This article appeared in a journal published by Elsevier. The attached
copy is furnished to the author for internal non-commercial research
and education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elsevier’s archiving and manuscript policies are
encouraged to visit:
http://www.elsevier.com/authorsrights
Author's personal copy
Research Policy 42 (2013) 1780– 1792
Contents lists available at ScienceDirect
Research  Policy
j o ur nal homep age: www.elsev ier .com/ locate / respol
Reassessing  patent  propensity:  Evidence  from  a  dataset  of  R&D
awards,  1977–2004
Roberto  Fontanaa,∗,  Alessandro  Nuvolarib,  Hiroshi  Shimizuc, Andrea  Vezzullid
a Department of Economics ad Management, University of Pavia, Via San Felice 5, 27100, Pavia & CRIOS – Bocconi University, Via Sarfatti 25, 20139 Milano,
Italy
b LEM – Sant’ Anna School of Advanced Studies, Piazza Martiri della Libertà 33, 56127 Pisa, Italy
c Institute of Innovation Research – Hitotsubashi University, Tokyo, Japan
d UECE-ISEG, Universitade Técnica de Lisboa, Rua Miguel Lupi, 20, 1249-078 Lisboa, Portugal
a  r  t  i  c  l  e  i  n  f  o
Article history:
Received 28 November 2011
Accepted 25 May  2012
Available online 16 September 2013
Keywords:
Innovation
Patent propensity
R&D awards
a  b  s  t  r  a  c  t
It  is  well  known  that  not  all innovations  are  patented,  but  the exact  volume  of  innovative  activities
undertaken  outside  the  coverage  of  patent  protection  and,  relatedly,  the  actual  propensity  to  patent  an
innovation  in  different  contexts  remain,  to  a  major  degree,  a matter  of speculation.  This  paper  presents
an  exploratory  study  comparing  systematically  patented  and  unpatented  innovations  over  the period
1977–2004  across  industrial  sectors.  The  main  data  source  is  the  ‘R&D  100  Awards’  competition  organized
by the  journal  Research  and  Development.  Since  1963,  the  magazine  has  been  awarding  this  prize to  the
100 most  technologically  significant  new  products  available  for  sale  or  licensing  in the  year  preceding  the
judgments.  We  match  the  products  winners  of  the  R&D  100  awards  competition  with  USPTO  patents  and
we  examine  the  variation  of  patent  propensity  across  different  contexts  (industries,  geographical  areas
and organizations).  Finally  we  compare  our  findings  with  previous  assessments  of patent  propensity
based  on  several  sources  of data.
© 2013 Elsevier B.V. All rights reserved.
1. Introduction
Within the Economics of Technical change and Innovation Stud-
ies (ETIS) literature it is today widely acknowledged that many
innovations are not patented. In principle, there may  be three types
of explanation accounting for the inventor’s decision of not taking
a patent (Basberg, 1987). The first explanation is that the innova-
tion is simply not patentable. In this case, the inventor believes
that the innovation in question does not represent suitable patent
matter (e.g. the patentability of ‘pure’ software program was  still a
matter of contention in many jurisdictions not so long ago). Alter-
natively, the innovation is in principle patentable but the inventor
may  anticipate that the inventive step embodied in her innovation
is not ‘high’ enough to be deemed worthy of patent protection by
patent examiners. In both these two examples the decision of not
patenting is determined by the fact that this is not actually possible
(or believed possible). The third possibility is that the inventor, even
when conceiving taking a patent as a fully feasible course of action,
∗ Corresponding author. Tel.: +39 0258363037; fax: +39 0258363399.
E-mail addresses: roberto.fontana@unibocconi.it (R. Fontana),
alessandro.nuvolari@sssup.it (A. Nuvolari), shimizu@iir.hit-u.ac.jp (H. Shimizu),
andreav@iseg.utl.pt (A. Vezzulli).
decides not to patent the innovation because she actually prefers
to do so. In this case, even though the innovation is patentable and
worth patenting, the inventor prefers industrial secrecy or other
alternative strategies to extract some economic returns from her
innovation. This third case is the most interesting one from the
viewpoint of innovation scholars.
The existence of ‘appropriability strategies’ that are alternative
to patenting was  initially documented by early economists of inno-
vation (Kuznets, 1962; Schmookler, 1966; Taylor and Silberston,
1973). Later on, the survey studies by Mansfield (1986) and Levin
et al. (1987) during the 1980s highlighted that, in most industries,
patent protection was  not the typical tool adopted by firms for the
extraction of economic returns from innovations, a finding further
corroborated by subsequent research both in US (Cohen et al., 2000)
and Europe (Arundel and Kabla, 1998). All these research results are
frequently cited and surely represent important pieces of evidence
discussed in the innovation literature. However, as aptly noted by
De Rasenfosse (2010), on closer inspection, it is difficult to avoid
the impression that the major implication of these findings (i.e.
that a sizeable share of innovations is never patented) has gone
completely neglected. To be sure, many empirical investigations
acknowledge the limitations of patents as innovation indicators.
However, once these limitations are gauged against their advan-
tages (i.e. availability and richness of information they provide)
0048-7333/$ – see front matter ©  2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.respol.2012.05.014
Author's personal copy
R. Fontana et al. / Research Policy 42 (2013) 1780– 1792 1781
the final choice is to rely on patents if anything because of the
sheer difficulty of constructing suitable indicators using alternative
sources.1
This state of affairs is deeply unsatisfactory as we incur the risk
that even the most carefully designed empirical studies will provide
us with a partial, and sometimes even distorted, representation
of innovative activities. Furthermore, since our understanding of
patent propensity in different contexts is still rudimentary, in many
cases, it is also difficult to formulate a sound assessment of the
margin of error and of the biases involved in the adoption of
patents as innovation indicators. Consider, for instance, the con-
cept of ‘propensity to patent’ usually defined in the literature as
the ratio between patents and R&D expenditures (Scherer, 1983;
Hall and Ziedonis, 2001). Though surely legitimate, we  should note
that this definition of patent propensity is simply describing the
overall relationship between patents and innovative efforts and it
is only indirectly linked to the actual decision to patent or not a
specific innovation (see De Rasenfosse, 2010 for a more extensive
discussion).
Interestingly enough, economic historians and historians of
technology have instead adopted a more ‘straightforward’ defini-
tion of patent propensity, namely the share of patented innovations
in the total number of innovations occurring in a given time
period (Sullivan, 1989; Moser, 2005, 2012).2 This conceptualiza-
tion of patent propensity, although intuitively appealing, is not of
immediate empirical operationalization because it requires some
form of direct assessment of the total amount of innovations
occurring in a given time period. Still, historians have displayed
considerable ingenuity both in the identification of sources (alter-
native to patents) that could be used for formulating quantitative
assessments of overall innovative output in different contexts and
periods, and in connecting these sources with the patent evidence
for constructing estimates of patent propensity. In this respect,
the recent contributions of Moser (2005, 2012) can be regarded
as among one of the most successful examples of this approach.
This paper argues that these historical investigations suggest a
framework of inquiry that can, and should, be fruitfully extended
within the field of ETIS. In the paper we present an application of
this method using a database of ‘important’ industrial innovations
occurred between 1977 and 2004. Our source of data is the ‘R&D
100 Awards’ competition organized by the journal Research and
Development. Since 1963, this journal (which at that time was called
Industrial Research) has been awarding a prize to 100 most techno-
logically significant new products available for sale or licensing in
the year preceding the judgement. The potential of this source was
already reckoned by Carpenter et al. (1981) and Scherer (1989).
This source has also been more recently used by Block and Keller
(2009) to document the increasing role of public institutions and
public funding in the generation of innovations in the US economy
in the period 1971–2006.
Though recent and therefore not ‘historical’ in a strict sense,
the database covers 30 years of innovations, several manufac-
turing industries, and different types of economic actors, both
corporations and Universities and Public Research Organizations
(PROs). These data seem particularly appropriate for studying the
propensity to patent for the following reasons: (i) the data con-
sider innovations that have been recognized by a jury of experts
1 In this respect, we  would argue that a large bulk of the most recent research
on  innovation seems implicitly to apply to patents Winston Churchill’s famous quip
on  democracy: “Democracy is the worst form of government, except all the others”.
In  the case of patents, the jibe would probably sound like: “patents are the worst
innovation indicator, except all the others”.
2 Moser (2005, 2012) uses the term ‘patenting rate’ to define the share of patented
inventions in the total number of inventions.
as significant and they should be commercially feasible at the time
of the awards; (ii) most of the awards have been granted to large
corporations accounting for a sizeable amount of total R&D invest-
ments; (iii) the data cover a relatively long time period allowing
us to take into account changes in the determinants of the propen-
sity to patent over time. Using these data we are able to assess
systematically the relative influence of sector, organization, and
inventor specific characteristics on the actual decision of taking or
not a patent.
Our study is based on a sample of about 3000 innovations that
have received an award. For each innovation in our dataset we
have retrieved information concerning: years of the award, descrip-
tion of the innovation, type and name of applicant organization(s),
application domain of the innovation, country, and name of inven-
tor(s). The first step of our analysis is to match awarded innovations
with patents using the search engine of the USPTO website. Then,
on the basis of the invention description contained in the journal,
we  classify all the awarded innovations in thirty different sectors of
activity. In this way, the data allow a thorough comparison between
patented and not patented innovations across different industrial
sectors, countries, types of organization and types of innovation.
Our results highlight the following patterns. First, a large share of
innovations is not protected by means of patents. Second, we  are
able to point out the existence of systematic significant differences
in patenting propensity across sectors, geographical areas, types of
organization and types of innovation.
The structure of the paper is as follows. Section 2 reviews the
empirical literature on the effectiveness of patents as appropriabil-
ity tools. Section 3 describes in detail our data source, our matching
procedure and the limitations of the dataset. Section 4 presents
our analysis of patent propensity across different dimensions. Sec-
tion 5 compares our findings with those of previous assessments
of patent propensity carried out using different types of data. Sec-
tion 6 concludes and draws some methodological implications with
particular reference to the possibility of extending the framework
of inquiry adopted by economic historians and historians of tech-
nology to contemporary studies of patent propensity.
2. Patents as indicators of innovation and their limitations
Scholars within the ETIS tradition have relied intensively on
patents to investigate the sources, nature, and the effects of
innovative activities. Innovative activities are inherently elusive
phenomena which almost by definition are bound to defy system-
atic attempts of (quantitative) measurement. It is not surprising
then that the existence of patent records has been regarded for
a long time, mostly by economists, but also by other scholars of
innovation with different disciplinary backgrounds, as an almost
unique source of insights into the nature of inventive activities.
The main merits of patent records as a source for measuring inno-
vation are easy to summarize: (i) they are by definition related to
innovative activities;3 (ii) they are readily available (allowing to
economize efforts of data collection);4 (iii) they are available for
relatively long periods of time; (iv) they contain a significant depth
of information (inventors’ names and addresses, ownership of the
3 In the words of Griliches (1990, p. 1169): “[A] patent represents a minimal quan-
tum of invention that has passed both the scrutiny of the patent office as to its novelty
and the test of the investment of effort and resources by the inventor and his organi-
zation into the development of this product or idea, indicating thereby the presence
of  a non-negligible expectation as to its ultimate utility and marketability”.
4 The ‘accessibility’ of patent as a source has greatly increased over the last 20
years or so thanks to the creation of on line search engine such as ESPACENET
and the efforts of construction of data-bases containing information gathered by
patent records such as the NBER-US patent data-set (Hall et al., 2001) and the
OECD-PATSTAT dataset.
Author's personal copy
1782 R. Fontana et al. / Research Policy 42 (2013) 1780– 1792
innovation, description of the innovation and its relation with pre-
vious ones, as represented by patent citations). These factors have
made patents the most adopted indicator for scholars interested
especially in measuring the output of innovation activities.
Although much progress has been achieved in this way, it is
well known that indicators of innovation based on patents suffer
from several limitations. These limitations can be summarized cit-
ing again from Griliches (1990, p. 1169): “Not all inventions are
patentable. Not all inventions are patented and the inventions that
are patented differ greatly in their ‘quality’, in the magnitude of
inventive output associated with them”.
The first limitation is clearly the most obvious one and prob-
ably the easiest to tackle. Some domains of inventive activity do
not constitute patentable subject matter. The solution is to resort
to alternative indicators for assessing inventive output in these
areas. The second limitation is that not all patentable innovations
are actually patented. This means that in contexts characterized
by low patent propensity, i.e. in environments in which firms
prefer to adopt alternative appropriability strategies, the use of
patents as innovation indicator may  result in a biased assessment
of the volume of innovative activities.5 The existence of a different
propensity to patent across industries has indeed been the most
important finding of studies based on surveys of the attitudes of
R&D personnel towards the use of patents. Mansfield (1986) exam-
ined how many patentable innovations were actually patented
in a random sample of large US firms in different industries. His
results highlighted that in sectors where patents are not regarded
as particularly effective ‘appropriability’ mechanisms (i.e. primary
metals, electrical equipment, instruments, office equipment, motor
vehicles, rubber and textiles) around 34% of potentially patentable
inventions were not patented. This percentage is around 16% in
those sectors where patents are considered to be more impor-
tant (i.e. pharmaceuticals, chemicals, petroleum, machinery, and
fabricated metal products). Results from subsequent surveys cor-
roborate these early findings. In some industries, secrecy and lead
times seem to be more important than patenting for appropriating
the returns from innovation (Levin et al., 1987). Moreover, patent
propensity varies depending on the type of innovation, with firms
more likely to apply for a patent for product innovation than for pro-
cess innovation (Cohen et al., 2000).6 Arundel and Kabla (1998) also
find that the (sales weighted) propensity to patent differs across
innovation type with relatively lower rates for process innovation
(24.8%) than for product innovations (35.9%).
Among the available alternatives, secrecy seems to play an
important role in protecting innovation. Looking at a sample of
European innovative firms, Arundel (2001) finds that secrecy is
generally rated as more valuable than patenting. This is particularly
true in the case of product innovation, though the probability of
being considered more valuable declines with firm size. Hussinger
(2007) carries out a similar analysis using sales figures to assess
the importance of alternative means of appropriation rather than
5 The fact that firms in some contexts prefer alternative appropriability strategies
to  patenting does not imply that patents are completely irrelevant as a measure
of  innovation. For example, in semiconductors, appropriability strategies are based
more on secrecy and lead times than on patents, but firms are increasingly resorting
to  patent protection in order to use patents as “bargaining chips” in negotiation
with  other firms (Hall and Ziedonis, 2001). As a result, even though in this industry
patents are taken for strategic motives rather than for reaping economic returns
from a specific innovation, they still provide a useful measure of innovative activities.
6 The survey questionnaire of Levin et al. (1987) did not contain a specific ques-
tion asking what percentage of innovation a firm typically patented, but simply
contained a question asking to assess the relative effectiveness of different appropri-
ability strategies. The survey questionnaire of Cohen et al. (2000) instead contained
a  specific question asking respondents what percentage of their innovations was
patented.
individual evaluations. Her findings suggest that secrecy is rela-
tively more important for innovations that are not commercialized.
The third limitation is that patents and innovations differ greatly
in their technological and economic significance.7 In particular,
several studies have shown that the ‘size’ distribution of innova-
tions is sharply skewed with the majority of innovations of little or
limited technical and economic significance and a restricted num-
ber of highly significant innovations (Silverberg and Verspagen,
2007). Innovation scholars have attempted to deal with this prob-
lem by weighting patents using citations or other information such
as claims, and family size (Trajtenberg, 1990). Still, it is acknowl-
edged that these methods represent only imperfect proxies of the
quality of the innovation underlying the patent in question. In fact,
the most sensible use of these proxies of patent value is to use them
as ‘probabilistic markers’ of the underlying economic value of the
patents and employ them for the identification of groups of poten-
tially valuable patents (van Zeebroeck, 2011). Clearly not taking
properly into account these variations in the underlying value of
patents may  again lead to biased assessments of inventive output.
Economic historians seem to have been more sensitive to the
limitations of patents as innovation indicators and have explored
the potentialities of alternative sources. Moser (2005, 2012) has
constructed a dataset of innovations on the basis of the catalogues
of nineteenth century industrial exhibitions (in her case she has
used the Crystal Palace exhibition of 1851 and the Centennial exhi-
bition in Philadelphia in 1876). According to Moser (2002, pp. 1–2),
this type of data covers “economically useful innovations” (i.e. the
commercial introduction of new products or processes), rather
than “inventions” (i.e. the additions to the stock of technologi-
cal knowledge). Furthermore, exhibition data measure innovations
regardless of whether they are patented or not. Moser’s findings
have produced novel insights on the sources of innovation across
countries and sectors during the second half of the nineteenth
century. Her findings show that in 1851, 89% of British innova-
tions on display at the Crystal palace exhibition were not patented.
Even among prize winning innovations, 84% were not patented.
Moreover, she finds that patent propensity is affected by the char-
acteristics of the sectors where the innovation occurs, the location
of the invention (urban vs. rural), and the quality of the invention
(incremental vs. radical).
Brunt et al. (2012) have instead relied upon information on
prizes and awards. Using a dataset of awards for inventions pro-
moted by the Royal Agricultural Society of England from 1839 to
1939, they have studied, among other aspects, whether and how
prizes affect innovation and patenting. Their findings point to the
presence of a positive relationship between prizes and patenting. In
particular, the propensity to patent in the technology category tar-
geted by the award increased by 42% for those inventions awarded
a gold medal. Moreover, patents are more likely to be renewed
when they were taken out of awarded inventions thus suggesting a
positive relationship between awards and the quality of the patent.
These recent contributions by historians clearly illustrate the
potentialities of assessments of innovative output using data
sources that are alternative to patents.8 In particular, we would
like to suggest that innovation scholars should consider with par-
ticular interest prize and exhibition data, because some of their
intrinsic characteristics are likely to render them less prone to the
7 In the words of Kuznets (1962, p. 37): “[T]he main difficulty with patent statistics
is,  of course, the enormous range in the magnitude of the inventions covered [. . .]
patented inventions do differ widely in their potential economic magnitude”.
8 Interestingly, Schmookler, one of the pioneer of the use of patent statistic in the
field  of ETIS, was  also one of the early scholars to argue in favour of cross-checking
the assessments of inventive output based on patent with data on ‘important inno-
vations’ (Schmookler, 1966).
Author's personal copy
R. Fontana et al. / Research Policy 42 (2013) 1780– 1792 1783
pitfalls that typically affect patent data. First, both prize and exhi-
bition data refer to valuable, or to use Moser’s words “economically
useful”, innovations. In the case of prizes this is almost a tautology
given that they have been recognized by experts in the field as supe-
rior to alternative available solutions and probably also to existing
practices. For the case of innovations displayed at industrial exhi-
bitions, their economic and technological significance will depend
on the exact criteria that an artefact must satisfy for being included
in the exhibition. Second, and most importantly, both types of data
typically comprise innovations with and without patents (Moser,
2012). In other words, using this type of data allows an assessment
of the share of innovations occurring outside of the patent system
as well as the construction of a more ‘direct’ indicator of propensity
to patent. This is a crucial advantage with respect to other sources
of data.
3. The ‘R&D 100 Award’ database
This paper presents an extension of the method recently
employed by historians to the field of ETIS. Using a source of data
that so far has received little attention we provide new estimates
of patent propensity across industries and over time. Our source of
data is the ‘R&D 100 Award’ competition organized by the magazine
Research and Development (previously called Industrial Research).
The magazine was founded in 1959 and it represents probably one
of the most authoritative regular publications for R&D practitioners.
Currently it has an estimated monthly readership of over 80,000.
It is estimated that about 75% of the readers works in high-tech
industries, whereas the remaining 25% works for government lab-
oratories, universities, and similar organizations. Over 60% of the
readers have managerial or executive type of jobs. The ‘R&D 100
Award’ competition has been running since 1963. Each year the
magazine awards with a prize the 100 most technologically signifi-
cant products available for sale or licensing in the year preceding the
judgement. Throughout the years, key breakthroughs inventions
such as Polacolor film (1963), the flashcube (1965), the automated
teller machine (1973), the halogen lamp (1974), the fax machine
(1975), the liquid crystal display (1980), the printer (1986), the
Kodak Photo CD (1991), the Nicoderm antismoking patch (1992),
Taxol anticancer drug (1993), lab on a chip (1996), and HDTV (1998)
have received the prize. In order to apply for the prize inventors, or
their employees, must fill an application form providing a detailed
description of the product in question. The prize consists of a plaque
which is presented in a special ceremony. There is no sum of money
involved. The prize is awarded by a jury composed of university
professors, industrial researchers and consultants with a certified
level of competence in the specific areas they are called to assess.
The members of the jury are selected by the editor of the mag-
azine. The main criteria for assessment are two: (i) technological
significance (i.e., whether the product can be considered a major
breakthrough from a technical point of view); (ii) competitive sig-
nificance (i.e., how the performance of the product compares to
rival solutions available on the market). R&D 100 awards are acco-
lades comparable to the Oscars for the motion picture industry as
“they carry considerable prestige within the community of R&D
professionals” (Block and Keller, 2009, p. 464).
The technological significance requirement is to be understood
in fairly broad terms:9
“[. . .]  products and processes that can change people’s lives for
the better, improve the standard of living for large numbers of
9 All the following quotes concerning the rules and organization of the R&D 100
competition have been retrieved from the magazine website, www.rdmag.com,
accessed 23 July, 2010 and on 7 April, 2012.
people, save lives, promote good health, clean up the environ-
ment, etc. [. . .]  A cure for cancer or AIDS. An engine that runs
on water. A safe, cheap method for cleaning up toxic waste. A
vehicle that can fly 800 passengers from New York to Tokyo
in two  hours. A device that would cut automotive accidents
or one that would reduce workplace injuries. A pollution-free
herbicide that would increase crop production in Third World
countries”.
Accordingly products with a wide potential of application are
preferred to those catering to very specific sets of user needs:
“Products or processes that solve very specialized or circum-
scribed problems could be judged less significant than those
that meet larger, more broad-based needs. For example, a new
scientific instrument that only benefits a few scientists in a nar-
row field of interest would have difficulty competing against
a device with much broader application. It would depend on
how significant the two fields of interest were and how much
the technical improvements contributed to the success of each
device.”
Furthermore, for attaining the prize there should be a proven
link between the effect of the innovation and an improvement in
technology:
“[. . .]  these improvements must be attributed to significant
breakthroughs in technology. In general, this means your prod-
uct should exhibit multiple levels of improvement - 53 times
faster, 103 greater throughput, 503 times more accurate - or,
preferably, orders of magnitude improvement over existing
technology. Again, we’re looking for ‘leapfrog’ gains in perfor-
mance, not expected, incremental improvements.”
Additionally the product should also represent a major improve-
ment in comparison with alternative solutions already existing
on the market. For this reason, the applicant is requested
to provide a ‘competitive matrix’ illustrating how the prod-
uct compares with rival solutions already available on the
market:
“The competitive matrix should show how your product com-
pares to existing products in terms of the crucial factors involved
in the technology. This is your opportunity to give the judges
a quick overview of how your product beats the competition.
[. . .]  Include only factors crucial to the technology. Don’t waste
space (and the judges’ time) throwing in every conceivable fac-
tor, just to pad your entry. However, you must list all factors
that are indeed crucial to the technology, even if you don’t
‘win’ that particular point. For example, if you fail to include
‘hardness’ in an entry involving a new alloy, your entry may  be
looked upon with suspicion by the judges. Some typical factors
you might want to include: signal-to-noise ratio, weight, speed,
reliability, resolution, cost, accuracy, life expectancy, mean
time between failures, sensitivity, reproducibility, strength,
power consumption, production yield, environmental operat-
ing, intensity, efficiency, size, output rate, bandwidth, number
of materials tested, stability”.
The product must exist in marketable form, i.e. it “must have
been first available for sale or licensing during the calendar year
preceding the judging”. Applicants are not restricted to firms, but
also governmental laboratories, universities, public research cen-
tres are allowed to compete. In case of products resulting from
research collaborations, the application form requires to include
all the organizations that have provided a “significant contribution
to [the] creation of the product” and to provide a description of their
precise role in the project. Hence, the rules of the competition make
sure that all the parties that have participated to the innovation are
Author's personal copy
1784 R. Fontana et al. / Research Policy 42 (2013) 1780– 1792
properly acknowledged.10 Finally an organization may  submit as
many products as it wishes at each yearly competition.
There are a number of characteristics of the R&D 100 awards
competition that, at least prima facie, appear particularly promis-
ing for using this data source to measure innovative output. First,
the R&D 100 awards competition seems to represent a good oppor-
tunity for companies, government laboratories, etc. to showcase
the outcome of their innovative activities. Thus, we can expect that
the awards will provide us with a fairly reliable sample of inno-
vations attained by R&D performers. Second, R&D 100 awards are
granted to innovations that, at least in principle, should embody
a significant improvement over the existing state-of-the-art that
is clearly documented. In other words awarded innovations should
represent, at least in principle, a technological breakthrough. Third,
the selection of the awards is made by what appears a competent,
authoritative jury of experts. Fourth, R&D awards may  be assigned
both to patented and not-patented innovations. Finally, there
seems to be limited space for strategic behaviour and attempts
to conditioning the jury, because the nature of the prize is sim-
ply honorific. Alongside these advantages, some biases exist which
prevent us from considering these awarded innovations as a fully
representative sample of innovative activities. For example, Scherer
was struck by the fact that awards covered a very limited number of
new weapon systems and a relatively few pharmaceutical products,
both sectors notoriously characterized by high R&D investments.
Nevertheless, he still regarded the source as capable of providing
useful insights on the nature of innovative processes (Scherer, 1999,
pp. 67–68).
Retrieving the information from different issues of the maga-
zine, we have constructed a data set of all the R&D 100 awards
granted in the period 1977–2004. Our dataset contains 2802 inven-
tions. The total is not equal to 2800, because the requirement of
awarding 100 inventions was apparently interpreted with some
degree of flexibility. Thus, the amount of awards given in each year
in the period we are considering ranges from 97 to 109. A major lim-
itation of the data set is that we do not have information on applying
innovations that were not awarded the prize. For this reason, we
cannot control whether specific factors, besides the specific tech-
nical and economic merits of the innovation, affected the selection
of the awards.
3.1. Matching awards with patent data
In order to assess the propensity to patent for the awarded
innovations included in our sample, we had to look for a possi-
ble match between each awarded innovations and one or more
USPTO patents.11 We  do not expect to find an exact match between
each awarded innovation and one patent. As noted above, awarded
innovations represents ‘products’ available for commercialization
or license, so it is possible that, in certain areas, individual compo-
nents of a specific product may  be protected by different patents.
We regarded the awarded innovation as ‘patented’ also in cases in
which one or more components of the innovation in question were
actually patented. To carry out the matching exercise we relied
upon the following information contained in the R&D 100 database:
(i) name of award winning organization(s), (ii) the name of the
innovation, (iii) the year of award, (iv) the name of developer(s)
and (v) the description of the innovation. We  have searched USPTO
10 In this respect it is also important to note that the rules of the competition state
explicitly that: “existing technologies purchased by third parties who then conduct
sales, [marketing and other commercialization] efforts” are considered eligible for
the  award only if the original developer is included in the application.
11 Given the nature of our data source, the most obvious choice was  to match R&D
100  innovations with patents taken in US.
patents granted in a time interval ranging from 3 years before to
3 years after the award. The criteria for ascertaining a ‘positive’
match were the name of the inventors, the name of the organi-
zation and the consistency between the description of the ‘R&D
100’ innovation and the title and abstract of the patent (also taking
into account the possibility that one or more components of the
awarded innovation could have been patented as a separate item).
In particular, the patent search procedure entailed the following
steps. First, the name of developer as Inventor and the name of
organization as Assignee were used to search patents in the USPTO
online database. If any patents were found, the patent title and
abstract were checked by looking at the information provided by
the R&D 100 to see if the patent was corresponding to the award
winning invention. Second, if the name of developer was  not avail-
able, abstract key word search with the name of organization was
carried out. The key words were selected from the technological
information of the innovations contained in the R&D 100 list. If a
match was found at this stage, a further check was carried out to
see whether the patent was related with the award winning inno-
vation by cross-checking the information of the patent and the R&D
100 innovation.
We  should note that the matching process may be subjected to
errors. More specifically, there may  be two  limitations in the patent
searching procedures we have adopted. The first is the time span of
the searching. The search considers as relevant to the innovation a
patent obtained in plus/minus 3 years from the year of the award.
It means that this procedure can overlook the relevant patent(s)
that were granted more than 3 years before or after the year of the
award. Second, the product name and description contained in the
R&D 100 database may not always provide enough information for
the identification of one or more possible underlying patents. These
limitations notwithstanding, we  are confident that our matching
procedure provided reliable results in most of the cases. Never-
theless, in uncertain cases, the ‘benefit of the doubt’ was given
to a positive match in the sense that we considered the awarded
invention as covered by a patent.12 For this reason, if anything,
the adopted matching procedure does not contain any in-built bias
leading to a systematic underestimation of patent propensity.
3.2. The ‘quality’ of R&D 100 awards
As highlighted by Moser (2012), one of the chief advantages
of employing data on awards and prizes as indicators of inno-
vative output is that, with respect to patents, this type of data
should in principle contain only relatively ‘important’ innovations,
namely those inventions deemed worthy of receiving the prize or
of being put on display. Accordingly, the first exercise we carried
out was  an attempt of checking whether our R&D 100 dataset con-
tains inventions that are above a certain quality threshold. This is
done by replicating an exercise originally performed by Carpenter
et al. (1981). Carpenter et al. (1981) used the 1969 and 1970 R&D
awards list and matched these inventions with the corresponding
US patents. In this way, they obtained a set of 100 patents whose
technological significance had been ‘certified’ by the granting of the
award. They then compared the citations received by this group of
patents with the citations received by a random sample of patents
distributed within the same time cohort. Their results showed that
the patents covering the R&D 100 awards received a significantly
higher number of citations than the control group. This obviously
suggests that R&D 100 innovations are on average of better quality
than the ‘average’ patent.
12 In this respect our matching procedure was  also robust to changes in the ‘time
range’ before and after the award was  received.
Author's personal copy
R. Fontana et al. / Research Policy 42 (2013) 1780– 1792 1785
Table 1
Patent citations received by awarded innovations and by a random sample of patents (matched by granted year and technology class).
Number Mean Median Standard deviation Min  Max
R&D 100 patents 535 12.88037 7 16.17822 0 137
Random sample 5331 8.483024 4 14.11133 0 329
Note: Mann–Whitney test rejects the Null Hypothesis of equal populations.
Table 2
Total awarded innovations and patents.
Awarded innovations Patented innovations Share not patented (%)
All the sample (1977–2004) 2802 255 90.9
Non-corporate 886 25 97.16
Corporate only 1751 220 87.44
Our results for the period 1977–2004 confirm the early findings
of Carpenter et al. (1981). For each R&D 100 innovation with one or
more USPTO patents we constructed a ‘matched random’ sample of
ten granted patents of the same granted year and of the same Inter-
national Patent Classification (IPC) class and then compared the
number of citations received by patents in this random sample with
the citations received by the patents covering an invention within
the R&D100 award.13 The results of this exercise are reported in
Table 1.14
The non-parametric Mann–Whitney test confirms that the
median number of citations of patents associated with a R&D 100
invention is significantly not lower than the median of the random
matched sample. Overall, this exercise confirms that the innova-
tions which received a R&D 100 award are more significant from a
technological or economic viewpoint than the ‘average’ patent in
their technological class.
4. R&D 100 awards and patent propensity
This section presents our estimates of patent propensity defined
as the share of patented innovations in the total number of
inventions that have received an R&D 100 award. We  compute
our estimates across different dimensions of innovative activities
(industry, geographical area, organization and type of invention).
4.1. Most awarded innovations are not patented
Table 2 reports the number of awarded innovations and the
percentage of not patented ones.
As highlighted above, the awarded innovations contained in
the R&D 100 list refer to products that are available on the market
or for licence when the application is submitted. Hence, it is
possible that the data will contain a bias against organizations
such as universities and PROs that lack ‘downstream’ assets for
the commercialization of a product. However, it is interesting to
note that results in Table 2 are consistent with the results of Block
and Keller (2009) showing that a significant share of awarded
innovations (more than 30%) are generated by non-corporate
type of organizations. So it would seem that, the presence of bias
notwithstanding, our data cover also a significant segment of
the population of non-corporate organization involved in R&D
13 Our control group is made of granted patents and not of patent applications. At
the USPTO a sizeable percentage of patent applications are not granted. In addition
to  this, many patent applications are discontinued. For these reasons it would have
been more rigorous to compare the sample of patented awarded innovations with
patent applications. However, these data are not fully available for US, thus the need
to  rely upon granted patents.
14 The random matched sample includes 5331 patents and not 5350 because for
some specific years in some technology classes it was  not possible to collect enough
patents to create the match.
activities.15 Overall, we found that 269 awarded innovations
(slightly less than 10%) were patented according to our matching
criteria suggesting that the great majority of innovations were
not patented. This percentage is slightly higher (12.56%) when we
consider only innovations that have been made by firms.
This estimated patent propensity is in line with the findings of
Moser who reports total patenting rates between 11% and 14% for
the inventions displayed at the Crystal Palace exhibition of 1851
(Moser, 2005, p. 1221). Of course, this finding should be interpreted
keeping in mind the inherent limitations the data set discussed in
Section 3. However, even if we  consider possible errors that may
have led us to underestimate patenting rates, the result that such a
sizeable share of major innovations is not patented is remarkable. In
particular, if we consider that ‘The R&D 100 Award’ is a competition
aimed at acknowledging the output of formalized R&D efforts, which
is notoriously one of the contexts with the highest propensity to
patent and that, we are in principle dealing with breakthrough
innovations, our findings reveal that patent protection, even in this
context, is actually a much less used appropriability strategy than
it is generally believed. In this respect, our findings are actually
a powerful corroboration of the findings of Moser (2005). Addi-
tionally, they are not inconsistent with the results of both the Yale
(Levin et al., 1987) and the Carnegie Mellon survey (Cohen et al.,
2000) indicating that only in a very restricted number of contexts
patents are considered as effective tools for protecting innovation.
The obvious policy implication is that the recent developments
towards the strengthening of IPR regimes may actually represent
a step going in the wrong direction, as it would appear consider-
ing the predominant share of innovative activities which is actually
carried out without resorting to patent protection.16
Fig. 1 displays the evolution over time of the propensity to patent
for all our sample of innovations.
Our estimated propensity to patent is never higher than 20%.
Moreover, contrary to what has been suggested by other studies
(Kortum and Lerner, 1999), our evidence does not seem to indi-
cate the existence of significant structural breaks in the time period
15 The requirement of availability for sale or licensing implies that the awarded
product must be available either for purchase or licensing during the year preced-
ing  the award and not necessarily already launched in the market on an extensive
scale. Clearly, this allows also organizations with limited capabilities in the commer-
cialization and in the “downstream” development of new products to compete, as
also prototypes ready to be licensed are eligible for the award. Some notable exam-
ples of awarded products that were developed only by academic institutions are:
the  “Cost Thin Film Solar Cell” (University of Delaware, 1979); the “Low-Palmitic
Soybean Oil” (Iowa State University, 1991); the “SC-54 Oral Vaccine” (Iowa State
University, 1996); the “Nanoruler” (M.I.T., 2004); the “Netsolve 1.2” (University of
Tennesse & U.C.S.D, 1999) and the “Chromium” (University of Virginia & Stanford
University, 2004) software.
16 For a more elaborate discussion of this point see Boldrin and Levine (2008).
Author's personal copy
1786 R. Fontana et al. / Research Policy 42 (2013) 1780– 1792
All industries and applicants
0%
2%
4%
6%
8%
10%
12%
14%
16%
18%
20%
2004200119981995199219891986198319801977
% Inventions Patented
Fig. 1. Pattern of change in the propensity to patent. All the samples.
0%
5%
10%
15%
20%
25%
30%
2004200119981995199219891986198319801977
All industries by applicant type
% Only corporate At le ast 1 academic At least  1 governamental
Fig. 2. Pattern of change in the propensity to patent by type of inventor.
considered.17 In our case, it is suggested that the propensity to
patent has been remarkably stable and possibly characterized by
a fluctuating behaviour around what seems to be a constant level
of 10%.
Fig. 2 shows the evolution of the propensity to patent broken
down by type of inventors, distinguishing between corporate (i.e.
firms) and non-corporate (i.e. PRO and universities) organizations.
As expected Fig. 2 suggests that the propensity to patent is
higher for firms than for PROs and universities. In several years
the propensity to patent of PROs and universities is equal to zero
indicating that no innovation has been patented. From early 1990s
onwards the propensity to patent of universities seems to increase.
It would be tempting to interpret this evidence as a consequence
of the strengthening of IPRs following the introduction of the
Bayh–Dole Act in US (Mowery and Ziedonis, 2002; Mowery et al.,
2001) though an analysis of the mechanisms underlying this trend
falls beyond the scope of the present paper.18
17 Hall (2005) finds several structural breaks in patent application series at USPTO
during the 1967–1997 time period. In our case, both the Phillips–Perron (PP) and the
Augmented Dickey–Fuller (ADF) tests for unit root reject the null hypothesis that
our patent propensity series contains a unit root at the 1% significance level, both
after  including a time trend and lagged terms (up to the 4th order) in the associated
regression. The Andrews (1983) test for structural breaks at unknown points (sup.
F  = 8.6538, T = 28) does not reject the null hypothesis of ‘no structural change’ at the
5%  significance level.
18 A preliminary analysis along these lines actually suggests that both the number
of  patents and the number of awarded inventions attributable to University and
PROs in our sample have increased.
4.2. Patenting rates vary across sectors
Awarded innovations are classified by the magazine in several
categories on the basis of their technological content. The classifica-
tion is not consistent over time and in some cases the innovations
were not even assigned to a specific category. Thus, in order to
examine the distribution of awarded innovations across different
technological fields, we  have proceeded as follows. First we reclas-
sified each awarded innovation according to a technology-oriented
classification of 30 different sectors based on the co-occurrence of
the IPC codes proposed by the Observatoire des Sciences et des Tech-
niques (OST).19 In a few doubtful cases, we have relied both upon
the classification in product categories of the R&D100 awards and
on the innovation description. It is important to note that we have
assigned each awarded innovation to only one of the 30 OST  sectors.
These sectors have been further aggregated into 5 ‘macro’ techno-
logical classes (called ‘OST5’ henceforth) defined according to the
ISI-INIPI-OST patent classification based on the EPO IPC technolog-
ical classes.
Table 3 displays the shares of awarded innovations that have
been patented (patenting rates) classified by both 5 and 30 OST
sectors.
The share of patented innovation varies considerably across
sectors. In terms of macro-sectors, the sector with the highest
propensity to patent is chemical/pharmaceuticals, a result which is
also in line with the results of the Yale and Carnegie Mellon surveys
on the effectiveness of patents for protecting innovations in these
fields. Finally, the macro-sector with the lowest patenting rate is
instruments.20 In this case we  should remember that many organi-
zations active in this sector are non-corporate institutions such as
universities and public research centres which traditionally display
a very low patenting attitude.
4.3. Patenting rates vary across geographical areas
Table 4 compares patent propensities across different world
regions.
The awarded innovations with at least one applicant from US
show a lower patent propensity with respect to the average level
of the whole sample (−1.38%, statistically significant at 10% signifi-
cance level), whereas applicants from Asia tend to patent more their
innovations than the average level (+15.53%, statistically significant
at 1% significance level). We  should note that the large bulk of Asian
innovations are awarded to Japanese companies. Hence our results
pointing to a significantly higher propensity to patent of Asian (and
especially Japanese) firms are consistent with previous research
on the aggressive patterns of foreign patenting of Japanese firms
in a comparative international perspective (Granstrand, 1999, pp.
134–175).
It may  be argued that this finding is the consequence of dif-
ferences in the institutional mix  of US and Asian prize winners. In
order to understand whether this is the case with our data we have
compared the propensities across subsets of similar organizations.
Table 5 confirms that these overall differences in patent propensity
rates across countries are not driven by different institutional mix
of US prize-winners with respect to other geographical areas.
In fact the higher propensity to patent of Asian prize-winners is
19 See Hinze et al. (1997).
20 A statistical test for binary variables was  carried out to check whether the dif-
ference in proportion between each industry and the total sample is statistically
significant. Results rejected the null hypothesis of equality in two cases: Chemical
and Pharmaceuticals (+4.6% with respect to the average patent propensity of the
whole sample, statistically significant at 5% significance level); Instruments (−2.8%
with respect to the average patent propensity of the whole sample, statistically
significant at 1% significance level).
Author's personal copy
R. Fontana et al. / Research Policy 42 (2013) 1780– 1792 1787
Table 3
Patenting rates of ‘R&D 100’ innovations.
OST5 OST30 All applicants Only corporate
No. of innovations Share patented No. of innovations Share patented
Electrical
engineering
1 Electrical engineering & devices 274 0.1350 177 0.1751
2  Audiovisual technology 19 0.1053 12 0.1667
3  Telecommunications 32 0.1563 20 0.25
4  Information Technologies 255 0.0824 157 0.1210
5  Semiconductors 148 0.1149 90 0.1333
728 0.1126 728 0.1513
Instruments 6  Optics 198 0.1111 123 0.1545
7  Control technology 629 0.0493 384 0.0729
8  Medical technology 125 0.1120 91 0.1209
27  Nuclear engineering 75 0.0400 41 0.0732
1027 0.0682 639 0.0954
Chemistry,
Pharma
9  Organic chemistry 0 – 0 –
10  Polymers 47 0.1489 41 0.1463
11  Pharmaceutics 0 – 0 –
12  Biotechnology 87 0.0690 51 0.0588
14  Food chemistry 0 – 0 –
15  Basic materials chemistry 42 0.2857 31 0.3870
176 0.1420 123 0.1703
Process
engineering
13  Materials metallurgy 240 0.1458 167 0.  2036
16  Chemical engineering 220 0.1091 118 0.  1525
17  Surface technology 0 – 0 –
18  Materials processing 8 0 0 –
20  Environmental technology 154 0.0714 81 0.  1235
24  Handling & printing 0 – 0 –
25  Food processing 0 – 0 –
622 0.1125 366 0.1694
Mechanical
engineering
19  Thermal processes 34 0.0882 23 0.1304
21  Machine tools 77 0.1169 47 0.1702
22  Engines 0 – 0 –
23  Mechanical elements 43 0.0465 30 0.0333
26  Transport 27 0.0370 0 –
28  Space technology 9 0.1111 0 –
29  Consumer goods 59 0.1017 44 0.1364
30  Civil engineering 0 – 0 –
249 0.0884 144 0.125
Total  2802 0.0960 1728 0.1336
confirmed even when considering awarded innovations with at
least one corporate applicant (columns 3–4) or only corporate
applicants (columns 5–6).
4.4. Multivariate regression analysis
Though interesting, the previous results only account for the
effect of a single characteristic at a time (i.e. technological sector,
country of origin, applicant type) on the propensity to patent. In
order to study the joint effect of all these variables in an unified
framework we  perform a probit multivariate regression analysis
using the probability to patent a given innovation Pr(PAT=1) as
dependent variable and a set of applicant and innovation specific
characteristics as independent and control variables respectively.
The list of additional variables includes: dummies for the tech-
nological sector (both at OST 5 and OST 30 aggregation level),
Table 4
Patenting rates by industry across countries.
Sector (OST 5) Full Sample USA Europe Asia Other
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14)
Total  %Pat Total %Pat Diff. (4)–(2) Total %Pat Diff. (7)–(2) Total %Pat Diff. (10)–(2) Total %Pat Diff. (13)–(2)
Elec. Eng. 728 11.2 597 8.21 −2.99%* 16 0 −11.2% 86 31.40 +20.2%*** 29 3.45 −7.75%
Instruments 1027 6.8 876 5.58 −1.22% 56 8.92 +2.12% 50 22 +15.2%*** 45 4.44 −2.36%
Chemistry 176 14.2 158 13.92 −0.28% 8 0 −14.2% 5 20 +5.8% 5 0 −14.2%
Proc.  Eng. 622 11.2 559 11.09 −0.11% 17 11.76 +0.56% 25 12 +0.8% 21 0 −11.2%
Mech. Eng. 249 8.8 207 7.73 −1.07%* 9 11.11 +2.31% 21 23.81 +15.01%** 12 0 −8.8%
All  2802 9.6 2397 8.22 −1.38%* 106 7.55 −2.05% 187 25.13 +15.5%*** 112 2.68 −6.92%**
Innovations with multiple applicants from different industries are double counted in the table.
* Difference is statistically significant at 10% significance level.
** Difference is statistically significant at 5% significance level.
*** Difference is statistically significant at 1% significance level.
Author's personal copy
1788 R. Fontana et al. / Research Policy 42 (2013) 1780– 1792
Table 5
Patenting rates by geographical area and organization type.
Type of applicant Full sample USA Europe Asia Other
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14)
Total  %Pat Total %Pat Diff. (4)–(2) Total %Pat Diff. (7)–(2) Total %Pat Diff. (10)–(2) Total %Pat Diff. (13)–(2)
At least 1 corporate 1916 12 1557 11.11 −0.89% 102 7.84 −4.16% 183 25.68 +13.68%*** 74 2.70 −9.3%**
Only corporate 1728 13.36 1389 11.74 −1.62% 94 7.45 −5.91%* 182 25.82 +12.46%*** 63 3.28 −10.08%**
ALL 2802 9.6 2397 8.22 −1.38%* 106 7.55 −2.05% 187 25.13 +15.5%*** 112 2.68 −6.92%**
* Difference is statistically significant at 10% significance level.
** Difference is statistically significant at 5% significance level.
*** Difference is statistically significant at 1% significance level.
Table 6
Variable description.
Description Type
Dependent variable
PAT = 1 if the innovation has been patented (see Section 3.1) Dummy
Independent variables
MAPPL = 1 for multiple applicant organizations, = 0 otherwise Dummy
NINV  = number of inventors Count
USA  = 1 if at least one applicant is a U.S. organization, = 0 otherwise Dummy
EUROPE = 1 if all the applicants are from continental Europe, = 0 otherwise Dummy
ASIA  = 1 if all the applicants are from Asia, = 0 otherwise Dummy
PRO = 1 if at least one applicant is a public research organization, = 0 otherwise Dummy
Controls
dum1986 1995 = 1 the innovation has been awarded in the 1986–1995 decade, = 0 otherwise Dummy
dum1996 2005 = 1 the innovation has been awarded in the 1996–2005 decade, = 0 otherwise Dummy
Electrical Eng = 1 if the innovation belongs to the Electrical Engineering OST5 macro sector, = 0 otherwise Dummy
Instruments = 1 if the innovation belongs to the Instruments OST5 macro sector, = 0 otherwise Dummy
Chemistry Pharma = 1 if the innovation belongs to the Chemistry & Pharma OST5 macro sector, = 0 otherwise Dummy
Process Eng = 1 if the innovation belongs to the Process Engineering OST5 macro sector, = 0 otherwise Dummy
dummies for the geographical macro areas (i.e. USA, EUROPE,
ASIA), and for the time period decades (1976–1985, 1986–1995,
1996–2005) to capture the year of award. Another set of regressors
includes: a dummy  (PRO) to account for innovations with at
least one public research organization applicant (i.e. either an
academic or a governmental organization), a dummy  (MAPPL) for
collaborative innovations to indicate the presence of more than
one applicant, and a count variable (NINV) that reports the number
of applicants. All the variables used in the regressions analysis
are summarized in Table 6. Table 7 provides instead their main
descriptive statistics.
Results from the multivariate probit regression are reported in
Table 8.
Column (1) reports the estimated coefficients for the most par-
simonious model which does not include sector and time-period
dummies. Column (2) includes both time-period and OST 5 macro-
sector dummies. Columns (3–4) report the estimated coefficients
and average marginal effects (AMEs) for the full model with both
Table 7
Descriptive statistics.
Variable Obs Mean Std. Dev. Min  Max
PAT 2802 0.096 0.295 0 1
MAPPL 2802 0.256 0.437 0 1
NINV 2802 1.665 0.902 1 5
USA  2802 0.877 0.329 0 1
EUROPE 2802 0.038 0.191 0 1
ASIA 2802 0.067 0.250 0 1
PRO  2802 0.375 0.484 0 1
dum1986 1995 2802 0.356 0.479 0 1
dum1996 2005 2802 0.322 0.467 0 1
Electrical Eng 2802 0.260 0.439 0 1
Instruments 2802 0.366 0.482 0 1
Chemistry Pharma 2802 0.063 0.243 0 1
Process Eng 2802 0.222 0.416 0 1
time-period and OST 30 macro-sector dummies (not reported for
the sake of clarity).
In terms of geographical and sectoral effect, our regression
analysis confirms the previous findings from the univariate analy-
sis. Ceteris paribus, awarded innovations with all applicants from
Asia show a larger probability to be patented with respect to
the excluded category (which includes non-US, non-European and
non-Asian countries), followed by awarded innovations with at
least one US applicant. No significant difference in the average
patent propensity is detected between the OST5 macro sectors rep-
resented by the four dummy  variables included in the multivariate
regression and the excluded one (Mechanical Engineering).
In terms of differences in the propensity across applicants
and/or between collaborative and non-collaborative innovations,
we find that innovations with at least one PRO as applicant show
a lower patent propensity with respect to the excluded category
(only corporate applicants). Finally patent propensity increases
with the number of inventors (NINV) whereas having multiple
applicants (MAPPL) does not seem to exert a significant effect on
the propensity to patent an awarded innovation.
The former finding confirms our previous point about the lower
propensity to patent for non-corporate inventors as a consequence
of their relative lack of downstream capabilities when compared
to firms. The latter finding seems to be in contrast with previous
studies, such as Brouwer and Kleinknecht (1999) and Kleinknecht
and van der Panne (2011), who instead found a positive correlation
between R&D collaborations and propensity to patent. This discrep-
ancy can be explained in term of different research design. These
two  studies employed firm level data (the Community Innovation
Survey) whereas our analysis is carried out at the innovation level.
Moreover these studies adopted a completely different depend-
ent variable (i.e. whether the firm applied or not for a patent at
the European Patent Office or the number of firm’s EPO patents)
and a different proxy for collaborative innovations (i.e. the firm’s
Author's personal copy
R. Fontana et al. / Research Policy 42 (2013) 1780– 1792 1789
Table 8
Estimation results for the propensity to patent.
Dependent variable: Pr(PAT = 1) (1) (2) (3) (4)
Coefficients Coefficients Coefficients Marginal effects
USA 0.406** 0.387** 0.407** 0.045***
(0.165) (0.169) (0.169) (0.014)
EUROPE 0.165 0.209 0.247 0.040
(0.216) (0.221) (0.222) (0.042)
ASIA 0.971*** 0.965*** 1.006*** 0.240***
(0.196) (0.201) (0.201) (0.067)
PRO −0.635*** −0.617*** −0.605*** −0.077***
(0.083) (0.084) (0.085) (0.009)
MAPPL −0.109 −0.087 −0.094 −0.013
(0.091) (0.091) (0.092) (0.012)
NINV 0.153*** 0.144*** 0.138*** 0.019***
(0.037) (0.038) (0.038) (0.005)
dum1986 1995 −0.157** −0.128 −0.017
(0.079) (0.082) (0.011)
dum1996 2005 −0.144 −0.128 −0.017
(0.093) (0.099) (0.013)
Electrical Eng 0.115
(0.129)
Instruments −0.120
(0.128)
Chemistry Pharma 0.259
(0.169)
Process Eng 0.189
(0.137)
OST30 Sector Dummies Yes Yes
Constant −1.816*** −1.755*** −5.281***
(0.177) (0.206) (0.275)
Observations 2802 2802 2802 2802
Log  likelihood −811.436 −796.360 −792.076
Probit Maximum Likelihood estimates with clustered robust standard errors (in parenthesis).
** Statistically significant at 5% level.
*** Statistically significant at 1% level.
past collaborations with other R&D partners). Results from other
innovation-level studies such as Mäkinen (2007) find instead a non-
significant effect of collaborations on the propensity to patent a
given innovation. Moreover, our result can be understood on the
basis of the two effects suggested by Peeters and Van Pottelsberghe
de la Potterie (2006) to explain the patenting of collaborative inno-
vations. On the one hand there is a ‘need effect’ which refers
to “[. . .]  a higher need for patent protection resulting from the
mutual access to the partners’ knowledge bases” (Peeters and Van
Pottelsberghe de la Potterie, 2006, p. 127). On the other hand,
there is a ‘novelty effect’ which refers to a “[. . .]  potentially more
fundamental and breakthrough knowledge generated by R&D col-
laborations compared to in-house R&D alone, which would result in
more patents” (Peeters and Van Pottelsberghe de la Potterie, 2006,
p. 127). In our case the ‘novelty effect’ is probably better proxied by
the number of inventors (NINV), which has a positive and statisti-
cally significant coefficient, and dominates the ‘need effect’ which
instead is better proxied by the dummy  variable MAPPL.
5. Reassessing patent propensity: a reappraisal of the
empirical evidence
At this point, it is instructive to compare our findings with prior
estimates of patent propensity. This is done in Table 9.
Table 9 is based on an extensive recognition of the existing liter-
ature and summarizes the findings of all the studies we have been
able to identify, that provide direct or indirect empirical estimates
of the importance of patent protection for product innovations in
different industries using different types of methods. The last row
of the table reports the findings of this paper. It has to be stressed
that this exercise has important limitations, because the approach
used to assess patent propensity is not consistent across studies.
For instance, in some cases the estimation of patent propensity
was  not the main goal of the research (this is, in particular, the
case for Acs and Audretsch, 1990). In other cases (Mansfield, 1986),
the survey asked how many patentable innovations were actually
patented which is different from our focus on the ratio between the
number of patented innovations and the total number of innova-
tions. Levin et al. (1987) asked respondents to assess the relative
effectiveness of patents in comparison to other alternative appro-
priability strategies. Cohen et al. (2000) asked firms to report the
share of innovations for which they have applied for a patent which,
again, is somewhat different from the variable we  have constructed
in this paper. These limitations notwithstanding, we believe that it
is useful to compare our results with the main findings concerning
patent propensity emerging from the literature.
It is immediately interesting to note that, consistently with what
we have highlighted in the introduction, four of the studies listed
in the table (Thomson, 2009; Moser, 2012; Meisenzahl and Mokyr,
2011; Nicholas, 2011) are contributions of economic historians or
historians of technology.
In terms of research strategies, we can draw a distinction
between studies using a survey approach (Mansfield, 1986; Levin
et al., 1987; Arundel and Kabla, 1998; Brouwer and Kleinknecht,
1999; Cohen et al., 2000) and other contributions that instead
estimate patent propensity using indicators of overall innovative
output not based on patents. Although providing very detailed
snapshots on innovative activities, it is well known that also the
data collected by means of innovation surveys suffer from several
shortcomings. The main one is that this type of data, unavoidably,
reflects the personal judgement of actors that are required to moni-
tor and self-assess their own  innovative activities and performance
(Mairesse and Mohnen, 2010).
Turning our attention to the research strategies and sources
employed by the studies that are not based on surveys, Acs and
Author's personal copy
1790 R. Fontana et al. / Research Policy 42 (2013) 1780– 1792
Ta
b
le
 
9
Es
ti
m
at
es
 
of
 
p
at
en
t 
p
ro
p
en
si
ty
 
in
 
d
if
fe
re
n
t 
st
u
d
ie
s.
A
ve
ra
ge
 
p
at
en
t
p
ro
p
en
si
ty
 
(%
)
1s
t  
h
ig
h
es
t
p
ro
p
en
si
ty
 
to
p
at
en
t
2n
d
 
h
ig
h
es
t
p
ro
p
en
si
ty
 
to
p
at
en
t
1s
t 
lo
w
es
t
p
ro
p
en
si
ty
 
to
p
at
en
t
2n
d
 
lo
w
es
t
p
ro
p
en
si
ty
 
to
p
at
en
t
In
d
ic
at
or
 
of
 
p
at
en
t
p
ro
p
en
si
ty
D
at
a 
&
 
m
et
h
od
s 
Pe
ri
od
 
G
eo
gr
ap
h
ic
al
sc
op
e
M
an
sfi
el
d
 
(1
98
6)
 
77
 
Pe
tr
ol
eu
m
 
(8
6%
) 
M
ac
h
in
er
y 
(8
6%
) 
Pr
im
ar
y 
m
et
al
s
(5
0%
)
M
ot
or
 
ve
h
ic
le
s
(6
5%
)
Pr
op
en
si
ty
 
to
 
p
at
en
t 
(s
al
es
w
ei
gh
te
d
)
Su
rv
ey
 
of
 
a 
ra
n
d
om
sa
m
p
le
 
of
 
la
rg
e
fi
rm
s 
(o
ve
r  
1
m
il
li
on
 
$ 
R
&
D
 
in
19
81
)
19
81
–1
98
3 
U
SA
Le
vi
n
 
et
 
al
. (
19
87
) 
D
ru
gs
 
O
rg
an
ic
 
ch
em
ic
al
s 
Pu
lp
 
an
d
 
p
ap
er
 
C
om
p
u
te
rs
 
Pe
rc
ei
ve
d
 
ef
fe
ct
iv
en
es
s 
of
p
at
en
ts
 
as
 
ap
p
ro
p
ri
at
io
n
to
ol
Y
al
e 
Su
rv
ey
 
(6
50
R
&
D
 
ex
ec
u
ti
ve
s
fr
om
 
13
0 
bu
si
n
es
s
li
n
es
)
19
81
 
U
SA
A
cs
 
an
d
 
A
u
d
re
ts
ch
(1
99
0)
Lu
m
be
r 
an
d
fu
rn
it
u
re
Fo
od
 
an
d
 
to
ba
cc
o 
Pe
tr
ol
eu
m
 
R
u
bb
er
 
an
d
 
p
la
st
ic
s 
Pa
te
n
ts
/I
n
n
ov
at
io
n
s 
Sm
al
l B
u
si
n
es
s
A
d
m
in
is
tr
at
io
n
In
n
ov
at
io
n
 
D
at
as
et
19
82
 
U
SA
A
ru
n
d
el
 
an
d
 
K
ab
la
(1
99
8)
35
.9
0 
Ph
ar
m
ac
eu
ti
ca
ls
(7
9.
2%
)
C
h
em
ic
al
s 
(5
7.
3%
) 
Te
xt
il
es
 
(8
.1
%
) 
B
as
ic
 
m
et
al
s
(1
4.
6%
)
Pr
op
en
si
ty
 
to
 
p
at
en
t
p
ro
d
u
ct
 
in
n
ov
at
io
n
s 
(s
al
es
w
ei
gh
te
d
)
PA
C
E 
Su
rv
ey
 
(6
04
la
rg
e 
fi
rm
s)
19
93
 
Eu
ro
p
e
B
ro
u
w
er
 
an
d
K
le
in
kn
ec
h
t 
(1
99
9)
25
.4
0 
R
u
bb
er
 
an
d
 
p
la
st
ic
p
ro
d
u
ct
s 
(3
6.
4%
)
Ph
ar
m
ac
eu
ti
ca
ls
,
C
h
em
ic
al
s,
Pe
tr
ol
eu
m
 
(3
6.
3%
)
B
as
ic
 
m
et
al
s 
(9
.9
%
) 
C
la
ss
, c
la
y 
an
d
ce
ra
m
ic
s 
(1
1.
8%
)
Pe
rc
en
ta
ge
 
of
 
fi
rm
s 
fo
r
w
h
ic
h
 
p
at
en
t 
ar
e 
“v
er
y
im
p
or
ta
n
t”
 
or
 
“c
ru
ci
al
” 
fo
r
p
ro
d
u
ct
 
in
n
ov
at
io
n
s
C
IS
 
Su
rv
ey
 
19
92
 
N
et
h
er
la
n
d
s
C
oh
en
 
et
 
al
. (
20
00
) 
49
.1
2 
D
ru
gs
 
(9
5.
5%
) 
M
in
er
al
 
p
ro
d
u
ct
s
(7
9.
2%
)
M
et
al
s 
(2
.9
7%
) 
St
ee
l (
4.
46
%
) 
Pr
op
en
si
ty
 
to
 
p
at
en
t
p
ro
d
u
ct
 
in
n
ov
at
io
n
s
C
M
S  
Su
rv
ey
, 1
16
5
la
rg
e 
fi
rm
s
19
91
–1
99
3  
U
SA
Th
om
so
n
 
(2
00
9)
 
60
.2
 
El
ec
tr
ic
it
y 
(8
7.
5%
) 
A
gr
ic
u
lt
u
re
:
h
ar
ve
st
in
g 
(8
5.
7%
)
C
lo
ck
s  
(1
1.
1%
) 
M
et
al
 
w
or
ki
n
g
(4
0%
)
Sh
ar
e 
of
 
ex
h
ib
it
or
s
d
is
p
la
yi
n
g 
p
at
en
te
d
in
n
ov
at
io
n
s
C
at
al
og
u
e 
of
 
N
ew
Y
or
k 
G
re
at
Ex
h
ib
it
io
n
18
53
 
U
SA
M
os
er
 
(2
01
2)
 
11
.1
0 
M
an
u
fa
ct
u
ri
n
g
m
ac
h
in
er
y 
(2
9.
8%
)
En
gi
n
es
 
(2
4.
6%
) 
M
in
in
g 
an
d
m
et
al
lu
rg
y 
(5
%
)
C
h
em
ic
al
s 
(5
.1
%
) 
N
u
m
be
r 
of
 
p
at
en
te
d
in
n
ov
at
io
n
s 
in
 
th
e 
to
ta
l
n
u
m
be
r 
of
 
in
n
ov
at
io
n
s
ex
h
ib
it
ed
C
at
al
og
u
e 
of
C
ry
st
al
 
Pa
la
ce
G
re
at
 
Ex
h
ib
it
io
n
18
51
 
U
K
M
ei
se
n
za
h
l a
n
d
 
M
ok
yr
(2
01
1)
60
.0
0 
Te
xt
il
es
 
(8
1%
) 
Sh
ip
s 
(7
0%
) 
In
st
ru
m
en
ts
,
C
on
st
ru
ct
io
n
 
(3
5%
)
M
in
in
g  
(4
1%
) 
N
u
m
be
r 
of
 
“g
re
at
in
ve
n
to
rs
” 
w
it
h
 
1  
p
at
en
t 
or
m
or
e
A
n
al
ys
is
 
of
bi
og
ra
p
h
ic
al
d
ic
ti
on
ar
ie
s
16
60
–1
83
0 
U
K
N
ic
h
ol
as
 
(2
01
1)
 
44
.6
 
M
ac
h
in
er
y 
an
d
m
ac
h
in
e 
to
ol
s
(7
0.
8%
)
A
u
to
 
(6
7.
7%
) 
Fo
od
 
(3
3.
1%
) 
C
h
em
ic
al
s 
(3
4.
9%
) 
N
u
m
be
r 
of
 
R
&
D
 
p
er
fo
rm
in
g
fi
rm
s 
w
it
h
 
1 
p
at
en
t o
r 
m
or
e
M
at
ch
in
g 
p
at
en
ts
to
 
fi
rm
s 
in
 
N
R
C
su
rv
ey
s
19
21
–1
93
8 
U
SA
Th
is
 
p
ap
er
 
9.
6 
C
h
em
is
tr
y,
 
p
h
ar
m
a
(1
4.
2%
)
Pr
oc
es
s
en
gi
n
ee
ri
n
g
(1
1.
2%
)
In
st
ru
m
en
ts
 
(6
.8
%
) 
M
ec
h
an
ic
al
en
gi
n
ee
ri
n
g 
(8
.8
%
)
N
u
m
be
r 
of
 
p
at
en
te
d
in
n
ov
at
io
n
s 
in
 
th
e 
to
ta
l
n
u
m
be
r 
of
 
in
n
ov
at
io
n
s
aw
ar
d
ed
 
th
e 
p
ri
ze
R
&
D
 
10
0
co
m
p
et
it
io
n
19
77
–2
00
4 
W
or
ld
Author's personal copy
R. Fontana et al. / Research Policy 42 (2013) 1780– 1792 1791
Audretsch (1990) is a study of innovation in small firms in the US
in 1982 based on the Small Business Administration Innovation
dataset. This dataset was constructed by retrieving innovations in a
large number of technical and trade journals. The results obtained
may  seem somewhat perplexing, because ‘petroleum’ and ‘rubber’,
sectors that in innovation surveys usually appear characterized by
a relative high propensity to patent, exhibit a very low propensity.
In this respect, two considerations are in order. First, the Acs and
Audretsch (1990) study is focused on small firms. Second, the indi-
cator of patent propensity is constructed by comparing innovations
with patents at industry level, but these quantities do not refer to
the same groups of firms.
The studies by Moser (2005, 2012) estimate patent propensity
as the share of patented inventions in the total number of inven-
tions presented at the great nineteenth century world fairs.21 The
most important result emerging from these studies is the relatively
low levels of patent propensity. In this respect, as we  have already
noted, the findings of our paper are intriguingly in line with this
type of evidence. The study of Thomson (2009) is an application of
the research strategy of Moser (2005, 2012) to the case of the New
York exhibition of 1853. Interestingly enough, Thomson finds lev-
els of patent propensity that are much higher than those estimated
by Moser (2012).
Meisenzahl and Mokyr (2011) is a study based on the ‘prosopo-
graphical’ analysis of a sample of 759 British inventors, engineers,
mechanics and skilled craftsmen active in the period 1660–1830
constructed using biographical dictionaries, an approach originally
pioneered by Khan and Sokoloff (1993). In this case the propen-
sity to patent was defined as the share of inventors that have been
granted at least one patent in the total number of inventors. The
study of Nicholas (2011) adopts a similar approach, but in this case
the actors are not individuals, but firms. Here propensity to patent
is defined as the share of firms with at least 1 patent in the total
number of firms with R&D establishment that were included in the
surveys of the National Research Council in the US in the period
1921–1938.
Overall, the main result emerging from Table 9, holding in dif-
ferent periods and locations, is that a sizeable share of inventive
activities is not covered by patents. Hence, the obvious recommen-
dation is that it is crucial, whenever possible, to assess inventive
output combining evidence from patents with evidence from other
sources. However, the wide range of the estimated levels of patent
propensity (even in studies such as Moser (2012) and Thomson
(2009) that adopt similar research designs), also suggest that we
are still very far from a robust understanding of patent propen-
sity in different contexts and that more research on this theme is
needed.
6. Concluding remarks
This paper has reassessed the propensity to patent using a novel
data source on R&D awards that so far had received little attention.
Our analysis has provided two types of evidence. First, a rela-
tive low number of important innovations are patented. Second,
patent propensity tends to vary across industrial sectors and types
of organizations. While these results mirror, to a certain extent,
some of the earlier findings in the literature, an important differ-
ence exists concerning the method. In particular, inspired by the
recent works of economic historians (Moser, 2005, 2010), we  have
adopted a straightforward definition of patent propensity in terms
of the share of innovations patented over the total innovative out-
put, a choice that marks an important departure with respect to
21 In Table 9 we  have reported only the results of Moser (2012) for the UK which
may be taken as representative of the overall findings.
investigations carried out in the field of ETIS which typically assess
innovative output using indicators based on patents or innovation
surveys.
In this respect, we believe that recent contributions in the field
of economic history and history of technology have probably some
important methodological lessons to teach to the ETIS field. In our
judgement, the main lesson is that much more effort ought to be
devoted to the exploration of the potentialities of new sources
of data for constructing indicators of innovative output. More
specifically, the contributions of economic historians suggest that,
with some ingenuity the literature produced by technologists and
engineering practitioners (e.g. technical journals and magazines,
exhibition and fair catalogues, engineering prizes and awards) can
be successfully employed for constructing indicators of innovative
output that can usefully integrate the patent evidence.
We would like to argue that this is precisely one of the method-
ological lessons emerging from some of the early contributions of
Nick von Tunzelmann. Although still not fully appreciated, impor-
tant insights of his famous analysis of the development and impact
of steam power technology in the British economy were based on
the evidence gathered from a contemporary engineering journal
published in Cornwall called Lean’s Engine Reporter.22 Exploiting
the information contained in the journal, von Tunzelmann was
able to provide new estimates for the rates of technical progress
as well as for the rate of diffusion of new technical practices in the
field of steam power technology (von Tunzelmann, 1970, 1978, pp.
252–264). It would be wrong to think that similar endeavours per-
tain exclusively in the domain of historical research. One of the aims
of this paper was precisely to show that similar sources are prob-
ably available also for modern studies of innovation and technical
change. At the same time, our plea for the use of new sources should
not be read as invoking the complete abandonment of patents as
indicator of inventive output. Rather, we  believe that real progress
can be only attained by systematically combining different types of
innovation indicators constructed using different types of sources
and measurement techniques.
Acknowledgements
Comments from participants to the conference “Technical
change: History, Economics and Policy” (SPRU, University of Sussex,
29–30 March 2010) are gratefully acknowledged. We  are also par-
ticularly grateful to Fred Block, Maria Savona, Jonathan Sapsed and
three anonymous referees for very helpful suggestions on subse-
quent drafts. Finally, we  would like to thank Nick von Tunzelmann
who, throughout the years, has been a constant source of inspira-
tion and encouragement.
References
Acs, Z.J., Audretsch, D.B., 1990. Innovation and Small Firms. MIT Press, Cambridge,
MA.
Arundel, A., 2001. The relative effectiveness of patents and secrecy for appropriation.
Research Policy 30, 611–624.
Arundel, A., Kabla, I., 1998. What percentage of innovations are patented? Empirical
estimates for European firms. Research Policy 27, 127–141.
Basberg, B.L., 1987. Patents and the measurement of technological change: a survey
of  the literature. Research Policy 16, 131–141.
Block, F., Keller, M.R., 2009. Where do innovations come from? Transformations in
the US economy 1970–2006. Socio-Economic Review 7, 459–483.
Boldrin, M.,  Levine, D., 2008. Against Intellectual Monopoly. Cambridge University
Press, Cambridge, UK.
Brouwer, E., Kleinknecht, A., 1999. Innovative output and a firm’s propensity to
patent. An exploration of CIS micro data. Research Policy 28, 615–624.
22 For a recent discussion of the historical significance of this engineering journal
see  Nuvolari and Verspagen (2007, 2009).
Author's personal copy
1792 R. Fontana et al. / Research Policy 42 (2013) 1780– 1792
Brunt, L., Lerner, J., Nicholas, T., 2012. Inducement prizes and innovation. Journal of
Industrial Economics 60, 657–696.
Carpenter, M.P., Narin, F., Woolf, P., 1981. Citation rates to technologically important
patents. World Patent Information 3, 160–163.
Cohen, W.M.,  Nelson, R.R., Walsh, J.P., 2000. Protecting their intellectual assets:
appropriability conditions and why  US manufacturing firms patent (or not). In:
NBER working paper no. 7552.
De Rasenfosse, G., 2010. How much do we know about firms’ propensity to patent
and  should we  worry about it? In: Paper presented at the 2010 Druid Summer
Conference.
Granstrand, O., 1999. The Economics and Management of Intellectual Property:
Towards Intellectual Capitalism. Edward Elgar, Aldershot.
Griliches, Z., 1990. Patents statistics as economic indicators. A survey. Journal of
Economic Literature 28, 1661–1707.
Hall, B.H., 2005. Exploring the patent explosion. Journal of Technology Transfer 30,
35–48.
Hall, B.H., Jaffe, A., Trajtenberg, M.,  2001. The NBER-Citations datafile: lessons,
insights and methodological tools. In: NBER working paper no. 8498.
Hall, B.H., Ziedonis, R., 2001. The patent paradox revisited. An empirical study of
patenting in the US semiconductor industry, 1979–1995. Rand Journal of Eco-
nomics 32, 101–128.
Hinze, S., Reiss, T., Schmoch, U., 1997. Statistical analysis on the distance between
fields of technology. In: Report for the European Commission, TSER project.
Hussinger, K., 2007. Is silence golden? Patent versus secrecy at the firm level. Eco-
nomics of Innovation and New Technology 15, 735–752.
Khan, B.Z., Sokoloff, K.L., 1993. “Schemes of Practical Utility”: entrepreneurship and
innovation among ‘Great Inventors’ in the United States, 1790–1865. Journal of
Economic History 53, 289–307.
Kleinknecht, A., van der Panne, G., 2011. The propensity to patent an innovation:
comparing entrepreneurial to routinized innovators. In: Audretsch, D.B., Falck,
O.,  Heblich, S. (Eds.), Handbook of Research on Innovation and Entrepreneurship.
Edward Elgar, Northampton, MA.
Kortum, S., Lerner, J., 1999. What is behind the recent surge in patenting? Research
Policy 28, 1–22.
Kuznets, S., 1962. Inventive activity: problems of definition and measurement. In:
Arrow, K.J, Nelson, R.R (Eds.), The Rate and Direction of Inventive Activity. Prince-
ton  University Press, Princeton.
Levin, R.C., Klevorick, A.K., Nelson, R.R., Winter, S.G., 1987. Appropriating the returns
from industrial research and development. Brookings Papers on Economic Activ-
ity 3, 783–831.
Mairesse, J., Mohnen, P., 2010. Using innovation surveys for econometric analysis.
In:  Hall, B.H, Rosenberg, N (Eds.), Handbook of Economics of Innovation, vol. II.
Elsevier, Amsterdam.
Mäkinen, I., 2007. To Patent or not to patent? An innovation-level investigation of
the  propensity to patent. In: VTT Publication Series No. 646.
Mansfield, E., 1986. Patents and innovation: an empirical study. Management Sci-
ence 32, 173–181.
Meisenzahl, R., Mokyr, J., 2011. The rate and direction of invention in the British
Industrial Revolution. Incentives and institutions. In: NBER Working paper, no.
16993.
Moser, P., 2002. The determinants of innovation: new evidence from nineteenth
century world fairs. University of California at Berkeley (PhD thesis).
Moser, P., 2005. How do patent laws influence innovation? Evidence
from nineteenth century world’s fairs. American Economic Review 95,
1214–1236.
Moser, P., 2012. Innovation without patents. Evidence from world fairs. Journal of
Law  and Economics 55, 43–74.
Mowery, D.C., Ziedonis, A.A., 2002. Academic patent quality and quantity before
and after the Bayh-Dole act in the United States. Research Policy 31,
399–418.
Mowery, D.C., Nelson, R.R., Sampat, B., Ziedonis, A., 2001. The growth of patenting
and licensing by U.S. universities: an assessment of the effects of the Bayh-Dole
Act of 1980. Research Policy 30, 99–119.
Nicholas, T., 2011. Did R&D firms used to patent? Evidence from the first innovation
surveys. Journal of Economic History 71, 1032–1059.
Nuvolari, A., Verspagen, B., 2007. Lean’s Engine Reporter and the development of
the  Cornish engine. A reappraisal. Transactions of the Newcomen Society 77,
167–189.
Nuvolari, A., Verspagen, B., 2009. Technical choice, innovation and British steam
engineering, 1800–1850. Economic History Review 62, 685–710.
Peeters, C., Van Pottelsberghe de la Potterie, B., 2006. Innovation strategy and the
patenting behavior of firms. Journal of Evolutionary Economics 16, 109–135.
Scherer, F.M., 1983. The propensity to patent. International Journal of Industrial
Organization 1, 107–128.
Scherer, F.M., 1989. Comments on Z. Griliches patents: recent trends and puzzles.
Brookings Papers on Economic Activity 9, 291–330.
Scherer, F.M., 1999. New Perspectives on Economic Growth and Technological Inno-
vation. Brookings Institution, New York.
Schmookler, J., 1966. Invention and Economic Growth. Harvard University Press,
Cambridge, MA.
Silverberg, G., Verspagen, B., 2007. The size distribution of inventions revisited: an
application of extreme value statistics to citation and value measure of patent
significance. Journal of Econometrics 139, 318–339.
Sullivan, R.J., 1989. England’s “age of invention”: the acceleration of patents and
patentable invention during the Industrial Revolution. Explorations in Economic
History 26, 429–443.
Taylor, C.T., Silberston, Z.A., 1973. The Economic Impact of the Patent System. A
Study of the British Experience. Cambridge University Press, Cambridge, UK.
Thomson, R., 2009. Structures of Change in the Mechanical Age. Technological
Innovation in the United States. John Hopkins University Press, Baltimore, pp.
1790–1865.
Trajtenberg, M.,  1990. A penny for your quotes: patent citations and the value of
innovations. Rand Journal of Economics 21, 172–187.
van Zeebroeck, N., 2011. The puzzle of patent value indicators. Economics of Inno-
vation and New Technology 20, 33–62.
von Tunzelmann, G.N., 1970. Technological diffusion during the Industrial Revolu-
tion. The case of the Cornish pumping engine. In: Hartwell, R (Ed.), The Industrial
Revolution. Basil Blackwell, Abingdon.
von Tunzelmann, G.N., 1978. Steam Power and British Industrialization to 1860.
Clarendon, Oxford.