Building on a FAIRly Strong Foundation to Connect Academic Research to Transl...
130-Ince Business patenting and publishing
1. The current database project aims at developing science and innovation indicators through patent and publication data
for world top business R&D spenders.
The EU Industrial R&D Investment Scoreboard of European Commission (The Scoreboard) represents around 90% of
total world business R&D spending in 2014. 1 The project firstly matches this highly representative firm sample with their
patent and publication data. . The core constructed database combine firms’ science and innovation indicators with firms’
financials data.
Furthermore, cleaning of researcher names and matching between inventors and authors allow for identification of two
important information about the private sector researchers: (i) Firm-level author-inventor pairs among researchers that
are active in publishing and patenting and (ii) Accurate information about the inventor teams of patents and author teams
of publications such as their size, geographical dispersion and affiliations.
The following sections describe the database construction, main results of matching of the Scoreboard with PATSTAT
and SCOPUS and then give further insights developed by the cleaning and matching of the researchers’ data of patent
and publications.
2. Data and Methodology
4. Conclusion
1. Introduction
3. Results
5. References
3.Results (Continued)
Business patenting and publishing:
Evidence from world top R&D spenders
Ela INCE
ULB, SBS – EM
European Center for Advanced Research in Economics and Statistics (ECARES)
The International Centre for Innovation, Technology and Education Studies (iCite)
Arora, A., Belenzon, S., & Patacconi, A. (2015). Killing the Golden Goose? The Decline of Science in Corporate R&D,
NBER Working Paper No. 20902.
Czarnitzki, D., & Kraft, K. (2010). On the profitability of innovative assets. Applied Economics, 42(15), 1941–1953.
Debackere, K., Magerman, T. & Van Looy, B., (2011). In search of anti-commons: patent paper pairs in biotechnology. An
analysis of citation flows. ZEW Conference on Economics of Innovation and Patenting edition:4.
Jones, B. F., Wuchty, S., & Uzzi, B. (2008). Multi-university research teams: shifting impact, geography, and stratification
in science. Science (New York, N.Y.), 322(November), 1259–1262.
Kaiser, U. (2009). Patents and profit rates. Economics Letters, 104(2), 79–80.
6. Acknowledgements
3.4. University – industry links in scientific publications
3.2. Firm-level author – inventor pairs
3.1. Patenting and publishing propensities
3.3. Technological applicability and team size in patenting
3.1.1. Patenting and publishing of top 10
Dividing number of patents and publications by R&D
spending for top R&D spenders exhibit clustering
depending on the main sector of the firm.
Samsung and Siemens mainly active on industrial goods
& services sector hold the highest patenting propensities.
However, Samsung, being active on more than one
sector, publishes more and patents less on average with
respect to its overall R&D spending.
Firms active on automobiles sector have higher patenting
but lower publishing propensities than firms active in
health care sector except for Johnson & Johnson.
Microsoft is the only software-technology firm shown
among the sample.
3.1.2. Patenting and publishing by sector
To ease the reading of graphics, sectors are aggregated
to Industry Classification Benchmark (ICB) 2-digit sector
codes.
The results exhibit differences in terms of patenting and
publishing propensities across sectors.
While some sectors are more intensively patenting such
as industrial goods & services and automobiles & parts,
other sectors such as food & beverage and utilities are
more intensive in publishing.
Except for financial services, bank and insurance sectors,
top business R&D spenders are substantially publishing
and patenting as a result of their R&D activities.
2.1. Matching the Scoreboard with PATSTAT & SCOPUS
The Scoreboard allows for building an unbalanced firm
panel with capital expenditure, operational profit, R&D
spending and employment information for the years 2000-
2013.
With the use of HAN database for PATSTAT, 79 % of the
Scoreboard firms is matched with patent applicant names.
The matching rate of the Scoreboard firms with SCOPUS is
70%.
The matching procedure allows constructing firm, sector,
patent and publication level variables for the project.
2.2. Cleaning and matching of author–inventor names
Researchers that patent are called inventors and
researchers that publish articles in scientific journals are
called authors.
Author-inventor names are cleaned and harmonised. The
first letter of first name and the full last name matching
between the author and inventor names for each
Scoreboard firm is applied for the author-inventor name
matching.
Firm-level author-inventor pairs construct on average 9.7%
of the total number of authors and 8.5% of the total
inventors of the firms.
3.3.1. Team behaviour
Firm - level inventor names’ harmonisation is implemented
for the inventor names of European Patent Office (EPO)
patents.
Cleaning and taking into account the commonness of the
first names and the surnames of inventors, inventor names
are harmonised within a firm.
The results show that the size of the patenting teams of
firms is significantly rising over time on average, leaving a
small sample of travel & leisure sector firms as an
exception.
3.3.2. Specialization vs generalization
The extent of technological applicability in patenting are
measured by the number of 4-digit international patent
classification (IPC) codes given to each patent.
IPC codes of patents represent the different areas of
technology to which they pertain.
Significant decline in the number of IPC codes of firms’
patents between 2000 and 2010 points out specialisation of
firms’ patenting over time. The only exception for this trend
is, as for the size of patenting teams, the travel and leisure
sector firms.
The percentage of academia affiliated authors within the
team of authors per publication is constructed through the
cleaning and harmonisation of author names per firm. This
gives possibility to look at the evolution of percentage of
authors from academia per publication over time.
The graphic shows the evolution for countries where the
most of the private sector publishing takes place among
the headquarter countries of top business R&D spenders.
Chinese firms have experienced the most significant rise
in university - industry links for the period 2000 – 2010.
European firms lag their American and Asian counterparts.
Correcting the EU sample size
The Scoreboard provides a separate list of the
top European business R&D spenders which leads to
a bias in the sample in terms of size between
the EU and the non - EU firms.
Excluding the medium size of the EU firms from the
full sample, the results for the low levels of university –
industry links remain significant between
the EU and the non - EU firms.
0
.1.2.3.4
Average%ofauthorsfromacademia
2000 2002 2004 2006 2008 2010
year
USA France Germany
UK Japan China
Contact Information:
SBS-EM, Université Libre de Bruxelles
Avenue F.D. Roosevelt 50, Brussels, Belgium
Phone: +32 (0)650 3929
Email: ela.ince@ulb.ac.be
3.2.1. Publishing of inventors
An inventor that doesn’t appear in scientific articles as an
author is called “Only Inventor” whereas the one that
publishes scientific articles in journals is called “Author
Inventor”.
Inventor – level patent quality and quantity variables are
constructed as follows through patent data:
i. Patent quality: forward citations from 3 - year window
corrected by patent family size and technology codes.
ii. Patent quantity: patent divided by the size of the
inventor team
Testing the difference between the average patenting
quality and the patenting quantity suggests that inventors
that are also active in publishing are significantly making
better and more patents than the inventors that are
concentrated only on patenting:
3.2.2. Patenting of authors
An author of a firm that is not listed among the listed
inventors of firm patents is called “Only Author” whereas a
firm author that is also active in firm patenting is called
“Inventor Author”.
Author - level publication quality and quantity variables
are constructed as follows through publication data:
i. Publication quality: publication weighted by their
journal impact factor.
ii. Publication quantity: publication divided by the size of
the publisher team.
Testing the difference between the average publishing
quality and quantity per author suggests that authors that
are also active in patenting are significantly making better
and more publications than their counterparts that only
publish:
The database is a joint project with Lauriane Dewulf, Prof. Michele Cincera and Prof. Nicolas van Zeebroeck.
I thank all members of iCite for the advice on the data.
0 .2 .4 .6 .8 1
Insurance
Banks
Financial Services
Utilities
Technology
Media
Travel & Leisure
Retail
Food & Beverage
Real Estate
Telecommunications
Oil & Gas
Automobiles & Parts
Basic Resources
Personal & Household Goods
Construction & Materials
Chemicals
Health Care
Industrial Goods & Services
Sectors
patent propensity publication propensity
The developed science and innovation indicators for the world top business R&D spenders allow for investigating various
questions in the literature with a highly representative sample.
First of all, as for Kaiser (2009) and Czarnitzki & Kraft (2010), the database generates profitability variable measured by
operational profit divided by sales and allows for looking at the impact of the innovative assets on the firm performance.
Furthermore, through patent and publication data and indicators generated by their combination, it is possible to make a
comparative analysis on their contribution or on firms’ preferences on the disclosure of innovation through patents and
publications as Arora et al (2015).
Second, following Jones et al (2008), research team organisations - technological behaviour in innovation could provide
evidence for the literature from private sector patenting. Furthermore, publication data enables to observe linkages
between industry and other institutions through authors’ affiliations.
Lastly, author - inventor pairs may lead to contribute to current strand of literature on author - inventor and patent –
publication pairs. Following Debackere et al., (2009), the project adds one more level to the matching procedure which is
firm - level matching and may generate firm-level patent-publication pairs. This extends the literature concentrated on
sector level analysis, on small sample of firms or on academic patent-publication pairs.
1The 2015 EU Industrial R&D Investment Scoreboard: http://iri.jrc.ec.europa.eu/scoreboard.html
SAMSUNG
GSK PFIZER
FORD
MICROSOFT
DAIMLER
GM
VW
SIEMENS
J&J
0
.2.4.6.8
0 .05 .1 .15 .2
Publishing propensity (Number of publications/€m)'
Top Business R&D Spenders
Automobiles
Basic Resources
Chemicals
Construction
Food & Beverage
Health Care
Industrials
Media
Oil & Gas
HH Goods
Retail
Technology
Telecomm.
Travel & Leisure
Utilities
Automobiles
Basic Resources
Chemicals
Construction
Food & Beverage
Health Care
Industrials
Media
Oil & Gas
HH Goods
Retail
Technology Telecommunications
Travel & Leisure
Utilities
1
1.5
2
2.5
3
2 2.5 3 3.5 4 2 2.5 3 3.5 4
year 2000 year 2010
Applicability-IPCcodesperpatent
Team size - Number of inventors per patent
Variable Obs.
Only
Inventor
Obs.
Auth.
Inventor
Mean
Diff.
Quality 1914119 0.643 277124 0.697 -0.053***
Quantity 1914162 1.670 277134 2.034 -0.365***
(*) if p < 0.05, (**) if p < 0.01, and (***) if p < 0.001.
Variable Obs. Only
Author
Obs. Invt.
Author
Mean
Diff.
Quality 2337323 1.771 227285 2.557 -0.786***
Quantity 2337319 1.307 227285 1.848 -0.542***
(*) if p < 0.05, (**) if p < 0.01, and (***) if p < 0.001.