Firm-Level Evidence on Innovation in Spain

Abstract

Using a new database based on a survey on innovation this paper aims at iden tifying the determinants of firms' R&D investment and innovation decisions. The descriptive statistics confirm the evidence that the percentage of firms investing in R&D is substantially lower than the percentage of firms innovating. The econometric analysis allows toidentify the geographic distribution of sales as an important determinant of both R&D investment and innovation. Parallel session 7. FDI, spillovers and innovation Presented at GLOBELICS 2009, 7th International Conference, 6-8 October, Dakar, Senegal.

Full text

Page 1

Firm-Level Evidence on Innovation in Spain∗
Lionel Artige
HEC - Universit´ e de Li` ege
Rosella Nicolini
Universitat Aut` onoma de Barcelona
September 15, 2009
Abstract
Using a new database based on a survey on innovation this paper aims at iden-
tifying the determinants of firms’ R&D investment and innovation decisions. The
descriptive statistics confirm the evidence that the percentage of firms investing in
R&D is substantially lower than the percentage of firms innovating. The economet-
ric analysis allows to identify the geographic distribution of sales as an important
determinant of both R&D investment and innovation.
Keywords: innovation, investment, microeconometrics
JEL Classification numbers: D21, 031.
∗We thank Bernard Lejeune for valuable comments. Responsibility for all errors is our own. Address
for correspondance: HEC - Universit´ e de Li` ege, Economics department, Boulevard du Rectorat,7, B31,
4000 Li` ege, Belgium. Email: lionel.artige@ulg.ac.be

Page 2

1Introduction
On the theoretical and policy points of view, innovation has become a central theme for
growth at the level of the firm and for the economy as a whole. In growth theory innovation
had not always been recognized as an important engine of growth because innovation did
not fit in the competitive general equilibrium framework. Schumpeter was one of the
rare economists at the time to consider innovation at the center of economic activity but
his ideas remained for long out of mainstream economics. Two decades ago endogenous
growth theory proposed new growth models integrating the old ideas of Schumpeter and
considered innovations, such as the introduction of new goods, new production processes
or qualitative changes in assets, as a necessary condition for long-term growth (Romer
(1990), Lucas (1988) or Aghion and Howitt (1992)). However, the enormous development
of the theoretical literature on this subject still remains unmatched by the development
of empirical evidence. The main reasons are the difficulties in defining innovation and
measuring innovative inputs and outputs (Rogers 1998). In the last twenty years surveys
at the firm level have been suggested as a complementary method to the R&D and patents
data to identify other potential measures of innovative inputs and outputs. Despite all
the legitimate caveats on the accuracy and objectivity of qualitative data surveys have
contributed to investigating innovation activities of firms that do not invest in R&D such
as small firms or even a sector like the services which accounts for a large part of industrial
countries’ economic activity.
The present paper adopts the survey method to study the determinants of innovation
at the firm level. A questionnaire was sent to more than 26 000 firms of more than 10
employees in Spain in 2006 and 2007. This survey resulted in a usable random sam-
ple of 516 enterprises of all sectors. Traditionally empirical studies have focused on the
manufacturing sector. Even though this paper’s sample contains a large fraction of man-
ufacturing firms it also include firms of other sectors. The main questions were whether
the enterprises realized R&D investment in 2005 and introduced innovations in 2004 or
2005. In the survey, innovation was defined as ”the implementation of a new or signif-
icantly improved product (good or service), or process, a new marketing method, or a
new organisational method in business practices, workplace organisation or external re-
lations.” (OECD (2005), p.46) This definition is broad and generally lead to a high rate
of innovating firms in empirical studies as is the case in this paper. However, the use
of R&D or patents data discards too many firms which do not need them to introduce
innovations such as enterprises in services.
Although innovation has been difficult to define and measure some stylized evidence have
been produced in the empirical literature. The frequent cross-section result is the pos-
itive relationship between productivity and R&D activity at the firm level (Hall (1996)
and Griliches (1998)). However, this result is not statistically significant in longitudinal
analyses. The role of firm size has been also very much investigated. It seems that R&D
expenditure per employee is independent of enterprise size but large firms are reported
to display a more than proportionally R&D effort (Cohen (1995) and Cohen and Klep-
1

Page 3

per (1996)). Griliches (1990) notes that the latter result could be influenced by the fact
that small firms tend not to report R&D investment as they realize informal R&D. Acs
and Audretsch (1988) argue that small and large firms respond differently to innovation
determinants. Another important innovation determinant that will not be covered in this
paper due to the lack of data is market structure. The empirical studies have produced
mixed results. Theoretically, strong competition may incite firms to innovate to gain a
competitive advantage but competition tends to reduce market power that is necessary
to invest in R&D. The Schumpeterian hypothesis states that monopolistic competition is
needed for firms to appropriate the returns of their innovation investments. The empiri-
cal studies have not produced a consensus over this issue (Cohen (1995) and Symeonidis
(1996)). The relationship between competition and market structure could be in fact in
the shape of an inverted-U (Aghion, Bloom, Blundell, Griffith, and Howitt 2005).
The objective of this section is to identify the relevant variables of our database that are
likely to account for the probability for an enterprise to invest in R&D and to introduce
new products. This objective is motivated by two facts. First, it is difficult to measure
innovation in terms of output units and evaluate its impact on enterprise performance.
Thus, we do not know whether innovating enterprises grow faster than non-innovating
enterprises or have a higher probability of surviving. Second, no all firms declare innovat-
ing and some claim that they need not innovate to make profits. Then, why to innovate?
Both facts incite the investigator to look into variables that may affect firm behavior in
order to identify innovating and non-innovating profiles by their determinants.
The paper is organized as follows. Section 2 presents the survey and the data. Some
descriptive statistics are discussed in Section 3. Section 4 analyzes the econometric results.
Finally, section 5 concludes.
2Survey and data
Our sample consists of 516 Spanish enterprises that responded to a survey we realized
in Spain in 2006-2007. Although there are existing data coming from Spanish surveys
on innovation such as the Encuesta sobre las Estrategias Empresariales (EEE) of the
Spanish statistical institute (INE) or the Encuesta sobre Innovaci´ on Tecnol´ ogica of the
SEPI foundation, we conducted our own survey to focus as much on non-innovating as
innovating enterprises and include information that is not covered by the existing surveys.
Our survey was sent electronically to all the Spanish enterprises of the Kompass database
(around 26,600 enterprises with at least 10 employees) from November 2006 until April
2007. The Kompass database is a random sample of all the Spanish enterprises that have
at least one enterprise as a client or as a supplier. All enterprises of the Kompass sample
had an email address and thus received the survey. These enterprises had one month to
voluntarily respond to the survey on a dedicated website on the internet. At the end
of the process, the survey resulted in a cross-section dataset of 598 enterprises from all
sectors. After removing the enterprises that responded partially to the questionnaire, we
ended up with the sample of 516 enterprises.
2

Page 4

Due to the selection method of our sample (random sampling), we used the INE’s pop-
ulation database on Spanish firms (DIRCE database) and stratified sample focusing on
innovation (EEE), depending on the variable under study, to identify potential biases of
our sample (S1). We performed the Wilcoxon test for three characteristics (firm size,
geographical location of firms and sectors of activity) to check whether the hypothesis
of equality between the distribution of our sample and those of the INE’s population
and stratified sample could be statistically rejected. The results of the test show that
the equality of distributions cannot be statistically rejected for all three characteristics
(Tables 1, 2 and 3).
Table 1: Distribution of Spanish firms by size in the population (DIRCE database) and
our sample (S1) and Wilcoxon test
Number of employeesDIRCE database (%)Sample S1 (%)
[10 − 50[
[50 − 200[
[200 − 500[
[500 − 1000[
> 1000
85.4
11.7
2.1
0.5
0.4
68.7
22.9
6.0
1.7
0.7
Wilcoxon signed-rank test
H0: Var(DIRCE) = Var(S1)
z = −0.674 Prob > |z| = 0.5002
Source: authors’ database
Variables
3 Descriptive statistics
Our sample include 598 Spanish enterprises localized throughout continental Spain, in the
Balearic islands and in the Canary islands. Among these firms, 6% belonged to multi-
national enterprises at the date of the survey and 22% had at least another geographical
location beyond the headquarters, either in Spain or abroad.
3

Page 5

Table 2: Distribution of Spanish firms by region in the population (DIRCE database) and
our sample (S1) and Wilcoxon test
Region in Spain DIRCE database (%)Sample S1 (%)
La Rioja
Cantabria
Extremedura
Navarra
Asturias
Baleares
Arag´ on
Murcia
Castilla-La Mancha
Castilla y Le´ on
Canarias
Galicia
Pa´ ıs Vasco
Comunidad Valenciana
Andaluc´ ıa
Madrid
Catalua
0.8
1.1
1.5
1.6
1.8
2.5
2.9
3.3
3.6
4.2
4.3
5.2
5.6
11.7
14.1
15.6
20.2
1.8
0.7
1.7
2.0
2.0
2.5
4.7
2.2
3.0
4.5
0.7
5.2
12.7
12.7
9.7
10.4
27.0
Wilcoxon signed-rank test
H0: Var(DIRCE) = Var(S1)
z = −0.142 Prob > |z| = 0.8869
Source: authors’ database
3.1Enterprises and R&D
Among the 598 firms that responded to the survey, 58.5% declared having invested in
R&D in 2005 while 41.5% declared having not. Some summary statistics on the whole
sample and on the two subsamples (enterprises that did invest in R&D and enterprises
that did not invest in R&D in 2005) are presented in Tables 4, 5 and 6. The mean year of
foundation is 1978 for the whole sample and is slightly more recent for the group of firms
without R&D. However, the median year (1982) is exactly the same for both subsamples
(Table 4). The descriptive statistics on the distribution of enterprises by labor force size
show a clear difference between both groups of firms. The mean or the median size of
firms investing in R&D is much larger than those of firms without R&D investment (Table
5). This result is in line with all empirical studies on this issue. Other statistics stress the
4

Page 6

Table 3: Distribution of Spanish firms by sector in the population (DIRCE database) and
our sample (S1) and Wilcoxon test
Region in SpainDIRCE database (%) Sample S1 (%)
Mining and quarrying (NACE 10-14)
Manufacturing (NACE 15-37)
Electricity, gas and water (NACE 40-41)
Construction (NACE 45)
Wholesale, retail trade and accomodation (NACE 50-55)
Transportation (NACE 60-63)
Financial and insurance activities (NACE 65-67)
Computing (NACE 72)
Scientific research and development (NACE 73)
Other service activities (NACE 71, 74)
Public administration (NACE 80, 85, 90-95)
0.5
35.2
0.5
16.4
21.1
4.2
1.4
2.2
0.4
10.8
7.3
1.7
75.2
1.2
3.6
7.4
1.7
0.0
2.4
1.4
4.1
1.2
Wilcoxon signed-rank test
H0: Var(DIRCE) = Var(S1)
z = −0.142Prob > |z| = 0.8869
Source: authors’ database
relevance of R&D as a criterion to identify firm behavior. Enterprises investing in R&D
export much more and invest more in physical and human capital than enterprises that do
not invest in R&D (Table 6). More than two-third of the enterprises that invest in R&D
export part of their production and less than 10% of them sell more than 50% in the local
market. For the enterprises that do not invest in R&D, the corresponding figures are less
than 50% and 36%. Another distinctive feature between both subgroups of enterprises
is the distribution of qualifications in the labor force. The enterprises investing in R&D
hire three times more PhD graduates and have a more qualified labor force than the
other subgroup. Finally, the last three paragraphs of Table 6 are devoted to innovation.1
The enterprises of our sample were asked whether they introduced new products or/and
new production process or/and new organization methods in the last two years from
the date of the survey. The answers to these questions must be treated with care since
respondents may not be accurate or objective. In fact, innovation possesses a qualitative
dimension for which there is no objective way to identify it. Due to this qualitative
nature, innovation has been hard to measure. The traditional measurement method has
relied on quantities such as R&D expenditures or patents but they are innovative inputs
rather than innovative outputs. Even if innovative output must be correlated in some
way with innovative inputs, the uncertain result of an innovation process should make us
cautious about the interpretation of measures of innovative inputs. Moreover, enterprises
can introduce innovations without spending on R&D and patents. They are likely small
or incremental innovations but can be as profitable as breakthrough innovations for the
1Innovation is defined as ”the implementation of a new or significantly improved product (good or
service), or process, a new marketing method, or a new organisational method in business practices,
workplace organisation or external relations.” (OECD 2005, p. 46)
5

Page 7

Table 4: Descriptive statistics on the distribution of firms by age (year of foundation)
All firmsFirms with R&D
Mean1978
Median1982
Maximum 2006
Minimum 1858
Standard deviation20.9
Bias-2.0
Kurtosis9.2
Jarque-Bera test 1289.9
Prob0.00
Observations562
Firms without R&D
1980
1982
2006
1911
16.6
-1.4
6.0
144.7
0.00
203
1976
1982
2006
1858
23.8
-2.1
8.9
636.9
0.00
288
Source: authors’ database
Table 5: Descriptive statistics on the distribution of firms by size
All firmsFirms with R&D
Mean90.3
Median30
Maximum4340
Minimum 10
Standard deviation 308.4
Bias 10.4
Kurtosis 123.0
Jarque-Bera test329434.5
Prob0.00
Observations533
Firms without R&D
42.5
24.5
800
10
76.0
7.1
62.7
32947.9
0.00
210
125.7
40
4340
10
403.2
8.0
72.8
63585.4
0.00
297
Source: authors’ database
individual firm.2The large spectrum of innovations has led the scientific community to use
survey to identify innovative and non-innovative firms. This is the method we adopted for
this work, adding qualitative information to quantitative data. Table 6 summarizes some
of this information. The results show that 56% of the firms having declared not investing
in R&D nonetheless declared having introduced new goods or services in 2004 or in 2005.
The figure is even higher for the organization methods (60%) and the production processes
(62%). These results tend to confirm that innovation can occur without spending on R&D.
Another explanation for these results is given by Griliches (1990) who stresses that small
firms, due to financial or organizational constraints, realize informal R&D that is not easy
to account for quantitatively and therefore do not report R&D expenditures. Although a
majority of our sample’s firms which do not report R&D spending declares having made
innovations, the frequencies are still substantially lower than for the firms investing in
R&D.
2The impact of an innovation on an individual firm should be distinguished from the global impact.
A firm may introduce an innovation that improves its market share without affecting the technological
frontier of the entire economy.
6

Page 8

Table 6: Descriptive statistics on enterprises and R&D
Firms without R&D
(%)
47.2
9.3
53.7
36.0
Firms with R&D
(%)
70.3
27.7
29.7
7.6
Exporters
Exports > 30% of total sales
100% national sales
Local sales > 50% of total sales
Gross Fixed Capital Formation (GFCF) in 2005
GFCF in 2005 but not in 2002-2005
GFCF in 2002-2005
57.9
3.3
72.4
72.6
1.3
78.5
Investment type:
CAD
Control instruments
Robotic
Flexible production systems
Other technology (internet)
30.4
43.5
14.0
34.6
50.5
48.8
67.0
31.4
57.4
62.7
Human capital:
Enterprises with PhD graduates
Enterprises with university graduates
Enterprises with more than 10% of University graduates
4.7
65.0
26.2
15.2
81.2
35.0
Training:
Enterprises that organized training
Computing training
Language training
Technical training
77.1
55.6
27.6
38.3
92.7
75.6
58.7
70.0
Enterprises that introduced new goods and services (NGS)
NGS realized by the enterprise
NGS realized in collaboration
NGS Realized by a provider
56.1
32.2
11.2
12.6
87.5
58.7
23.6
5.6
Enterprises that introduced new production processes (NPP)
NPP realized by the enterprise
NPP realized in collaboration
NPP Realized by a supplier
62.6
42.5
15.9
9.3
87.1
61.7
22.4
3.0
Enterprises that introduced new firm organization (NFO)
NFO realized by the enterprise
NFO realized in collaboration
NFO realized by a provider
58.9
42.5
13.6
2.8
80.2
59.4
18.5
2.3
Source: authors’ database
7

Page 9

3.1.1Enterprises investing in R&D
Focusing on the subgroup of firms investing in R&D, some summary statistics allow to
identify interesting features of their behavior (Table 7). Most expenditure in R&D aims
at developing new products (60% of the firms) and production processes to a lesser extent
(30%). The clear expected benefit of their R&D investment is an increase in productivity
for 70% of the firms and an improvement in product quality for 17%. Our data show that
R&D investment is a risky business since only 25% of the firms declare that it is always
successful. Among these R&D investments yielding successful innovations, 30% of the
firms decide to patent them. Regarding the financing of R&D investments, a very large
majority of firms use internal financial resources and 70% of them use their own cash to
finance more than 50% of their R&D activity. A non-negligible fraction of firms (40%)
use subsidies as external finance but for a marginal share of their R&D expenditures.
Very few firms use banking credit but our survey does not allow to know whether the
firms did not obtain banking credit or did not want to use banking credit to finance R&D
investments. Another interesting information is the substantial fraction of firms (45%)
that declare collaborating with other firms in their R&D activity. These firms may be
joint ventures, suppliers or consultancy firms.
3.1.2Enterprises that do not invest in R&D
Regarding the subgroup of firms that did not invest in R&D in 2005, 80% of them declared
that they even never made R&D investment in the past (Table 8). Only 26% declare to
plan to do it in the future. This gives some interesting information about firm behavior
with respect to R&D activity: either the enterprise invest in R&D continuously (see Table
7) or never. The issue is then to identify the variables likely to explain this observed
contrast in firm behavior. The descriptive statistics of Table 8 provide some insights
about the factors that seem to explain why some firms decide not to invest in R&D. The
main reasons cited by these firms are the large costs of R&D for 40% of them, the absence
of market incentives to engage in R&D for 39%, the lack of qualified workers for 30% and
the lack of financial resources for 25% of them. The lack of market incentives as one of
the main factors for not investing in R&D is corroborated by the answer to the question
whether R&D is related to firm’s competitiveness. Only 20% of these firms agree with
the statement that a lack of R&D implies a lack of competitiveness.
3.2Enterprises and innovation
The enterprises of the survey were asked to answer the question whether they introduced
in the last two years each of the three main types of innovations (product innovation,
innovation in production process in innovation in organization methods). The frequencies
of responses are presented in Table (9). Three main results may be highlighted. First, the
percentage of innovating firms is large which is a confirmation of the existing literature
8

Page 10

Table 7: Descriptive statistics on enterprises investing in R&D
%
R&D type:
production processes
goods and services
organization
31.4
59.7
6.6
Expected benefits:
increase in productivity
improvement in product quality
cost reduction
reduction in labor cost
reduction in energy use
improvement in environmental impact
70.3
17.2
4.0
2.0
0.3
2.0
Realization of R&D activity:
continuously
intends to invest in the future
74.9
93.7
Result of R&D activity:
always positive
almost always
sometimes positive
almost never positive
25.4
49.2
20.5
1.7
Positive results that lead to patents28.7
Labor force in R&D:
enterprises with researchers (PhD)
enterprises with more than 30% of PhD in R&D
enterprises with technicians
enterprises with research assistants
32.7
18.2
78.2
44.9
R&D financing:
Internal resources
more than 50% from internal resources
External financing
more than 10% from external resources
more than 30% from external resources
more than 50% from external resources
Subsidies
more than 10% from subsidies
more than 30% from subsidies
more than 50% from subsidies
84.8
68.0
20.1
18.8
11.2
6.9
40.3%
31.0
14.9
5.3
Enterprises that used following financial support:
subsidies
credit
desgravaciones
capital participation
public institutions’ support
awards
no support
55.4
7.3
5.9
0.3
0.3
0.3
27.4
Source of financial support:
local government
national Government
European Union
72.6
18.5
5.6
Exchange of information with:
other enterprises
customers
suppliers
competitors
experts
labs
universities
public institutions
20.1
38.0
4.6
1.7
15.5
3.6
8.6
5.0
R&D activity with other enterprises45.5
Source: authors’ database
9

Page 11

Table 8: Descriptive statistics on enterprises that do not invest in R&D
%
Enterprises did not invest in R&D in 2005 because:
they did not plan to invest in R&D that year
they did invest in R&D in the past
they plan to do it in the future
30.4
19.6
26.6
Reasons why enterprises did not invest in R&D:
business risk
cost
lack of financial resources
organization difficulties
lack of qualified labor
lack of technological information
lack of market information
lack of flexibility
lack of market demand
market conditions do not require R&D
old R&D still good enough
70.3
40.2
24.8
22.0
29.9
15.4
12.1
12.6
24.3
38.8
15.4
Lack of R&D implies lack of competitiveness 20.1
Source: authors’ database
and the fact that innovation is generally viewed in a broad sense. This means that a
sizeable fraction of firms (56%) declares that they innovate while declaring that they do
not invest in R&D. A firm may thus innovate without investing in R&D. Second, the
frequency of firms innovating is much higher when they invest in R&D rather than not.
Finally, the decision to innovate does not seem very sensitive to firm size as it is the case
with the decision to invest in R&D.
3.3 Descriptive statistics by firm size
We already mentioned that the R&D investment decision depends a lot on firm size as
documented by the literature. The larger the enterprise, the higher the probability that
this enterprise will engage in R&D investment. This result is confirmed by the frequency
rates shown in Table (10). As the firm size increases, the percentage of firm declaring R&D
investment in 2005 also increases. A slight majority of firms with less than 50 employees
are R&D investors while 90% of large firms are. However, firm behavior characterized
by the few questions that were asked is strikingly invariant across group sizes. For firms
investing in R&D, the percentage of firms relying on internal resources to finance R&D
investment is very similar across group sizes. As for the enterprises that do not invest
in R&D, their behavior and the reasons mentioned by them are also very similar across
group sizes. One exception is the percentage of firms mentioning the business risk as an
obstacle to R&D investment. This reason is cited by 26% of the firms with less than 50
employees and only 13% of the firms hiring between 50 and 250 employees. ?????????%
of fimrs that never invested in R&D:
10

Page 12

Table 9: Descriptive statistics on innovation by firm size
All firms Firms with R&DFirms without R&D
Firms having introduced new products (%)
Firms < 50 employees
Firms [50,250] employees
Firms > 250 employees
All firm sizes
72.4 (348)
76.3 (127)
87.0 (31)
74.6 (516)
87.7 (179)
86.5 (89)
89.3 (28)
87.7 (302)
56.2 (169)
52.6 (38)
66.6 (3)
56.0 (214)
Firms having introduced new production processes (%)
Firms < 50 employees
Firms [50,250] employees
Firms > 250 employees
All firm sizes
74.7
81.1
93.5
77.1
85.4
88.7
96.4
87.4
63.3
63.1
66.6
62.6
Firms having introduced new organization methods (%)
Firms < 50 employees
Firms [50,250] employees
Firms > 250 employees
All firm sizes
70.6
72.4
80.6
71.5
81.0
78.6
85.7
80.4
59.7
57.9
33.3
58.8
(..) Total number of observations for each firm size in the database
Table 10: Descriptive statistics on enterprises by firm size
All firmsFirmsFirms Firms
< 50 employees [50,250] employees> 250 employees
Enterprises and R&D:
Enterprises that invested in R&D in 2005
(% of each group size’s observations)
58.5% 51.5%70.0%90.3%
Enterprises that did not invest in R&D in 2005
(% of each group size’s observations)
because:
they did not plan to invest in R&D that year
they did invest in R&D in the past
they plan to do it in the future
41.5%48.5%30%9.7%
30.4%
19.6%
26.6%
33%
16.5%
26%
18%
21%
31%
n. a.*
n.a.
n. a.
R&D Financing:
internal resources
more than 50% from internal resources
84.8%
68.0%
85%
67%
88%
68%
78%
71%
Reasons for not investing in R&D
business risk
cost
market conditions do not require R&D
old R&D still good enough
Observations
23.8%
40.2%
38.8%
15.4%
516
26%
42%
40%
17%
348
13%
37%
37%
9%
127
n. a.
n. a.
n. a.
n. a.
31
* Not applicable
Source: authors’ database
11

Page 13

Table 11: Descriptive statistics on enterprises by sector
All firms(1)(2)(3)(4) (5)(6)(7)
Enterprises and in R&D:
Enterprises that invested in R&D in 200558.5%59.7%64.3% 76.5%55.9% 66.6%44.1%62.8%
Enterprises that did not invest in R&D in 2005
because:
they did not plan to invest in R&D that year
they did invest in R&D in the past
they plan to do it in the future
41.5%40.3%35.7%23.5%44.1% 33.3%55.9% 37.2%
30.4%
19.6%
26.6%
36%
14.5%
30%
n.a.*
n.a.
n.a.
n.a.
n.a.
n.a.
40%
11.1%.
27%
n.a.
n.a.
n.a.
n.a.
n.a.
n.a.
n.a.
n.a.
n.a.
R&D Financing:
internal resources
more than 50% from internal resources
84.8%
68.0%
88%
72%
89%
74%
92%
80%
88%
72%
92%
77%
73.1%
72.7%
83%
52%
Reasons for not investing in R&D
business risk
cost
market conditions do not require R&D
old R&D still good enough
Observations
23.8%
40.2%
38.8%
15.4%
26%
46%
40%
19%
390
n.a.
n.a.
n.a.
n.a.
27
n.a.
n.a.
n.a.
n.a.
51
31%
53%
22%
18%
102
n.a.
n.a.
n.a.
n.a.
39
n.a.
n.a.
n.a.
n.a.
34
n.a.
n.a.
n.a.
n.a.
43
* Not applicable
Source: authors’ database
(1) Manufacturing (1993 NACE: 15-37)
(2) Food (1993 NACE: 15-16)
(3) Chemical industry (1993 NACE: 24-25)
(4) Metallurgy (1993 NACE: 27-28)
(5) Machinery (1993 NACE: 29)
(6) Wholesale, retail and accomodation (1993 NACE: 50-55)
(7) Services to enterprises (1993 NACE: 71-74)
12

Page 14

4Econometric analysis
The objective of this section is to identify the relevant variables of our database that are
likely to account for the probability for an enterprise to invest in R&D and to introduce
new products. This objective is motivated by two facts. First, it is difficult to measure
innovation in terms of output units and evaluate its impact on enterprise performance.
Thus, we do not know whether innovating enterprises grow faster than non-innovating
enterprises or have a higher probability of surviving. Second, no all firms declare innovat-
ing and some claim that they need not innovate to make profits. Then, why to innovate?
Both facts incite the investigator to look into variables that may affect firm behavior in
order to identify innovating and non-innovating profiles by their determinants.
4.1The econometric model
Innovation is widely considered as an important engine of output growth even though it
is tricky to identify and measure its effect on enterprise performance. Formally, at the
firm level, it is considered that
dy(t)
dt
=d
dt
?n
1
fi,t[z(t)]di,(1)
wheredy(t)
i at time t, z(t) is a vector of inputs and n ? 1 is the number of varieties produced by the
firm. The innovation may occur in the number of varieties produced, n, in the substitution
of one variety by another one (fjreplacing fi) or in the improvement in the technology
(fi,t=1(.) > fi,t=0(.)). Even though the costs of innovations may be evaluated, the return
of innovations are difficult to identify especially if they occur in the various possible areas,
if they are incremental and if they yield returns in the long run. Due to these difficulties,
researchers have focused on z(t), i.e. on measurable variables supposedly associated with
innovation such as R&D investment and patents. The latter can be considered as both
an innovation input and output while the former is an innovation input. The drawback
of both indicators is its limited scope. No all innovating firms invest in R&D and even
less patent their inventions. Moreover, they may signal innovation behavior but they are
not necessarily proxies for innovation output. In fact, investments realized by firms to
innovate may be unsuccessful. The uncertainty and the innovation costs may explain
why some firms renounce to innovation. Another way to study innovation is by using
surveys to collect qualitative information on firm behavior with respect to innovation.
The drawback of this method is the subjectivity of respondents and the too large number
of firms declaring innovating. In this paper, we use both types of indicators: R&D
investment decision and innovation decision. Formerly, we assume that
dtis the real output growth rate of the firm, fi,tis the production function of good
I(t) = g[x(t)], (2)
13

Page 15

where I(t) ? 0 is the innovation output, x(t) is a vector of factors including innovation
inputs and other relevant variables and g(.) is a function relating these innovation factors
and the innovation output. The firm will decide to introduce I(t) on the market if it
expects a positive return. Formally, I(t) > 0 if
E[π(t)] = E?y(t)[I(t)]?− C(t) ? 0, (3)
where E[π(t)] is the expected profit of the firm, E?y(t)[I(t)]?
firm can be interpreted as the expected enterprise performance or its survival probability.
The expected profit may depend on different factors, such as firm characteristics, market
structure, or sectoral idiosyncracies. As we already mentioned, it is hard to measure this
expected profit which will be considered here as the latent variable. If we consider the
innovation decision as a dichotomous variable, the decision rule can be defined as
is the expected sales of
the innovative output and C(t) is the cost of production. The expected profit of the
I(t) = 1 ⇒dE[π(t)]
dX
> 0 (4)
I(t) = 0 ⇒dE[π(t)]
dX
? 0. (5)
The objective of this paper is to identify the vector of variables X, if any, in order
to account for firm’s innovation behavior. The general econometric model can thus be
written as
I = Xβ + ?,(6)
where I is the firm’s decision on R&D investment or on the introduction of a new product,
i.e., a binary variable that take the value of 1 if the decision appears to be yes in the
survey and 0 otherwise; The matrix X represents the vector of explanatory variables that
we selected from the ones available in the database; and β is the vector of the coefficients
and ? is an error term i.i.d. following a normal distribution with zero mean and variance
equal to σ2.
Both dependent variables of this exercise are dichotomous qualitative variables and are
generally estimated with the maximum likelihood method. For both types of estimation
we will use the logit model implying the application of a logistic transformation of the
linear specification Xβ.
4.2Estimation results
4.2.1Variables
The econometric analysis consists in estimating the conditional probability of innovating
in 2004 or in 2005 and the conditional probability of investing in R&D in 2005. The
14