CASE Network Studies and Analyses 410 - The Impact of innovation activities on firm performance using a multi-stage model: evidence from the Community Innovation Survey 4

CASE Network Studies & Analyses No.410 The impact of innovation activities on firm perfo…
Materials published here have a working paper character. They can be subject to further
publication. The views and opinions expressed here reflect the author(s) point of view and
not necessarily those of CASE Network.
This paper was produced in the framework of the international research project 'MICRODYN-The
competitiveness of firms, regions and industries in the knowledge-based economy: What
room for job-rich growth in Europe?' (www.micro-dyn.eu), coordinated by the Vienna
Institute for International Economic Studies (WIIW) and funded by the European
Commission's Framework VI Programme, Grant No. 028868(CIT4).
Keywords: innovation, firm performance, multi-stage model, Community Innovation Survey 4
JEL Codes: O310, L25
© CASE – Center for Social and Economic Research, Warsaw, 2010
Graphic Design: Agnieszka Natalia Bury
EAN 9788371785207
Publisher:
CASE-Center for Social and Economic Research on behalf of CASE Network
12 Sienkiewicza, 00-010 Warsaw, Poland
tel.: (48 22) 622 66 27, 828 61 33, fax: (48 22) 828 60 69
e-mail: case@case-research.eu
http://www.case-research.eu
1

2
Contents
Abstract...................................................................................................................................3
1. Introduction ........................................................................................................................5
2. Theoretical framework .......................................................................................................6
3. Literature review.................................................................................................................8
4. Data....................................................................................................................................14
5. Model specification ..........................................................................................................16
5.1. General specification of the system model.............................................................16
5.2. Definition of variables and specification of the model ..........................................18
6. Interpretation of findings.................................................................................................20
6.1. Decision to innovate..................................................................................................20
6.2. Innovation investment...............................................................................................22
6.3. Innovation output ......................................................................................................23
6.4. Productivity (performance) and innovation output ................................................26
7. Sensitivity analysis ..........................................................................................................28
8. Conclusion........................................................................................................................31
References............................................................................................................................33
Appendix...............................................................................................................................37

Iraj Hashi is a CASE Fellow and Professor of Economics at Staffordshire University, United
Kingdom. He is also the Director of the Centre for Research on Emerging Economics at
Staffordshire University Business School. His research has focused on microeconomic
problems of transition to a market economy in Central and Eastern Europe, especially in
areas such as: privatisation, corporate governance, enterprise restructuring,
competitiveness, bankruptcy and reorganisation, employee financial participation and
competition policy. He has worked with research and consultancy institutions in Albania,
Bosnia-Herzegovina, Croatia, Czech Republic, Kosovo, Macedonia, Poland, Romania and
Russia.
Nebojsa Stojcic is an Assistant in the Department of Economics, University of Dubrovnik,
Croatia where he obtained his Diploma in Economics. He is currently completing his PhD on
the impact of restructuring on the competitiveness of firms in transition economies at
Staffordshire University, United Kingdom. Nebojsa has worked on several research projects
on sustainable development, technology transfer, enterprise restructuring and innovations.
His main research interests are competitiveness, industrial economics, firm behaviour in the
process of transition, innovations and agglomeration externalities.
3

4
Abstract
The impact of innovation on firm performance has been a matter of significant interest to
economists and policy makers for decades. Although innovation is generally regarded as a
means of improving the competitiveness of firms and their performance on domestic and
foreign markets, this relationship has not been supported unambiguously by empirical work.
Innovative activities of firms influence their performance not necessarily directly but through
the production of useful innovations and increased productivity. Therefore, in recent years,
the relationship between innovation and firm performance has been modelled by a multi-stage
approach. However, the findings from existing studies differ in many respects which
suggests that there is the need for further research. In this paper we employ firm level data
from the fourth Community Innovation Survey (CIS4), covering some 90,000 firms in 16
West and East European countries in order to assess the drivers of the innovation process
in two different institutional settings, a number of mature market economies of Western
Europe and a number of advanced transition economies from Central and Eastern Europe.
A four-equation model, originating in the work of Crepon et al., (1998), has been used to link
the innovation decision of firms to their performance through the impact of innovation input
on innovation output and the innovation output on productivity and better performance. Our
findings confirm the positive relationship between innovation activities and productivity at the
firm level and provide further evidence on the relationship between size and innovation
activities.

5
1. Introduction
The traditional economic theory predicts that in the long run all firms will converge to their
long run steady state equilibrium position and optimum size. However, the evidence from
different industries suggests that firms which perform better today are more likely to perform
better tomorrow as well. The main explanation for this non-transitory feature of firm
behaviour is the different capabilities of firms to generate and implement new knowledge
which determine their relative position in the industry. In last few decades a large number of
studies have attempted to map the channels and mechanisms through which new knowledge
is transformed into better performance. However, the evidence from this literature is
inconclusive thus calling for further research.
The interest in innovation spans from the firm level to the national level. It is argued that
countries can achieve higher rates of growth and favourable terms of trade by specialising in
knowledge intensive products containing higher added value (OECD/Eurostat 1997). This is
the reason why policy makers across the globe have been struggling to develop policies
which would stimulate spending on R&D activities and increase the efficiency of the
innovation process. In 2000 the EU set itself the goal of becoming the most competitive
knowledge based economy in the world. The failure to achieve this goal can be traced to
many factors including the inability to stimulate R&D spending and enhance the innovation
activities of firms in EU countries, particularly the new members from Central and Eastern
European Countries (CEECs) which are seriously lagging behind.
Our study presents one of the first attempts to compare the determinants of the innovation
process in mature market economies of Western Europe and the transition economies which
have recently joined the EU. A multi-stage approach to innovation is applied to the firm-level
data collected by the Fourth Community Innovation Survey (CIS4) in order to identify factors
which drive each stage of the innovation process in different institutional settings. The rest of
paper is structured as follows. In section 2 we present the theoretical framework for the
relationship between innovation and firm performance using a multi-stage framework.
Section 3 reviews major findings from the innovation literature. The main characteristics of
dataset in our two samples are analysed in section 4 while section 5 presents the
methodology and estimative form. The results of our baseline specification are presented in
section 6. The sensitivity of the model is examined in section 7. Finally, section 8 concludes.

6
2. Theoretical framework
Innovation refers to all scientific, technological, organisational, financial and commercial
activities which lead to, or are intended to lead to, the implementation of technologically new
or improved products or services (OECD/Eurostat, 1997 p. 39). Hence, an innovation
contains new ideas which influence the behaviour of economic agents in a previously
unknown way. The introduction of new technology, human capital and the improvements in
the organisation of production increases firm’s efficiency and enables it to produce at lower
costs than its rivals. Similarly, the introduction of new products provides consumers with new
goods and services which, in turn, lead to the expansion of firms in new segments of the
market (OECD/Eurostat, 1997 p. 31). From here it follows that innovations enable firms to
differentiate themselves from their rivals (by new products, processes, costs or
organisational improvements).
The traditional model of the firm behaviour postulates that innovations can have only
transitory effect on firm performance as the new knowledge will soon be diffused and
imitated by rivals. Hence, in the long run all firms will converge to the steady-state equilibrium
(Knight 1921). However, there is a vast amount of evidence about some firms in different
industries and in different institutional settings remaining superior to their rivals for a
considerable period of time, irrespective of the measures of firm performance used -
employment, sales, productivity or profitability (Klomp and Van Leeuwen 2001, Loof,
Heshmati and Asplund, et al. 2002, Kemp, et al. 2003).
The above findings are more consistent with the conceptualisations offered in other schools
of thought, particularly Schumpeterian, evolutionist and the endogenous growth theory.
According to Schumpeter’s thesis of creative destruction, the introduction of new goods, new
methods of production, opening of new markets, discovery of new sources of supply and
organisational changes are elements within the system which regularly result in the
destruction of the existing economic structures and their replacement with new ones. In early
Schumpeter’s work, the need of active entrepreneurs to constantly move boundaries and to
change the existing organisational forms was considered the main generator of innovation
(Schumpeter 1934). However, in his later work (Schumpeter 1942), he argued that large
firms operating in concentrated industries are the main source of innovative activity. It is
suggested that the development of innovation requires the accumulation of knowledge and
financial means; thus the small entrepreneur can no longer be the principal driving force of
innovation. His role is relegated to large firms and their R & D laboratories which are more

likely to posses the necessary human and financial capital. In the Schumpeterian literature,
innovation is regarded as a central feature of a market economy.
Another set of explanations are offered by the evolutionary model of the firm (Nelson and
Winter 1982) which maintains that the behaviour of any firm consists of, and is based on, a
set of learned principles or routines. The quality of individual firm’s routines determines its
position in relation to rivals, analogous to the position of species in the evolutionary chain.
Firms cannot, of course, maintain their superiority permanently on the basis of their existing
routines. Innovations, which enable firms to develop new and upgrade existing routines, drive
the continuous changes in the economic system.
The endogenous growth literature introduces the simultaneity in the relationship between
innovation and performance. In this model the growth of an economy is determined by the
level of technology and innovation which, in turn, depend on the share of GDP devoted to
these activities (Romer 1990, Grossman and Helpman 1994, Aghion and Howitt 1998). In
such setting innovation is seen as a non-rivalrous input in the production process. It is also
emphasised that the incentive to innovate is closely linked to the functioning of institutional
framework as the innovators would not be able to acquire rents from their invention in an
unsuitable institutional environment.
Building on these foundations, Klette et al., (2000) developed the multi-stage model of firm
behaviour in which they argue that the growth of a firm is determined by the quality and price
of its own and its competitors’ products and that the quality of its products can be improved
through innovation. The intensity of innovation, however, is postulated to be independent of
the firm’s size – rather similar to the Gibrat's law. Instead, it is related to the profit margin of
firm which, in turn, depends on the degree to which the firm can differentiate its products
from its rivals’ products. The model also identifies some industry characteristics as the
determinants of R&D intensity as the firms in industries with higher demand for high quality
products and higher innovative opportunities are more likely to have higher R&D intensity.
In recent years, the above mentioned insights on innovation process have been synthesised
in a number of papers through the multi-stage model of innovation process (Crepon, Duguet
and Mairesse 1998, Loof and Heshmati 2002, Loof and Heshmati 2006). This stream of
literature traces innovation process from firm’s decision to innovate to its performance
bringing together features of innovation recognised in earlier models such as reverse
causality, individual heterogeneity etc. The present paper is based on this approach to
modelling the innovation behaviour of firms.
7

8
3. Literature review
Studies from the early period of research on innovation have typically reported a positive
relationship between innovation and measures of firm performance. Most of these studies
used innovation expenditure as the principal measure of innovation at the firm level. Using
cross-sectional data for US firms between 1972 and 1977 Griliches (1986) finds that the
higher R&D investment leads to higher rates of productivity growth among firms. Basic
research appears to be a more important determinant of productivity than other types of R&D
and privately financed R&D expenditure appears to be more effective than those financed by
the state. These findings were later confirmed by Lichtenberg et al., (1991) who used
longitudinal data on US firms between 1972 and 1985. Similar findings have also been
reported for other countries. Goto and Suzuki (1989) using a sample of Japanese
manufacturing firms in 1982, find that the growth of productivity is positively related to the
growth of R&D investment in firm's core activity and also to the growth of R&D investment in
supplying industries. Also, Wakelin (1998) find that for a sample of UK firms between 1988
and 1992, R&D intensity had a positive and significant effect on productivity growth.
However, it has been suggested in several studies that R&D expenditure suffers from several
shortcomings when used as the measure of innovation activity. The Oslo Manual
(OECD/Eurostat, 1997) notes that measures of innovation input, although related to technical
change, are not its direct measures. It has also been emphasised that the R&D expenditure
does not encompass all innovative efforts of firms such as learning by doing or the
knowledge embodied in its investment in new machinery and also its human capital. Kemp et
al. (2003) add that studies based on R&D expenditure are not informative on the actual
process of innovation. Moreover, the expenditure approach to the innovation may be
misleading as the lower amounts of own expenditure on the innovations may simply reflect
the fact that the innovation is being developed in cooperation with universities or other firms
where the outside agency covers the cost of R&D expenditure. Also, many small firms do not
clearly separate R&D investment (or department) from their other activities. Finally, the fact
that only a small proportion of the innovation efforts lead to innovation output and better
performance means that firms may spend on R&D for many years without reaping the
potential benefits of such spending (Bessler and Bittelmeyer 2008).
In a new generation of models studying the impact of innovative activities on firm
performance, the focus has shifted to the complex innovation process and channels through
which the innovation inputs are transformed into better performance (Crepon, Duguet and

Mairesse 1998, Hall and Kramarz 1998, Loof and Heshmati 2002, Kemp, et al. 2003, Loof
and Heshmati 2006, Bessler and Bittelmeyer 2008). In recent years the four-equation model
originally developed by Crepon et al. (1998) has become the dominant model within this
strand. The model portrays innovation process as consisting of four stages: the decision to
innovate, the decision on how much to spend on innovation activities, the relation between
expenditure on innovation and innovation output, and the relation between innovation output
and performance. These four stages are estimated in a sequential way and it is assumed
that the causality runs from the decision to innovate to the firm performance. However, it has
also been argued that there is reverse causality from firm performance to innovation output
stage. The four stages are modelled in a way to incorporate various factors identified in the
literature as determinants of the innovation process such as firm characteristics, industry
specific factors and the institutional background. The range of factors used in individual
studies depends on the quality and coverage of the dataset used.
The first two stages of the innovation process, i.e. the decision to innovate and the decision
on how much to invest in innovation (innovation input) are usually estimated jointly in
systemic approach. In the majority of studies, the innovation input is defined as investment in
R&D, either as the absolute amount of investment (Loof and Heshmati 2002, Kemp, et al.
2003, Loof and Heshmati 2006) or as the ratio of innovation expenditure to total sales
turnover, i.e. innovation intensity (Klomp and Van Leeuwen 2001, Stoevsky 2005,
Chudnovsky, Lopez and Pupato 2006). The main advantage of these studies is that they
take somewhat broader definition of innovation investment by including expenditures on
organisational changes, machinery or marketing in addition to expenditure on R&D.
Most studies have adopted the practice of including same variables as determinants of the
decision to innovate and the decision on how much to invest in innovation: the size of the
firm, its export intensity, human capital, the possibility of cooperation with customers and
research institutions or universities, the existence of public support for innovation, and
previous research activity of the firm (Loof and Heshmati 2002, Klomp and Van Leeuwen
2001, Loof and Heshmati 2006). Sometimes the factors which hamper or facilitate
innovation such as the access to financing are also included. The evidence from different
studies differs in many respects which suggests that there is a need for further investigation
of these relationships.
One of the factors frequently related to the decision to innovate and the innovation input is
the size of the firm with the number of employees as the most commonly used measure. As
noted in section 2, the theory postulates that under different conditions firm size could be
positively or negatively related to innovation. This postulate is supported by a large number
9

of empirical studies which have reported positive, negative or even insignificant relationship
between the firm’s size and its decision to innovate or the innovation input decision (Klomp
and Van Leeuwen 2001, Loof and Heshmati 2002, Kemp, et al. 2003, Loof and Heshmati
2006). The main reason for such ambiguous findings may be industry-specific
characteristics. As noted by Hall and Kramarz (1998) in industries abundant with low skilled
labour, increasing innovation efforts could be the result of the decision of firms to reduce the
existing stock of low skill labour and keep only the high skill workers while the opposite may
hold for industries with high concentration of skilled labour. Similarly, Acs and Audretsch
(1987) find that larger firms are more likely to innovate in industries which are more
concentrated and have higher entry barriers while the opposite holds for firms in industries
characterised by low entry barriers and higher degree of competition. Finally, Cohen and
Klepper (1996) introduce four stylised facts about the relationship between innovation and
firm size. First, the probability of a firm undertaking innovation increases with firm size.
Second, within an industry, innovation efforts and firm size are positively related across all
firm size groups. Third, R&D expenditure rises proportionately with firm size in most
industries. Fourth, the number of patents and innovations per unit of innovation investment
decreases with firm size.
Another frequently employed determinant of innovation process is export intensity. The
previous literature held that higher export intensity facilitates the decision of firms to innovate.
The logic behind this thesis is that foreign competition is more intense than domestic and it
requires continuous upgrading of the firm’s products and processes. Studies for firms in
developed countries have reported the existence of a positive relationship between the
export intensity and firm’s decisions to engage in innovation as well as the amount of
investment in innovation (Loof and Heshmati 2002, Kleinknecht and Oostendorp 2002,
Kemp, et al. 2003). However, it should be noted that the main trade partners of these
countries are other developed countries. While it is true that the export orientation of firms
towards the more advanced foreign markets is likely to encourage the decision to innovate,
the effect may be non-existent, or even the opposite, when a firm exports to less developed
markets. Hence, the direction of relationship between the export intensity and the decision of
firm to innovate depends on the character of the foreign market on which they compete.
Another group of factors which are likely to influence the input stage of innovation are the
characteristics of socio-economic environment in which the firm operates. These include
access to finance, the quality of institutional setting and the country specific cultural values.
Various studies have reported that access to finance, particularly the availability of public
subsidies for innovation activities, is a crucial determinant of the innovation process (Klomp
10

and Van Leeuwen 2001, Kemp, et al. 2003). The likely reason for this is that in the presence
of high levels of uncertainty and information asymmetry (market failures), firms will focus only
on those projects which are profitable. Hence, by providing subsidies public authorities
motivate firms to undertake also those innovations which would otherwise be abandoned.
Apart from subsidies Kemp et al. (2003) find that the innovative input is positively influenced
by the contacts and cooperation with research institutes. Finally, Loof et al. (2002) find that
the cooperation with domestic rivals, customers and some sources of information is positively
related to higher innovative efforts by firms.
An important stage of innovation process is the innovation throughput which encompasses
the transformation of innovation inputs into innovation output. In the four-equation model, this
stage is modelled through factors which are expected to influence the efficiency of the
innovation process such as cooperation with universities and other institutions, or the
previous experience in innovation. Besides these, various studies have also included factors
which hamper innovative activities such as the financial constraints or the previously
abandoned innovations (Klomp and Van Leeuwen 2001, Loof, Heshmati and Asplund, et al.
2002, Kleinknecht and Oostendorp 2002, Kemp, et al. 2003). As with other variables the
findings based on different datasets exhibit considerable variations. Klomp and van Leeuwen
(2001) report that permanent involvement in R&D has a positive and significant effect on the
innovation output while Kemp et al. (2003) add to this also the subsidies and changes in the
organisation of firms. The opposite finding is reported by Loof et al., (2002) as in this study
none of process related variables is statistically significant determinant of innovation output
while the internal sources of information and the objective of extending the product range are
significant and positively related to the innovation input. Finally, Klomp et al., (2002) find that
interactions between the universities, firms and research institutions have a positive effect on
the efficiency of the innovation process.
The innovation output can manifest itself as product and/or process innovations. In the
majority of the studies, product innovations are employed as the measure of innovation
output and process innovations are neglected (Kemp, et al. 2003). The most commonly
employed measures of product innovation are the share of turnover generated by new
products, number of patents and new product announcements (Acs and Audretsch 1987,
Klomp, et al. 2002, Loof and Heshmati 2002). The proportion of sales attributable to new
products is particularly useful as it directly relates the innovation process with the commercial
success of firm. The number of patents, on the other hand, is not a very useful measure of
innovation output as it presents only an intermediate measure of innovation output and it also
suffers from several shortcomings (Kemp, et al. 2003). Finally, there is the new product
11

announcement which measures the innovation output directly (Acs and Audretsch 1987,
Kemp, et al. 2003). However, the problem with this measure occurs in cross-country
comparisons and the selection of the relevant sources in which new products are
announced. Moreover, by confining strictly to this measure it is likely that process innovations
remain under-reported. The overall conclusion from the literature is that sales from new
products is the most robust measure of innovation output as it includes the entire innovation
process (Kemp, et al. 2003).
The determinants of the innovation output have mainly been looked for in the innovation
input and process-related variables. Although most studies have hypothesised that the
innovation inputs positively influence the innovation output, the empirical evidence has been
found only in few cases (Klomp and Van Leeuwen 2001, Loof and Heshmati 2002). Similarly,
the relationship between process-related factors and the innovation output has been found
only for some factors as we have mentioned earlier in this section. Loof et al., (2002) suggest
that these mixed results could be caused by the shortcomings of existing models to properly
handle the complexity of the innovation and by the shortcomings of the available datasets. In
addition to these determinants, various firm characteristics such as the firm size have been
included as the determinants of the innovation output but mostly they have been insignificant
(Loof and Heshmati 2002).
The most commonly employed measures of the performance are productivity, sales, export
revenues and profits although sometimes financial measures such as the returns on the
assets are also employed (Loof and Heshmati 2002, Bessler and Bittelmeyer 2008). Most
studies have reported a positive relationship between innovation and firm performance. Loof
(2000) tests the existence of a positive relationship between the innovation output measured
by sales of new products per employee and five different measures of firm performance
(employment growth, value added per employee, sales per employee, operating profit per
employee and return on assets). A positive relationship was confirmed for all five indicators.
However, not all studies have confirmed the existence of this relationship. Klomp and van
Leeuwen (2001), for example, have found a positive relationship between innovation output
and sales growth but no evidences of a relationship between the innovation output and
employment growth.
Kemp et al. (2003) have found a positive relationship between the innovation output
(measured by the share of sales from new products in total turnover) and the growth of
turnover and employment and no significant with profit. Bloom and Van Reenen (2002) find
that the impact of innovation output on the firm performance appears to be contemporaneous
when performance is measured by market value but it occurs with a lag when performance is
12

measured by productivity. Bessler and Bittelmeyer (2008) report that innovations bestow on
firms only temporary advantage in the short run and their effect appears to be diminishing in
the long run. This finding is consistent with Schumpeterian thesis of creative destruction.
Innovations provide competitive advantage for a limited period of time after which knowledge
is diffused across the market. As new products enter the market the competitive advantage
of the firm diminishes and it will suffer loss and eventually will be forced to exit the market
unless it develops even better product.
Loof et al. (2002) find that the innovation output, firm size, and the share of non-R&D
engineers and administrators in total employment are significant factors affecting the firm’s
productivity. Adamou and Sasidsharan (2007) have also found the positive effect of R&D
intensity on sales growth. Apart from innovation intensity other significant variables include
size and age which have a negative effect on the firm’s sales growth Finally, in one of rare
studies encompassing both product and process innovations, Nguyen et al., (2007) have
found a positive and statistically significant relationship between the different measures of
innovation output and export intensity.
Several authors have documented the innovation process in new EU members and some
other transition countries as well. Masso and Vahter (2007) find that decisions of Estonian
firms to innovate and spending on innovation are positively related with the orientation to
international market, existence of formal legal institutions protecting innovations and access
to national subsidies. It has been also found that the larger firms have higher propensity to
engage in innovative activities while the obstacles to financing are a statistically significant
impediment to innovations. In innovation output stage the study distinguishes between the
process and product innovations. It is found that the latter is positively related to the
protection of innovations and access to subsidies while the former is positively related to the
size of the firm. Finally, the performance of firm measured by productivity is positively
influenced by process innovations but not by product innovations. The findings for other
transition countries are rather similar. Roud (2007) reports results for a sample of Russian
firms that the firm’s size and the availability of public subsidies positively influence the
decision of firms to innovate and the amount of expenditure on innovation while innovation
output is influenced by the innovation input and productivity is positively influenced by the
innovation output but negatively by the firm’s size.
There is much evidence on the benefits which subsidiary firms in developing countries can
receive from their mother companies in developed economies. In this context, one may
expect to find a difference in the innovation behaviour of domestic and foreign owned
companies. Domadenik et al., (2008) find for a sample of Slovenian firms that the firms
13

owned by domestic owners invest significantly more in the R&D than firms owned by other
types of owners. However, Chudnovsky et al., (2006) and Stoevsky (2005) do not find any
evidence that the innovation efforts are different between domestic and foreign owned
companies in Argentina and Bulgaria respectively. Finally, Raffo et al., (2008) conclude that
while in general it can be said that foreign owned firms invest more in R&D, the findings on
the innovation output and firm performance are very heterogeneous and no general
conclusion can be drawn.
14
4. Data
The empirical analysis in this paper is based on the data from fourth Community Innovation
Survey (CIS4), conducted in 2004. This survey covered some 90,000 firms from 16
European countries, including seven which had joined the EU in that year (Czech Republic,
Estonia, Hungary, Lithuania, Latvia, Slovenia, Slovakia) and two which were official
candidates at that point (Bulgaria and Romania), five EU countries (Spain, France, Italy,
Luxembourg and Portugal) and two countries which are not EU members (Iceland and
Norway).1 The surveyed firms are distributed across all major sectors of economic activity.
Table 1 provides descriptive statistics on the innovation behaviour and productivity of firms,
grouped according to several criteria, in three groups of countries: total sample, the sub-sample
of countries from CEECs and the sub-sample of old EU members together with
Norway and Iceland.2 The comparison of average productivity of firms in different sub-groups
across the two sub-samples shows substantial differences, with firms based in old EU group
of countries having much higher levels of productivity.
1 The anonymised data was made available to the Microdyn project on CDROM by Eurostat. Access to the raw
data was provided at the Eurostat Safe Centre in Luxembourg. We are grateful to Sergiu-Valentin Parvan for
facilitating access to data and confirming the regression results produced at the Centre.
2 For expositional convenience, hereafter, we will refer to the two groups as Central and East European countries
(CEECs) and West European countries.

Table 1: Average productivity of different groups of firms in the database
Sample All countries CEECs Western Europe
Criterion Yes No Yes No Yes No
Innovation activity in past 3yrs 211.16 141.05 118.09 81.33 248.71 162.45
Market orientation in past 3 yrs
Domestic market 186.72 91.31 94.68 56.29 236.79 120.81
EU market 193.33 125.41 84.24 73.43 255.18 161.83
Other foreign markets 239.77 126.84 114.42 71.76 280.67 168.555
Type of cooperation
Part of a group 338.37 76.04 211.02 49.54 378.38 98.05
Cooperation on innovation activities with
other enterprises or institutions in past 3yrs 306.47 161.27 101.72 126.43 408.31 174.96
Factors hampering innovation activities
Costs 97.88 171.47 44.42 98.70 140.07 213.83
Knowledge 116.58 152.41 49.46 83.40 155.70 198.56
Market factors 110.17 155.22 54.75 83.02 144.01 203.07
Access to subsidies
Received subsidies from local sources 120.91 208.91 74.41 117.99 123.25 240.89
Received subsidies from national sources 172.02 203.18 92.02 120.17 190.77 230.39
Received subsidies from EU sources 153.29 201.2 58.27 120.28 186.13 226.15
Note: The ‘Yes’ and ‘No’ columns refer to the average productivity of firms with each of the characteristic listed
compared with the productivity of firms that do not have that characteristic.
Source: CIS Database, Eurostat
Turning to the most important issue, the difference in performance between innovating and
non-innovating firms, it is evident that in both sub–samples, firms undertaking some
innovation activities in the three previous years, performed much better than their rivals
which did not engage in any innovation activities. However, here again we observe
differences among firms in the two groups. Market orientation appears to be another
determinant of firm performance. Firms that export to the EU and countries outside the EU,
have much higher productivity than firms which sell their products on the domestic market.
One explanation for this finding may be that greater competition on the EU and foreign
markets forces firms to be more innovative and efficient – but it also reflects the fact that
firms withstanding foreign competition are likely to be more efficient in the first place. Other
firm characteristics suggest that firms which have some kind of cooperation with rivals,
customers, universities or research laboratories and firms which are part of groups have
much higher productivity in both sub-samples. This implies that knowledge spillovers play an
important role in firm’s performance. With respect to factors hampering innovation, cost
obstacles seem to be the most prominent. Firms that reported some kind of obstacle in the
innovation process have lower levels of productivity. Contrary to the common wisdom, firms
which had access to local, state or EU funded subsidies have much lower levels of
productivity.
15

In summary, it is evident that there are considerable differences in performance, measured
by levels of productivity among innovators and non-innovators. Moreover, innovating firms in
different sub-samples vary considerably with respect to their performance. This is particularly
evident when performance of firms from West European countries is compared with that of
firms from CEECs. It appears that apart from innovation, market orientation and creation of
networks with other enterprises and research institutions have positive effect on firm’s
performance while the factors hampering innovation such as cost and knowledge obstacles
as well as the use of subsidies are negatively related to the firm performance. The relative
weight of these and some additional factors influencing the innovation process and firm
performance will be explored in the remainder of this paper.
16
5. Model specification
The literature on innovation and firm performance identifies two major problems with the
econometric specification of this relationship, namely the selectivity bias and simultaneity bias. The
selectivity bias arises from the fact that not all firms engage in innovation and some innovations are
not successful. In addition, we have already argued that there are many factors which can influence
both firms’ decision to innovate, its level of expenditure on innovation as well as its final
performance. This creates simultaneity bias. The four‐stage model discussed earlier (Crepon, Duguet
and Mairesse 1998) is capable of addressing these problems. In this model the decision to innovate
and the decision on how much to invest in innovation are linked to their determinants in the first two
stages of the innovation process. The third stage is a knowledge production function linking
innovation input and output. Finally, in the fourth stage the productivity of firm is related to the
innovation output.
5.1. General specification of the system model
If is unobserved decision variable of whether or not a firm invests in innovation and the
unobserved level of firm’s investment in innovation, with and being their observable
counterparts, the first two stages of systemic approach can be defined as follows:
(1)

17
and
(2)
In these expressions are vectors of independent variables and their
corresponding unknown parameters which reflect the impact of certain determinants on the
firm’s decision to engage in investment in innovation and the actual level of expenditure on
innovation. The and are random error terms with zero mean, constant variances and
not correlated with the explanatory variables. However, it is assumed that the two error terms
are correlated with each other on the basis of unobservable characteristics of firms. In the
innovation literature these two stages of the system approach are estimated in two ways.
One way of estimation involves a probit equation in the first step and OLS estimation in the
second step using inverse Mills ratio from the probit equation to correct for possible selection
bias. Alternatively, the first two stages of the systemic approach can be estimated jointly by a
generalised tobit model with the maximum likelihood estimation method.
The third stage of the estimation is represented by the following equation:
(3)
where represents the observed level of innovation output, and are the innovation
input from previous equation and its corresponding unknown parameter, is the vector of
other explanatory variables which includes among others inverse Mill’s ratio from the first
stage and performance from the fourth stage to control for selection bias and feedback
effect. is the vector of corresponding unknown parameters while is the random error
term with mean zero and constant variance not correlated with explanatory variables.
Finally, the last equation of the model relates the innovation output with firm’s performance.
(4)
with indicating firm performance, and being a vector of independent variables and its
corresponding unknown parameters and being the error term. The third and fourth stages
are estimated jointly as a system. In the baseline specification we estimate the model with
three stage least squares as this methodology allows to control for feedback effect from
performance to innovation output stage. However, we also estimate the specification with two

stage least squares in which the above mentioned feedback effect is excluded. It is
assumed that the error terms are uncorrelated with explanatory variables but the model
allows the arbitrary correlation among four error terms which implies that there are some
common firm characteristics and unobserved variables influencing all four stages of the
innovation process. As a consequence it is possible that innovation input in the third stage
and innovation output in fourth are endogenous. This fact is acknowledged and controlled for
with proper instrumentation.
5.2. Definition of variables and specification of the model
The four dependent variables of the model and the explanatory variables in each stage are
defined as follows.3 A firm is considered as having taken the decision to innovate if in three
years prior to survey, it has undertaken any of the following activities: invested in intramural
or extramural activities in R&D, purchased new machinery, equipment and software,
purchased or licensed patents, know-how and similar forms of knowledge from other
organisations, engaged in training of staff for development of new or significantly improved
products or processes and undertook activities for the market introduction of new or
significantly improved goods and services. The decision to innovate is modelled as a function
of: firm size measured as natural logarithm of employment; three dummy variables for market
orientation (national, EU and other); a dummy variable for firm being part of a group; a
dummy variable for firms that had in the previous three years ongoing or abandoned
innovations; four dummy variables for factors hampering innovation (costs, knowledge,
market and other reasons); and two dummy variables for firms that undertook organisational
or marketing innovation in previous three years.4
Innovation input is indicated by innovation expenditure measured by the natural logarithm of
overall amount spent on innovations in 2004. This variable encompasses spending on all
innovation activities mentioned above (intramural and extramural R&D expenditure,
acquisition of machinery, equipment and software -excluding the machinery, equipment and
software for R&D- and other acquisitions of external knowledge).5 In comparison to selection
equation, the innovation input equation includes as explanatory variables also three dummy
variables for receiving different types of subsidies (from local, national and EU sources) and
four dummy variables for important sources of information on innovation activities (internal
sources, market sources, institutional sources and other sources) while market and other
3 For a fuller definition of variables, see Table A1 in Appendix.
4 In CIS4 questionairre, the „other factors hampering innovation“ refer to lack of need for innovations due to prior
innovations or due to lack of demand for innovations.
5 This broader definition of innovation input responds to the criticism that many firms (especially smaller ones)
would not include R&D expenditure explicitly in their accounts and therefore R&D expenditure would
underestimate the actual amounts spend on innovation inputs.
18

factors hampering innovation are omitted. Hence, we assume that innovative activities of
enterprises at this stage are hampered primarily by costs of these activities as well as the
knowledge factors such as lack of qualified personnel, lack of information about technology
or market.
Innovation output is measured by the natural logarithm of the share of sales of new products
and services (new to a firm and new to the firm’s market) in total turnover of the firm. The
explanatory variables in this equation are: firm size; innovation input from the second stage;
inverse mills ratio from the first stage; the natural logarithm of labour productivity; sources of
information on innovation activities; organisational and marketing innovations; receiving
different types of subsidies; a dummy variable for cooperation with any of suppliers,
customers, universities, research laboratories, etc.; and three dummy variables indicating
whether the firm perceives its innovation to be highly important for improvement of its
products, processes or other aspect of its performance.
The dependent variable in the fourth stage of the model is the natural logarithm of firm’s
labour productivity defined as the ratio of firm’s total turnover by total employment in 2004. It
is specified as a function of: firm size; innovation output from third stage; organisational and
marketing innovations; factors hampering innovation (all mentioned earlier); two dummy
variables for sources of innovation indicating if the improvements in products and processes
in the previous three years were developed within enterprise or in the wn efforts cooperation
with other firms and institutions. All four models contain three dummy variables for industry
specific effects for manufacturing, services and trade sectors.6 In addition, we include in
regressions for full sample a dummy variable for firms located in one of CEECs.
The first two stages include all firms in the sample and are estimated jointly with the
maximum likelihood method of generalised tobit. The estimation in third and fourth stage is
undertaken on firms that have reported sales from new products. As we already mentioned,
these two equations are estimated as a system in a framework of simultaneous equations
where the feedback is allowed from productivity to the innovation output. In our baseline
specification the last two equations are estimated using three stage least squares system of
simultaneous equations.
19
6 The base group is ‘all other industries’

20
6. Interpretation of findings
In this section we report the results of the estimation procedure. The analysis was conducted
for the full sample and the two sub-samples (Western Europe and CEECs) separately. For
expositional convenience the results for each stage of the innovation process will be
presented in separate sub-sections.
6.1. Decision to innovate
Table 2 presents the results of the estimation of the first stage for three samples. The results
are very similar for the three samples. In general it can be concluded that the probability of
engagement in innovation for a typical firm increase with firm size. Firms that are oriented
towards national, EU and other foreign markets are more likely to innovate than the firms
oriented towards local/regional markets. This implies that the intensity of competition
motivates firms to innovate. Being part of the group and having previously abandoned
innovation activities also increases the probability of deciding to innovate. This suggests that
knowledge accumulated from previous innovative activities (even when they were not
successful) as well as knowledge transfer from other parts of group motivates firms to
engage in new innovations.

Table 2: Results from the selection equation
21
Total CEECs WE
Firm size 0.074*** 0.077*** 0.086***
Market orientation
National market 0.169*** 0.099*** 0.189***
EU market 0.142*** -0.003 0.206***
All other countries 0.219*** 0.208*** 0.182***
Part of a group 0.098*** 0.315*** 0.025*
Abandoned or ongoing innovations 2.767*** 2.870*** 2.729***
Highly important factors hampering innovation
Cost factors 0.086*** 0.049** 0.099***
Knowledge factors 0.069*** -0.010 0.098***
Market factors -0.034*** -0.007 -0.046***
Other reasons for not to innovate -0.517*** -0.329*** -0.560***
Organisational and marketing innovations
Organisational innovation 0.782*** 0.806*** 0.761***
Marketing innovation 0.532*** 0.705*** 0.462***
Industry specific characteristics
Manufacturing 0.196*** 0.284*** 0.178***
Trade -0.142*** -0.134*** -0.124***
Services -0.004 0.219*** -0.058***
Institutional setting
CEECs 0.081 - =
No. of observations 85777 25527 62244
Note: ***, **, * indicate statistical significance at 1%, 5% and 10% levels respectively.
A somewhat surprising finding is the positive and significant coefficient on cost and
knowledge factors hampering innovation. Loof et al. (2002) report the same finding for
Sweden suggesting that an innovating firm has a pool of ideas but the financial and
knowledge restrictions force it to be selective, hence selecting only those ideas which have
high probability of success. The market factors and other factors hampering innovation (such
as the existence of previous innovations) have negative impact on the probability of the firm’s
decision to engage in innovation activities. Organisational and marketing innovations have a
positive impact on the probability of the firm’s decision to engage in innovation in all three
samples. This suggests that changes in the structure of management, introduction of new
skills or improvement of the relations with clients and suppliers as well as design
improvements increase the probability of innovation decision.

6.2. Innovation investment
The results from the estimation of the innovation investment equation are presented in Table
3. The findings from all three samples indicate that innovation investment increases with firm
size and that the effect of size is similar in all three regressions. Firms oriented towards all
markets except the local/regional market invest more in innovation. The only exception is the
negative coefficient for the domestic market in the group of new EU countries. One likely
explanation is that domestic markets of these countries are not very demanding with respect
to knowledge intensive products. Investment in innovation is higher in firms which are part of
a group and which previously had abandoned some innovation efforts. In terms of
organisational and marketing innovations, the coefficient on organisational innovations is
highly significant and positive while the marketing innovations coefficients are negative and
only marginally significant in the full sample and in the sub-sample of West European firms
(but insignificant for firms in new EU members). A likely explanation for the first two samples
could be that marketing innovations and the penetration into new markets involve
considerable amount of firm’s financial means thus having a negative effect on investment in
product and process innovations.
National and EU subsidies have positive and statistically significant effect on the innovation
investment in all three regressions. However, local subsidies have negative sign in the full
sample and the old EU sample while the coefficient for firms in new EU countries is
statistically insignificant. The likely explanation for this is that when deciding on provision of
subsidies local administrations may have other objectives and may base their decision on
political factors. Another likely explanation could be that the amount of national and EU
subsidies is larger and firms must meet stricter criteria to obtain them. Finally, in all three
regressions the internal market and institutional sources of information about innovation are
statistically significant and positive. The only exception is the group of new EU countries
where the coefficient for institutional sources of information is statistically insignificant
possibly reflecting insufficient cooperation between firms and universities and other public
institutions in these countries. In all three regressions other sources of information
(conferences, scientific journals and professional and industry associations) are statistically
insignificant.
22

Table 3: Results from the innovation investment equation
23
Total CEECs WE
Firm size 0.681*** 0.635*** 0.693***
Market orientation
National market 0.814*** -0.082** 1.295***
EU market 0.259*** 0.097** 0.219***
All other countries 0.301*** 0.282*** 0.271***
Part of a group 0.691*** 0.532*** 0.745***
Abandoned or ongoing innovations 1.051*** 0.505*** 1.320***
Highly important factors hampering innovation
Cost factors 0.031 -0.271*** 0.134***
Knowledge factors 0.182 -0.064 0.316***
Organisational and marketing innovations
Organisational innovation 0.116*** 0.189*** 0.141***
Marketing innovation -0.090** -0.021 -0.075*
Access to subsidies
Local/regional subsidies -0.356*** 0.020 -0.414***
National subsidies 0.863*** 0.787*** 0.834***
EU subsidies 0.821*** 0.562*** 0.856***
Highly important sources of information about
innovation
Internal sources 0.886*** 0.411*** 1.040***
Market sources 0.209*** 0.210*** 0.225***
Institutional sources 0.514*** -0.045 0.727***
Other sources -0.018 -0.010 0.076
Industry
Manufacturing 1.253*** 0.171** 1.435***
Trade 1.047*** -0.035 1.224***
Services 1.327*** 0.486*** 1.444***
CEECs 0.974*** - -
No of observations 85777 25527 62244
Note: ***, **, * indicate statistical significance at 1%, 5% and 10% levels respectively
6.3. Innovation output
The third stage consists only of firms that have reported positive amount of innovation output.
Hence, we include inverse Mill’s ratio, calculated from the first equation to control for
potential selectivity bias. The results of the estimation are presented in Table 4. The
presence of selectivity is indicated only in equation for the full sample and even there the
coefficient is only marginally significant, at the 10% level. In the other two cases the
coefficient of the Mill’s ratio is statistically insignificant.

Table 4: Results from the innovation output equation
24
Total CEECs WE
Firm size -0.416*** -0.546*** -0.376***
Mill’s ratio 0.211* 0.080 0.187
Innovation input 0.308*** 0.621** 0.253***
Productivity 0.065** -0.552*** 0.026
Cooperation 0.158*** 0.166*** 0.192***
Highly important effects of innovation
Product effects 0.176*** 0.275*** 0.182***
Process effects 0.006 0.096*** 0.006
Other effects 0.050*** 0.146*** 0.047**
Organisational and
marketing innovations
Organisational innovation 0.133*** 0.047 0.158***
Access to subsidies
Local/regional subsidies -0.035 -0.063 -0.037
National subsidies -0.216** -0.376** -0.143*
EU subsidies -0.242*** -0.364* -0.189**
Highly important sources of information
about innovation
Internal sources -0.174** -0.117 -0.159*
Market sources -0.018 -0.047 -0.003
Institutional sources -0.082 -0.436 -0.046
Other sources 0.017 0.002 0.038
Manufacturing -0.195 0.056 -0.234
Trade -0.381*** 0.567*** -0.331***
Services -0.312** -0.083 -0.302**
CEECs 0.225*** - -
Sample size 15644 5444 10200
The coefficient on firm size is negative and statistically significant in all three regressions.
This means, ceteris paribus, a certain amount of innovation investment produces less output
in larger firms than in smaller firms (i.e., that larger firms are less efficient than smaller firms
in converting innovation input to input output). This finding, and that of the previous sections,
appear to confirm the stylised observations put forward by Cohen and Klepper (1996),
discussed earlier. In all three regressions we find a positive and statistically significant
relationship between innovation input and output. In the full sample an increase of 1% in
investment in innovation activities leads to a higher innovation output of about 0.31 percent.
In the sub-sample of firms in the new EU countries the effect is even larger and the same

increase in innovation investment would yield 0.62 percent increase in the innovation output.7
The findings with respect to feedback effect from productivity to innovation output are rather
ambiguous. While for the full sample we find a positive and statistically significant coefficient,
for the sub-sample of firms in CEECs countries the effect of productivity on innovation output
is negative and for the sub-sample of West European firms it is insignificant. The finding for
CEECs is particularly interesting as it implies that more efficient firms have lower proportion
of sales from new products in their total revenues. Having in mind that throughout the
transition period firms from these countries largely specialised in labour intensive products
produced with standardised technologies, the above finding probably reflects the risk-aversion
of their firms. Accordingly, the introduction of new products or services increases
the risk of failure which is the reason why these firms transform improvements in efficiency
into competitive advantages in production of existing products.
The results from Table 4 include several variables which could be defined as implicit
measures of innovation throughput, i.e. the transformation of innovation inputs into
innovation output. They indicate that the cooperation with other firms, universities and other
institutions influences the innovation output positively. In addition the product oriented effects
of innovation such as improvements in quality or the wider range of goods and services are
positively correlated with the innovation output. In the case of firms in the new EU countries it
is also suggested that the process effects of innovation are also positive and significant. In all
three regressions, other effects (that encompass meeting of regulatory requirements and
environmental care) are positive and statistically significant.
The two dummy variables for marketing and organisational innovation are statistically
significant and positive. This implies that improvements in organisational efficiency and
differentiation in terms of design, packaging or delivery help firms to achieve higher sales
from new products. With respect to subsidies we find statistically significant coefficients only
for the national and EU subsidies, but with negative coefficient. Having in mind the positive
effect of these variables on innovation investment, this questions the ability of existing
schemes for the allocation of subsidies at both EU and national levels to facilitate innovation
process. Among sources of information we find only weak significance for internal sources of
information in the full sample and the sample of firms from the old EU countries. In all other
cases the sources of information appear to be insignificant.
7 These coefficients are large but in the Loof, et al.’s work (2001 and 2006) with CIS3, they are also larger (and
sometimes larger).
25

6.4. Productivity (performance) and innovation output
The result of the estimation of the fourth stage of the model is presented in Table 5. The
table indicates that the firm’s productivity increases significantly with innovation output. In all
three regressions the coefficients are statistically significant and positive. The size of the firm
also positively and significantly affects its productivity in all three regressions, i.e., that the
same level of innovation output has a larger impact on productivity in larger firms than in
smaller firms. With respect to the variables indicating sources of innovation we find that firms
which develop their product innovation internally or in cooperation with other firms and
institutions have higher productivity than firms that implement innovations developed by other
firms or institutions. However, with respect to process innovations we find positive
coefficients for innovations developed internally and together with other enterprises and
institutions for the full sample and for the sub-sample of West European firms but not for the
sub-sample of East European firms. For these firms the coefficient for innovations developed
within the enterprise is statistically significant and negative. This probably reflects the inability
of firms in these countries to undertake process innovations on their own and their
dependency on other sources of innovation.
26

Table 5: Results from the productivity equation
27
Total CEECs WE
Sample size 15644 5444 10200
Firm size 0.405*** 0.178*** 0.422***
Innovation output 1.638*** 0.644*** 1.463***
Source of product innovations
Within enterprise 0.361*** 0.161** 0.760***
Together with other enterprises or institutions 0.113 0.058 0.348***
Sources of process innovations
Within enterprise 0.693*** -0.280*** 1.203***
Together with other enterprises or institutions 0.368*** 0.011 0.563***
Organisational and
Organisational innovation -0.177*** 0.157*** -0.215**
Marketing innovation -0.099* -0.075* 0.008
Factors hampering innovation
Cost factors -0.156*** -0.398*** -0.111**
Knowledge factors -0.021 -0.069 0.031
Market factors 0.131*** -0.071 0.240***
Other reasons not to innovate -0.169** 0.262*** -0.231**
Manufacturing 0.111 -0.789*** 0.667***
Trade 1.295*** 0.516*** 1.478***
Services 0.154 -0.368*** 0.436***
CEECs -0.637*** - -
Somewhat puzzling is the negative coefficient on organisational innovations found for the full
sample and the sub-sample of firms in Western Europe. The coefficient for marketing
innovations is only marginally significant for the full sample and the sub-sample of firms in
new EU countries and insignificant for the second sub-sample. The likely explanation is that
organizational innovations take some time to be implemented. Therefore, change in
organization results in lowering performance of firms in the short run. Finally, among factors
hampering innovation we find cost, market and other factors as statistically significant with
negative signs. This implies that difficulties of access to external finance, domination of
established enterprises and the existence of previous innovations reduce the firm’s
productivity in the current period.
As a separate group of factors we have also included dummy variables for industry specific
effects in all equations of the model. The coefficients are significant in a majority of cases
which suggests that there is some degree of individual heterogeneity present. In addition, the
coefficient for New EU countries is statistically significant in all but the first stage equation.

28
7. Sensitivity analysis
To check the sensitivity of above results we exclude feedback effect from the third stage,
productivity to innovation output, equation and estimate the system for innovation output and
productivity with two-stage least squares method. This method allows only one way causality
from the innovation output to productivity. Table 6 presents the results of this estimation. As
we can see the results for the main variable of interest, i.e. innovation input, do not differ
considerably for the full sample and the sub-sample of West European firms. However, the
coefficient on innovation input becomes insignificant in the second sub-sample. With respect
to the selection bias coefficient it is now significant in all three regressions suggesting that
there are some characteristics which distinguish innovators from non-innovators once we
take out the feedback from performance on innovation output. The coefficient on firm size
becomes insignificant for the sub-sample of firms in CEECs but it retains the negative sign in
all three regressions.
The changes in coefficients are noticeable with respect to other variables as well. In the full
sample, the variables for cooperation, other effects of innovation and EU subsidies lose their
significance but keep their sign. At the same time, the variables for process effects of
innovation, local subsidies, and institutional and other sources of innovation become highly
significant. The same findings hold when we restrict our sample to firms in West European
countries. In the sub-sample of firms in CEECs, national subsidies lose their significance. In
addition, the EU subsidies change their sign from negative to positive while all other
variables retain their signs and significance as in previous regressions (Tables 4 and 5).

Table 6: Results from the innovation output equation 2sls
29
Total CEECs WE
Sample size 15644 5444 10200
Firm size -0.391*** -0.096 -0.366***
Innovation input 0.330*** -0.065 0.296***
Mill’s ratio 0.427*** -0.131** 0.486***
Cooperation 0.004 0.089*** -0.013
Highly important effects of innovation
Product effects 0.184*** 0.286*** 0.114***
Process effects 0.069*** 0.068** 0.070***
Other effects 0.026 0.095*** 0.008
Organisational and
Organisational innovation 0.218*** 0.042 0.248***
Access to subsidies
Local/regional subsidies 0.219*** 0.092 0.224***
National subsidies -0.185** 0.042 -0.096
EU subsidies -0.114 0.209*** -0.119
Highly important sources of information
about innovation
Internal sources -0.309*** 0.041 -0.317***
Market sources -0.053** -0.001 -0.039
Institutional sources -0.124** -0.006 -0.143**
Other sources 0.098*** 0.061* 0.078**
Manufacturing -0.129 0.473*** -0.206
Trade -0.298** 0.217** -0.322***
Services -0.261** 0.370*** -0.295**
CEECs 0.223*** - -
Table 7 presents the results for the productivity equation when the feedback effect from
productivity to innovation output is excluded. We can see that the new specification does not
bring about any major changes with respect to the significance of coefficients or their signs
(compare to Table 5). The only exceptions are coefficients for sources of innovation being
cooperation with other enterprises or institutions in the full sample, which becomes
significant, and the cost factors hampering innovation whose effect in the Western Europe
sub-sample diminishes.
In sum, the findings in the productivity equation are unaffected by the exclusion of the
feedback effect from productivity to innovation output. However, the findings in the innovation
output equation are considerably changed which implies that many variables in this equation

are correlated with the excluded variable. These results confirm our expectation about the
two-way relationship between firm performance and innovation.
Table 7: Results from the productivity equation 2sls
30
Total CEECs WE
Sample size 15644 5444 10200
Firm size 0.403*** 0.179*** 0.415***
Innovation output 1.632*** 0.646*** 1.380***
Source of product innovations
Within enterprise 0.528*** 0.156* 0.904***
Together with other enterprises or institutions 0.266** 0.052 0.581***
Within enterprise 0.534*** -0.301*** 1.171***
Together with other enterprises or institutions 0.242** -0.011 0.509***
Organisational and
Organisational innovation -0.176** 0.159*** -0.206**
Marketing innovation -0.100* -0.074* 0.006
Factors hampering innovation
Cost factors -0.208*** -0.394*** -0.080
Knowledge factors -0.017 -0.080 0.092
Market factors 0.224*** -0.071 0.372***
Other reasons not to innovate -0.230** 0.279*** -0.511***
Manufacturing 0.106 -0.788*** 0.686***
Trade 1.297*** 0.516*** 1.496***
Services 0.156 -0.367*** 0.458***
CEECs -0.634*** - -

31
8. Conclusion
In recent years the multi-stage approach to the innovation process has become the
predominant method for modelling the relationship between innovation and firm performance.
It has been suggested that the generation of new knowledge and its transformation to
improved performance of firms occur through various stages with different factors affecting
different stages of this process. However, the evidence on factors influencing various stages
of innovation process differs in many respects, requiring further investigation to clarify these
differences. The four stage model, originally developed by Crepon et al. (1998), is capable of
capturing the complex nature of the innovation process by highlighting the different stages,
from the decision to innovate to the impact of innovation output on firms’ productivity. This
model makes it possible to take into account the joint inter-dependence of the four stages of
the innovation process.
In this paper we employed firm level data from the fourth Community Innovation Survey
(CIS4), covering around 90,000 enterprises in 16 European countries, in order to assess the
drivers of the innovation process in two different institutional settings, the mature market
economies of Western Europe and transition economies from Central and Eastern Europe. In
addition, the results for the two groups were compared with those obtained for the full
sample.
The empirical results show that investment in innovation activities positively influences the
sales of new products. This in turn contributes to the better productivity of firms. The positive
feedback effect from productivity to innovation output was found only when firms from all
countries are brought together. In mature market economies the feedback is not significant
while in case of transition countries we observe negative effect from productivity to
innovation output, probably reflecting the strong specialisation of firms from these countries
in labour intensive products. However, the exclusion of feedback from the estimation process
yields ambiguous results not only with respect to the four stages of innovation process but
also with respect to the other variables.
The results for determinants of four stages of innovation process seem to follow general
predictions from earlier theoretical and empirical literature on innovation. It was shown that
firms engage in innovation and decide how much to invest under the pressure of domestic
and international competition (the belief held strongly since the work of Schumpeter). Our
evidence indicates that in making this decision firms rely on knowledge accumulated from
previously abandoned innovations and use resources from other members of their group or

their associates and collaborators. An important conclusion, in comparison with earlier
studies, is the confirmation of the relationship between innovation and firm size as put
forward by Cohen and Klepper., (1996) . Larger firms are more likely to engage in innovation
activities and invest more in innovation but innovation output decreases with firm size.
There is some evidence that the organisational and marketing innovations facilitate
successful completion of the innovation process. However, our evidence also indicates that
firms in mature market economies implement changes in organisational routines slowly
which lead to their lower productivity. It is also evident that high cost of innovation acts as an
important obstacle for the decision on how much to spend on innovation in CEECs. The high
cost of innovation was also shown to lead to lower productivity of firms in both sub-groups
studied. From other factors hampering innovation we find that the lack of knowledge
motivates firms to engage in innovation while the existence of established innovations and
market structure negatively influence the firm’s decision to innovate and its productivity.
Among the factors facilitating the transformation of innovation input into innovation output
national and EU subsidies should be highlighted, particularly as the recipient firms were also
found to have lower levels of innovation output. This questions the validity of existing national
and EU schemes for subsidizing innovation. Product oriented effects of innovation seem to
be another important determinant of the innovation process. In addition, regulatory
requirements and environmental issues appear to contribute to higher levels of innovation
output. Among the sources of information about innovation we find evidence of significance
for internal, institutional and market sources of information in the investment stage of the
innovation process. However, at the output stage, only internal sources of information are
significant and positive for mature market economies. There is also evidence that, in all
samples, the more productive firms undertake product innovations in house. The same
finding holds for process innovations but only for firms in mature market economies.
32

33
References
Acs, Zoltan, and David Audretsch. “Innovation, Market Structure and Firm Size.” The Review
of Economics and Statistics, 1987: 567-574.
Adamou, Adamos, and Subash Sasidharan. “The Impact of R&D and FDI on Firm Growth in
Emerging-Developing Countries: Evidence from Indian Manufacturing Industries.” Vers.
Working Paper. Available at SSRN: http://ssrn.com/abstract=987024. April 2007. (accessed
06 26, 2010).
Aghion, P., and P. Howitt. Endogenous Growth Theory. Cambridge: MIT Press, 1998.
Bessler, Wolfgang, and Claudia Bittelmeyer. “Patents and the Performance of Technology
Firms: Evidence from Initial Public Offerings in Germany.” Financial Markets and Portfolio
Management, 2008: 323-356.
Bloom, Nicholas, and John Van Reenen. “Patents, Real Options and Firm Performance.” The
Economic Journal, 2002: C97-C116.
Chudnovsky, D., A. Lopez, and G. Pupato. “Innovation and Productivity in Developing
Countries: A Study of Argentine Manufacturing Firms’ Behaviour (1992–2001).” Research
Policy, 2006: 266-288.
Cohen, W., and S. Klepper. “A Reprise of Size and R&D.” The Economic Journal, 1996: 925-
951.
Crepon, Bruno, Emmanuel Duguet, and Jacques Mairesse. “Research, Innovation, and
Productivity:.” NBER Working Paper No. 6696, 1998.
Domadenik, Polona, Janez Prasnikar, and Jan Svejnar. “How to Increase R&D in Transition
Economies? Evidence from Slovenia.” Review of Development Economics, 2008: 193-208.
Goto, Akira, and Kazuyuki Suzuki. “R&D Capital, Rate of Return on R&D Investment and
Spillover of R&D in Japanese Manufacturing Industries.” The Review of Economics and
Statistics, 1989: 555-564.
Griliches, Zvi. “Productivity, R&D and Basic Research at the Firm Level in the 1970s.”
American Economic Review, 1986: 143-154.

Grossman, Gene M., and Elhanan Helpman. “Endogenous Innovation in the Theory of
Growth.” The Journal of Economic Perspectives, 1994: 23-44.
Hall, Bronwyn, and Francis Kramarz. “Introduction: Effects of Technology and Innovation on
Firm Performance, Employment and Wages.” Economics of Innovation and New
Technologies, 1998: 99-107.
Kemp, R.G.M, M. Folkeringa, J.P.J. de Jong, and E.F.M Wubben. Innovation and Firm
Performance. Scales Research Reports, Zoetermeer: EIM Business and Policy Research,
2003.
Kleinknecht, A., and R. Oostendorp. “R&D and export performance: Taking account of
simultaneity.” In Innovation and Firm Performance: Econometric Exploration of Survey Data,
edited by A. Kleinknecht and P. Mohnen, 310-320. Basingstoke: Palgrave, 2002.
Klette, Tor J., and Zvi Griliches. “Empirical Patterns of Firm Growth and R&D Investment: A
Quality Ladder Model Interpretation.” Economic Journal, 2000: 363-87.
Klomp, L., G.W. Meinen, A. Meurink, and M. Roessingh. “Knowledge Based Economy 2001:
R&D and Innovation in Netherlands.” Statistics Netherlands, 2002.
Klomp, Luuk, and George Van Leeuwen. “Linking Innovation and Firm Performance: A New
Approach.” International Journal of the Economics of Business, 2001: 343-364.
Knight, Frank H. Risk, Uncertainty and Profit. 1. Boston: Hart, Schaffner & Marx; Houghton
Mifflin Co, 1921.
Lichtenberg, F., and D. Siegel. “The Impact of R&D Investment on Productivity: New
Evidence Used Linked R&D-LDR Data.” Economic Inquiry, 1991: 203-229.
Loof, Hans. »Outsourcing, innovation and performance in service and manufacturing
industries.« Conference innovation and enterprise creation: Statistics and indicators. 23-24.
11 2000.
Loof, Hans, Almas Heshmati, Rita Asplund, and Svein-Olav Naas. “Innovation and
Performance in Manufacturing Industries: A Comparison of Nordic Countries.” SSE/EFI
Working Paper. 2002.
Loof, Hans, and Almas Heshmati. “On The Relationship Between Innovatoin and
Performance: A Sensitivity Analysis.” Economics of Innovation and New Technology, 2006:
317-344.
34

“Knowledge Capital and Performance Heterogeneity: A Firm-Level Innovation Study.”
International Journal of Production Economics, 2002: 61-85.
Masso, Jaan, and Priit Vahter. “Innovation and Firm Performance in a Catching-up
Economy.” Vers. Paper from Conference on Micro Evidence on Innovation and Development
(MEIDE). Maastricht Economic and social Research and training centre on Innovation and
Technology (UNU-MERIT). 2007.
http://www.merit.unu.edu/MEIDE/papers/2007/MASSO_VAHTER_Innovation%20and%20fir
m%20performance%20in%20a%20catching-up%20economy.pdf (accessed 06 26, 2010).
Nelson, Richard, and Sydney Winter. An Evolutionary Theory of Economic Change.
Cambridge, Massachusetts: Belknap Press, 1982.
Nguyen, Ngoc Anh, Quang Ngoc Pham, Dinh Chuc Nguyen, and Duc Nhat Nguyen.
“Innovation and Export of Vietnam’s SME Sector.” Vers. Paper No. 3256. Munich Personal
Repec Archive (MPRA). 07 11 2007. http://mpra.ub.uni-muenchen.de/3256/ (accessed 06
26, 2010).
OECD/Eurostat. OECD Proposed Guidelines for Collecting and Interpreting Technological
Innovation Data - Oslo Manual. Paris: OECD, 1997.
Raffo, Julio, Stephane Lhuillery, and Luis Miotti. “Northern and Southern Innovativity: A
comparison across European and Latin-American countries.” The European Journal of
Development Research, 2008: 219-239.
Romer, David. “Endogenous Technological Change.” The Journal of Political Economy,
October 1990: S71-S102.
Roud, Vitaliy. “Firm‐level Research on Innovation and Productivity: Russian Experience.”
Higher School of Economics, Institute of Statistical Studies and Economics of Knowledge,
Moscow, 2007.
Schumpeter, Joseph A. Capitalism, Socialism and Democracy. New York: Harper & Row,
1942.
The Theory of Economic Development: An Inquiry into Profits, Capital, Credit, Interest, and
the Business Cycle. New York: Oxford University Press, 1934.
35

Stoevsky, G. »Innovation and Business Performance of Bulgarian Companies (structural
econometric analysis at firm level).« Economic Research and Projections Directorate
Working Paper. Bulgarian National Bank, 2005.
Wakelin, Katharine. “Innovation and Export Behaviour at the Firm Level.” Research Policy,
1998: 829-841.
36

37
Appendix
Table A1: Explanation of variables
Dependent variables
Eq1: Decision to innovate Dummy variable; 1 if firm in 3 years prior to survey engaged in
intramural or extramural R&D, purchased new machinery,
equipment, software or other external knowledge, engaged in
training of personnel, market research or did any other
preparations to implement new or significantly improved products
and processes
Eq2: Innovation input (natural
logarithm)
Amount (in Euro) of expenditure on intramural or extramural R&D,
acquisition of machinery, equipment and software or acquisition
of other external knowledge in year of survey.
Eq3: Innovation output (natural
logarithm)
Percent of firm’s turnover in year of survey coming from goods or
services that were new to market or to enterprise in 3 years prior
to survey
Eq4: Labour productivity (natural
logarithm)
Turnover divided by number of employees in year of survey
Independent variables
Firm size (natural logarithm) Number of employees in 2004
National market Dummy variable; 1 if firm in past 3 years sold goods on national
market
EU Market Dummy variable; 1 if firm in past 3 years sold goods on EU, EFTA
or EU candidate countries marketsa
All other countries Dummy variable; 1 if firm in past 3 years sold goods on markets
of other countries
Part of a group Dummy variable; 1 if firm is part of an enterprise group
Abandoned or ongoing
innovations
Dummy variable; 1 if firm in past 3 years had any abandoned or
ongoing innovations
Cost factors Dummy variable; 1 if firm perceives the lack of funds, finance
from sources outside the enterprise and high costs of innovation
as highly important factors hampering its innovation activities,
projects or decision to innovate
Knowledge factors Dummy variable; 1 if firm perceives the lack of qualified
personnel, information on technology or markets or difficulties in
finding cooperation partners for innovation as highly important
factors hampering its innovation activities, projects or decision to
innovate
Market factors Dummy variable; 1 if firm perceives the domination over market
by established enterprises or the uncertainty of demand for
innovation goods and services as highly important factors
hampering its innovation activities, projects or decision to
innovate
Other reasons not to innovate Dummy variable; 1 if firm perceives a lack of need to innovate
due to prior innovations or low demand for innovations as highly
important factors hampering its innovation activities, projects or
decision to innovate
Organisational innovation Dummy variable; 1 if firm in past 3 years introduced new or
improved knowledge management system, changed
management structure, integrated different activities or introduced
changes in its relations with other enterprises or public institutions
(alliances, partnerships or subcontracting)
Marketing innovation Dummy variable; 1 if firm in past 3 years introduced significant
changes to packaging of goods or services or changed its sales
or distribution methods
Local/regional subsidies Dummy variable; 1 if firm in past 3 years received financial

38
Dependent variables
support for innovation activities from local/regional authorities
National subsidies Dummy variable; 1 if firm in past 3 years received financial
support for innovation activities from central government
EU subsidies Dummy variable; 1 if firm in past 3 years received financial
support for innovation activities from EU authorities
Internal sources of information
about innovation
Dummy variable; 1 if firm perceives sources of information within
enterprise or group as highly important
Market sources of information
about innovation
Dummy variable; 1 if firm perceives suppliers, customers,
competitors, consultants or R&D labs as highly important sources
of information on innovation
Institutional sources of
information about innovation
Dummy variable; 1 if firm perceives universities or government as
highly important sources of information on innovation
Other sources of information
about innovation
Dummy variable; 1 if firm perceives conferences, trade fairs,
exhibitions, publications, professional and industry associations
as highly important sources of information on innovation
Inverse Mill’s ratio Inverse Mill’s ratio from selection equation
Product oriented effects of
innovation
Dummy variable; 1 if firm perceives increased range of goods and
services, increase in market share or improved quality of goods
and services as highly important effects of innovations introduced
in 3 years prior to survey
Process oriented effects of
innovation
Dummy variable; 1 if firm perceives improved flexibility of
production, increased capacities, reduced costs of labour,
material or energy as highly important effects of innovations
introduced in 3 years prior to survey
Other effects of innovation Dummy variable; 1 if firm perceives reduced environmental
impacts, improved health and safety or meeting of regulatory
requirements as highly important effects of innovations introduced
in 3 years prior to survey
Cooperation Dummy variable; 1 if firm cooperated on innovations with other
enterprises or institutions in 3 years prior to survey
Sources of product innovations
Within enterprise Dummy variable; 1 if enterprise developed product innovations in
3 years prior to survey alone (base category: developed by other
enterprises or institutions)
In cooperation with other
enterprises or institutions
Dummy variable; 1 if enterprise developed product innovations in
3 years prior to survey in cooperation with other enterprises or
institutions (base category: developed by other enterprises or
institutions)
Within enterprise Dummy variable; 1 if enterprise developed process innovations in
3 years prior to survey alone (base category: developed by other
enterprises or institutions)
In cooperation with other
enterprises or institutions
Dummy variable; 1 if enterprise developed process innovations in
3 years prior to survey in cooperation with other enterprises or
institutions (base category: developed by other enterprises or
institutions)
Manufacturing Dummy variable; 1 if firm operates in manufacturing sector (base
category: other activities)
Service Dummy variable; 1 if firm operates in service sector (base
category: other activities)
Trade Dummy variable; 1 if firm operates in trade sector (base category:
other activities)
CEECs Dummy variable; 1 if firm belongs is one of the CEECs (base
category: West European countries)
a Includes following countries: Austria, Belgium, Bulgaria, Croatia, Cyprus, Czech Republic, Denmark, Estonia,
Finland, France, Germany, Greece, Hungary, Iceland, Italy, Ireland, Latvia, Liechtenstein, Lithuania, Luxembourg,
Malta, Netherlands, Norway, Poland, Portugal, Romania, Slovenia, Slovakia, Switzerland, Turkey, Spain,
Sweden, United Kingdom

39

CASE Network Studies and Analyses 410 - The Impact of innovation activities on firm performance using a multi-stage model: evidence from the Community Innovation Survey 4

More Related Content

What's hot (20)

Similar to CASE Network Studies and Analyses 410 - The Impact of innovation activities on firm performance using a multi-stage model: evidence from the Community Innovation Survey 4 (20)

More from CASE Center for Social and Economic Research (20)

Recently uploaded (20)

CASE Network Studies and Analyses 410 - The Impact of innovation activities on firm performance using a multi-stage model: evidence from the Community Innovation Survey 4