The Royal Institute of Technology
Master’s program in Economics of Innovation and Growth
(Professor Stefan Fölster) Master Thesis in Economics
“An analysis of the effectiveness of
government R&D policies on business
R&D expenditure”

Written by
Kaifeng Li ([email protected])
June 2010
II

ABSTRACT
This thesis investigates the effectiveness of government R&D policies in 13 OECD


Special thanks go to the Professor Hans Lööf for the data analysis guidance, without
his generous support and unlimited patience, it is surely that I cannot go this far.

I would also like to thank Professor Kristina Nyström for the grammar checking and
constructive suggestions, I am sincerely appreciated to her kindness.

At last, I would like to show my deepest gratitude to my mom, whom extremely
eliminates my pressure with her love, and her tough support strongly encouraged my
thesis working.

In a word, I would like to extend my sincere gratitude to everyone for kind helps.
IV Table of Contents
ABSTRACT II
ACKNOWLEDGEMENTS III

was defined by Frascati Manual as “Research and experimental development
(R&D) comprise creative work undertaken on a systematic basis in order to
increase the stock of knowledge, including knowledge of man, culture and society
and the use of this stock of knowledge to devise new applications”. Despite the
recognition of the importance attached to R&D now for long-run economic
growth and living standards, it is commonly argued that social optimal R&D level
cannot be reached without government intervention. Schumpeter (1942), Nelson
(1959) and Arrow (1962) firstly argued the rationale of government R&D
intervention. They hold such a conceptual idea that knowledge is non-rival good.
Therefore, the private return on R&D investment will hardly be appropriated,
which leads to an under-provision of R&D investment in the economy (Lööf and
Hesmati, 2004). Guellec and van Pottelsberghe (2000) argued that imperfect
appropriability and diffusion of knowledge uncontrolled caused innovators cannot
fully appropriate the benefits of their innovations, which implied that the rate of
private return to R&D is lower than its social return. Becker and Pain (2003)
emphasized that market failures can provide a rational for government
intervention to support private R&D. They mentioned that the expenditure on
R&D should be lower than social optimal level if the private rate of return is
lower than social rate of return, and if firms experience the significant external
financial constrains, the R&D expenditure will also be lower than social optimal
level. Streicher, Schibany and Gretzmacher (2004) claimed that pure markets will
not be efficient in stimulating innovation due to the inherent characteristics of
2

R&D. “In most situations the market will fail to provide sufficient incentives to
invest in R&D since firms face appropriability problems. The reason is that R&D
has some characteristics of a public good, so that the private returns on innovation
will be lower than its social return”.

Since the government R&D policy performed, lots of researchers have been


The remaining part of this thesis is structured as follows: The next section will
provide the theoretical background, and section 3 reviews the empirical literature
of the effectiveness of R&D policies. In section 4, the thesis will introduce the
economic background of the 13 OECD countries. Moreover, in the following two
sections, this thesis will describe the data set and the analytical model, and state
the empirical results. The final conclusion will be provided in the section 7.
2. THE THEORETICAL BACKGROUND
Historically, there are various policy tools available for government to stimulate
business-funded R&D. The empirical literature summarizes those policy tools into
three main categories:

Firstly, government can encourage business R&D activities through favorable tax
treatment. The R&D tax may increase R&D that is marginally profitable for the
firm, but only if the elasticity of R&D with respect to costs is high. Government
implements this policy tool to decrease firms’ R&D risk through tax breaks based
on the level of R&D expenditure. Currently, there are many forms of tax treatment
of R&D, like accelerated depreciation of investment, tax credit. In contrast to the
other R&D policy tools, this one is more transparent, and this policy tool is a more
market-oriented approach. The policy tool leaves decisions on the level and timing
of R&D expenditure to the private sector, and meanwhile, it gives a government
the option to pay for R&D that is otherwise not profitable at all for the firm, but
may be socially worthwhile. Guellec and van Pottelsberghe (2000) argued that tax
concessions were not conditional on the type of recipient’s R&D performance.
4

Therefore, tax incentives will not affect the R&D composition.

Secondly, government can directly fund business R&D through granting or/and
procuring private R&D projects. The previous mentioned Frascati Manual

will be complementary, which means increase the intensity of one will enhance
the other. However, empirically, those policy tools have been challenged by four
main grounds: full crowding-out effect, partial crowing-out effect, no influence,
and allocative distortion. Streicher, Schibany and Gretzmacher (2004) argued that
R&D expenditure and the reaction of R&D subsidies were the result of firms’
internal decisions. Therefore, government policy tools cannot (or only partially)
influence private R&D directly. The figure 1 is used to graphically describe part of
the five main effects on public support to business-funded R&D.
Figure 1. The effects of R&D subsidies on total R&D expenditures
(Source: Input Additionality Effects of R&D Subsidies in Austria. 2004)

Firstly, for crowding-out effects, the full crowding-out effects implies firms may
use government money as “windfall gains”. They just use that money simply to
substitute their own spending. Moreover, government spending may increase the
cost of R&D to crowd out private money. Goolsbee (1998) and David and Hall
6

(1999) both had been observed that government funding significantly raised the
wage of researchers. For instance, government funding may increase the salary of
researchers, although the total amount of R&D costs looks higher, but nothing
actually changes, and the real amount of R&D may be even lower than before.
The partial crowding-out effect means that firms may raise their R&D expenditure,
but less than the amount of government support.

Secondly, public support has no influences on private R&D occurs when firms
maintain the level of their R&D expenditures, but by use of full amount of the
subsidy extends total research. Because of firms would like to do more R&D than

spending which is in the neighborhood of unity in the short run, and
concluded the existence of positive relationship between tax credits and
private R&D spending. The author also argued that the R&D tax credits had
the intended effect, and although the high correlation over time of R&D
spending at the firm level makes it difficult to estimate long run effects
precisely, but the same high correlation makes it probable that these effects
are large. McCutcheon (1993) analyzed the response of the strategic groups
in the pharmaceutical industry to the credit. He formed four strategic groups
using different levels of research intensity and relative cash flow margin, and
concluded that tax credits stimulated firms’ competitive R&D expenditures in
the pharmaceutical industry. The author contended that a 1.6% R&D increase
was attributable to the credit. Hall and Van Reenen (1999) argued that the
effectiveness of fiscal incentives for R&D based on the tax system in OECD
countries on the user cost of R&D, and they concluded that a dollar in tax
credit for R&D stimulated a dollar of additional R&D. Bloom, Griffith and
Reenen (2000) examined the impact of fiscal incentives on the level of R&D
investment with an econometric model of R&D investment. They used a
8

panel of data on tax changes and R&D spending in 9 OECD countries from
1979 to 1996. They concluded that tax incentives were effective in increasing
R&D intensity. This is true even after allowing for permanent country
specific characteristics, world macro shocks and other policy influences.
They concluded that a 10% fall in the cost of R&D stimulates just over a 1%
rise in the level of R&D in the short-run, and just under a 10% rise in R&D
in the long-run. Guellec and van Pottelsberghe (2000) used a first-difference
auto-regressive model to analyze a panel data which collected from 17
OECD countries over 1983-1996. They concluded that tax incentives have
positive effect on business-financed R&D, The short-term (long-term) private
R&D elasticities is -0.29 (-0.33) for tax incentives. Mulkay and Mairesse

industry data
Firm level
McCutcheon
Tax credits stimulated firms’
competitive R&D expenditures, and
a 1.6% R&D increase was
attributable to the credit.
9

1999
26 OECD
countries’ data
Country level
Hall and Van
Reenen
A dollar in tax credit for R&D
stimulated a dollar of additional
R&D.
2000
9 OECD
countries data
Country level
Bloom,
Griffith and
Van Reenen
Tax incentives were effective in
increasing R&D intensity. A 10%
fall in the cost of R&D stimulates
just over a 1% rise in the level of
R&D in the short-run, and just

financed R&D. One dollar give to firms results in 1.70 dollars of private
research. However, the stimulating effect will increases up to a certain
threshold and then decreases beyond, which the threshold is about 13% of
business R&D. Becker and Pain (2003) estimated an econometric model of
R&D expenditure to analyze a panel of UK manufacturing industries. Their
results highlighted the importance of industry characteristics, and they found
that government funding appeared to play an important role on the total
industry R&D expenditures. 1 percentage point in the share of business R&D
expenditure funded by the government is estimated to raise the level of R&D
10

expenditure by 1.8%. Streicher, Schibany and Gretzmacher (2004) argued
that leverage effect of public subsidies to private R&D based on firm-level
data from the Austrian Industrial Research Promotion Fund. They concluded
that the public subsidies to private R&D have a crowding-in effect. With 1
additional euro of funding will induce firms to contribute an additional 40
cents of their own money. They also mentioned that very small and large
firms seem to exhibit higher leverage effect, but small and medium-sized
firms appeared to have smaller leverage effect. Lööf and Hesmati (2004)
investigated the effectiveness of a public innovation policy aimed at
stimulating private R&D investment at firm level. However, their result is
somewhat difference. They used the data from the Community Innovation
Survey (CIS) III for Sweden, and evaluated whether firms receiving public
funds have a higher R&D intensity on average compared to those not
receiving any such support. They concluded that there were additive effects
of public R&D financing on private research expenditure, but only for small
firms. Benavente (2003) compared and analyzed the Chilean manufacturing
firms under a situation that whether firms received or did not receive R&D
subsidies from 1995 to 1998. The author found a positive relationship
between public funding and private R&D. There was a 0.3 dollars

countries’ data
Country level
Guellec and
Van
Pottelsberghe
Direct government funding of
business R&D has a positive impact
on business financed R&D. One
dollar give to firms results in 1.70
dollars of private research.
However, the stimulating effect will
increases up to a certain threshold
and then decreases beyond, which
the threshold is about 13% of
business R&D.
2000
SBIR data
Firm level
Wallsten
Government grants crowded out
firm-financed spending R&D dollar
for dollar.
2003
11 UK
manufacturing
industries’ data
Industry level
Becker and
Pain
Government funding played an

Gretzmacher
will induce firms to contribute an
additional 40 cents of their own
money. Very small and large firms
seem to exhibit higher leverage
effect, but small and medium-sized
firms appeared to have smaller
leverage effect.
2004
CIS III data for
Sweden
Firm level
Lööf and
Hesmati
There were additive effects of
public R&D financing on private
research expenditure, but only for
small firms.
2009
Finnish data
Firm level
Einiö
Government funding induced
additional private R&D
expenditures. One subsidy euro
induced additional R&D worth at
least 1.5 euro

3.3 Empirical Literature Studies on Public Research
In 1996, Kealey argued that public research activities are irrelevant on

Public research activities are
irrelevant on business R&D
expenditure, and free commerce
will virtually and automatically
generate technological innovation
and economic growth even without
government intervention.
2000
17 OECD
countries’ data
Country level
Guellec and
Van
Pottelsberghe
Defence research performed in
public labs and universities crowds
out private financed R&D. The
short-term (long-term) private R&D
elasticities are -0.07(-0.08) for
government research and
-0.04(-0.05) for university search.
The negative effect of university
research is mitigated when
government funding of business
R&D increases.
2003
Chilean
manufacturing
firms’ data
Firm-level


Figure 2. Business R&D trends by area, 1993-2007
Japan
United States
Total OECD
EU27
0
100
200
300
400
500
600
1993
1995
1997
1999
2001
2003
2005
2007
15

(Source: OECD Science, Technology and Industry Scoreboard 2009)

From the figure 2 we can see that the business R&D expenditure has been
roughly keeping grow over 1997-2007. In this period, business R&D in
OECD-area grew by USD 160 billion (in PPP of 2000), in which United
States accounted for almost 40% of the growth, and Japan accounted for
around 20%. Among the EU27 countries, Portugal experienced strong growth

In 2008, France and Spain provided the largest subsidies and made no
distinction between large and small firms. Japan, United Kingdom and the
Netherlands appeared to be more generous for small firms. Finland and
Germany displayed negative tax subsidy rate
1
both for large firms and SMEs.
(The value of tax subsidy rate for R&D in 2008 is provide in appendix B)

The tax subsidies for R&D to large firms increased significantly
between 1999 and 2008 in France, and to a lesser extent in Italy, Portugal,
United Kingdom, Belgium and Japan. However, the OECD science,
technology and industry scoreboard 2009 indicated that Italy had been
experiencing the largest decreases in R&D tax subsidies for small and
medium-sized enterprises.

4.3 Changes of Government R&D Budget
For public research, government R&D budget data indicates the relative
importance in public R&D spending of various socioeconomic objectives,
such as defense, health and the environment. The following figure displayed
the government R&D budget changes in the 13 OECD countries over
1998-2008. 1
Citizens for tax justice (1996),”A negative tax rate indicates that rather than paying taxes, a corporation was able
to use excess deductions to get a refund for taxes paid in earlier years.”

17

a share of GDP are in Spain, Portugal and the United States. In 2008, defense
accounted for 57% of the total government R&D budget in the United States,
30% in France and 24% in the United Kingdom. However, Portugal, Spain,
Denmark and Finland spend the largest government R&D budget for civil
projects.
5. DATA AND METHODOLOGY
This thesis analyze annual data from 13 OECD countries over 1985-2007. In the
collected data, the values of business sector gross value-added were collected
from OECD STAN database in annual national accounts, and their units in US
dollars for current prices and current PPPs. The B-index was picked from the
existing literatures respectively. The appendix C indicates the trends of B-index
from 1985 to 2007. All the rest of the data were collected from eurostat database
which with the unit of Millions of euro (from 1.1.1999)/Millions of ECU (up to
31.12.1998).

As previous mentioned, this thesis implement a previous analytical model
published by Gullec and van Pottelsberghe (2000). Their R&D investment model
considered business-funded R&D as a function of output, government funded
R&D performed by business, tax incentives, public research, time dummies, and
country-specific fixed effects. In the model, government funding of business R&D
is composed of procurement and grants or subsidies, and public research is broken
down into two components, which are government research and university
research. The model can be written as follows:

ΔRPi,t = λΔRPi,t-1 + βVAΔVAi,t + βRGΔRGi,t-1 + βBΔBi,t-1 + βGOVΔGOVi,t-1 +
βHEΔHEi,t-1 + τt +ｅi,t (1)

This equation is a first-difference auto-regressive model, where RP is
19

advantage of Extended Instrument Variable estimation. The signs of the

2
The motivation of first-difference operator is that it controls for all country differences that are constant over
the observation period
3
Hall (1992) argued three reasons for R&D investments are subjects to important adjustment costs. First,
compared with ordinary investment, R&D tends to have a low variance. Second, at least half of R&D
investment consists of payments to scientists and engineers who embody the firm’s stock of knowledge and
contribute to its increase. At last, R&D investment usually needs time to contribute profits, and it is costly to
stop.
20

parameters RG, B, GOV, and HE can be either positive or negative, depending on
whether the stimulating and spillover effects outweigh the crowding-out and
crowding-in effect.
6. EMPIRICAL RESULTS
In the data analysis, except the analysis of United Kingdom at country-level study,
all the rest of analysis performed Extended Instrument Variable (EIV) estimation
under GMM framework. The validity of the regression results is confirmed using
Anderson canon. corr. LR statistic (identification test), Hansen J statistic
(overidentification test), Pagan-Hall general test statistic (heteroskedasticity test),
and Pagan-Hall test w/assumed normality (heteroskedasticity test). In the analysis
of United Kingdom at country-level study, the regression technique performed
Extended Instrument Variable only, because of the insufficient observations. But
except the overidentification test performed by Sargan statistic in the analysis of
United Kingdom, all the rest of tests are remain same as the other analysis.
Furthermore, in the country-level analysis, country dummies and time dummies
are both excluded to fix the colinearity issues.


R&D
expenditure
ΔGOV
Higher
education
R&D outlays
ΔHE
T
0.6414268
-0.0009006
0.1198406
0.2558809
0.4112655

(0.044)**
(0.973)
(0.355)
(0.000)***
(0.000)***
T-1
0.6703866
0.0043491
-0.204128
-0.0265447
0.0934961

(0.074)*
(0.884)
(0.154)
(0.757)

-0.1152112

(0.107)
(0.430)
(0.034)**
(0.186)
(0.114)
T-5
-0.0269330
-0.0595957
-0.3628577
-0.1539247
-0.1571469

(0.940)
(0.007)***
(0.379)
(0.005)***
(0.077)*
T-6
-0.9174692
-0.0358072
-0.4054725
-0.1134951
-0.0122706

(0.010)**
(0.320)
(0.170)
(0.163)

0.0183271

(0.849)
(0.523)
(0.097)*
(0.390)
(0.768)
Note: The estimates cover 13 OECD countries over 1985-2007, each variable are
expressed in first differences of logarithms. In the table, the values displayed in
parentheses are p-values. The notation of “*”, “**”, and “***” separately
indicates the estimates that are significantly different from zero at 10%, 5%, and
1%.

The table 4 analyzed the effectiveness of 13 OECD countries’ R&D policy
instruments for nine-year lag to try to capture longer term effects of basic research.
As seen from table 4, business sector value-added has its contemporaneous impact
which with an elasticity of about 0.64. But this impact becomes negative after few