American Economic Association
Real Effects of Academic Research: Comment
Author(s): Zoltan J. Acs, David B. Audretsch and Maryann P. Feldman
Source:
The American Economic Review,
Vol. 82, No. 1 (Mar., 1992), pp. 363-367
Published by: American Economic Association
Stable URL: http://www.jstor.org/stable/2117624 .
Accessed: 30/07/2014 10:40
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
.
American Economic Association is collaborating with JSTOR to digitize, preserve and extend access to The
American Economic Review.
http://www.jstor.org
This content downloaded from 152.23.75.172 on Wed, 30 Jul 2014 10:40:55 AM
All use subject to JSTOR Terms and Conditions
Real Effects of Academic Research:
Comment
By
ZOLTAN J.
Acs,
DAVID B.
AUDRETSCH,
AND
MARYANN
P. FELDMAN*
A
fundamental
issue which remains unre-
solved in the economics of
technology
is
the
identification and measurement
of
R&D
spillovers,
or
the
extent to which a firm
is
able
to
exploit economically
the
investment
in
R&D
made by another
company.
In a
1989 paper
in this
Review,
Adam
Jaffe
ex-
tended
his
pathbreaking
1986
study
measur-
ing
the total R&D
"pool"
available
for
spillovers
to
identify
the
contribution
of
spillovers
from
university
research
to "com-
mercial
innovation" (Jaffe, 1989
p. 957).
Jaffe's
findings
were the
first
to
identify
the
extent
to which
university
research
spills
over
into
the
generation
of
inventions
and
innovations
by private
firms.
To measure
technological
change,
Jaffe
relies upon
the number of
patented
inven-
tions
registered
at the
U.S.
patent
office,
which
he
argues
is "a
proxy
for new eco-
nomically useful
knowledge"
(Jaffe,
1989
p.
958).
In
order to relate the
response
of this
measure
to R&D
spillovers
from
universi-
ties,
Jaffe
modifies the
"knowledge produc-
tion function" introduced
by
Zvi
Griliches
(1979)
for
two
inputs:
(1)
log(Pik)
=
/31k
log(Iik)
+
I32k
log(Uik)
+
fl3k[lOg(Uik)
Xlog(Cik)]
+
eik
where P is
number of patented
inventions,
I
represents the private
corporate expendi-
tures on
R&D, U represents
the research
expenditures
undertaken at
universities, C
is a
measure of the geographic
coincidence
of
university and
corporate
research,
and e
represents
stochastic
disturbance.
The unit
of
observation is
at
the level of
the
state, i,
and what
Jaffe
terms the
"technological
area,"
or the
industrial
sector, k. In
addi-
tion,
Jaffe
includes
the
state
population
(Popik)
in
his
estimating
equation in
order
to
control for
the size
differential
across the
geographic units of
observation.
Jaffe's
(1989)
statistical results
provide ev-
idence that
corporate
patent
activity re-
sponds
positively
to
commercial
spillovers
from
university
research.
Not only
does
patent
activity
increase in
the
presence of
high
private
corporate
expenditures
on
R&D, but
also as a
result
of
research ex-
penditures
undertaken
by universities
within
the
state. The results
concerning
the role of
geographic
proximity
in
spillovers
from uni-
versity research
are
clouded,
however, by
the lack
of
evidence that
geographic proxim-
ity
within the
state
matters as
well. Accord-
ing
to
Jaffe
(1989 p.
968), "There is
only
weak
evidence that
spillovers
are
facilitated
by
geographic coincidence
of
universities
and
research
labs within
the
state."
While Jaffe's
(1989)
model
is
constructed
to
identify
the contribution of
university
re-
search to
generating
"new
economically
useful
knowledge" (p.
958),
F.
M. Scherer
(1983), Edwin Mansfield
(1984),
and
Griliches
(1990)
have all
warned
that
mea-
suring
the
number
of
patented
inventions
is
not
the
equivalent
of a
direct measure of
innovative
output.
For
example,
Ariel
Pakes
and Griliches
(1980
p.
378) argued
that
"patents
are
a
flawed
measure
(of
innova-
tive
output);
particularly
since
not
all
new
innovations
are
patented
and
since
patents
differ
greatly
in
their economic
impact."
In
addressing
the
question
"Patents as indica-
tors of
what?"
Griliches
(1990
p.
1669)
con-
cludes that
"Ideally,
we
might hope
that
patent
statistics would
provide
a
measure
of
the
(innovative)
output...
.
The
reality,
however,
is
very
far from
it. The dream of
*Acs:
Department
of
Economics and
Finance, Uni-
versity
of
Baltimore,
Baltimore,
MD;
Audretsch:
Wis-
senschaftszentrum
Berlin
fur
Sozialforschung,
Berlin,
Germany;
Feldman:
Carnegie
Mellon
University and
Goucher
College.
We
thank
two
referees
for
their
helpful
comments
and
suggestions.
All
errors
and
omissions
remain
our
responsibility.
363
This content downloaded from 152.23.75.172 on Wed, 30 Jul 2014 10:40:55 AM
All use subject to JSTOR Terms and Conditions
364
THE
AMERICAN
ECONOMIC
REVIEW
MARCH
1991
TABLE 1-COMPARISON
AMONG PATENT, UNIVERSITY
RESEARCH,
AND
INNOVATION
MEASURES
Number of innovations
Standard
yielded
per unit
Measure
Mean
deviation
Minimum Maximum
of
input
University research
expenditures
98.8 144.0
12.0 710.4 1.3
(millions of dollars)
Drugs
28.5
35.3
2.2 142.3
3.3
Chemicals
5.7
9.7 0.5
46.7
1.9
Electronics
21.0 49.2 0.3
239.0
2.8
Mechanical
12.7
25.6
0.9
126.1
3.5
Corporate patents
879.4
975.7
39.0
3,230.0
0.148
Drugs
71.7
99.4 1.0
418.0 0.132
Chemicals
201.2
249.0 6.0 908.0 0.054
Electronics
225.0 295.3 7.0 1,142.0
0.263
Mechanical
300.8 319.9
20.0 993.0
0.146
Innovations
130.1 206.4
4.0 974.0
Drugs
9.5
16.0 0.0 75.0
Chemicals
10.9 17.7 0.0 80.0
Electronics
59.2 100.5
1.0 475.0
Mechanical
44.5 79.7
0.0
416.0
Notes:
All
dollar
figures are
millions of 1972 dollars. Data on
university research
funds by state are available
for the
four broad technical
areas
of
drugs
and
medical
technology;
chemicals;
electronics, optics
and nuclear
technology;
and mechanical
arts. These groups, along
with the data for
university research expenditures
and corporate
patents,
are
from
Jaffe (1989).
getting hold of an
output indicator of inven-
tive activity is one
of the strong
motivating
forces for economic
research
in
this
area."
The use of patent
counts
to
identify
the
effect of spillovers
from
university
research
might be expected
to be particularly
sensi-
tive to what Scherer
(1983 p. 108) has
termed
the "propensity
to
patent."
Just
as
Albert
N. Link and John
Rees (1990) found that
small new entrepreneurial
firms tend
to
benefit more than
their established larger
counterparts from
university research
spillovers, Griliches
(1990) and Scherer
(1983)
both
concluded
that
the
propensity
to patent does not
appear to be
invariant
across
a
wide
range
of firm
sizes.
A
different
and more direct measure
of
innovative output
was introduced
in Acs
and Audretsch (1987),
where the measure
of innovative
activity
is the
number of inno-
vations recorded
in
1982 by
the
U.S.
Small
Business Administration
from the leading
technology, engineering,
and trade journals
in each manufacturing
industry. A detailed
description and
analysis
of the data can be
found
in Acs
and
Audretsch
(1988,
1990).
Because
each innovation
was recorded sub-
sequent to
its introduction
in the market,
the
resulting
data base
provides a
more
direct
measure of
innovative activity than
do
patent counts. That is,
the innovation
data base
includes inventions
that were not
patented
but
were
ultimately introduced
into
the market
and excludes inventions that
were
patented but never
proved to be eco-
nomically
viable enough to
appear
in
the
market.
The
extent to
which
university-research
spillovers
serve as a catalyst for
private-cor-
poration
innovative
activity
can be
identi-
fied
by
using
the direct measure of innova-
tive
activity
in
the
model
introduced by Jaffe
in equation
(1). This
enables a direct com-
parison
of
the influence
of
university
R&D
spillovers
on
innovation with
the results that
Jaffe reported
using
the
patent measure.
Table
1
compares
the
mean measures of
university research
expenditures and corpo-
rate patents
for all 29
states used
by Jaffe
with
the
mean
number of innovations
per
state. It should
be noted
that, while Jaffe's
university-research
and
patent
measures are
This content downloaded from 152.23.75.172 on Wed, 30 Jul 2014 10:40:55 AM
All use subject to JSTOR Terms and Conditions
VOL.82 NO.
1
ACS ETAL.: REAL
EFFECTS OFACADEMIC
RESEARCH,
COMMENT
365
TABLE 2-A
COMPARISON BETWEEN
REGRESSION RESULTS
USING JAFFE'S PATENT
MEASURE AND
THE
INNOVATION
MEASURE
All
areas
Electronics Mechanical arts
Patents
Innovations Patents Innovations
Patents Innovations
Independent variable
(i)
(ii)
(iii)
(iv)
(v) (vi)
Log(I) 0.668
0.428 0.631 0.268
0.643
0.649
(8.919)
(4.653) (5.517)
(1.370)
(6.712) (4.720)
Log(Uik)
0.241 0.431 0.265 0.520
0.059
0.329
(3.650)
(6.024) (2.598)
(2.977)
(0.490) (1.999)
Log(Uik)xLog(Ci)
0.020 0.173 .063 0.272
-0.046
0.224
(0.244)
(1.914) (0.531)
(1.331)
(-0.406) (1.436)
Log(Popj)
0.159
-
0.072 0.076 0.076
0.177
-
0.143
(1.297)
(-1.287) (1.263)
(0.742)
(3.767) (-2.051)
S:
0.444
0.451 0.203 0.348
0.181 0.247
R2: 0.959 0.902 0.992 0.951 0.994
0.974
N: 145
125
29 29 27
27
Note: Numbers
in
parentheses
are
t
statistics.
based upon
an
eight-year
sample
(1972-
1977, 1979,
and 1981),
the innovation
mea-
sure
is
based upon
a single year, 1982.
Both
the
number of innovations per
university
research dollar
(millions)
and the
number of
innovations
per patent
vary considerably
across the four
industrial
sectors
included
in
Jaffe's sample.
The
number of innovations
yielded per
dollar
of
university
research
is
apparently
highest
in
the
mechanical
indus-
tries and
lowest
in the chemical
industries.
As
in Acs and Audretsch
(1988), the
amount
of
innovative
activity yielded
per patent
is
highest
in the electronics
sector and
lowest
in chemicals.
While
Jaffe (1989)
was able to
pool
the
different
years across each
state observation
in
estimating
the production
function
for
patented
inventions,
this is not
possible
us-
ing the
innovation measure, due
to data
constraints.
Thus,
it
is
important
first to
establish
that Jaffe's
(1989)
results
do
not
differ greatly
from
estimates
-for
a
single
year.
This is done in
equation
(i)
of Table
2,
where Jaffe's
(1989) patent
measure
for 1981
is used in
the same
estimating equation
found
in
his
table
4B,
based
on all
(techno-
logical) areas.
All of the
data sources
and a
detailed
description
of
the
data and mea-
sures
can
be
found
in
Jaffe (1989).
Using
the
patent
measure
for a
single year
yields
virtually
identical
results
to
those based on
the pooled estimation reported
in
Jaffe's
article. That
is,
both
private corporate ex-
penditures
on
R&D and
expenditures by
universities on research are found to exert a
positive
and
significant
influence
on
patent
activity. Similarly,
both
the geographic coin-
cidence effect and the population variables
have positive coefficients. The estimated co-
efficient of 0.668 for
log(Ij)
in
equation (i)
of Table 2 is
remarkably close to the coef-
ficient
of
0.713 estimated by Jaffe using the
pooled sample. We
conclude
that using
a
single
estimation
year does
not
greatly alter
the results obtained
by Jaffe (1989) using
several years
to
measure the extent of patent
activity.
The number
of
1982 innovations is substi-
tuted for
the number of
registered patents
as the
dependent
variable
in
equation (ii)
of
Table
2,
which
estimates
the
impact
of
spillovers
on
all
technological
areas
com-
bined.1
There are
two
important
differences
that
emerge
when the innovation measure is
used instead
of
the
patent
measure.
First,
the
elasticity
of
log(Uik)
almost
doubles,
'The sample
sizes
differ
between
the
patent
and
innovation
estimations
because
the
observations
with
the
value of zero
had
to be
omitted.
This content downloaded from 152.23.75.172 on Wed, 30 Jul 2014 10:40:55 AM
All use subject to JSTOR Terms and Conditions
366
THE
AMERICAN
ECONOMIC
REVIEW
MARCH
1991
from 0.241 when the patent measure is used
in
equation (i) to
0.431
when
the
innovation
measure is used
in
equation (ii).
That
is,
the
impact of university spillovers is apparently
greater on innovations than on patented
inventions. Second, the impact of the geo-
graphic coincidence effect also is much
greater on innovation activity than on
patents,
suggesting that spillovers from geo-
graphic proximity may be more important
than
Jaffe (1989) concluded.
Jaffe
(1989)
also
estimated knowledge-
production functions for what he calls spe-
cific
technical areas.2 Equations (iii) and (iv)
in
Table
2
compare the estimations based
on
the patent
and
innovation measures for
the electronics area,
and
equations (v)
and
(vi) compare the estimations based on the
two
measures for the mechanical-arts area.
The patent and innovation measures yield
somewhat different results. For the elec-
tronics
area,
expenditures on R&D by pri-
vate corporations
are found to
have
a
posi-
tive and significant influence on patents but
not
on
innovative
activity. By contrast,
in
the
mechanical-arts area, both patent and
innovative
activity respond positively
to
pri-
vate R&D spending. This may reflect the
difference
in what
Sidney
G.
Winter (1984)
termed
the
"technological regime"
between
the electronics and
mechanical-arts areas.
That is, under the "entrepreneurial
regime,"
the
underlying technological information re-
quired
to
produce
an innovation
is
more
likely
to
come
from basic
research
and from
outside
of the
industry. By contrast,
under
the "routinized
regime,"
an
innovation
is
more
likely
to
result
from
technological
in-
formation
from
an
R&D
laboratory
within
the
industry.
Since
the
electronics
area more
closely corresponds
to Winter's
notion
of
the
entrepreneurial regime,
while
the
me-
chanical-arts
area
more
closely
resembles
the routinized
regime,
it is
not
surprising
that
company
R&D
expenditures
are
rela-
tively
less
important
and
university expendi-
tures
on research are
relatively
more
impor-
tant
in producing
innovations
in
electronics
but
not in
the
mechanical
arts.
Further,
as
Mansfield
(1984
p.
462)
noted,
innovations
may
have
a
particular
tendency
not
to
result
from
patented
inventions
in industries
such
as
electronics:
"The
value
and
cost
of
indi-
vidual
patents
vary
enormously
within
and
across
industries....
Many
inventions
are
not
patented.
And
in some
industries,
like
electronics,
there
is considerable
specula-
tion
that
the
patent
system
is
being
by-
passed
to a
greater
extent
than
in
the
past."
Substitution
of
the
direct
measure
of
in-
novative
activity
for
the
patent
measure
in
the
knowledge-production
function
gener-
ally strengthens
Jaffe's
(1989)
arguments
and
reinforces
his
findings.
Most
importantly,
use of
the
innovation
data provides
even
greater
support
than
was found
by
Jaffe:
as
he
predicted,
spillovers
are
facilitated
by
the geographic
coincidence
of
universities
and research
labs within
the
state.
In addi-
tion,
there
is at
least some
evidence
that,
because
the
patent
and
innovation
mea-
sures capture
different
aspects
of
the
pro-
cess
of
technological
change,
results
for spe-
cific sectors
may
be,
at least
to
some extent,
influenced
by
the
technological
regime.
Thus,
we
find that
the importance
of
univer-
sity
spillovers
relative
to
private-company
R&D
spending
is
considerably
greater
in
the electronics
sector
when
the
direct
mea-
sure
of
innovative
activity
is
substituted
for
the
patent
measure.
REFERENCES
Acs,
Zoltan
J.
and
Audretsch,
David
B.,
"In-
novation,
Market
Structure
and
Firm
Size,"
Review
of
Economics
and
Statistics,
November
1987,
69,
567-75.
______ and
, "Innovation
in
Large
and
Small
Firms:
An
Empirical
Analysis,"
American
Economic
Review,
September
1988, 78,
678-90.
and
, Innovation
and
Small
Firms,
Cambridge,
MA:
MIT
Press,
1990.
Griliches,
Zvi,
"Issues
in
Assessing
the
Con-
tribution
of
R &
D
to
Productivity
Growth,"
Bell Journal
of
Economics,
Spring
1979, 10,
92-116.
,"
4Patent
Statistics
as
Economic
In-
2The technological areas are based on a technologi-
cal classification and not on
an
industrial classification.
For further explanation see appendix
A in
Jaffe (1989).
This content downloaded from 152.23.75.172 on Wed, 30 Jul 2014 10:40:55 AM
All use subject to JSTOR Terms and Conditions