Reliance on science: Worldwide front‐page patent citations to scientific articles

• DOI: http://doi.org/10.1002/smj.3145
• Date: 01 September 2020
• Published date: 01 September 2020
• Author: Matt Marx,Aaron Fuegi

RESEARCH ARTICLE

Reliance on science: Worldwide front-page

patent citations to scientific articles

Matt Marx

1

| Aaron Fuegi

2

1

Boston University Questrom School of Business, Boston, Massachusetts

2

Boston University Information Services & Technology, Boston, Massachusetts

Correspondence

Matt Marx, Boston University Questrom

School of Business, 595 Commonwealth

Avenue, Boston, MA 02215, room 643A.

Email: mattmarx@bu.edu

Funding information

National Science Foundation, Grant/

Award Number: 1735669

Abstract

Research summary:To what extent do firms rely on

basic science in their R&D efforts? Several scholars

have sought to answer this and related questions, but

progress has been impeded by the difficulty of matching

unstructured references in patents to published papers.

We introduce an open-access dataset of references from

the front pages of patents granted worldwide to scien-

tific papers published since 1800. Each patent-paper

linkage is assigned a confidence score, which is charac-

terized in a random sample by false negatives versus

false positives. All matches are available for download

at http://relianceonscience.org. We outline several ave-

nues for strategy research enabled by these new data.

Managerial summary:To what extent do firms rely

on basic science in their R&D efforts? Several scholars

have sought to answer this and related questions, but

progress has been impeded by the difficulty of matching

unstructured references in patents to published papers.

We introduce an open-access dataset of references from

the front pages of patents granted worldwide to scien-

tific papers published since 1800. Each patent-paper

linkage is assigned a confidence score, and we check a

random sample of these confidence scores by hand in

order to estimate both coverage (i.e., of the matches we

Received: 8 March 2019 Revised: 10 January 2020 Accepted: 14 January 2020 Published on: 16 April 2020

DOI: 10.1002/smj.3145

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and

reproduction in any medium, provided the original work is properly cited.

© 2020 The Authors. Strategic Management Journal published by John Wiley & Sons, Ltd.

1572 Strat. Mgmt. J. 2020;41:1572–1594.wileyonlinelibrary.com/journal/smj

should have found, what percentage did we find) and

accuracy (i.e., of the matches we found, what percent-

age are correct). We outline several avenues for strategy

research enabled by these new data.

KEYWORDS

basic science, dataset article, patent citations, research &

development

1|INTRODUCTION

This paper details the construction of a publicly-available set of citations from U.S. patents

(1947–2018) and non-U.S. patents (1782–2018) to scientific articles (1800–2018). We establish

approximately 22 million patent citations to science. The patent-paper linkages, as well as

selected metadata on the articles (whether cited or not), and the source code are publicly avail-

able at http://relianceonscience.org.

Patent citations to science (hereafter, PCS) are of interest to strategy researchers who seek to

understand innovation in firms: the nature of research and development, how inventors and sci-

entists search for commercializable basic science, and the process by which university inventions

are exploited by firms. Despite these advantages, PCS have only sometimes been used in strategy

research, for at least two reasons. First, PCS are difficult to work with given that they appear in

patent records as unstructured text strings. Thus researchers must either match patents and sci-

entific articles by hand (for small samples) or (for large samples) build algorithms that are possi-

bly error prone. Second, even when research teams have invested the effort to link patents and

scientific articles at scale, they have typically done so using proprietary databases such as Scopus

or the Web of Science. Thus the matched PCS cannot be shared with other research teams, who

must license the databases for themselves and/or develop algorithms from scratch.

As other research teams have (Gaetani & Li Bergolis, 2015; Fleming, Greene, Li, Marx, &

Yao, 2019), we link data from the U.S. Patent & Trademark Office to a broad set of scientific arti-

cles not limited by industry or field. Our linkages involve not only proprietary article databases,

which cannot be shared, but also a newly-available, open-source database from Microsoft (Sinha

et al., 2015) which permits us to post the resulting PCS for public use. Based on third-party

assessment, we estimate that our algorithm can capture up to 93% of patent citations to science

with an accuracy rate of 99% or higher. We believe this to be the longest panel of patent-to-paper

citations (spanning more than seven decades) that is publicly available and is accompanied by

rigorous performance metrics. We also provide matches fromworldwide patents to PubMed.

The paper is organized as follows. We begin by motivating the use of PCS in strategy research and

review prior approaches. Second, we detail our patent-paper linking algorithm. Third, we describe

both the private and publicly-available data products as well as our methods for assessing their effi-

cacy. We conclude by sketching research avenuesopenedupbythebroadavailabilityofPCS.

2|MOTIVATION

Innovation is a key source of sustainable differentiation for firms and thus a longtime focus of

strategy researchers. The lottery-based nature of research & development (R&D) has long

MARX AND FUEGI 1573