni, Networks of Inventors and the Role of Acade-

mia: An Exploration of Italian Patent Data, 33 Res. Pol’y 127 (2004).

615

individual level.

4

In these studies, patent data have served a wide variety of purposes.

5

For

example, citations have been used as a proxy for knowledge inputs or measure of a pat-

ent’s value,

6

patents themselves have been used as proxy measures for innovation more

generally,

7

and the structure of the prior art citation network has been used to infer the

existence of thickets of intellectual property rights.

8

This project engages with the tradition of providing patent-related data that can

enrich future research on patent law and innovation. We begin by briefly reviewing the

state of available patent data, and the research that relies on it. We subsequently intro-

duce the Patent Similarity Dataset, which uses a vector space model to compute pairwise

distances between a large number of patents. After introducing vector space models gen-

erally, and explaining how the Patent Similarity Dataset was created, this article’s final

section goes on to describe the Patent Similarity Dataset’s qualities and demonstrate how

it can be used to generate a wide variety of metrics that provide new perspective on pat-

ent law and innovation.

A. The Growth in Patent Data Availability, and Patent-Data-Driven Research

One of the functions of patent law is to incentivize the disclosure of information relating

to innovation.

9

As a result of this, the patent system generates a large amount of data,

much of which is publicly available.

10

For decades now, researchers have been drawing

on this increasingly large body of available patent data to help better understand innova-

tion, science, and intellectual property law. Because the universe of patent data is quite

large and data are available in varying formats, many of these projects require substantial

data cleaning and preparation work. Researchers thus often publish their datasets both

4

See, e.g., Martin G. Moehrle, Lothar Walter, Anja Geritz & Sandra Mu

¨ller, Patent-Based Inventor Profiles as a

Basis for Human Resource Decisions in Research and Development,” 35 R&D Mgmt. 513 (2005).

5

For a review of patent data as an economic indicator, see Zvi Griliches, Patent Statistics as Economic Indicators: A

Survey,” in R&D and Productivity 287 (Univ. of Chicago Press 1998); Sadao Nagaoka, Kazuyuki Motohashi& Akira

Goto, Patent Statistics as an Innovation Indicator, 2 Handbook of the Economics of Innovation 1083 (B. H. Hall &

N. Rosenberg, eds., North-Holland 2010).

6

See Manuel Trajtenberg, A Penny for Your Quotes: Patent Citations and the Value of Innovations,” Rand J. of

Econ. 172 (1990).

7

Daron Acemoglu, Ufuk Akcigit & William R. Kerr, Innovation Network, 113 PNAS 11483 (2016).

8

George von Graevenitz, Stefan Wagner & Dietmar Harhoff, How to Measure Patent Thickets—A Novel Approach,

111 Econ. Letters 6 (2011).

9

Jeanne C. Fromer, Patent Disclosure, 94 Iowa L. Rev. 539 (2008–2009).

10

Indeed, the Patent Act requires that the PTO make patent data available. 35 U.S.C § 41(i). For an overview of IP

data, see David L. Schwartz & Ted Sichelman, Data Sources on Patents, Copyrights, Trademarks, and Other Intel-

lectual Property, in Research Handbook on the Economics of Intellectual Property Law (Edward Elgar Publish-

ing 2019).

616 Whalen et al.