Revisiting the Impact of Academic Research on Inventive Output

Much of the early research of the
economics of science and technology is centered around the search for variables
to be used as some sort of indicators of input to innovation and some measure
of the output of innovation. Researchers have found it difficult to find any
catch-all measure of innovation. Perhaps the best measure of the input of
innovative activity is Research and Development (R&D) expenditure, and the
best measure of innovative output is patents. Data on patents and R&D
expenditure is readily available for most developed countries, collected at
aggregate and firm level allowing for a wide scope analysis.


Early research focuses on
determining whether patents are a valuable measure of inventive output as well
as the causality between patents (output) and R&D (input). An early survey
written by Zvi Griliches (1990) is an excellent compilation of the research
devoted to determining the value of patents as an economic indicator. Patent
statistics were being used as early as the 1950’s and 1960’s as an index of
inventive output.


In the following decades, an
“attempt to ‘validate’ patents as an economic indicator,” (Griliches, 1990) led
to increased research into the relationship between R&D and patent
statistics. Major research contributions contemporary with Griliches survey
include Bound et. al (1984), Hall, Griliches, and Hausman (1986), Pakes and
Griliches (1984), Scherer (1983), and Acs and Audretsch (1989).[1]
These studies mostly focus on cross-sectional analysis between countries or
states. This prevailing literature came to the main conclusion that there is a
strong relationship between patenting and R&D expenditure, paving the way
for extensive use of patent and R&D statistics in economic analysis. 


the conclusion that patent and R&D statistics can provide valuable insights
into the economy, more modern research efforts focus on the underlying legal
and regulatory aspects on world patent systems. This part of the literature
focuses questions such as, the way specific patent laws and restrictions and
intellectual property rights affect the inventive output of firms and
countries. An important survey by Hall and Harhoff (2012) provide an detailed
explanation of these ideas. This is an important area of research in this
field, but is not the focus of this research project, so I will leave this as a
note for interest in future research.


of the many areas of interest within the patent-R&D nexus is the idea of
R&D spillovers. A spillover being, the ability of firms to exploit the
investment in R&D made by other firms or entities. Adam B. Jaffe, starting
with his thesis work (1985) and subsequent papers (1986, 1989), was one of the
first to research the effects of R&D investment spillover for inventive
output of firms. Of interest to this research proposal is Jaffe’s work
involving R&D spillover of academic research (1989). This paper provides
the inspiration for the research question that is asked in this proposal: What
is the impact of higher education research on inventive output? Where Jaffe
asks what this impact is at the state level (in the United States), this
proposal asks what this impact is at the national level, using cross country
data for OECD member countries.


impact of education research on inventive output will measure the real effects
that university level research has on the inventive output of private
corporations. This is what Jaffe, and others in the literature call
“spillover”. The degree to which university research spills over into the
private domain to help promote innovation. The motivation behind the research
question is to determine if universities have a positive effect on promoting
innovation, and if they do it is worthwhile to invest more research into
understanding how to best maximize universities research potential. This paper
hopes to update the work done by Jaffe, with a model inspired by his, using a
significantly different dataset, and therefore explain this spillover effect at
a national level instead of state-level.


I will propose a methodology similar to Jaffe’s. The paper will be empirical in
nature and use regression analysis to explain the impact of university R&D
on innovation. The dependent variable will be some measure of innovative output.
Although somewhat contended (Griliches, 1990), patents provide a readily
available measure of innovative output. One such argument against patents as a
measure of output is that not all inventions are patented, but this still may
be the best measure available for such a study. The main independent variables
will be R&D expenditure. Firstly, R&D expenditure follows the early
literature that shows strong linkages between patenting and R&D, therefore
it is expected that R&D will have a positive relationship with total
patents. Secondly, this R&D expenditure will be split into private
corporation R&D and higher education R&D. Business R&D is expected
to have a positive impact on innovative output. Based on Jaffe’s results,
higher education R&D is also expected to have a positive impact on
innovative output, this being the main interest of the research proposed. It
will also be interesting to compare the impacts of business R&D and
university R&D. Also, following Jaffe and previous literature other
independent variables will be considered as controls, and are likely to include
variables such as population, as a control for size of the potential research
population, and also some measure of the amount of universities in the country,
as a control for the research potential of higher education. It is expected
that as population and university size grow the amount of innovative output
will grow. Therefore, to summarize, the model can be visualized as follows,


= BusinessR&Di,t + UniversityR&Di,t + POPi,t
+ UniSizei,t + ei,t


data is available through the OECD “Main Science and Technology Indicators”
database which provides various measures of patenting, R&D, technology
trading and other statistics for OECD member countries. The data is also
available for the last six available years, which allows for either a
cross-sectional analysis or to extend the analysis to include the time
dimension using panel data estimation methods. It is likely that both the
cross-sectional and panel analysis will be done to provide a clear analysis of
the available data and increase the number of observations.