Content uploaded by Jutta Günther
Author content
All content in this area was uploaded by Jutta Günther
Content may be subject to copyright.
Institut für Halle Institute for Economic Research
Wirtschaftsforschung Halle
IWH-Diskussionspapiere
IWH-Discussion Papers
Does Qualification Drive Innovation?
A Microeconometric Analysis
Using Linked-employer-employee Data
Bianca Brandenburg, Jutta Günther, Lutz Schneider
September 2007 No. 10
Does Qualification Drive Innovation?
A Microeconometric Analysis
Using Linked-employer-employee Data
Bianca Brandenburg, Jutta Günther, Lutz Schneider
September 2007 No. 10
IWH __________________________________________________________________
IWH-Diskussionspapiere 10/2007 2
Authors: Bianca Brandenburg
Federal Ministry of Finance (BMF), Germany
Bianca.Brandenburg@bmf.bund.de
Dr. Jutta Günther
Halle Institute for Economic Research (IWH), Germany
Jutta.Guenther@iwh-halle.de
Lutz Schneider
Halle Institute for Economic Research (IWH), Germany
Lutz.Schneider@iwh-halle.de
The responsibility for discussion papers lies solely with the individual authors. The views ex-
pressed herein do not necessarily represent those of the IWH. The papers represent preliminary
work and are circulated to encourage discussion with the author. Citation of the discussion pa-
pers should account for their provisional character; a revised version may be available directly
from the author. Comments and suggestions on the methods and results presented are well-
come.
IWH-Discussion Papers are indexed in RePEc-Econpapers and in ECONIS.
This research has been partially financed by the EU Commission, in Framework Program 6,
Priority 7 on “Citizens and Governance in a knowledge based society”, contract no. CIT5-
028519. The authors are solely responsible for the contents which might not represent the opin-
ion of the Community. The Community is not responsible for any use that might be made of
data appearing in this publication.
Herausgeber:
INSTITUT FÜR WIRTSCHAFTSFORSCHUNG HALLE – IWH
Prof. Dr. Ulrich Blum (Präsident), Dr. Hubert Gabrisch (Forschungsdirektor)
Das IWH ist Mitglied der Leibniz-Gemeinschaft
Address: Kleine Märkerstraße 8, 06108 Halle (Saale)
PO box: Postfach 11 03 61, 06017 Halle (Saale)
Phone: (0345) 77 53-60
Fax: (0345) 77 53-8 20
Internet: http://www.iwh-halle.de
__________________________________________________________________ IWH
IWH-Diskussionspapiere 10/2007 3
Does Qualification Drive Innovation?
A Microeconometric Analysis
Using Linked-employer-employee Data
Abstract
Degree-level science and engineering skills as well as management and leadership skills
are often referred to as a source of innovative activities within companies. Broken down
by sectoral innovation patterns, this article examines the role of formal education and
actual occupation for product innovation performance in manufacturing firms within a
probit model. It uses unique micro data for Germany (LIAB) that contain detailed in-
formation about innovative activities and the qualification of employees. We find sig-
nificant differences of the human capital endowment between sectors differentiated ac-
cording to the Pavitt classification. Sectors with a high share of highly skilled employees
engage in product innovation above average (specialized suppliers and science based in-
dustries). According to our hitherto estimation results, within these sectors the share of
highly skilled employees does not, however, substantially increase the probability to be
an innovative firm.
Key words: innovation, human capital, qualification, sectoral innovation system
JEL classification: O31, J 24
IWH __________________________________________________________________
IWH-Diskussionspapiere 10/2007 4
Zusammenfassung
Natur- und ingenieurwissenschaftliche Fähigkeiten sowie Management- und Führungs-
kompetenzen werden häufig als Quelle von betrieblichen Innovationsaktivitäten betra-
chetet. Der vorliegende Artikel untersucht die Rolle von Humankapital im Sinne des
formalen Bildungsabschlusses und des tatsächlich ausgeübten Berufes für die betriebli-
che Innovationstätigkeit im Rahmen eines Probit-Ansatzes, wobei zwischen sektoralen
Innovationsregimen unterschieden wird. Die Analyse basiert auf einem Mikrodatensatz
deutscher Betriebe (LIAB), welcher detailierte Informationen über die Innovationsakti-
vitäten und die Qualifikation der Beschäftigten enthält. Es zeigen sich signifikante Un-
terschiede der Humankapitalausstattung zwischen Sektoren, welche nach der Pavitt-
Klassifikation unterschieden wurden. Sektoren mit einem hohen Anteil hochqualifizier-
ter Beschäftigter sind überdurchschnittlich oft unter den Produktinnovatoren zu finden
(spezialisierte Zulieferer und wissenschaftsbezogene Branchen). Indes lassen die reali-
sierten Regressionen keine signifikanten Effekte der Beschäftigtenqualifikation auf die
Innovationstätigkeit innerhalb dieser Branchen erkennen.
Schlüsselwörter: Innovation, Humankapital, Qualifikation, sektorale Innovationssysteme
__________________________________________________________________ IWH
IWH-Diskussionspapiere 10/2007 5
Does Qualification Drive Innovation?
A Microeconometric Analysis
Using Linked-employer-employee Data
1 Introduction1
Education, R&D and innovation rank very high in today’s policy agendas. Frequently
citied in this context is the Lisbon strategy of the EU. Although the originally ambigu-
ous goals have been revised recently (European Commission 2005), the agenda is still
recognized as a political milestone in support of the knowledge based economy. Com-
plementary to this, Germany, like many other EU countries, launched a national master
plan, too. The “High-tech Strategy for Germany” (BMBF 2006a) emphasises the need to
focus on the creation of new knowledge and particularly on the translation of new
knowledge and inventions into marketable products.
The need for action is obvious. On the one hand, we face an ongoing structural change
towards a knowledge based society (Heidenreich 2003). On the other hand, Germany
faces decreasing numbers of university students (Statistisches Bundesamt 2005), a
demographic change towards an aging society (Statistisches Bundesamt 2006), and a
lack of qualified workers (Reinberg and Hummel 2004, BMBF 2006b, pp. 61 sqq.). As
shown by innovation survey data, the lack of qualified employees as a hampering factor
for innovation is even stronger in Germany compared to most other EU countries (Luck-
ing 2004, p. 18).
In this paper we take a closer look on the relationship between human capital and inno-
vation. This relationship is not one-dimensional. Concerning the heterogeneity of la-
bour, new technologies often require organizational changes and different qualifications.
A wide range of literature addresses the skill biased technological change and empirical
findings indicate that innovation is generally associated with an increase in high-skilled
and a decline in low-skilled employment (e.g. Pianta 2005, pp. 575 sqq.; Blechinger and
Pfeiffer 1999).
On the other hand, we can regard human capital as a central determinant or input of in-
novation. This paper explicitly considers the impact of the human capital endowment in
terms of qualification on product innovation processes in manufacturing firms. From a
1 The paper has been presented at the “Workshop on Economics of Knowledge and Innovation” on
July 11, 2007 at the Halle Institute for Economic Research (IWH). Thanks go to the participants of
the workshop who gave us helpful comments and recommendations. Furthermore the authors thank
the Institute for Employment Research Nuernberg (Institut für Arbeitsmarkt- und Berufsforschung,
IAB) for the provision of data.
IWH __________________________________________________________________
IWH-Diskussionspapiere 10/2007 6
scientific point of view the role of human capital as an important input for innovation is
well recognized and documented in the new growth theory. Nevertheless, current em-
pirical studies explicitly investigating the relationship between human capital and firms’
innovation performance are rare.
A recent study of Dakhli and De Clercq (2003) finds evidence for the importance of
human capital as a determinant of innovation. The results are based on a cross-country
analysis, where innovation is proxied through patents, R&D expenditure, and high-tech
exports. But direct innovation measures and firm level data should be preferred in our
context. Rammer et al. (2005, pp. 214 sqq.) use innovation survey data for Germany and
provide evidence for the importance of human capital as a determinant of firms’ overall
innovation activity. When looking at particular types of product innovation, however,
the coefficient turns insignificant or even significantly negative. An empirical study
from Günther and Gebhardt (2005) provides similar results. Using micro data for estab-
lishments (local business units) in East Germany they find no significant impact of hu-
man capital on establishments’ innovation activity. In both analyses, human capital is
measured as the share of employees with a higher education degree.
To sum up, existing studies show different results, and they use education degrees as a
measure for human capital. In this paper, we make use of alternative micro data for
Germany – a linked employer-employee data set – which allows us to consider the ac-
tual occupation of employees instead of just the formal qualification and the duration of
employment. Furthermore, based on the idea of sectoral innovation systems, we con-
sider branch differences within manufacturing industry, too. The following chapter pre-
sents the theoretical considerations followed by the introduction of the econometric
model and the data in Chapter 3. Finally, estimation results are presented (Chapter 4)
and conclusions drawn (Chapter 5).
__________________________________________________________________ IWH
IWH-Diskussionspapiere 10/2007 7
2 Innovation in Skill Related and Sectoral Perspective
2.1 Human Capital and Innovation
New Growth Theory
The technological progress in common growth models often refers to innovation as an
important source of economic growth. Different from traditional growth theory, models
of endogenous growth relate the human capital stock to a country’s ability to innovate
and catching-up with more advanced economies and specify technological change or the
growth of total factor productivity as a function of human capital. Thereby investments
in human capital and R&D lead to technological change (innovation) and increase the
productivity of labour and capital at firm level constantly (Romer 1986; Aghion and
Howitt 1998). Due to the public good character of technology, spillovers occur between
firms, and the economy faces increasing returns to investment and thus long run growth
(Grossman and Helpman 1997). There are various specifications in new growth models
that particularly stress the role of human capital (Barro and Sala-i-Martin 1998, pp. 200
sqq.), but the wide spectrum would go beyond the scope and the need of this paper.
What remains important for the purpose of this paper is the fact that the new growth
theory underlines the importance of qualified employees as an input for the R&D sector
where new knowledge is created and subsequently introduced in the form of new prod-
ucts etc. Basically, one can assume that an increasing supply of human capital leads to a
better performance of innovation.
While endogenous growth theory takes a macro perspective, we might also assume that
the central message of new growth theory – human capital as an important determinant of
innovation – applies at the firm level, too. However, the mechanisms through which
highly qualified people contribute to innovation remain an unexplored topic in economic
theory. In search for a stronger theoretical backing at the firm level, we additionally con-
sult theoretical approaches concerning the interdisciplinary field of innovation studies.2
Into the Black Box: Innovation Studies
Contributions, usually assigned to the area of innovation studies or systemic innovation
theory take a holistic view and contribute to a better understanding of the nature of the
innovation process as such.3
The traditional model of science-push (Bush 1945) stressed the importance of R&D as
well as science and engineering skills in the sense of a small elite group for innovation
processes. Later on, this so-called linear model has been extended by the perspective
2 For an insightful discussion of “innovation studies” as a discipline, see Fagerberg and Verspagen (2006).
3 For an overview, see Fagerberg et al. (2005).
IWH __________________________________________________________________
IWH-Diskussionspapiere 10/2007 8
that innovation is an interactive processes that largely involves inter-personal as well as
inter-organizational learning too (e.g. Kline and Rosenberg 1986).
As regards human capital more specifically, Nelson and Phelps (1966) documented in a
simple growth model that better educated people fulfil regular activities more effec-
tively, and that they are more competent in the use and exploitation of new technologies.
The latter aspect has been proven empirically by showing that high-educated farmers in-
troduce new technologies quicker and with better results than average. Similar findings
have been documented by Schultz (1975), who refers to the exploitation competence as
an ‘entrepreneurial’ capability.
Lundvall (2007) picks up this topic and develops it further in the context of the ‘learning
economy’ (Lundvall and Johnson 1994). He states that there are two ways by which
higher education impacts on innovation: On the one hand, higher education graduates
can operate as basic innovators for instance by inventing and developing new technolo-
gies. On the other hand, they might serve as second stage innovators, who rather exploit
the technological progress and assure the ‘equilibrium’ between technological change
and daily business. According to this differentiation, he concludes that engineers and
scientists are particularly active as basic innovators while people with a management
and social sciences degree are important as second stage innovators.
Human capital covers knowledge, embodied skills, and expertise that people bring into or-
ganizations and society. One important component of human capital is the formal qualifi-
cation, and as indicated above, especially tertiary education is viewed as a crucial determi-
nant for innovative activity. Accordingly, we formulate our first hypothesis as follows.
Hypothesis 1: The higher the human capital endowment in terms of engineers, scientist,
and managers, the higher the company’s innovation output.
2.2 Sectoral Innovation Regimes
In the tradition of evolutionary theory, the theoretical concept of ‘sectoral innovation
system’ starts from the idea that firms are not homogenous regarding their innovation
processes. Instead, sectors largely differ with respect to their innovation processes.
Malerba (2005) explains this along three dimensions: i.e. knowledge and technological
domain, actors and networks and institutions.
From the literature, we know different approaches to make distinctions among sectors
regarding their technological or innovation regime.4
The simplest classification, frequently used in international comparative studies, is the
one made by the OECD, developed by Hatzichronoglou (1997). According to R&D in-
4 For a recent overview of industry classifications in general, see e.g. Peneder (2003).
__________________________________________________________________ IWH
IWH-Diskussionspapiere 10/2007 9
tensity one can distinguish high-technology, medium-high-technology, medium-low-
technology, low-technology industries.
An expanded industry classification, frequently used in empirical innovation studies,
was introduced by Pavitt (1984). In his view, several sources matter for innovation, not
only own R&D but also aspects like supplier-customer relations, learning-by-doing or
learning-by-interacting, etc. Based on a very extensive data set on innovation for the
UK, he distinguishes four categories according to different innovation patterns, which
have their own requirements for skill-sets.5
(1) Science-based industries are characterized by much organized R&D with a strong
link to university or other publicly funded basic research. These industries require high-
level science and engineering skills, such as in chemical industry or electronics.
(2) Specialized suppliers are characterized by a close relationship with frequent users.
Firms in this category strongly focus on product innovations and require skills of interactive
learning as well as the capacity to develop highly client specific solutions and vocational,
practical development skills. A typical example for specialized suppliers is machinery.
(3) Scale intensive industries are production intensive companies with rather simple
production, and often with mass products. Innovation is mostly process oriented. R&D
activities predominantly serve internal purposes. Economies of scale require scientific
managers with cross-functional skills, specialists in product design, development skills
as well as a qualified workforce that is able to adapt new technologies (e.g. transport
equipment, steel industry).
(4) Supplier dominated industries tend to be oriented towards process innovation. Op-
erators in this category are mostly defined in terms of their professional skills, design,
brand and advertising. Technological innovations, however, mainly come from outside
the companies. In-house R&D and engineering capabilities are considered to be weak
(e.g. textile industry).
According to Pavitt (1984), the science-based industries and the specialized suppliers
serve the rest of the economy with new technology. Thereby, scale intensive industries
mostly take over and adapt external technology while supplier dominated industries
hardly fulfil own development activities. With respect to human capital, we formulate
the following hypothesis.
Hypothesis 2: The higher the original innovation activity of a sector, the stronger the
importance of a highly qualified workforce.
5 The assignment of industries (three digit level) to the four Pavitt categories based on International
Standard Industrial Classification of All Economic Activities, Revision 3 (1990) is shown in Appen-
dix (see Table 4).
IWH __________________________________________________________________
IWH-Diskussionspapiere 10/2007 10
3 Model and Data
3.1 Econometric Model
Besides bivariate descriptive analysis, the hypotheses are tested on the basis of a micro-
econometric probit model, since the firm’s innovation activity as dependent variable is
measured by a binary variable. The innovation variable is regressed on variables repre-
senting the firm’s human capital endowment in terms of qualification. In order to avoid
regression biases due to the problem of omitted variables, almost every central impact
on the innovation behavior of the firm – additional to the primarily interesting level of
qualification – has to be included in the estimation. In accordance to the empirical litera-
ture focusing on determinants of innovation activity the following exogenous variables
are taken into account:6
Firm size: The size of an enterprise is assumed to facilitate the innovation activity due
to more favorable conditions to finance innovations, the availability of real and human
capital resources, and the exploitation of scale effects.7
R&D activities: According to the ‘science push model’ of innovation, R&D is a central
source of innovation. Although this one-dimensional perspective has been extended in
the meantime, we still have to assume that enterprises should be particularly innovative
if they employ resources for the development of new products.
Job tenure & experience: On the one hand, a longer work experience within the same
firm should drive innovation, since experienced employees have learned from past in-
novation problems. Therefore, the risk of innovation failure is reduced. On the other
hand, experience might cause technological inertia limiting the scope and intensity of
innovations. Hence, from a theoretical point of view, the impact of experience on inno-
vation output is quite ambiguous and the empirical literature has largely neglect this
topic so far.
Further training: This variable tells whether the firm invests in further education of the
employees. In the sense of life-long-learning, such activities add to the knowledge and
capabilities of the workforce and are associated with a positive impact on the innovation
behavior.
6 A discussion of the variables selected here can be found in Günther and Gebhardt (2006); Gottschalk
and Janz (2003); Rammer et al. (2005).
7 This assumption originally dates back to Schumpeter, the pioneer of innovation research. Recent em-
pirical studies indicate that a linear relationship between size and innovation cannot clearly be con-
firmed any longer (Gottschalk and Janz 2003).
__________________________________________________________________ IWH
IWH-Diskussionspapiere 10/2007 11
Profitability: The financing of innovation activities predominantly comes from internal
resources of the firm since banks are usually reluctant to provide capital for risky pro-
jects like innovation. Accordingly, a profitable firm will be more likely to generate the
monetary resources needed for innovations.
Export intensity: Firms selling their products on foreign markets are subject to global
competition forces – a survival under strong competition should require persistent inno-
vation efforts.
Age of the firm: A high age of a firm might indicate the ability to meet market chal-
lenges sufficiently, thus ample adaptation capacities could be expected. From this point
of view the age of a firm should be positively correlated to its innovation activities.
However, one reason for the emergence of enterprises might refer to the fact that exist-
ing (older) firms will resist radical types of innovations – e.g. due to path dependencies.
Therefore, the impact of the age of the firm is not clear cut.
Equipment: A sufficient technological standard is a precondition for the feasibility of
elaborate innovation types. Moreover, the technical equipment complements the absorp-
tive abilities of an enterprise. Hence, a high level of technology should promote the in-
novation propensity of firms.
Foreign ownership: In order to control for different access to non-market knowledge
flows, a dummy variable measuring a majority foreign ownership is implemented. Due
to an easier import of advanced technology from the multinational enterprise group, a
foreign owned firm should face advantages in innovation processes.
East-location: Due to regional distinctions resulting from the transition period, a
dummy is included controlling for an unexplained East-effect, thereby expecting a lower
innovativeness of firms located in the Eastern part of Germany.
Thus, the estimation equation has the following general form:
* ' X
i i i i
y HK e
α β γ
= + + +
1 if yi* >0
with yi=
} 0 if yi* ≤0 and e ~ N(0,1)
Where yi denotes our binary outcome, which takes the value of 1 if firm i is active in
product innovation, and y* is a latent variable. HK is our qualification variable, denoting
the share of high-qualified employees respectively in terms of formal education or occu-
pational characteristics alternatively. γ denotes a vector of coefficients for the above de-
scribed exogenous control variables in Xi, α represents the constant, and e denotes the
error term. The estimations are limited to the manufacturing sector (without construc-
IWH __________________________________________________________________
IWH-Diskussionspapiere 10/2007 12
tion) and to firms with at least 10 employees.8 The model is estimated for the entire sam-
ple and separately for each of the four sub-samples according to the Pavitt categories.
3.2 Data
The analysis is carried out on the basis of the linked-employer-employee dataset (LIAB),
provided by the Institut für Arbeitsmarkt- und Berufsforschung Nuremberg, Germany.9
The dataset contains firm-level data from the IAB-Betriebspanel, an annual panel survey
of about 15,000 German firms, and individual data of the employees working in the
panel firms. The individual statistics covers all workers, which are in the scope of the
national social insurance system. For the topic of this paper, the LIAB dataset is an ap-
propriate data base since the firm-level data about innovation activity and other relevant
firm characteristics can be combined with information on the qualification level of the
firms’ employees. Hence, the question how the qualification of a firm’s workforce af-
fects its innovation behavior can appropriately be addressed. An advantage of the data
set consists in the rich information about the qualification structure. Qualification can be
measured not only in terms of formal education (degrees), but also in terms of the actual
occupational status. So, the data precisely allow for detecting the actual qualification
level within a firm.
The dependent variable stemming from the panel survey is binary coded.10 A value of 1
is assigned if the firm is engaged in product innovation. Three categories of product in-
novation are distinguished in the data set:
i) Improvement of an existing product (improvement)
ii) Introduction of a product new to the firm – extension of the product range
(new product)
iii) Creation of a market novelty (market novelty)
In addition, the aggregate variable product innovation is set to 1 if at least one of the
three types of product innovation was realized.
The collection of innovation data through the IAB-Betriebspanel largely corresponds to
the international standards of innovation surveys provided in the ‘Oslo Manual’ (OECD
8 Innovations in the other sectors – in particular regarding the service industries – are difficult to iden-
tify and factors driving innovation cannot be easily determined (Hempel 2003). Under this condi-
tions, an estimation runs the risk to neglect substantial impact of innovation behavior, the estimation
coefficients will therefore be biased.
9 For a description of the data set see Alda (2005) and Alda and Herrlinger (2005).
10 A detailed description of endogenous and exogenous variables is given in the Appendix (see Table 3).
__________________________________________________________________ IWH
IWH-Diskussionspapiere 10/2007 13
2005). Product innovations are subject to the survey every three years. Process innova-
tions are not subject to the survey at all.11
The qualification variable is based on occupational status, which is reported in the em-
ployee’s statistics of the LIAB. According to the typology of Blossfeld (1985), an em-
ployee is classified as high qualified if he or she performs a job as an engineer, scientist,
or manager. These occupations usually require formal education of the tertiary level. Al-
ternatively, the formal education (tertiary degree) is used as qualification variable.12
The second variable stemming from the LIAB is experience. To control for different
stocks of work experience, three categories of job tenure within the firm are distinguished
(up to 1 year, 1-5 years, above 5 years). The other variables are taken from panel survey,
so the information rely on the firm’s own assessment. Firm size is measured by the loga-
rithm of the number of employees. Export intensity is defined as the share of sales abroad.
Further training activities are measured by the ratio of further training participants to the
number of employees. The remaining control variables are implemented as dummy vari-
ables. The R&D variable is set to 1 if the firm is engaged in R&D activities or coopera-
tion. If the firm rates its profitability as at least ‘good’ the corresponding dummy is set to
1. Due to lacking differentiation, the age of the firm has to be implemented as binary vari-
able, too. A value of 1 is assigned if the enterprise was founded before 1990. Foreign
ownership is set to 1 if the majority of the firm is owned by foreigners. The value 1 is as-
signed to the equipment dummy if the firm rates its technological level as ‘state of the art’.
Of course, the East dummy is 1 if the firm is located in the area of the former GDR.
The probit estimation is performed for the most recent year available, which is 2004. After
the exclusion of non-manufacturing firms, firms with less than 10 employees and firms
with missing values, 1,307 firms remain in the sample. The data about the innovation ac-
tivity refer to the period of two years preceding the survey, which has been carried out in
June 2004. The exogenous variables relate to 2002, i.e. the year before the innovation.13
The implementation of lagged variables is necessary to address the problem of endoge-
neity. Because innovation may itself lead to adjustments of the production system, the
exogenous variables should measure the inputs before innovation took place. The use of
a lagged model meets – at least to some degree – the problem of causality.
11 Information on organizational innovations, related to management, labor organization, quality control
etc., is available. But since organizational innovations follow a very different logic, especially in the
sectoral perspective (Lam 2005), we exclude them from our analysis.
12 As to be seen in Chapter 4, the qualification variable based on occupational status is a more suitable
concept since the operationalisation via formal education includes employees with a tertiary degree
though, performing jobs being not classified as highly qualified.
13 Due to data availability, only the further training variable refers to 2001. Values of the R&D variables
are taken from 2004, because earlier surveys do not contain information about R&D cooperation.
IWH __________________________________________________________________
IWH-Diskussionspapiere 10/2007 14
4 Empirical Results
4.1 Descriptive Analysis
Regarding innovation activities in the four sectoral groups, we can confirm the innova-
tion patterns described by Pavitt (1984). Science based industries and specialized sup-
pliers make up for 84 percent and 77 percent of product innovators respectively while
supplier dominated and scale intensive industries only account for 67 and 55 percent re-
spectively (see Table 1). The same pattern is found for the different types of innovation.
Especially market novelties are primarily developed within the group of science based
industries. Among companies of the supplier dominated sector, only 5 percent develop
market novelties, whereas 30 percent of companies in the science based sectors are ac-
tive in this field.14
Table 1:
Sector specific share of innovators (%) by types of innovation
Sector
Share of innovators
Supplier dominated in-
dustries Scale intensive in-
dustries Specialized
suppliers Science
based industries
Product innovation 54.7% 67.2% 77.3% 83.8%
Improvement 52.2% 64.6% 73.1% 79.1%
New product 16.4% 25.1% 32.0% 32.4%
Market novelty 4.7% 11.8% 19.1% 29.1%
Organizational innovation 63.4% 66.8% 70.6% 70.9%
Sample size 232 618 309 148
Source: LIAB 2001-2004.
A similar picture arises from the qualification structure (see Table 2, first row). The
share of employees with a tertiary education ranges between 5 and 15 percent. Supplier
dominated and scale intensive industries employ relatively few formally high-qualified
employees, whereas specialized suppliers and science based industries employ more
people with a higher education degree (12 and 15 percent respectively).
According to the occupational status (see Table 2, second row), the share of higher
qualified employees (engineers, scientists, and managers) ranges between 4 and 11 per-
14 Furthermore, Table 1 shows, that organizational innovations differ much less across the four groups.
As mentioned before, they are subject to a different pattern of innovation behaviour. Thus, the deci-
sion to restrict the analysis to product innovations is supported by the data.
__________________________________________________________________ IWH
IWH-Diskussionspapiere 10/2007 15
cent.15 The highest share of high-qualified employees arises in such industries that are
above average active in product innovation. Thus, our first hypothesis can be confirmed
through the descriptive data analysis.
Table 2:
Sector specific share of high qualified employees
Sector
Qualification indicator
Supplier dominated
industries Scale intensive
industries Specialized sup-
pliers Science based in-
dustries
High qualification -
measured by formal education 5.1% 7.0% 11.9% 15.1%
High qualification -
measured by occupational status 4.3% 5.7% 9.6% 11.1%
Source: LIAB 2001-2004.
We now look at the qualification structure of innovative and non-innovative firms
within the Pavitt categories (see Table 3). For the low innovation sectors (supplier
dominated and scale intensive industries) the employment of high-qualified people is
not or only slightly higher in companies that are active in product innovation. Remark-
able differences arise among innovators and non-innovators in specialized suppliers and
science based industries. In both sectoral groups the level of qualification is obviously
higher for innovators than for non-innovators.16 Thus, there is some descriptive evi-
dence that qualification is more important in companies concerned with original innova-
tions than in low innovation sectors that mostly take over and adapt external technology
or hardly fulfil own development activities. Therefore, Hypothesis 2 seems to be sup-
ported by the descriptive analysis too.
One can assume that the higher share of qualified employees, especially among innova-
tors in the group of specialized suppliers and science based industries, is an expression
of the fact that these firms employ more R&D personnel than others. Thus, in a further
step we look at the differences in the share of high qualified employees according to the
R&D participation of firms (see Table 4). Apart from supplier dominated industries, the
share of high qualified employees is higher in firms with own R&D activities compared
to firms without R&D activities. This effect in especially visible in the group of special-
ized suppliers (12.4% versus 5.7%).
15 As indicated above (Footnote 11), there are employees with a tertiary degree, but not working in po-
sitions that are classified as high qualified. This is shown by the fact that the share of high-qualified
employees – measured by formal education is higher than the share of high qualified employees –
measured by occupational status (see Table 2).
16 Within the science based industries one exception occurs: The share of high-qualified employees in compa-
nies, which upgrade their product range (‘new product’) is lower than in non innovative companies.
IWH __________________________________________________________________
IWH-Diskussionspapiere 10/2007 16
Table 3:
Sector specific share of high qualified employees (occupational status) according to in-
novators and non-innovators and type of innovation
Sample
Type of Innovation
Supplier
dominated industries Scale intensive indus-
tries Specialized suppliers
Science based indus-
tries
Innovation ( yes/no) yes No Yes no yes no yes no
Product innovation 4.0% 4.7% 5.7% 5.6% 10.4% 6.7% 11.5% 9.2%
Improvement 4.0% 4.6% 5.7% 5.5% 10.8% 6.4% 11.3% 10.3%
New product 4.4% 4.3% 5.9% 5.6% 10.8% 9.0% 10.2% 11.6%
Market novelty 6.8% 4.2% 6.3% 5.6% 14.0% 8.6% 12.5% 10.5%
Source: LIAB 2001-2004.
Table 4:
Sector specific share of high qualified employees (occupational status) according to
R&D activity
Sample
R&D participation
Supplier dominated in-
dustries Scale intensive
industries Specialized sup-
pliers Science based in-
dustries
R&D existent 4.2% 6.5% 12.4% 12.3%
R&D nonexistent 4.3% 5.0% 5.7% 9.1%
Source: LIAB 2001-2004.
Obviously, a high share of high qualified employees and R&D activities are intercon-
nected. Therefore, we run the regression analyses also with an interaction term of human
capital and R&D, expecting a positive impact on innovation.
4.2 Estimation Results
Firstly, we run the regression analysis with the full sample (see Table 5). When includ-
ing qualification and R&D without interaction term (Model I), the qualification variable
does not turn out to be significant.17 Other commonly estimated effects stemming from
R&D activity, firm size, and export intensity appear to be significant with the anticipated
17 For all estimations we present the coefficients for the linear relationship of the underlying latent va-
riable. The coefficients indicate the sign and significance of influence, but are not interpretable in
terms of magnitude.
__________________________________________________________________ IWH
IWH-Diskussionspapiere 10/2007 17
positive direction of influence. Furthermore, the two dummies for the science based and
the specialized supplier industries have a significant positive impact on the probability of a
firm’s product innovation activity which corresponds to our expectations.18
Table 5:
Regression results of the probit estimation without interaction between R&D and quali-
fication (Modell I) and with interaction (Modell II) (full sample)
Model I
(no interaction of qualification
and R&D)
Model II
(interaction of qualification and
R&D)
Dependent Variable:
Product innovation
Coefficient z-value Coefficient z-value
High qualification (occupational status) -0,938 -1.61 -2.2565*** -2.82
R&D activities 1.201*** 11.85 0.9711*** 7.14
Interaction R&D – high qualification - - 3.2868** 2.41
Further training 0.150 0.66 0.1304 0.57
Job tenure max. 1 year 0.621 1.38 0.5893 1.31
Job tenure 1-5 years 0.292 1.51 0.2505 1.29
Firm size 0.187*** 4.76 0.1963*** 4.95
Export intensity 0.941*** 4.62 0.9516*** 4.63
Profitability 0.138 1.55 0.1497* 1.65
Equipment 0.019 0.19 0.0068 0.07
Age of the firm 0.021 0.19 0.0243 0.22
East 0.046 0.45 0.0562 0.55
Foreignness 0.015 0.11 -0.0005 -0.00
Scale intensive industry 0.063 0.56 0.0646 0.58
Specialized supplier 0.230* 1.70 0.2118 1.56
Science based industry 0.311* 1.75 0.3195* 1.78
Constant -1.293*** -5.10 -1.2436*** -4.88
Sample size 1,307 1,307
LR-Test 410.96*** 417.78***
McFadden R2 0.255 0.259
Significance levels: *** 1%, ** 5%, * 10%.
Source: LIAB 2001-2004.
18 The results with respect to the qualification variable do not change when we exclude the R&D variable.
IWH __________________________________________________________________
IWH-Diskussionspapiere 10/2007 18
When we include an interaction term of qualification and R&D (Model II), the qualifica-
tion variable turns out to be significant, but with a negative sign while the interaction
term exhibits a significantly positive impact. This means that if R&D and qualification
occur together in a firm, they clearly have a positive impact on the firm’s propensity to
carry out a product innovation. The negative sign of the qualification variable implies
that high qualified personnel in a firm without R&D rather hinders innovation. How-
ever, this somehow surprising effect might stem from firms in the sample, which have a
high share of qualified people (engineers, scientists and managers), but do not engage in
any product innovation activity.
As we have seen in the descriptive part, the correlation between qualification and sec-
toral innovation patterns is quite high. Thus, the impact of the human capital variable
could possibly be covered by the dummy variables for the Pavitt categories. In order to
control for this, we run the regressions separately for sectoral sub-samples according to
Pavitt’s industry categories (see Table 6). But here again, the qualification variable does
not appear to have a significantly positive (basic) effect. The interaction term exhibits a
significantly positive impact only in the group of specialized suppliers while the basic
effect of qualification is significantly negative here. This finding might be related to the
fact that in specialized supplier firms the R&D and production activities are closely
connected (e.g. production of special equipment in small charges or single-unit accord-
ing to particular customer order).
A similar picture arises if the dependent innovation variable is disaggregated into the
three types of product innovations: improvement, new product, or market novelty (see
Appendix, Tables 1-2).
One explanation for the sector specific findings might be the occurrence of differences
in the qualification level especially between and not within the sectoral innovation cate-
gories. Within the Pavitt categories, firms differ only slightly in respect of the share of
high skilled employees, and thus, innovation activity is not affected. This might indicate
that in terms of the employment of high-skilled persons, the qualitative characteristics
could be more important than quantitative ones. Although the quantity of high-skilled
employees differs only slightly within the sectoral groups, highly qualified staff could
differ in terms of their specific discipline, university background, and respective im-
parted knowledge and skills. Descriptive statistics, however, reveal that this is only
partly true. At least for the specialized suppliers, there are clear variations in the share of
highly qualified employees.
__________________________________________________________________ IWH
IWH-Diskussionspapiere 10/2007 19
Table 6:
Regression results of the Probit estimation of Model II (with interaction)
Supplier domi
nated
industries Scale intensive
industries Specialized sup-
pliers Science based indus-
tries
Dependent Variable:
Product innovation Coefficient
(z-value) Coefficient
(z-value) Coefficient
(z-value) Coefficient
(z-value)
High qualification
(occupational status) -1.7072
(-0.92) -1.9408
(-1.45) -2.6019*
(-1.66) -1.6265
(-0.70)
R&D activities
1.5566***
(3.57) 0.8332***
(4.24) 0.8402***
(2.71) 1.1902**
(2.04)
Interaction R&D – high qualifica-
tion 1.5856
(0.22) 3.5061
(1.40) 4.1660*
(1.80) 6.5115
(1.18)
Further training
2.0535**
(2.41) -0.0103
(-0.03) 0.0589
(0.12) -0.5668
(-0.77)
Job tenure max. 1 year
-0.1666
(-0.17) 0.8692
(1.37) 1.2513
(1.17) -4.5096*
(-1.71)
Job tenure 1-5 years
0.8625*
(1.76) 0.4721*
(1.64) -0.2133
(-0.51) -0.8964
(-1.42)
Firm size
0.1705*
(1.67) 0.2032***
(3.69) 0.2928***
(2.98) 0.1913
(1.37)
Export intensity
1.9098***
(3.19) 1.2821***
(3.96) 0.7381*
(1.78) -0.5538
(-0.89)
Profitability
0.1181
(0.55) 0.1527
(1.21) 0.1748
(0.87) 0.3375
(0.98)
Equipment
-0.0898
(-0.38) -0.0831
(-0.62) 0.0146
(0.07) 0.3287
(0.85)
Age of the firm
-0.0780
(-0.28) 0.1726
(1.13) -0.2492
(-1.03) 0.2349
(0.58)
East
-0.1017
(-0.42) 0.1454
(1.02) 0.0245
(0.10) 0.4897
(1.21)
Foreignness
-0.1920
(-0.51) 0.1510
(0.73) -0.3559
(-1.09) 0.6049
(1.27)
Scale intensive industry - - - -
Specialized supplier - - - -
Science based industry - - - -
Constant -1.4308**
(-2.06) -1.4101***
(-4.10) -1.0296*
(-1.87) -0.4630
(-0.49)
Sample size 232 618 309 148
LR-Test 90.12** 176.85*** 95.29* 47.74
McFadden R2 0.282 0.226 0.288 0.364
Significance levels: *** 1%, ** 5%, * 10%, z-Values in Parentheses.
Source: LIAB 2001-2004.
IWH __________________________________________________________________
IWH-Diskussionspapiere 10/2007 20
5 Conclusions
The descriptive analysis reveals significant differences with respect to the share of
highly qualified employees between sectors distinguished according to the classical in-
novation patterns described by Pavitt. Sectors with a high share of highly qualified em-
ployees are characterized by product innovation activities clearly above average (spe-
cialized suppliers and science based industries). Furthermore, within the sectoral clus-
ters qualification seems to be particularly important for companies that are engaged in
original innovations. Thus, descriptive findings support our hypotheses.
However, the regression results for the tested specifications do not reveal significantly
positive coefficients for the qualification variables. Instead, we observe a significantly
negative effect of qualification when we introduce an interaction term of R&D and
qualification. This indicates that a high share of qualification as such is not enough as a
driving force for product innovation. The findings suggest that qualification drives in-
novation only when the qualified people focus on innovative activities (R&D)
– indicated by the significantly positive sign of the interaction term.
However, these are preliminary conclusions. The results call for further specification
and correlation tests to be carried out. One further step could be an alternative opera-
tionalisation of the qualification variable (inclusion of technical assistants, exclusion of
managers etc.).
To sum up, we find significant differences in the qualification levels between innovative
and non-innovative firms, but until now we cannot statistically verify a positive impact
of the share of highly qualified staff on the probability of product innovation. In the case
that further specification test do not reveal other results, further research should examine
the question whether there are rather qualitative than quantitative aspects determining a
firms’ innovative power.
__________________________________________________________________ IWH
IWH-Diskussionspapiere 10/2007 21
References
Aghion, P.; Howitt, P. W. (1998): Endogenous Growth Theory. The MIT Press: Cam-
bridge.
Alda, H. (2005): Betriebe und Beschäftigte in den Linked-Employer-Employee-Daten –
LIAB des Instituts für Arbeitsmarkt- und Berufsforschung. FDZ-Datenreport
01/2005. Nürnberg.
Alda, H.; Herrlinger, D. (2005): LIAB-Datenhandbuch. Version 1.0. FDZ Datenreport
07/2005. Nürnberg.
Barro, R. J.; Sala-i-Martin, X. (1998): Wirtschaftswachstum. Wolls Lehr- und Hand-
bücher der Wirtschafts- und Sozialwissenschaften. Oldenbourg Wissenschafts-
verlag: München.
Blechinger, D.; Pfeiffer, F. (1999): Qualifikation. Beschäftigung und technischer Fort-
schritt. Jahrbücher für Nationalökonomie und Statistik 218(1+2), 128-146.
Blossfeld, H.-P. (1985): Bildungsexpansion und Berufschancen. Campus: Frankfurt.
BMBF (2006a): Die Hightech-Strategie für Deutschland. Bundesministerium für Bil-
dung und Forschung: Berlin.
BMBF (2006b): Bericht zur technologischen Leistungsfähigkeit Deutschlands 2006.
Bundesministerium für Bildung und Forschung: Berlin.
Bush, V. (1945): Science: The Endless Frontier. US Government Printing Office: Wash-
ington.
Dakhli, M.; De Clercq, D. (2003): Human Capital, Social Capital, and Innovation: A
Multi Country Study. Vlerick Leuven Gent Working Paper 2003/18.
European Commission (2005): Communication from the Commission to the Council
and the European Parliament. Common Actions for Growth and Employment:
The Community Lisbon Programme. SEC(2005)981.
Fagerberg, J.; Mowery, D.; Nelson, R. (eds.) (2005): The Oxford Handbook of Innova-
tion. Oxford University Press: Oxford.
Fagerberg, J.; Verspagen, B. (2006): Innovation studies – an emerging discipline (or
what)? A study of the global network on innovation scholars. Paper presented at
the SPRU 40th Anniversary Conference. University of Sussex. SPRU. September
11-13, 2006.
Gottschalk, S.; Janz, N. (2003): Bestimmungsfaktoren der Innovationstätigkeit, in: Janz,
N.; Licht, G. (eds.): Innovationsforschung heute: Die Mannheimer Innovation-
spanels. Nomos-Verlagsgesellschaft: Baden-Baden, 17-39.
IWH __________________________________________________________________
IWH-Diskussionspapiere 10/2007 22
Grossman, G. M.; Helpman, E. (1997): Innovation and Growth in the Global Economy.
MIT Press: Cambridge.
Günther, J.; Gebhardt, O. (2006): Auswärtige Betriebe und Innovationstätigkeit, in:
IWH; IAB (eds.): Beschäftigungsanalysen mit den Daten des IAB-Betriebspanels
– Tagungsband zum Workshop des IAB und IWH. IWH: Halle, 107-129.
Hatzichronoglou, T. (1997): Revision of the High-Technology Sector and Product Clas-
sification. STI Working Papers 1997/2. OECD: Paris.
Heidenreich, M. (2003): Die Debatte um die Wissensgesellschaft, in: Böschen, S.;
Schulz-Schaeffer, I. (eds.): Wissenschaft in der Wissensgesellschaft. Westdeut-
scher Verlag: Wiesbaden, 25-51.
Hempel, T. (2003): Innovation im Dienstleistungssektor, in: Janz, N.; Licht, G. (eds.):
Innovationsforschung heute. Die Mannheimer Innovationspanels. Nomos-
Verlagsgesellschaft: Baden-Baden, 149-183.
Kline, S. J.; Rosenberg, N. (1986): An Overview of Innovation, in: Landau, R.;
Rosenberg, N. (eds.): The Positive Sum Strategy. Harnessing Technology for
Economic Growth. National Academy Press: Washington, 275-305.
Lam, A. (2005): Organizational Innovation, in: Fagerberg et al. (2005), 115-147.
Lucking, B. (2004): International Comparison of the Third Community Innovation Sur-
vey (CIS). TESE: London.
Lundvall, B.-Å. (2007): Higher Education. Innovation and Economic Development. De-
partment of Business Studies. Aalborg University. Denmark. Paper presented at
the World Bank’s Regional Bank Conference on Development Economics. Bei-
jing. January 16-17, 2007.
Lundvall, B.-Å.; Johnson, B. (1994): The Learning Economy. Journal of Industry Stud-
ies 1(2), 23-42.
Malerba, F. (2005): Sectoral Systems: How and Why Innovation Differs across Sectors,
in: Fagerberg et al. (2005), 380-406.
Nelson, R.; Phelps, E. (1966): Investment in humans, technological diffusion, and eco-
nomic growth. American Economic Review. Papers and Proceedings 51(2), 69-
75.
OECD (2005): Oslo Manual: Guidelines for Collecting and Interpreting Innovation
Data. OECD: Paris.
Pavitt, K. (1984): Sectoral patterns of technical change: Towards a taxonomy and a the-
ory. Research Policy 6, 343-373.
Peneder, M. (2003): Industry Classification: Aim, Scope, and Techniques. Journal of
Industry, Competition, and Trade 3(1-2), 109-129.
__________________________________________________________________ IWH
IWH-Diskussionspapiere 10/2007 23
Peters, B. (2003): Innovation und Beschäftigung, in: Janz, N.; Licht, G. (eds.): Innova-
tionsforschung heute. Die Mannheimer Innovationspanels. Nomos-
Verlagsgesellschaft: Baden-Baden, 113-148.
Pianta, M. (2005): Innovation and Employment, in: Fagerberg et al. (2005), 568-598.
Rammer, C.; Peters, B.; Schmidt, T.; Aschhoff, B.; Doherr, T.; Niggemann, H. (2005):
Innovationen in Deutschland. Ergebnisse der Innovationserhebung in der
deutschen Wirtschaft. Nomos-Verlagsgesellschaft: Baden-Baden.
Reinberg, A.; Hummel, M. (2004): Fachkräftemangel bedroht Wettbewerbsfähigkeit der
deutschen Wirtschaft. Aus Politik und Zeitgeschichte 28, 3-16.
Robinson, C.; Stokes, L.; Stuivenwold, E.; Van Ark, B. (2003): Industry Structure and
Taxonomies, in: European Commission: EU Productivity and Competitiveness:
An Industry Perspective. Can Europe Resume the Catching-up Process? Luxem-
bourg: Office for Official Publications of the European Communities, 37-71.
Romer, P. M. (1986): Increasing Returns and Long-run Growth. Journal of Political
Economy 94(5), 1002-1037.
Schultz, T. W. (1975): The Value of the Ability to Deal with Disequilibria. Journal of
Economic Literature 13(3), 827-846.
Statistisches Bundesamt (2005): Hochschulstandort Deutschland 2005. Statistisches
Bundesamt: Wiesbaden.
Statistisches Bundesamt (2006): Bevölkerung Deutschlands bis 2050. 11. Koordinierte
Bevölkerungsvorausberechnung. Statistisches Bundesamt: Wiesbaden.
Tether, B.; Mina, A.; Consoli, D.; Gagliardi, D. (2005): A Literature Review on Skills
and Innovation. ESRC Centre for Research on Innovation and Competition. Uni-
versity of Manchester: Manchester.
IWH __________________________________________________________________
IWH-Diskussionspapiere 10/2007 24
Appendix
Table 1:
Probit regression coefficients for qualification variable (Model I)
Dependent variable Improvement New product Market novelty
Sample Coefficient
z-value Coefficient
z-value Coefficient
z-value
Entire Sample -0.7747 -1.33 -0.1836 -0.36 1.0788* 1.94
Supplier dominated -0.8389 -0.46 -0.2792 -0.13 4.0116 1.45
Scale intensive -0.8564 -0.78 -0.0722 -0.07 0.8908 0.72
Specialized suppliers -0.0795 -0.07 0.5712 0.67 1.2654 1.33
Science based -1.2852 -0.88 -0.3261 -0.26 1.1706 1.12
Significance levels: *** 1%, ** 5%, * 10%.
Source: LIAB 2001-2004.
Table 2:
Probit regression coefficients for qualification variable (Model II)
Dependent
variable Improvement New product Market novelty
Sample High Qualifi-
cation Interaction
R&D/Qual. High Qualifi-
cation Interaction
R&D/Qual. High Qualifi-
cation Interaction
R&D/Qual.
Entire Sam-
ple -1.9197**
(-2.36) 2.5374**
(2.03) -0.4526
(-0.52) 0.3918
(0.38) -0.0881
(-0.08) 1.4975
(1.16)
Supplier
dominated -1.1430
(-0.61) 4.1388
(0.59) -1.9867
(-0.81)
8.9289*
(1.68)
4.5066
(1.55) -3.8882
(-0.51)
Scale inten-
sive
industry
-1.8769
(-1.38) 3.3249
(1.34) -0.8122
(-0.53) 1.3358
(0.66) 0.4073
(0.20) 0.7617
(0.31)
Specialized
suppliers -1.6026
(-1.02) 2.9326
(1.33) -0.6944
(-0.40) 1.6565
(0.85) -4.3317
(-1.22) 6.4348*
(1.74)
Science based
industry -2.8216
(-1.04) 2.2744
(0.67) 0.4588
(0.19) -1.0829
(-0.38) -2.2426
(-0.66) 3.8202
(1.07)
Significance levels: *** 1%, ** 5%, * 10%, z-Values in Parentheses.
Source: LIAB 2001-2004.
__________________________________________________________________ IWH
IWH-Diskussionspapiere 10/2007 25
Table 3:
Description of regression variables
Variable Scale Year of reference
Description
Endogenous variables
Product innovation 0/1 2002-2004 At least one product innovation (product im-
provement, new product or market novelty)
Improvement 0/1 2002-2004 At least one product improvement
New Product 0/1 2002-2004 At least one new product or extension of
product range
Market novelty 0/1 2002-2004 At least one market novelty
Exogenous variables
High qualification
(occupational status) % 2002 Share of engineers, scientists, and managers
within the firm
Job tenure max. 1 year % 2002 Share of employees with max. 1 year job ten-
ure within the firm
Job tenure 1-5 years % 2002 Share of employees with 1-5 years job
tenure
within the firm
R&D activities 0/1 2004 Engagement in R&D activities or coopera
tion
Firm size log 2002 Log. number of Employees
Export intensity % 2002 Share of sales abroad
Profitability 0/1 2002 At least good profitability
(Assessment better than 3 on a range of 1-5)
Equipment 0/1 2002 At least good technological standard (As-
sessment better than 3 on a range of 1-5)
Further training % 2001 Share of further training participants on total
employees
Age of the firm 0/1 2002 Firm foundation before 1990
East 0/1 2002 Firm located in East-Germany
Foreignness 0/1 2002 Majority of firm owned by foreigners
Scale intensive industry 0/1 2002
Specialized suppliers 0/1 2002
Science based industry 0/1 2002
According to Pavitt (1984) and Robinson et
al. (2003), see Appendix table 7.
IWH __________________________________________________________________
IWH-Diskussionspapiere 10/2007 26
Table 4:
Pavitt Taxonomy (producing sector without construction)
Category International Standard Industrial Classification of All Economic Ac-
tivities. Revision 3 (1990) ISIC (Rev. 3)
Supplier dominated
industries
Agriculture (01); Forestry (02); Fishing (05); Textiles (17); Cloth
ing
(18); Leather and footwear (19); Wood & products of wood and
cork (20); Pulp, paper & paper products (21); Printing & publishing
(22); Furniture, miscellaneous manufacturing, recycling (36-37).
Scale intensive industries
Mining and quarrying (10-14); Food, drink & tobacco (15-16);
Mineral oil refining, coke & nuclear fuel (23); Rubber & plastics
(25); Non-metallic mineral products (26); Basic metals (27); Fabri-
cated metal products (28); Motor vehicles (34); Building and repair-
ing of ships and boats (351); Aircraft and spacecraft (353); Railroad
equipment and transport equipment n.e.c. (352+359); Electricity,
gas and water supply (40-41).
Specialized suppliers
Mechanical engineering (29); Office machinery (30); Insulated wire
(313); Electronic valves and tubes (321); Telecommunication equip-
ment (322); Scientific instruments (331); Other instruments
(33-331).
Science based industries Chemicals (24); Other electrical machinery & apparatus (31-313);
Radio and television receivers (323).
Source: Robinson et al. (2003).
