the location of technological activities of mncs in european regions
TRANSCRIPT
The location of technological activities of MNCs in
European regions:
The role of spillovers and local competencies
John Cantwella, Lucia Piscitelloa,b,*
aUniversity of Reading, Reading, MA, USAbDipartimento di Economia e Produzione, Politecnico di Milano, Piazza Leonardo da Vinci 32,
Milan 20133, Italy
Abstract
This paper examines the relative attractiveness of the Italian, German and UK regions for the siting
of foreign-owned corporate technological development, controlling for the geographical distribution of
the equivalent innovative efforts of established indigenous firms in each country. The data used are
patents granted in the US to the world’s largest firms from 1969 to 1995, identifying the ultimate
nationality of ownership and the location of the research facility responsible for each patent. Through
econometric models based on count data techniques, it is shown that the relative attraction of a location
for foreign-owned research depends positively upon the local market size, the local scientific and
educational infrastructure and the potential for intra- and interindustry spillovers. D 2002 Elsevier
Science Inc. All rights reserved.
Keywords: Location; European regions; MNCs; Spillovers
1. Introduction
The existing knowledge base of a region plays an important role in the decisions of the
largest foreign-owned firms as to where to locate their technological activities (Cantwell and
1075-4253/02/$ – see front matter D 2002 Elsevier Science Inc. All rights reserved.
PII: S1075 -4253 (01 )00056 -4
* Corresponding author. Dipartimento di Economia e Produzione, Politecnico di Milano, Piazza Leonardo da
Vinci 32, Milano 20133, Italy. Tel.: +39-2-2399-2740; fax: +39-2-2399-2710.
E-mail address: [email protected] (L. Piscitello).
Journal of International Management
8 (2002) 69–96
Iammarino, 1998, 2000; Cantwell and Noonan, 2001) as well as other location-specific factors
mainly related to the market. Thus, in each country, the local technological efforts of foreign-
owned firms tend to be strongly agglomerated at a subnational and regional level. The present
paper analyses this agglomeration effect at the regional level (Braunerhjelm and Svensson,
1998; Barrel and Pain, 1999), having controlled for the cross-regional and cross-sectoral
distribution of technological activities in indigenous firms. Specifically, the purpose of this
paper is twofold: (1) to analyse the regional distribution of technological activities carried out
by large multinational corporations (MNCs) in Germany, the UK and Italy over the last 30
years and (2) to explain the locational preferences of foreign-owned firms across the regions in
each of the three countries considered.
The empirical investigation uses patents granted in the US to the world’s largest industrial
firms for inventions achieved in their European-located operations, classified by the host
European region in which the research facility responsible is located. We examine the
regional distribution of corporate research activity in Italy, Germany and the UK, distinguish-
ing between domestically owned and foreign-owned firms in each of these three countries.
The spatial patterns of activity in foreign-owned firms are considered while controlling for
indigenous firms’ patterns. Departures from a linear proportionality between the locational
distributions of these two sets of firms constitute differences in their locational behaviour.
These differences in the cross-regional distribution of technological development between
locally-owned and foreign-owned firms1 could be explained through variables related to the
local market size, to the knowledge base and to the potential for intra- and interindustry
spillovers. We discuss some explanations for these differences and propose an econometric
model based on count data techniques in order to explain these locational preferences of
MNCs across regions within each of the three European countries considered.
The paper is organised in the following way. Section 2 sets out the conceptual framework
for the analysis of the determinants of locational choice in the technological activities of
MNCs. Section 3 investigates the extent and evolution of the internationalisation of
technological activity in the German, UK and Italian regions in the period 1969–1995.
Section 4 reports the econometric model, the variables and the results obtained. Finally,
Section 5 presents some summarising and concluding remarks, draws out one of the policy
implications of our argument and indicates an agenda for future research.
2. The location of technological activities: a theoretical framework
At a general level, a firm’s operations may be dispersed across different types of
productive activity (the diversification of technologies or products) or over geographical
space (the internationalisation of the same). However, spreading the product markets in which
the firm is involved may be a matter of exploiting more effectively established competencies
while moving into new areas of technological development means creating new competence.
1 For a similar result in the UK, see also Cantwell and Iammarino (2000).
J. Cantwell, L. Piscitello / Journal of International Management 8 (2002) 69–9670
In order to not only exploit effectively but also to consolidate an existing capability, it is
generally necessary for a firm to extend that capability into new related fields of production
and technology and across a variety of geographical sites (Cantwell, 1995). The corporate
internationalisation and diversification of technological activity are indeed both ways of
spreading the competence base of the firm and of acquiring new technological assets or
sources of competitive advantage.2
Attention has been increasingly focused on the emergence of the trend for MNCs to
establish internal and external networks for innovation (Cantwell, 1995; Cantwell and
Iammarino, 2000; Kuemmerle, 1999; Zander, 1999), which are characterised by different
levels of territorial and social embeddedness with reference to both the motivations for
overseas R&D and the location that hosts them. Indeed, the localisation of research activities
is partially determined by the specific function, which this activity fulfills within a given firm
(Carrincazeaux et al., 2001). The internationalisation of R&D may be motivated by several
considerations and specifically (Kumar, 2001) to support foreign production by adaptation to
host country markets [home-base exploiting R&D (HBE) in the words of Kuemmerle, 1999]
and to tap into the capabilities available in host countries, thus benefiting from the localised
knowledge spillovers [home-base augmenting (HBA)]. Despite these suggestions of intrafirm
differences in the influences on the internationalisation of R&D, the bulk of analysis of
overseas R&D has been carried at the level of the corporate group as a whole rather than the
individual subsidiary or laboratory or at the level of collective groups of firms, and it has
focused on the determinants of intensity of overseas R&D in an interindustry, interfirm or
intercountry context (see Zejan, 1990; Kumar, 1996, 2001; Niosi, 1999 for a number of recent
studies). Specifically, firms’ locational choices and their location-specific determinants have
been mainly analysed at the country level (Hakanson, 1992; Fors, 1996; Kumar, 1996;
Odagiri and Yasuda, 1996), and only a minority of studies has recently started to investigate
their regional or subnational dimension (e.g. Cantwell and Iammarino, 1998, 2000; Carrin-
cazeaux et al., 2001). In fact, although some authors have recently suggested that regions are
increasingly becoming important milieu for the competitive-enhancing activities of mobile
investors (Porter, 1996; Scott, 1998; Dunning, 2000), thus replacing the nation state as the
principal spatial economic entity (Ohmae, 1995), there is still only quite a scant existing
empirical research on multinational location at this subnational level.
The development of cross-border corporate integration and intraborder intercompany
sectoral integration, as new forms of global governance, makes it increasingly important to
examine where and how innovative activity by MNCs is internationally dispersed and
regionally concentrated. To the multinational firm, the innovativeness of the corporate group
as a whole depends upon the extent of the locational diversity that it can manage to combine
and sustain in its technological efforts and the degree to which it can choose to site activity so
as to reduce overlapping duplication but enhance technological complementarity between the
2 The background to this study is the relationship between the diversification and internationalisation of the
technological competence of large MNCs, which have been explored extensively in our earlier work (Cantwell
and Janne, 1999; Cantwell and Piscitello, 1999, 2000).
J. Cantwell, L. Piscitello / Journal of International Management 8 (2002) 69–96 71
locations selected. Therefore, the locational choice of the MNC depends upon (1) the strategy
followed by the MNC (the extent to which it has developed a network with home-base
augmenting facilities, as opposed to the more traditional home-base exploiting) and (2) the
location-specific characteristics of alternative contexts in which research may be located. The
present paper specifically focuses on (2). In particular, we claim that the relevant location-
specific characteristics are the following.
2.1. Agglomeration and industry-specific spillovers
The firms of each country tend to embark on a path of technological accumulation that
has certain unique characteristics and sustains a distinct profile of national technological
specialisation (Rosenberg, 1976; Pavitt, 1987; Cantwell, 2000). The kinds of linkages that
grow up between competitors, suppliers and customers in any regional district or country are
also, to some extent, peculiar to that location and imbue the technology creation of its firms
with distinctive features (Mariotti and Piscitello, 2001). For these reasons, other MNCs often
need to be on-site with their own production and their innovatory capacity if they are to
properly benefit from the latest advances in geographically localised technological deve-
lopment to feed their innovation (Cantwell, 1989; Kogut and Chang, 1991). Moreover, due
to the complexity of technological learning and the significance of maintaining face-to-face
contacts, the localisation of technological contacts tends to occur at a regional level within
host countries (Jaffe et al, 1993; Almeida, 1997; Cantwell and Iammarino, 1998; Verspagen
and Schoenmakers, 2000). It is therefore typically when there is already a strong existing
domestic technological presence that the R&D of foreign-owned affiliates is most likely
to become substantial and to gain a creative role with respect to the global technological
development strategy of the MNC as a whole. However, where indigenous technological
development is highly concentrated in just one or two major local firms, any industry-
specific agglomeration effect may be offset by a competitive deterrence effect, both in terms
of bidding for local resources and in terms of the availability of potential local technologi-
cal spillovers.
2.2. External sources of knowledge and science–technology spillovers
Firms’ efforts to advance technology do not generally proceed in isolation but are strongly
supported by various external sources of knowledge: public research centres, universities,
industry associations, an adequate educational system and science base and other firms (Kline
and Rosenberg, 1986; Nelson, 1993; Nelson and Rosenberg, 1999; Rosenberg and Nelson,
1996; Breschi, 2000). There is growing evidence, so far mainly from the US, that these
science–technology or university–industry linkages tend to be geographically localised (Jaffe
et al, 1993; Audretsch and Feldman, 1996; Audretsch and Stephan, 1996; Acs et al., 2000;
Adams, 2001). This is especially likely to be true of foreign-owned firms in an economy,
which tend to have a greater degree of locational mobility when siting their corporate research,
and so pay greater attention to being close to relevant public research facilities (see Gorg and
Strobl, 2001 on the greater international locational mobility of MNCs). Thus, in an earlier
J. Cantwell, L. Piscitello / Journal of International Management 8 (2002) 69–9672
study, it was shown that foreign-owned firms in the UK are relatively more drawn (than are
UK-owned firms) to locate their research in regions such as Scotland and East Anglia, in which
the public research base and higher education infrastructure is also relatively good (Cantwell
and Iammarino, 2000). Therefore, we expect that compared to indigenous firms that have a
certain locational inertia around their major centre of technological activity, which is already
situated in their home base, foreign-owned research facilities are relatively more attracted
by external sources of knowledge and science–technology spillovers.
2.3. Localised intercompany spillovers
As knowledge is mainly tacit, geographical distance increases the difficulty in both
transmitting and absorbing it. In other words, tacit knowledge travels easily over small
distances but far less easily over longer distances (Caniels, 2000). This leads to the hypothesis
that the intensity of spillovers increases with geographical proximity (Verspagen and
Schoenmakers, 2000). Specifically, we distinguish three different types of interfirm spillovers
as follows.
(1) Cluster-based spillovers, associated with the presence of a wide-ranging collection of
technologically active firms within a given industry or sector, all concentrated in the same
geographical area (Baptista and Swann, 1998, 1999). The geographical concentration of firms
engaged in similar activities or within a common industry leads to further local clustering of
related firms and the local accumulation of relevant knowledge (Braunerhjelm et al., 2000).
(2) General purpose spillovers, entailing interindustry spillovers (Lipsey et al., 1998)
associated with the existence of firms working in several different fields of research but with a
common overlapping interest in certain general purpose technologies (GPTs), which are
relevant in most industries such as machinery or computing. Such spillovers are more likely to
occur in an all-round ‘‘higher order’’ centre of excellence, which facilitates a more favourable
interaction with indigenous firms, and greater opportunities for intercompany alliances for the
purposes of technological collaboration and exchange (Cantwell and Mudambi, 2000). Within
a host country, an all-round regional centre of excellence is likely to attract the research-based
investments of a wide variety of foreign-owned MNCs, as the attraction is related to spillovers
in core technological sectors (GPTs), which are of an interindustry kind. Moreover, there is
some evidence relating to cities in the US that diversity across industries may promote
innovation and knowledge spillovers (Feldman and Audretsch, 1999).
(3) Finally, we argue that, according with the bulk of the analysis on overseas R&D,
locational determinants related to the size of the local market have a positive and significant
influence on affiliate R&D location (Zejan, 1990; Kumar, 1996, 2001; Braunerhjelm et al.,
2000). While the other locational effects we have described relate especially to the attraction
of localised R&D in the newer competence-creating types of subsidiaries, the pull of local
market demand relates more to the attraction of R&D in the traditional competence-ex-
ploiting types of subsidiaries for the purpose of adapting products for local markets. How-
ever, when working at the level of firms as a whole or groups of firms, we must consider
these motives together so as to assess the relative significance of each of these pull factors
on average.
J. Cantwell, L. Piscitello / Journal of International Management 8 (2002) 69–96 73
All these considerations suggest that innovative activities tend in general to favour specific
locations, and the same rationale explaining location decisions holds in principle both for
foreign-owned and domestic firms. Nonetheless, as the former have a home base located
outside the country, they might be more mobile within the host country in terms of their
choice of location, thus responding more effectively to regional differences. Therefore, we
expect the locational choices of foreign-owned firms to be especially strongly influenced by
the specific features of the region as well as by the potential spillover benefits, as far as their
technological activities are concerned.
3. Evidence on the globalisation of corporate technology at the European regional and
sectoral level
The use of corporate patents as an indicator of advanced technological capacity and the
ability to develop innovation is one of the most established and reliable methods of estimating
the cross-sectional patterns of innovative activities. The advantages and disadvantages of
using patent statistics are well known in the literature (Schmookler, 1950, 1966; Pavitt, 1985,
1988; Griliches, 1990; Archibugi, 1992). The use of patent records provides information both
on the owner of the invention (from which the country of location of the ultimate parent firm
has been derived through a consolidation of patents at the level of international corporate
group) and separately the address of the inventor, thus allowing the identification of where the
R&D underlying the invention was carried out in geographical terms.
The database used for the study consists of patents granted in the US to the world’s 792
largest industrial firms as of 1982, derived from both the Fortune 500 US and the Fortune 500
non-US firms listings (Dunning and Pearce, 1985).3 Of these 792 companies, 730 had an
active patenting presence during the period 1969–1995. Another 54 historically significant
firms were added to these, making 784 corporate groups in all. The additions include (mainly
for recent years but occasionally historically) enterprises that occupied a prominent position
in the US patent records, some of which are firms that were omitted from Fortune’s listing for
classification reasons (e.g. RCA and AT&T were classified as service companies) and others
that reflect recent mergers and acquisitions or new entrants to the population of large firms.
Patents have been consolidated at the level of the international group of ultimate ownership,
allowing for changes due to mergers and acquisitions since 1982. For patents that are
attributable to research facilities located in selected European countries, we have identified
the precise regional location of research, as is explained further below.
Table 1 indicates the share of European host countries in the foreign-located research of
large firms. In particular, it is shown that overall the most attractive European host countries
for the technological activity of foreign-owned MNCs were Germany (29% in 1991–1995),
the UK (21%) and France (16%) and only to a lesser extent Italy (6%). Since 1969–1972, the
3 Fortune 500 provided two separate listings, one for the largest US and one for the largest non-US firms.
While we included all the 500 US firms, non-US firms were then included so long as they were larger than the
500th US firm. Hence, the original 792 includes 292 non-US firms.
J. Cantwell, L. Piscitello / Journal of International Management 8 (2002) 69–9674
UK has lost some of its earlier share (29%) to most other countries. Table 2 reports figures by
European host country on the share of foreign-owned firms in total corporate patents ema-
nating from locally based research. The proportion of European research activity undertaken
by foreign-owned companies has increased overall from 23% to 29%, having fallen slightly
during the 1970s and then risen during the 1980s before rising sharply in the 1990s. This is
consistent with the general increase in the internationalisation of technological development in
major firms already acknowledged elsewhere (e.g. Dunning, 1994; Cantwell, 1995).
In order to analyse the location of corporate R&D activities at a more detailed level of
geographical disaggregation, we focused on the subnational entities4 that derive from
normative criteria, as classified by Eurostat in the Nomenclature of Territorial Units for
Statistics (NUTS). The NUTS classification is based on the institutional divisions currently in
force in the member states, according to the tasks allocated to territorial communities, to the
sizes of population necessary to carry out these tasks efficiently and economically and to
historical, cultural and other factors.
To provide a single uniform breakdown of territorial systems, we referred to the NUTS 2
level for the three countries considered. The NUTS 2 level (206 Basic Regions) is generally
used by the EU members for the application of their regional policies and thus is the most
appropriate to analyse the regional distribution of technological activities. Indeed, although
other studies about various regional issues in the EU consider different subnational NUTS
levels for different countries in order to assure economic homogeneity,5 in the present
context, considering NUTS 2 assures a more uniform distribution of patent data across
regions in the period considered. The one exception is that in the case of Lombardia, which is
comfortably the largest region for technological development in Italy, we created a
4 There is evidence that it is in Europe in which cross-border MNC networks have reached their most
advanced state (Cantwell and Janne, 1999), and so examining the determinants of the geographical pattern of
MNC innovation in these regions offers a good test of our alternative hypotheses outlined above.
Table 1
Patenting activity attributable to European-located, foreign-owned research across host countries, 1969–1995 (%)
Total patents from foreign-owned facilitiesEuropean
host country 1969–1972 1973–1977 1978–1982 1983–1986 1987–1990 1991–1995
Germany 27.03 30.23 31.81 35.63 33.47 28.87
UK 29.34 26.78 25.03 22.63 21.00 21.15
Italy 4.34 4.94 4.37 4.50 5.97 6.46
France 13.21 14.95 14.52 14.21 14.92 15.60
Rest of Europe 26.08 23.10 24.27 23.03 24.64 27.92
Total Europe 100.00 100.00 100.00 100.00 100.00 100.00
Source: US patent database developed by John Cantwell at the University of Reading, with the assistance of the
US Patent and Trademark Office.
5 For example Paci (1997) considers 109 regions corresponding to NUTS 0 for Denmark, Luxembourg and
Ireland, NUTS 1 for Belgium, Germany, Netherlands and the UK and NUTS 2 for Italy, France, Spain, Portugal
and Greece. Likewise, Cantwell and Iammarino (1998) and Breschi (1999) consider NUTS 1 for the UK and
NUTS 2 for Italy.
J. Cantwell, L. Piscitello / Journal of International Management 8 (2002) 69–96 75
subdivision between Milano and the rest of Lombardia. The empirical investigation uses
patents granted to the world’s largest industrial firms for inventions achieved in their
European-located operations, classified by the host European region in which the responsible
research facility is located.
The regionalisation of our US patent database consists of attributing a revised, although
still compatible, NUTS 2 code for each patent record, according to the location of inventors in
the EU countries, with reference to the period 1969–1995 (Cantwell and Iammarino, 1998,
2000; Cantwell et al., 2001) that has been extended to cover Germany, UK and Italy. The
three host countries substantially differ each other in terms of the magnitude of the
phenomenon under investigation. The total number of corporate patents due to German-
located activity registered in the database over the period 1969–1995 (91,433) is more than
twice that registered for the UK (45,136), which in turn is more than six times that registered
for Italy (7030).
Tables 3–5 report the total number and the percentage share of patents granted to the
domestic firms and to foreign-owned firms in each region considered. Concerning Germany
(see Table 3), it is worth noting that the number of patents granted to domestic firms (76,215) is
about five times that for foreign-owned firms (15,218), while for both the UK and Italy, the
efforts of indigenous firms is about twice that of foreign-owned firms (23,350 as against 11,786
and 4490 as against 2540, respectively). However, in the UK, this is due to a high degree of
both inward and outward internationalisation, while in Italy, it is due in large part to the
comparative weakness of very large indigenous companies in the Italian industrial structure.
In Italy (see Table 5), just as for indigenous Italian firms, foreign-owned firms record the
highest concentration of research (40.16%) in Milano. Outside of this very striking
geographical agglomeration, however, as highlighted by Cantwell and Iammarino (1998),
foreign-owned research appears to be relatively more dispersed than that undertaken by their
indigenous counterparts. Whilst foreign-owned firms locate approximately 68% of their R&D
in the two core regions of Lombardia and Piemonte, 82% of patenting by indigenous firms is
located there.
Some variations in foreign-owned by comparison with indigenous R&D location patterns
are also recorded in the UK. Similarly to the case of Italy (Lombardia), foreign-owned firms
are more highly concentrated in the core region (the Southeast) than are their indigenous
Table 2
Patenting activity attributable to foreign-owned research, as a proportion of all patenting from the local research of
large firms, by European host country, 1969–1995 (%)
Proportion of patents from foreign-owned facilitiesEuropean
host country 1969–1972 1973–1977 1978–1982 1983–1986 1987–1990 1991–1995
Germany 16.32 15.57 15.16 18.77 18.09 17.37
UK 27.66 30.80 31.30 36.00 35.44 45.23
Italy 27.32 31.09 26.49 32.85 43.93 57.50
France 24.17 24.73 24.04 25.13 27.05 28.94
Total Europe 22.70 21.63 21.43 24.40 24.97 28.63
Source: as for Table 1.
J. Cantwell, L. Piscitello / Journal of International Management 8 (2002) 69–9676
Table 3
Patents granted for research inGermany in large foreign-owned and domestically owned firms by region, 1969–1995
Region
No. of foreign-
owned patents %
No. of domestically
owned patents %
Stuttgart 1851 12.16 7768 10.19
Karlrushe 992 6.52 2755 3.61
Freiburg 1342 8.82 808 1.06
Tubingen 599 3.94 1089 1.43
Baden-Wurttemberg 4784 31.44 12,420 16.30
Oberbayern 381 2.50 10,785 14.15
Niederbayern 90 0.59 819 1.07
Oberpfalz 283 1.86 559 0.73
Oberfranken 89 0.58 533 0.70
Mittelfranken 422 2.77 3806 4.99
Unterfranken 593 3.90 1238 1.62
Schwaben 284 1.87 1101 1.44
Bayern 2142 14.08 18,841 24.72
Berlin 141 0.93 1875 2.46
Brandenburg 24 0.16 56 0.07
Bremen 47 0.31 128 0.17
Hamburg 754 4.95 315 0.41
Darmstadt 1959 12.87 9195 12.06
Giessen 191 1.26 650 0.85
Kassel 47 0.31 174 0.23
Hessen 2197 14.44 10,019 13.15
Meckelburg-Vorpommern 24 0.16 94 0.12
Braunschweig 110 0.72 913 1.20
Hannover 495 3.25 1048 1.38
Luneburg 147 0.97 349 0.46
Weser-Ems 41 0.27 280 0.37
Niedersachsen 793 5.21 2590 3.40
Dusseldorf 951 6.25 9444 12.39
Koln 1484 9.75 9586 12.58
Munster 135 0.89 1345 1.76
Detmold 44 0.29 300 0.39
Amsberg 282 1.85 1268 1.66
Nordrhein-Westfalen 2896 19.03 21,943 28.79
Koblenz 351 2.31 585 0.77
Trier 86 0.57 253 0.33
Rheinhessen-Pfalz 322 2.12 6212 8.15
Rheinland-Pfalz 759 4.99 7050 9.25
Saarland 62 0.41 137 0.18
Sachsen 33 0.22 112 0.15
Dessau 1 0.01 19 0.02
Halle 4 0.03 15 0.02
Magdeburg 5 0.03 24 0.03
Sachsen-Anhalt 10 0.07 58 0.08
Scheleswig-Holstein 530 3.48 511 0.67
Thuringen 22 0.14 66 0.09
Total 15,218 100.00 76,215 100.00
J. Cantwell, L. Piscitello / Journal of International Management 8 (2002) 69–96 77
Table 4
Patents granted for research in the UK in large foreign-owned and domestically owned firms by region, 1969–1995
Region
No. of foreign-
owned patents %
No. of domestically
owned patents %
Cleveland, Durham 82 0.70 629 2.69
Cumbria 19 0.16 136 0.58
Northumberland, Tyne and Wear 164 1.39 166 0.71
North 265 2.25 931 3.99
Humberside 66 0.56 213 0.91
North Yorkshire 107 0.91 362 1.55
South Yorkshire 133 1.13 199 0.85
West Yorkshire 157 1.33 255 1.09
Yorkshire and Humberside 463 3.93 1029 4.41
Derbyshire, Nottinghamshire 143 1.21 921 3.94
Leics., Northamptonshire 319 2.71 503 2.15
Lincolnshire 71 0.60 61 0.26
East Midlands 533 4.52 1485 6.36
East Anglia 621 5.27 342 1.46
Bedfordshire, Hertfordshire 1353 11.48 1528 6.54
Berks., Bucks., Oxfordshire 1231 10.44 1669 7.15
Surrey, East–West Sussex 1250 10.61 1703 7.29
Essex 815 6.91 991 4.24
Greater London 1300 11.03 2487 10.65
Hampshire, Isle of Wight 801 6.80 463 1.98
Kent 432 3.67 574 2.46
South East 7182 60.94 9415 40.32
Avon, Gloucs., Wiltshire 370 3.14 1103 4.72
Cornwall, Devon 62 0.53 64 0.27
Dorset, Somerset 65 0.55 166 0.71
South West 497 4.22 1333 5.71
Hereford-Worcs., Warwicks 135 1.15 983 4.21
Shropshire, Staffordshire 90 0.76 620 2.66
West Midlands 187 1.59 2200 9.42
West Midlands 412 3.50 3803 16.29
Cheshire 235 1.99 1161 4.97
Greater Manchester 455 3.86 1202 5.15
Lancashire 98 0.83 516 2.21
Merseyside 132 1.12 1100 4.71
North West 920 7.81 3979 17.04
Clwyd, Dyfed, Gwynedd, Powys 66 0.56 153 0.66
Gwent, Mid-S-W Glamorgan 323 2.74 398 1.70
Wales 389 3.30 551 2.36
Bord.-Centr.-Fife-Loth-Tayside 187 1.59 175 0.75
Dumfr.-Galloway, Strathclyde 231 1.96 251 1.07
Highlands, Islands 3 0.03 2 0.01
Grampian 53 0.45 42 0.18
Scotland 474 4.02 470 2.01
Northern Ireland 30 0.25 12 0.05
Total 11,786 100.00 23,350 100.00
J. Cantwell, L. Piscitello / Journal of International Management 8 (2002) 69–9678
counterparts. Also, as in Italy (Piemonte), indigenous firms locate a substantial proportion of
their innovative activity outside of the core as well, in the West Midlands and the North-
west—regions that, relative to their overall shares, have failed to attract so much foreign-
owned activity. Interesting also is the ability of regions such as East Anglia or Scotland to
attract relatively higher foreign-owned firm innovative activity despite their low overall share
in the UK-owned figure. A similar and indeed stronger result is found in the German case.
Despite the fact that Baden Wuerttemberg is only the third most popular location for German-
owned research, hosting approximately 19%, this region represents the prime location for
foreign-owned firms, which undertake 31% of their research there. The same pattern is
recorded in the northwest region of Niedersachsen, despite the fact that it hosts a low overall
share of total activity (under 4%) but where foreign-owned firms locate over 5% of their
patenting activities.
Indeed, the German case contrasts with patterns recorded for both the UK and Italy on a
number of fronts. Whilst both foreign-owned and indigenous firms concentrate their research
Table 5
Patents granted for research in Italy in large foreign-owned and domestically owned firms by region, 1969–1995
Region
No. of foreign-
owned patents %
No. of domestically
owned patents %
Piemonte 287 11.30 1430 31.85
Valle d’Aosta 0 0.00 1 0.02
Liguria 170 6.69 16 0.36
North West 457 17.99 1447 32.23
Milano 1020 40.16 1986 44.23
Lombardia (excluding Milano) 431 16.97 274 6.10
Lombardia 1451 57.13 2260 50.33
Trentino Alto Adige 2 0.08 13 0.29
Veneto 86 3.39 127 2.83
Friuli Venezia Giulia 16 0.63 87 1.94
North East 104 4.09 227 5.06
Emilia Romagna 169 6.65 186 4.14
Toscana 36 1.42 107 2.38
Umbria 3 0.12 52 1.16
Marche 4 0.16 7 0.16
Centre 43 1.69 166 3.70
Lazio 188 7.40 148 3.30
Abruzzo 2 0.08 20 0.45
Molise 0 0.00 0 0.00
Abruzzo-Molise 2 0.08 20 0.45
Campania 47 1.85 16 0.36
Puglia 34 1.34 13 0.29
Basilicata 0 0.00 0 0.00
Calabria 2 0.08 0 0.00
South 36 1.42 13 0.29
Sicilia 42 1.65 6 0.13
Sardegna 1 0.04 1 0.02
Total 2540 100.00 4490 100.00
J. Cantwell, L. Piscitello / Journal of International Management 8 (2002) 69–96 79
in the same region in the UK (the Southeast) and Italy (Lombardia), the same does not hold
for Germany. Nordrhein Westfalen (which borders Belgium and the Netherlands in the west
of the country) hosts the highest concentration of indigenous activity (29%) but only
represents the second most popular location for foreign-owned research. Foreign-owned
firms, as noted above, record their highest concentration of innovative activity in the
southwest region of Baden Wuerttemberg. This differing pattern for Germany, we believe,
can be explained by considering the type of technological activity associated with Nordrhein
Westfalen. This region is the traditional home of the German chemical/pharmaceutical
industry and continues to record substantial technological advantage for indigenous firms
that base their research there (see Table 6 below). This strength is further reflected in the
research profiles of the universities and research institutes located in the region. We
tentatively suggest therefore that because foreign-owned chemical firms may experience
difficulty in trying to access the deeply entrenched technology networks and communication
channels that have evolved through time, they disperse their research more widely and
account for a relatively low share of total German research in chemicals. This deterrence
effect on foreign-owned firms within the industries of primary indigenous strength is quite
common in most countries (Cantwell and Kosmopoulou, in press), but in Germany, it is
distinguished by its strong locational influence, given the heavily regionally specific character
of the leading companies in domestic German industry.
The sectoral forms of innovative activities is shown in Tables 6–8, which examine the
contribution to local research of both foreign-owned and domestic firms by industry. In
Table 6
Patents granted for research in Germany in large foreign-owned and domestically owned firms by industry,
1969–1995
Industry
No. of foreign-
owned patents %
No. of domestically
owned patents %
Food, drink, tobacco and allied products 271 1.78 1 0.00
Chemicals 2234 14.68 33,759 44.29
Pharmaceuticals 803 5.28 5070 6.65
Metals 1008 6.62 6030 7.91
Mechanical engineering 1505 9.89 4352 5.71
Electrical equipment 5425 35.65 12,931 16.97
Office equipment 825 5.42 133 0.17
Motor vehicles 967 6.35 10,838 14.22
Aircraft 166 1.09 970 1.27
Other transport equipment 8 0.05 0 0.00
Textiles 4 0.03 16 0.02
Paper, printing and publishing 91 0.60 12 0.02
Rubber products 99 0.65 190 0.25
Nonmetallic mineral products 755 4.96 0 0.00
Coal and petroleum products 483 3.17 156 0.20
Professional and scientific instruments 510 3.35 1187 1.56
Other manufacturing 64 0.42 570 0.75
Total 15,218 100.00 76,215 100.00
J. Cantwell, L. Piscitello / Journal of International Management 8 (2002) 69–9680
Germany (Table 6), foreign-owned firms contribute relatively much in electrical and
computing equipment (35.65% vs. 17%) and in general engineering (9.9% vs. 5.7%) but
relatively little in chemicals (14.7%), the area of greatest indigenous strength (44.3%). This
helps to explain why foreign-owned firms may be less attracted to the main centre for
chemical research in Germany (in Nordrhein Westfalen). The most attractive macroregion for
foreign-owned R&D is Baden-Wurtemburg, which is a centre of engineering excellence in
the motor vehicle industry (in which sphere of technology creation is very highly specialised)
and which has proved to be a magnet for foreign-owned development efforts in the areas of
electrical and computing equipment and general engineering (Cantwell and Noonan, in
press). This area is also well known for the innovativeness of local small and medium-sized
firms (SMEs), whose expertise in developing specialised machinery, equipment and compo-
nents and in engineering may also provide a fruitful interaction with the R&D of large
foreign-owned firms.
Turning now to the British experience (Table 7), foreign-owned firms contribute most to
the UK research base again in mechanical engineering (7.12% vs. 4.22%), electrical (24.74%
vs. 16.4%) and office equipment (7.92% vs. 1.23%) and instruments (3.33% vs. 0.04%).
They have also participated well in the British success in pharmaceuticals research (16.99%
vs. 8.51%) and have made a roughly average contribution in chemicals (19.11% vs. 22.62%).
As a general consequence, the development efforts of foreign-owned firms in the UK are
most attracted as we have seen already the wider technology base and infrastructure of the
higher-order centre of London and the Southeast, and this is especially true in the fields of
Table 7
Patents granted for research in the UK in large foreign-owned and domestically owned firms by industry,
1969–1995
Industry
No. of foreign-
owned patents %
No. of domestically
owned patents %
Food, drink, tobacco and allied products 223 1.89 1225 5.25
Chemicals 2252 19.11 5281 22.62
Pharmaceuticals 2003 16.99 1987 8.51
Metals 545 4.62 1228 5.26
Mechanical engineering 839 7.12 985 4.22
Electrical equipment 2916 24.74 3826 16.39
Office equipment 933 7.92 287 1.23
Motor vehicles 508 4.31 3359 14.39
Aircraft 200 1.70 1737 7.44
Other transport equipment 1 0.01 35 0.15
Textiles 8 0.07 271 1.16
Paper, printing and publishing 78 0.66 119 0.51
Rubber products 123 1.04 581 2.49
Nonmetallic mineral products 199 1.69 507 2.17
Coal and petroleum products 312 2.65 1587 6.80
Professional and scientific instruments 392 3.33 10 0.04
Other manufacturing 254 2.16 325 1.39
Total 11,786 100.00 23,350 100.00
J. Cantwell, L. Piscitello / Journal of International Management 8 (2002) 69–96 81
electrical equipment and pharmaceuticals (Cantwell and Iammarino, 2000). Foreign-owned
efforts are relatively much less attracted to the intermediate centres of the Northwest and the
West Midlands than indigenous activity might suggest, but insofar as they are active there,
they match local specialisation in chemicals in the Northwest and in engineering and transport
equipment in the West Midlands.
In the Italian case as well, foreign-owned firms make their greatest contribution to the
domestic research base in general engineering (8.90% vs. 0.0%), electrical equipment
(25.59% vs. 1.40%) and pharmaceuticals (6.42% vs. 0.0%). We know that the development
efforts of foreign-owned firms are drawn even in relative terms to the major centre of
Lombardia due to the availability of general technological skills and wider infrastructure there
rather than for any particularly specialised expertise. However, it is Lombardia outside
Milano that is relatively most attractive for the siting of R&D by foreign-owned firms, while
Milano itself is ranked only moderately by foreign-owned firms. This may be consistent with
what we know of the lack of technological cospecialisation between indigenous and foreign-
owned firms in Lombardia as a whole (Cantwell and Iammarino, 1998). While foreign-owned
companies are keen to access the regional infrastructure, as latecomers (compared to the
established domestically owned firms), they wish to do so while avoiding the costs of
congestion within Milano itself.
Foreign investment in innovation has therefore as much a regional scope as it has a
national one. In particular, recent trends in the EU support the conjecture that a comparative
analysis at the subnational scale is the most appropriate way to identify the effects of
globalisation (Cantwell and Iammarino, 2000).
Table 8
Patents granted for research in Italy in large foreign-owned and domestically owned firms by industry, 1969–1995
Industry
No. of foreign-
owned patents %
No. of domestically
owned patents %
Food, drink, tobacco and allied products 35 1.38 0 0.00
Chemicals 738 29.06 1592 35.46
Pharmaceuticals 163 6.42 0 0.00
Metals 71 2.80 43 0.96
Mechanical engineering 226 8.90 0 0.00
Electrical equipment 650 25.59 63 1.40
Office equipment 171 6.73 609 13.56
Motor vehicles 71 2.80 924 20.58
Aircraft 111 4.37 0 0.00
Other transport equipment 0 0.00 0 0.00
Textiles 1 0.04 113 2.52
Paper, printing and publishing 3 0.12 0 0.00
Rubber products 75 2.95 508 11.31
Nonmetallic mineral products 66 2.60 0 0.00
Coal and petroleum products 82 3.23 638 14.21
Professional and scientific instruments 6 0.24 0 0.00
Other manufacturing 71 2.80 0 0.00
Total 2540 100.00 4490 100.00
J. Cantwell, L. Piscitello / Journal of International Management 8 (2002) 69–9682
Having illustrated the geographical and sectoral distribution of the technological activity
carried out by domestic and foreign-owned firms across the regions of the three countries
considered, the main issue that arises is the determinants of asymmetries between the
geographical distribution of foreign-owned firm activity compared to that of domestically
Fig. 1. Domestically owned and foreign-owned patents, Germany.
Fig. 2. Domestically owned and foreign-owned patents, United Kingdom.
J. Cantwell, L. Piscitello / Journal of International Management 8 (2002) 69–96 83
owned firms. In other words, we investigate whether foreign-owned labs mirror domestic
location, and if they do not (see Figs. 1–3), the question becomes why, i.e. whether a linear
proportional mapping from the geographical dispersion of indigenous company activity
(a linear agglomeration effect) would exhaustively explain foreign-owned firms’ locational
patterns or whether the effect is instead more complex and reinforced by the attraction exerted
by other location-specific factors.
4. The econometric model and specification of the variables used
The phenomenon under study is the locational preference of foreign-owned firms as
between alternative regions once companies have decided to locate their technological
activities in a given country. Therefore, the dependent variable is the number of patents
granted to foreign-owned firms in each region i and industry j, as follows:
NPAT�FORij = number of patents granted to foreign-owned firms in region i and industry
j over the period 1969–1995: i = 1,. . ., 38 for Germany; i = 1,. . ., 35 for the UK; i = 1,. . ., 21for Italy; j = 1,. . ., 17 industries.
The industrial dimension j allows us to take into account the sectoral disparities in the
innovation-related activities’ propensity to cluster as well as in the propensity to patent.6
Fig. 3. Domestically owned and foreign-owned patents, Italy.
6 The authors are grateful to an anonymous referee for this helpful suggestion.
J. Cantwell, L. Piscitello / Journal of International Management 8 (2002) 69–9684
Indeed, while innovative activities tend in general to agglomerate within specific locations,
the intensity of the geographical concentration and the spatial organisation of the innovative
processes may differ remarkably across sectors (Breschi, 1999). The classification of firms
into one of 17 primary industries of output is as shown in Tables 6–8.
As the dependent variable is clearly a count variable, a binomial regression model was
fitted to the data.7 This kind of linear exponential model offers an improved methodology for
count models for the cases of patents and innovation counts (Hausman et al., 1984; Blundell
et al., 1995; Cameron and Trivedi, 1998).
According to the conceptual model developed in Section 2, the independent variables
relate to:
(1) The agglomeration effect or industry-specific spillovers has been proxied by the
number of patents granted that are attributable to the research of domestically owned
firms in region i and industry j over the period 1969–1995 (IND�SPILLij).(2) Local knowledge externalities (external sources of knowledge). In order to capture
the complex character of local knowledge externalities, we considered proxies both
for non-corporate R&D activities and the size of tertiary education in each region.
The proxies used for the former are R&D expenditures and personnel engaged in
R&D in the government sector (RDEXP�Gi, RDPER�Gi) and in higher educa-
tion (RDEXP�Hi, RDPER�Hi). The commitment to higher and further educa-
tion has been proxied by the number of full-time students in total (EDUC�TOTi)and those in higher education (EDUC�Hi). All these data come from the Regio
dataset (Eurostat).
(3) Localised inter-company cluster-based spillovers have been proxied by a composite
variable (CLUST�SPILLij), taking into account the cumulative impact of a widespread
inter-company technological presence (CLUSTERi) in reinforcing industry-specific
spillovers (IND�SPILLij). Specifically, CLUSTERi is measured by the inverse of the
cross-firm concentration of activity in a region, which rises the more widely that
technological effort is dispersed across firms in each region. In particular, the
variable is the inverse of the coefficient of variation across firm shares of patenting
(CLUSTERi = mi/si, where mi and si are the mean and standard deviation, respectively,
over the shares of corporate patenting of all the individual firms active in region i). The
composite variable CLUST�SPILLij has been therefore calculated as CLUSTERi�IND�SPILLij.
(4) General purpose spillovers (GEN�SPILLi) relate to the breadth of technological
development in a region creating the opportunity for interindustry exchanges.
Therefore, they have been proxied by the inverse of the coefficient of variation over
7 The other possible model normally used for count data, the Poisson model, presents a major drawback
related to the fact that the conditional mean is assumed to be equal to the conditional variance, so that any cross-
sectional heterogeneity is ruled out. The negative binomial model provides a generalisation that solves the
problem by introducing an individual unobserved effect into the conditional mean (Greene, 1997).
J. Cantwell, L. Piscitello / Journal of International Management 8 (2002) 69–96 85
the profile of regional technological specialisation across technological fields
(GEN�SPILLi = mi/si). The profile of regional technological specialisation is measured
by the RTA index, RTAik in region i and technological field k (where k= 1,. . ., 56).8
(5) Local market size has been proxied by the GDP per capita (GDP�PC). Data for this
variable come from the Regio dataset (Eurostat).
The summary characteristics of the variables and the correlation matrix are reported in
Tables 9 and 10, respectively.
5. Empirical findings
Empirical findings obtained for the three countries are reported in Tables 11–13. As using
absolute numbers of patents as a dependent variable might pose difficulties associated with
differences in the propensity to patent in different industries, this has been circumvented by
using industry dummies.
The results confirm that the geographical agglomeration of innovation is remarkable
(Cantwell and Iammarino, 2000) and demonstrate statistically that foreign-owned firms are
even more sensitive than are indigenous companies to agglomeration potential. Nonetheless,
the estimated equations confirm the differences already highlighted in Figs. 1–3 among the
three countries considered.
Local external sources of knowledge show a positive uniform impact on the locational
choice of MNCs for all three countries. Specifically, a larger educational base is attractive to
the location of foreign-owned research (EDUC�TOT9 is always significant at P < .01) as are
region-specific public science externalities. Indeed, the positive and significant signs for
RDEXP�G10 (at P < .05 in the UK as a supplement to EDUC�TOT and at P< .05 in
Germany and Italy when considered separately rather than in addition to EDUC�TOT) bearfurther testament to the role played by the governments in strengthening the regional science
base by providing the core general funding. The two effects are jointly positive and
statistically significant for the UK, while in Germany and Italy they had to be considered
separately. These results confirm that lower-order regions can be highly relatively attractive
9 The same results hold when considering higher education instead of total tertiary education. As a matter of
fact, EDUC�TOT and EDUC�H are highly correlated (the correlation coefficient is always above .9).
8 The Revealed Technological Advantage (RTA) index is a proxy for specialisation across technological fields,
which fields are groupings derived from the US patent class system (for a discussion and a list of the 56 fields
used, see, e.g. Cantwell and Iammarino, 2000), and is calculated in the following way: RTAik = (Pik/Pwk)/(P
kPik/PkPwk) where Pik= number of patents granted in field k to firms for research in region i and Pwk = number of
world corporate patents granted in sector k. It should be noted that patents associated with some field k may be due
to firms in any industry (any j), and so widespread regional technological development across a broad range of
fields k is usually indicative of the existence of areas of technological overlap between industries and hence
indicates the scope for technological spillovers between industries and especially in GPTs, which are those
technologies that are relevant to more than one industry.
J. Cantwell, L. Piscitello / Journal of International Management 8 (2002) 69–9686
where they have a good local science base (Cantwell and Iammarino, 2000). Even more, they
also confirm the importance in Europe of co-location for science–technology linkages, as
demonstrated previously from US evidence (Jaffe et al., 1993).
Table 9
The dependent and independent variables
Mean S.E. Minimum Maximum
Germany
NPAT�FOR 23.557 77.111 0.000 761.000
IND�SPILL 117.980 635.766 0.000 8284.000
MAKT�SIZE 110.947 45.010 15.000 215.000
RDEXP�T 1076.103 583.570 183.900 2516.000
RDEXP�G 162.828 135.180 29.833 795.050
RDPER�G 4950.564 15,042.864 164.000 88,888.566
RDEXP�H 158.654 82.133 55.800 437.433
RDPER�H 3927.306 1718.195 1262.000 8937.500
EDUC�T 358.500 160.302 122.000 792.000
EDUC�H 55.447 33.765 9.000 160.000
CLUSTER 0.004 0.002 0.001 0.011
GEN�SPILL 0.006 0.002 0.003 0.012
UK
NPAT�FOR 19.808 61.407 0.000 825.000
IN�SPILL 39.244 114.793 0.000 1386.000
MAKT�SIZE 84.000 12.682 65.000 128.000
RDEXP�T 559.440 443.509 140.067 1240.829
RDEXP�G 73.054 78.244 3.050 237.525
RDPER�G 577.081 1497.594 0.000 6521.437
RDEXP�H 100.397 1.067 8.000 14.700
RDPER�H 3219.818 2301.895 1512.000 13,040.000
EDUC�T 364.028 101.315 226.333 533.429
EDUC�H 38.257 11.963 21.333 60.000
CLUSTER 0.004 0.002 0.002 0.110
GEN�SPILL 0.006 0.002 0.002 0.014
Italy
NPAT�FOR 7.864 29.614 0.000 323.000
IND�SPILL 13.900 76.877 0.000 881.000
MAKT�SIZE 80.152 43.080 13.100 127.000
RDEXP�T 498.573 550.196 9.000 1725.400
RDEXP�G 115.304 233.595 0.000 1125.700
RDPER�G 1890.190 3875.062 4.000 18,603.000
RDEXP�H 162.311 73.729 51.700 298.500
RDPER�H 3203.266 3037.054 123.000 11,300.000
EDUC�T 302.538 312.523 4.000 1156.238
EDUC�H 75.479 67.900 0.000 234.000
CLUSTER 0.005 0.001 0.003 0.006
GEN�SPILL 0.005 0.001 0.003 0.008
J. Cantwell, L. Piscitello / Journal of International Management 8 (2002) 69–96 87
Table 10
Correlation matrix
IND�SPILL MAKT�SIZE RDEXP�G RDPER�G RDEXP�H RDPER�H EDUC�T EDUC�H CLUSTER
Germany
MAKT�SIZE .12
RDEXP�G .01 .01
RDPER�G � .04 .36 � .1
RDEXP�H .04 .01 .94 � .22
RDPER�H .04 .06 .86 � .26 .98
EDUC�T .07 � .14 .67 � .32 � .11 .89
EDUC�H .06 � .05 .69 � .22 .92 .93 .94
CLUSTER � .17 � .32 � .26 .16 � .26 � .24 � .25 � .25
GEN�SPILL .14 .11 .05 � .24 .05 .04 .08 .02 � .24
UK
MAKT�SIZE .14
RDEXP�G .16 .27
RDPER�G � .07 � .23 � .21
RDEXP�H .09 .25 .89 � .09
RDPER�H .07 .23 .84 � .07 .99
EDUC�T .19 .24 .83 � .17 .54 .43
EDUC�H .17 .35 .83 � .12 .64 .56 .93
CLUSTER � .17 .08 � .14 .11 � .14 � .1 � .18 � .14
GEN�SPILL .23 .4 .63 � .09 .5 .44 .57 .43 � .23
Italy
MAKT�SIZE � .2
RDEXP�G � .06 .2
RDPER�G � .04 .12 .99
RDEXP�H � .07 .26 .74 .74
RDPER�H � .08 .27 .87 .86 .97
EDUC�T � .06 � .08 .52 .54 .65 .61
EDUC�H � .1 .12 .73 .74 .85 .84 .93
CLUSTER � .32 .48 .38 .35 .2 .32 .02 .2
GEN�SPILL .18 � .64 � .003 .04 .3 .2 .1 .07 � .43
J.Cantwell,
L.Piscitello
/JournalofIntern
atio
nalManagem
ent8(2002)69–96
88
Table 11
Estimation results, Germany
Model 1 Model 2 Model 3 Model 4
IND�SPILL 0.0001 (0.972) 0.0001 (0.656)
MAKT�SIZE 0.0220*** (9.516) 0.0235*** (10.000) 0.0217*** (8.932) 0.0231*** (9.344)
RDEXP�G 0.0013** (2.161) 0.0013** (2.146)
EDUC�T 0.0016*** (3.073) 0.0015*** (2.941)
GEN�SPILL 271.1435*** (6.239) 249.7371*** (5.646) 270.7478*** (6.212) 250.6148*** (5.636)
CLUST�SPILL 0.0916 (1.238) 0.0698 (0.946)
Food, drink and tobacco 1.3356*** (3.332) 1.3183*** (3.304) 1.2005*** (2.904) 1.1954*** (2.905)
Chemicals 3.0989*** (7.624) 3.1395*** (7.783) 3.0180*** (7.011) 3.0751*** (7.195)
Pharmaceuticals 2.5016*** (6.388) 2.5289*** (6.500) 2.5142*** (6.248) 2.5437*** (6.359)
Metals 2.3863*** (6.042) 2.3809*** (6.066) 2.3799*** (5.834) 2.3821*** (5.877)
Mechanical engineering 3.0457*** (7.806) 3.0884*** (7.950) 3.0526*** (7.601) 3.0999*** (7.747)
Electrical equipment 4.1346*** (10.538) 4.1333*** (10.586) 4.0234*** (9.953) 4.0330*** (10.017)
Office equipment 2.0936*** (5.236) 2.1085*** (5.333) 2.1419*** (5.221) 2.1629*** (5.331)
Motor vehicles 2.4900*** (6.355) 2.4681*** (6.331) 2.4978*** (6.208) 2.4738*** (6.175)
Aircraft 0.6187 * (1.516) 0.6809 * (1.690) 0.6545** (1.565) 0.7201 * (1.743)
Other transport equipment � 2.4010*** (� 4.244) � 2.4368*** (� 4.316) � 2.5484*** (� 4.322) � 2.5765*** (� 4.382)
Textiles � 2.8694*** (� 4.275) � 2.8435*** (� 4.247) � 2.8215*** (� 4.169) � 2.7933*** (� 4.137)
Paper, printing and publishing � 0.1016 (� 0.237) � 0.0998 (� 0.233) � 0.1702 (� 0.391) � 0.1672 (� 0.384)
Rubber products 0.0461 (0.112) � 0.0845 (� 0.204) 0.1041 (0.247) � 0.0240 (� 0.057)
Nonmetallic mineral products 2.5039*** (6.390) 2.4637*** (6.314) 2.5728*** (6.394) 2.5293*** (6.309)
Coal and petroleum products 2.0351*** (5.171) 1.9600*** (4.990) 2.0177*** (4.990) 1.9419*** (4.809)
Professional and scientific
instruments
1.6823*** (4.228) 1.7143*** (4.363) 1.6693*** (4.082) 1.7125*** (4.243)
Constant � 4.1255*** (� 9.010) � 4.5581*** (� 9.448) � 4.1059*** (� 8.672) � 4.5022*** (� 9.062)
Included observations 595 595 561 561
Log likelihood � 1735.072 � 1732.8122 � 1650.8422 � 1648.941
LR statistic 52,327.673*** 52,332.192*** 50,142.951*** 50,146.755***
LR index (pseudo-r2) .938 .938 .938 .938
Numbers in brackets are z-ratios.
* Significant at P< .10.
** Significant at P< .05.
*** Significant at P< .01.
J.Cantwell,
L.Piscitello
/JournalofIntern
atio
nalManagem
ent8(2002)69–96
89
The coefficients of the proxy used for localised industry-specific spillovers (IND�SPILL)act either autonomously or in harness with the variable built to proxy the existence of
localised cluster-based spillovers (CLUST�SPILL). In Germany, these two spillover effects
show a positive but not significant coefficient, while foreign-owned firms seem to be much
more attracted by the local market size. The latter is consistent with the observation above
that indigenous technological development is often highly regionally polarised in Germany
Table 12
Estimation results, UK
Model 1 Model 2 Model 3
IND�SPILL 0.0015*** (2.639) 0.0016*** (2.673)
MAKT�SIZE 0.0093 (1.556)
RDEXP�G 0.0028** (2.091) 0.0028** (2.084) 0.0027** (2.026)
EDUC�T 0.0048*** (5.009) 0.0049*** (5.029) 0.0049*** (5.061)
GEN�SPILL 42.8512 (1.185) 66.6064** (2.063) 65.9688** (2.036)
CLUST�SPILL 0.6143*** (2.835)
Food, drink and tobacco 0.1952 (0.600) 0.1899 (0.582) 0.1755 (0.538)
Chemicals 2.3530*** (7.047) 2.3354*** (6.971) 2.3123*** (6.852)
Pharmaceuticals 2.1644*** (6.850) 2.1502*** (6.791) 2.1255*** (6.704)
Metals 0.9622*** (2.996) 0.9672*** (3.003) 0.9378*** (2.905)
Mechanical engineering 2.1219*** (6.627) 2.1398*** (6.666) 2.1204*** (6.601)
Electrical equipment 2.0809*** (6.477) 2.0520*** (6.376) 2.0065*** (6.212)
Office equipment 1.4509*** (4.535) 1.4652*** (4.571) 1.4668*** (4.579)
Motor vehicles 0.8167*** (2.553) 0.7847** (2.448) 0.7482 * (2.326)
Aircraft 0.0861 (0.263) 0.0981 (0.299) 0.0774 (0.236)
Other transport
equipment
� 5.1721*** (� 4.850) � 5.1664*** (� 4.852) � 5.1606 (� 4.849)
Textiles � 2.9386*** (� 5.717) � 2.9446*** (� 5.727) � 2.9468*** (� 5.740)
Paper, printing and
publishing
� 0.9884*** (� 2.864) � 1.0006*** (� 2.892) � 0.9996*** (� 2.893)
Rubber products � 0.6100 * (� 1.817) � 0.5993 * (� 1.782) � 0.5976 * (� 1.779)
Nonmetallic mineral
products
0.7940*** (2.429) 0.7593** (2.322) 0.7507** (2.297)
Coal and petroleum
products
0.0521 (0.160) 0.1076 (0.332) 0.0858 (0.264)
Professional and
scientific instruments
0.0962 (0.299) 0.1337 (0.416) 0.1408 (0.438)
Constant � 1.6698*** (� 2.969) � 1.0497*** (� 2.654) � 1.0373*** (� 2.629)
Included observations 561 561 561
Log likelihood � 1650.012 � 1651.261 � 1650.952
LR statistic 36,042.526*** 36,040.028*** 36,040.646***
LR index (pseudo-r2) .916 .916 .916
Numbers in brackets are z-ratios.
* Significant at P< .10.
** Significant at P< .05.
*** Significant at P< .01.
J. Cantwell, L. Piscitello / Journal of International Management 8 (2002) 69–9690
Table 13
Estimation results, Italy
Model 1 Model 2 Model 3 Model 4
IND�SPILL 0.0050*** (3.238) 0.0050*** (3.250) 0.0051*** (3.097)
MAKT�SIZERDEXP�G 0.0000 (0.043) 0.0009** (1.844)
EDUC�T 0.0020*** (3.470) 0.0020*** (3.969) 0.0020*** (4.054)
GEN�SPILL 581.2624*** (6.003) 577.4209*** (5.984) 581.3784*** (6.006) 527.2761*** (4.760)
CLUST�SPILL 1.7068*** (3.264)
Food, drink and tobacco � 0.7540 (� 1.058) � 0.7568 (� 1.063) � 0.7546 (� 1.059) � 0.8225 (� (1.118)
Chemicals 0.5773 (0.823) 0.3867 (0.538) 0.5763 (0.822) 0.8731 (1.188)
Pharmaceuticals 0.4874 (0.711) 0.4914 (0.721) 0.4905 (0.720) 0.4289 (0.600)
Metals � 0.5134 (� 0.736) � 0.5238 (� 0.752) � 0.5149 (� 0.739) � 0.0115 (� 0.016)
Mechanical engineering 1.5371** (2.247) 1.5318*** (2.251) 1.5340** (2.254) 1.7366** (2.429)
Electrical equipment 1.8127*** (2.696) 1.8104*** (2.696) 1.8113*** (2.697) 2.0575*** (2.918)
Office equipment � 0.0750 (� 0.107) � 0.0763 (� 0.109) � 0.0750 (� 0.107) 0.0253 (0.035)
Motor vehicles � 1.6042 * (� 1.857) � 1.6370 * (� 1.878) � 1.6034 (� 1.857) � 1.5617 * (� 1.769)
Aircraft � 0.2790 (� 0.398) � 0.2785 (� 0.397) � 0.2803 (� 0.400) � 0.1876 (� 0.258)
Other transport equipment � 32.4333 (� 0.000) � 35.6421 (� 0.000) � 35.7371 (� 0.000) � 36.3427 (� 0.000)
Textiles � 4.7500*** (� 3.737) � 4.7641*** (� 3.750) � 4.7502*** (� 3.737) � 4.7633*** (� 3.593)
Paper, printing and publishing � 3.2660*** (� 3.535) � 3.2589*** (� 3.532) � 3.2644*** (� 3.536) � 3.1842*** (� 3.384)
Rubber products � 0.4253 (� 0.595) � 0.4179 (� 0.600) � 0.4184 (� 0.601) � 0.3739 (� 0.503)
Nonmetallic mineral products 0.1507 (0.217) 0.1459 (0.211) 0.1483 (0.215) 0.2618 (0.361)
Coal and petroleum products 0.0118 (0.016) � 0.0974 (� 0.134) 0.0123 (0.017) � 0.1150 (� 0.157)
Professional and scientific
instruments
� 3.1537*** (� 3.581) � 3.1439*** (� 3.610) � 3.1592*** (� 3.623) � 2.4018*** (� 2.828)
Constant � 2.1624*** (� 2.798) � 2.1235*** (� 2.756) � 2.1651*** (� 2.809) � 0.9709 (� 1.273)
Included observations 153 153 153 153
Log likelihood � 378.08471 � 378.143 � 378.086 � 385.0559
LR statistic 6673.6567*** 6673.540*** 6673.655*** 6659.7143***
LR index (pseudo r2) 0.900 0.898 0.898 0.896
Numbers in brackets are z-ratios.
* Significant at P< .10.
** Significant at P< .05.
*** Significant at P< .01.
J.Cantwell,
L.Piscitello
/JournalofIntern
atio
nalManagem
ent8(2002)69–96
91
and the qualification that the agglomeration effect can only work where there are a variety
of sources of spillovers and the absence of a single dominant firm that acts to competitively
deter its major rivals. In contrast, in the UK and in Italy, CLUST�SPILL and IND�SPILLare both highly significant (P < .01) attractors to foreign-owned activity. Conversely, the
local market size does not seem to particularly influence the foreign MNC’s relative lo-
cational choice. The combined industry-specific and cluster-based spillover effects show
their strongest impact in Italy. In Italy and the UK, there seems to be a general ag-
glomeration effect associated with the widespread dispersion of patenting firms in the same
region and in the same sector of activity. The presence of general purpose spillovers
(GEN�SPILL) shows a significant positive impact on the location of foreign-owned re-
search facilities especially in Germany and Italy and only to a slightly lesser extent in the
UK. The technological breadth of a region and the presence of innovative overlaps across
industries in the development of GPTs is an important factor in the attraction of foreign-
owned research facilities.
6. Summary and conclusions
Since the late 1970s (Cantwell and Piscitello, 2000), large MNCs have increasingly
extended or diversified their fields of technological competence through their use of
internationally integrated networks for technological development. In each location, in
such a network, MNCs tap into specialised sources of local expertise, and so differentiate
their technological capability, by exploiting geographically separate and hence distinct
streams of innovative potential. The recent emergence of internationally integrated MNC
networks is best observed in Europe, where the contribution of foreign-owned MNCs to
national technological capabilities is much greater than elsewhere. About one-quarter of
large firm R&D carried out within in Europe has been conducted under foreign ownership
(and this figure had risen to nearly 29% by the early 1990s), while the world average is
only just over one-tenth. Part of the reason is that European-owned MNCs are the most
internationalised in their strategies for technology development, while much of their fo-
reign-located R&D has remained within Europe and their European orientation has in-
creased (from a 30% share of foreign R&D in Europe in the late 1960s to a 40% share by
the 1990s).
Our results suggest that the relative attractiveness of regions in Europe to the technological
efforts of foreign-owned MNCs depends upon (1) the presence of external sources of
knowledge, (2) the presence of industry-specific and cluster-based spillovers and (3) the
breadth of local technological specialisation in the region, i.e. the opportunity to capture
general purpose spillovers. That has some implications in suggesting regional policy forms
mainly based on regional investments (rather than exclusively on regional incentives), which
enhance the attractiveness of the region as an appealing economic environment for potential
investors (Braunerhjelm et al., 2000).
Factors (1) and (3) seem to matter for regions throughout Germany, the UK and Italy. This
is consistent with other literature that has emphasised the growing importance of science–
J. Cantwell, L. Piscitello / Journal of International Management 8 (2002) 69–9692
technology spillovers in the current technoeconomic paradigm and which is now paying
increasing attention to the central role of GPTs. To these latter strands of recent literature,
what we add here is the dimension of corporate internationalisation: MNCs will develop
abroad in the appropriate international centre GPTs alongside the firms of other industries
and technologies that rely on linkages to a good local science base. Factor (2) depends
critically on the dispersion of technological development among a sufficient variety of local
actors to attract foreign-owned research to a localised cluster. This occurs quite often in the
UK and Italy but when, as is more frequently the case in Germany, local development is
heavily concentrated in just a few leading firms in a region (i.e. where the leading
domestically owned firms are strongly regionally separated and each have a clear regional
identity), then a crowding out effect is likely to outweigh any agglomeration attraction. In
Germany, each of the major companies, e.g. in the chemical industry, has ‘‘its own’’ region,
and so in a sense, the deterrence effect to technological entry in a region with an existing
dominant player is observed even among the large indigenous German firms themselves.
Naturally, it affects foreign-owned firms in the same industry (and hence which are
competitors of the dominant company in a region) just as much, and so there is much less
scope here for an agglomeration effect.
The present paper could be extended to consider the type of motivations of foreign
investment in each location. Countries and regions seek to attract MNC activity as a means of
improving their locational advantages through spillovers and linkages due to MNC activity.
However, the quality and the extent of the externalities due to MNC activities depends on the
motivation of their investment, which is itself dependent on the kinds of location advantages
available to them (Narula and Dunning, 2000; Cantwell and Narula, 2001).
References
Acs, Z.J., de la Mothe, J., Paquet, G., 2000. Regional innovation: in search of an enabling strategy. In: Acs, Z.J.
(Ed.), Regional Innovation, Knowledge and Global Change. Pinter, London.
Adams, J.D., 2001. Comparative localisation of academic and industrial spillovers, WP 8292, NBER.
Almeida, P., 1997. Knowledge sourcing by foreign multinationals: patent citation analysis in the US semicon-
ductor industry. Strategic Manage. J. 17, 155–165.
Archibugi, D., 1992. Patenting as an indicator of technological innovation: a review. Sci. Public Policy 19,
357–368.
Audretsch, D.B., Feldman, M.P., 1996. R&D spillovers and the geography of innovation and production. Am.
Econ. Rev. 86, 630–640.
Audretsch, D.B., Stephan, P.E., 1996. How localized are networks in biotechnology? Am. Econ. Rev. 86,
641–652.
Baptista, R., Swann, G.M.P., 1998. Do firms in clusters innovate more? Res. Policy 27, 527–542.
Baptista, R., Swann, G.M.P., 1999. A comparison of clustering dynamics in the US and UK computer industries.
J. Evol. Econ. 9, 373–399.
Barrell, R., Pain, N., 1999. Domestic institutions, agglomerations and foreign direct investment in Europe. Eur.
Econ. Rev. 43, 925–934.
Blundell, R., Griffith, R., Van Reenan, J., 1995. Dynamic count data models of technological innovation. Econ. J.
105, 333–344.
Braunerhjelm, P., Svensson, R., 1998. Agglomeration in the geographical location of Swedish MNEs. In:
J. Cantwell, L. Piscitello / Journal of International Management 8 (2002) 69–96 93
Braunerhjelm, P., Ekholm, K. (Eds.), The Geography of Multinational Firms. Kluwer Academic Publish-
ing, Dordrecht.
Braunerhjelm, P., Faini, R., Norman, V., Ruane, F., Seabright, P., 2000. Integration and the Regions of Europe:
How the Right Policies Can Prevent Polarization. Monitoring European Integration 10. Centre for Economic
Policy Research, London.
Breschi, S., 1999. Spatial patterns of innovation: evidence from patent data. In: Gambardella, A., Malerba, F.
(Eds.), The Organization of Economic Innovation in Europe. Cambridge Univ. Press, Cambridge.
Breschi, S., 2000. The geography of innovation: a cross sector analysis. Reg. Stud. 34, 213–229.
Cameron, A.C., Trivedi, P.K., 1998. Regression Analysis of Count Data. Cambridge Univ. Press, Cambridge.
Caniels, M.C.J., 2000. Knowledge Spillovers and Economic Growth: Regional Growth Differentials Across
Europe. Edward Elgar, Cheltenham.
Cantwell, J.A., 1989. Technological Innovation and Multinational Corporations. Blackwell, Oxford.
Cantwell, J.A., 1995. The globalisation of technology: what remains of the product cycle model? Cambridge J.
Econ. 19, 155–174.
Cantwell, J.A., 2000. Technological lock-in of large firms since the interwar period. Eur. Rev. Econ. Hist. 4,
147–174.
Cantwell, J.A., Iammarino, S., 1998. MNCs, technological innovation and regional systems in the EU: some
evidence in the Italian case. Int. J. Econ. Bus. 5 (3), 383–408.
Cantwell, J.A., Iammarino, S., 2000. Multinational corporations and the location of technological innovation in
the UK regions. Reg. Stud. 34, 317–332.
Cantwell, J.A., Janne, O.E.M., 1999. Technological globalisation and innovative centres: the role of corporate
technological leadership and location hierarchy. Res. Policy 28, 119–144.
Cantwell, J.A., Kosmopoulou, E., 2002. What determines the internationalisation of corporate technology? In:
Forsgren, M., Hakanson, H., Havila, V. (Eds.), Critical Perspectives on Internationalisation (in press).
Cantwell, J.A., Mudambi, R., 2000. The location of MNE R&D activity: the role of investment incentives.
Manage. Int. Rev. 40 (Special Issue 1), 127–148.
Cantwell, I.A., Narula, R., 2001. The eclectic paradigm in the global economy. Int. J. Econ. Bus. 8.
Cantwell, J.A., Noonan, C.A., 2001. The regional distribution of technological development—evidence from
foreign-owned firms in Germany. In: Feldman, M.P., Massard, N. (Eds.), Knowledge Spillovers and the
Geography of Innovation. Kluwer Academic Publishing, Dordrecht.
Cantwell, J.A., Piscitello, L., 1999. The emergence of corporate international networks for the accumulation of
dispersed technological capabilities. Manage. Int. Rev. 39 (Special Issue 1), 123–147.
Cantwell, J.A., Piscitello, L., 2000. Accumulating technological competence— its changing impact on corporate
diversification and internationalisation. Ind. Corporate Change 9, 21–51.
Cantwell, J.A., Iammarino, S., Noonan, C.A., 2001. Sticky places in slippery space— the location of innovation
by MNCs in the European regions. In: Pain, N. (Ed.), Inward Investment, Technological Change and Growth.
The Impact of Multinational Corporations on the UK Economy. Palgrave, Basingstoke.
Carrincazeaux, C., Lung, Y., Rallet, A., 2001. Proximity and localisation of R&D activities. Res. Policy 30,
777–789.
Dunning, J.H., 1994. Multinational enterprises and the globalisation of innovatory capacity. Res. Policy 23, 67–88.
Dunning, J.H., 2000. Regions, globalisation and the knowledge economy: the issues stated. In: Dunning, J.H.
(Ed.), Regions, Globalisation and the Knowledge Based Economy. Oxford Univ. Press, Oxford.
Dunning, J.H., Pearce, R.D., 1985. The World’s Largest Industrial Enterprises 1962–1983. Gower, Farnborough.
Feldman, M.P., Audretsch, D.B., 1999. Innovation in cities: science-based diversity, specialisation and localised
competition. Eur. Econ. Rev. 43, 409–429.
Fors, G., 1996. R&D and Technology Transfer by Multinational Enterprises. Almquist and Wiksell and
IUI, Stockholm.
Gorg, H. and Strobl, E. 2001. Footloose multinationals? University of Nottingham Leverhulme Centre Research
Paper Series, 2001/07.
Greene, W.H., 1997. Econometric Analysis, third ed. Prentice-Hall, New York.
J. Cantwell, L. Piscitello / Journal of International Management 8 (2002) 69–9694
Griliches, Z., 1990. Patent statistics as economic indicators: a survey. J. Econ. Lit. 28, 1661–1707.
Hakanson, L., 1992. Locational determinants of foreign R&D in Swedish multinationals. In: Granstrand, O.,
Hakanson, L., Sjolander, S. (Eds.), Technology Management and International Business: Internationalisation of
R&D and Technology. Wiley, Chichester.
Hausman, J., Hall, B., Griliches, Z., 1984. Econometric model for count data with an application to the patents—
R&D relationship. Econometrica 52, 909–938.
Jaffe, A., Trajtenberg, M., Henderson, R., 1993. Geographical localisation of knowledge spillovers, as evidenced
by patent citations. Q. J. Econ. 58, 577–598.
Kline, S.J., Rosenberg, N., 1986. An overview of innovation. In: Landau, R., Rosenberg, N. (Eds.), The Positive
Sum Strategy. National Academy Press, New York.
Kogut, B., Chang, S.J., 1991. Technological capabilities and Japanese foreign direct investment in the United
States. Rev. Econ. Stat. 73, 401–413.
Kuemmerle, W., 1999. The drivers of foreign direct investment into research and development: and empirical
investigation. J. Int. Bus. Stud. 30 (1), 1–24.
Kumar, N., 1996. Intellectual property protection, market orientation and location of overseas R&D activities by
multinational enterprises. World Dev. 24 (4), 673–688.
Kumar, N., 2001. Determinants of location of overseas R&D activity of multinational enterprises: the case of US
and Japanese corporations. Res. Policy 30, 159–174.
Lipsey, R.G., Bekar, C., Carlaw, K., 1998. What requires explanation? In: Helpman, E. (Ed.), General Purpose
Technologies and Economic Growth. MIT Press, Cambridge.
Mariotti, S., Piscitello, L., 2001. The role of territorial externalities in affecting internationalisation of production
by SMEs. Entrepreneurship Reg. Dev. 13, 65–80.
Narula, R., Dunning, J.H., 2000. Industrial development, globalisation and multinational enterprises: new realities
for developing countries. Oxford Dev. Stud. 28 (2).
Nelson, R.R., 1993. National Systems of Innovation. Frances Pinter, London.
Nelson, R.R., Rosenberg, N., 1999. Science, technological advance and economic growth. In: Chandler, A.D.,
Hagstrom, P., Solvell, O. (Eds.), The Dynamic Firm: The Role of Technology, Strategy, Organization, and
Regions. Oxford Univ. Press, Oxford.
Niosi, J. (Ed.), 1999. Special Issue on Globalisation of Technology. Res. Policy 28 (2–3).
Odagiri, H., Yasuda, H., 1996. The determinants of overseas R&D by Japanese firms: an empirical study at the
industry and company levels. Res. Policy 25, 1059–1079.
Ohmae, K., 1995. The End of the Nation State: The Rise of Regional Economies. Harper Collins, London.
Paci, R., 1997. More similar and less equal: economic growth in the European regions, Weetwiztschafteiches
Archiv 133, 609–634.
Pavitt, K.L.R., 1985. Patents statistics as indicators of innovative activities: possibilities and problems. Sciento-
metrics 7, 1–2.
Pavitt, K.L.R., 1987. International patterns of technological accumulation. In: Hood, N., Vahlne, J.E. (Eds.),
Strategies in Global Competition. Croom Helm, London.
Pavitt, K.L.R., 1988. Uses and abuses of patent statistics. In: van Raan, A.F.J. (Ed.), Handbook of Quantitative
Studies of Science and Technology. Elsevier, Amsterdam.
Porter, M.E., 1996. Competitive advantage, agglomerative economics and regional policy. Int. Reg. Sci. Rev. 19,
85–94.
Rosenberg, N., 1976. Perspectives on Technology. Cambridge Univ. Press, Cambridge.
Rosenberg, N., Nelson, R.R., 1996. The roles of universities in the advance of industrial technology. In: Rose-
nbloom, R.S., Spencer, W.J. (Eds.), Engines of Innovation: US Industrial Research at the End of an Era.
Harvard Business School Press, Boston, MA.
Schmookler, J., 1950. The interpretation of patent statistics. J. Pat. Off. Soc. 32, 123–146.
Schmookler, J., 1966. Inventions and Economic Growth. Harvard Univ. Press, Cambridge, MA.
Scott, A.J., 1998, March. Regional Motors of the Global Economy. Rutgers University, Newark (CIBER Dis-
tinguished Lecture Series No. 1).
J. Cantwell, L. Piscitello / Journal of International Management 8 (2002) 69–96 95
Verspagen, B., Schoenmakers, W., 2000. The spatial dimension of knowledge spillovers in Europe: evidence from
firm patenting data. MERIT Working paper 2/20-16.
Zander, I., 1999. How do you mean global? A taxonomy of innovation networks in the multinational corporation.
Res. Policy 28 (2–3), 195–213.
Zejan, M.C., 1990. R&D activities in affiliates of Swedish multinational enterprises. Scand. J. Econ. 92 (3),
487–500.
J. Cantwell, L. Piscitello / Journal of International Management 8 (2002) 69–9696