This article was downloaded by: [North Dakota State University]On: 15 September 2013, At: 07:46Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number:1072954 Registered office: Mortimer House, 37-41 Mortimer Street,London W1T 3JH, UK

Economic Systems ResearchPublication details, including instructions forauthors and subscription information:http://www.tandfonline.com/loi/cesr20

Comparing Patterns ofIndustrial Interdependencein National Systems ofInnovation - A Study ofGermany, the UnitedKingdom, Japan and theUnited StatesIna Drejer a ba IKE Group and DRUID, Department of BusinessStudies, Aalborg University, Denmarkb PLS Consult, DenmarkPublished online: 01 Jul 2010.

To cite this article: Ina Drejer (2000) Comparing Patterns of IndustrialInterdependence in National Systems of Innovation - A Study of Germany, theUnited Kingdom, Japan and the United States, Economic Systems Research, 12:3,377-399, DOI: 10.1080/09535310050120943

To link to this article: http://dx.doi.org/10.1080/09535310050120943

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of allthe information (the “Content”) contained in the publications on ourplatform. However, Taylor & Francis, our agents, and our licensorsmake no representations or warranties whatsoever as to the accuracy,completeness, or suitability for any purpose of the Content. Any opinionsand views expressed in this publication are the opinions and views ofthe authors, and are not the views of or endorsed by Taylor & Francis.The accuracy of the Content should not be relied upon and should beindependently verified with primary sources of information. Taylor andFrancis shall not be liable for any losses, actions, claims, proceedings,

demands, costs, expenses, damages, and other liabilities whatsoeveror howsoever caused arising directly or indirectly in connection with, inrelation to or arising out of the use of the Content.

This article may be used for research, teaching, and private studypurposes. Any substantial or systematic reproduction, redistribution,reselling, loan, sub-licensing, systematic supply, or distribution in anyform to anyone is expressly forbidden. Terms & Conditions of accessand use can be found at http://www.tandfonline.com/page/terms-and-conditions

Eco

nom

ic S

yste

ms

Res

earc

h 20

00.1

2:37

7-39

9.

Economic Systems Research, Vol. 12, No. 3, 2000

Comparing Patterns of Industrial Interdependence

in National Systems of InnovationÐ A Study of

Germany, the United Kingdom, Japan and the

United States

INA DREJER

(Received July 1999; revised March 2000)

Abst r ac t This paper presents a quantitatively based method for comparing the structureof National Systems of Innovation (NSI). The emphasis is on technological interdependen-cies at the industrial level in Germany, Japan, the United Kingdom and the UnitedStates. The mapping of the interdependencies, based on input± output tables, builds on agraph theoretical model (a minimal ¯ ow analysis). R&D expenses are used as thetechnology indicator. The NSI framework is taken as the point of departure. It isclaimed that `history matters’ , through relating historical descriptions and analyses ofindustrialization processes to the ® ndings of structural analyses of R&D interdependencieswithin the NSIs. The paper shows that the national systems tend to cluster in two main`bulks’ . One is centred around industrial chemicals and/or pharmaceuticals, and the otheris centred around communication equipment, electronics etc. In most cases these clustersdo not appear to be closely technologically related through embodied R&D ¯ ows, i.e. itseems appropriate to assume that two distinct technology bases are at play.

Keywor ds: National Systems of Innovation; minimal ¯ ow analysis

1. Introduction

This paper compares the structures of technological interindustrial interdependencein the national system of innovation (NSI) for four major OECD countries:Germany, the United Kingdom, Japan and the United States. The paper sets outto compare the structure of technological interdependencies between industries asthey are expressed by embodied R&D ¯ ows in the four countries. The nationalsystem of innovation approach is used as the point of departure, and the majoraim is to identify diþ erent structures and to illustrate how these diþ erences inclusters relate to national industrial trajectories. Thus, the paper relates historical

Ina Drejer, IKE Group and DRUID, Department of Business Studies, Aalborg University, Denmark,and PLS Consult, Denmark. E-mail: id@business.auc.dk. A previous version of this paper was presentedat the DRUID conference on National Innovation Systems, Industrial Dynamics and Innovation Policy, inRebild, Denmark, 9 ± 12 June 1999. The author is grateful to the two discussants, Bart Verspagen andRajneesh Narula, for giving valuable comments at that occasion. The usual disclaimer applies.

ISSN 0953-5314 print; 1469-5758 online/00/030377-23� 2000 The International Input± Output Association

Eco

nom

ic S

yste

ms

Res

earc

h 20

00.1

2:37

7-39

9.

378 I. Drejer

descriptions and analyses of industrialization processes to the ® ndings of structuralanalyses of R&D interdependencies.

The paper shows that a relatively simple graphical representation of major R&Drelations within a national system of innovation can illustrate some fundamentaldiþ erences between systems, which cannot be revealed from, for example, economickey ® gures. Such economic key ® gures or indicators for the four countries analysedare presented in Appendix A. It is claimed that an input± output based graphtheoretical model is a relatively simple way, by applying quantitative data, toillustrate some qualitative diþ erences between national innovation systems.

Of course, each country cannot be done justice in the limited space availablehere, but the admittedly super® cial stories of the development of each of the foursystems serve to illustrate the major point of this paper: that diþ erences in theindustrial development of each country, and the institutional factors in¯ uencingthis development, result in diþ erences in the overall structural relations of thenational systems of innovation. It should also be kept in mind, that the aim of thepaper is not to present a comprehensive or balanced historical explanation ofnational diþ erences. Thus, only historical factors, which can be seen to explainstructural diþ erences, are included in the paper.

A national system of innovation is constituted by the institutions and economicstructures aþ ecting the rate and direction of technological change in the society. Itincludes not only the system of technology diþ usion and the R&D system, butalso institutions and factors determining how technology aþ ects productivity andeconomic growth (Edquist & Lundvall, 1993, p. 267). Even though knowledgeand technology gets diþ used through several other channels rather than embodiedR&D ¯ ows, the identi® cation of these ¯ ows is an important ® rst step in understand-ing the structure of a national system of innovation. An analysis of embodied R&D¯ ows that uncovers major sources for the spread of technology in the economicsystem can point out sectors that have a widespread eþ ect on the whole systemthrough the diþ usion of technology as a result of transactions between industries.However, the patterns of interdependence might also help in understanding theimportance of the historical background for the present set-up of the system,claiming that the current structure of the systems is largely dependent on their pasthistory of industrialization.

R&D is only a proxy of the input eþ ort given to a technology creation anddevelopment process, and furthermore technology itself is only a subset of know-ledge. The analysis in the present paper is thus limited to the study of a reasonablywell de® ned corner of the total knowledge interdependence and diþ usion systemwithin a national system of innovation, keeping in mind that knowledge getsdiþ used through several channels, of which many are informal and diý cultÐ if notimpossibleÐ to measure. Thus, even though it is claimed that it is national systemsof innovation that are being studied in the present paper, it should be kept in mindthat the focus is on the R&D dimension, i.e. it is the R&D systemÐ a subset of theinnovation systemÐ that is the subject of the study.

The relations studied are national only in that the focus is on intra-countryrelations, i.e. trying to compare national diþ erences in the way the national econom-ies are integrated. This, of course, does not imply that international relations arenot crucial in understanding and explaining technological development in advancedopen economies, especially since some industries are more internationally orientedthan others. But a national system of innovation is characterized by historical speci-® city and a multiplicity of institutional con® gurations that aþ ect its outcomes, and,

Eco

nom

ic S

yste

ms

Res

earc

h 20

00.1

2:37

7-39

9.

Industrial Interdependence in National Systems of Innovation 379

although, globalization can change the nature of a national system of innovationsubstantially by, for example, adding new international linkages and by making thesystems more interactive, it is unlikely that globalization will eliminate national orlocal speci® cities completely (Saviotti, 1997, pp. 195± 196).

The analysis of the characteristics of each national system of innovation drawson the empirical analysis of national areas of strength and weakness in Porter(1990), since these national analyses are very rich in empirical and historical detail.Although Porter does not deal with the innovation system concept, but focuses onexplaining and analysing the competitive advantages and disadvantages of nations,there is a large degree of overlap between the two approaches, at least when itcomes to analysing dynamic and competitive characteristics of nation countries.Furthermore, Porter played an important role in introducing the linkage conceptin a policy perspective, which contributed to turning increased attention towardsthe systemicÐ interdependentÐ nature of economies. An example of a cluster-based industry and technology policy is presented in Drejer et al. (1999).

The analysis is related to the analysis presented in DuÈ ring & Schnabl (2000),which compares structural changes in Germany, Japan and the United Statesbetween 1980 and 1990. While DuÈ ring & Schnabl’s primary concern is thecomparison of the structure and the structural changes of interindustry R&D ¯ owswithin the three countries from a convergence point of view, i.e. stressing increasingsimilarities between national systems, the present paper focuses on relating thestructures to the historical past, thus stressing the diþ erences between thesesystems.

2. Why Study National Systems of Innovation?

Lundvall (1998) answers the question `why study national systems of innovation’by pointing to the importance of understanding diþ erent styles of innovation, anddiþ erences in how new knowledge is created, distributed and used for establishinga theoretical basis for the analysis of national systems of innovation.

Innovation and learning are cornerstones in the national system of innovationapproach. But the way that innovative activity is carried out, and the way learningÐwhich, in the national system of innovation approach, is perceived as interactivelearning’ Ð takes place in a system, is aþ ected by institutions. Both formal institu-tions and informal institutions perceived as norms, routines, habits etc., areimportant. Some of the types of informal institutions pointed to by Lundvall (1998,p. 409) as especially important in the context of learning and innovation are thefollowing.

(i) The `time horizon’ of agents: the Anglo-Saxon systems are characterizedby a shorter time horizon in corporate governance than the Japanese andGerman systems, which are known for working with quite long timehorizons in investment decisions.

(ii) The role of trust’ : the German and Japanese systems are perceived asbeing more trust oriented in business matters than the Anglo-Saxonsystems (see, for example, DeBresson et al., 1998).

(iii) The way that `authority’ is expressed: the expression of `authority’ inindustrial relations aþ ects the capability to learn. Here, Lundvall points toPolyani’s (1966) proposition that the learning of new skills typically takesplace in the context of a master± apprenticeship relationship, where a

Eco

nom

ic S

yste

ms

Res

earc

h 20

00.1

2:37

7-39

9.

380 I. Drejer

mixture of trust and authority is necessary in order for learning to takeplace eý ciently. The learning capabilities of Asian countries, representedhere by Japan, are, in certain areas, suggested to be rooted in the specialkinds of authority relations in these countries (Lundvall, 1998).

It is not possible to study the functioning of these diþ erent types of informalinstitutions in an analysis of the aggregated type, as performed here. However, itis worth remembering that these underlying factors might help explain the develop-ment leading to the current structure.

Patel & Pavitt (1993) also deal with the institutional in¯ uence on the set-up ofnational systems of innovation, but Patel & Pavitt primarily deal with formalinstitutions in the form of business ® rms, universities and other training institutionsas well as government. Thus, they de® ne a national system of innovation as:

the national institutions, their incentive structures and their competencies,that determine the rate and direction of technological learning (or thevolume and composition of change-generating activities) in a country.(Patel & Pavitt, 1993, pp. 5± 6)

The kinds of institutions in mind are mentioned above. The incentive structure,among other things, involves government support for basic research, but Patel &Pavitt also point to possible disincentives regarding investment in competenceenhancement in the form of mobility of employees (making the return of ® rm-based training uncertain), and the relation between competition and imitation. Aswe will show below, diþ erences in institutions, in particular higher educationinstitutions, as well as in incentive structures in¯ uencing, for example, militaryoriented research, have had a considerable in¯ uence on observed diþ erences in thestructural set-up of national systems of innovation.

3. Methodological Considerations

R&D expenses are used as a proxy of technology in the present paper. The use ofR&D expenses represents a very narrow perception of technology. R&D is oneinput factor to a technological creation process; a process that is, in fact, toocomplex to describe using one single factor only. As illustrated in Drejer (1999),the combination of diþ erent knowledge and technology indicators, including bothinput and output indicators, is an important step in de® ning technology intensiveindustries’ . However, due to restrictions in the form of data availability, it has notbeen possible to combine several indicators in the present paper, and thus thelimitations of just looking at one indicator of technological activity should be keptin mind.

The OECD STAN databases, which among other things consist of data cover-ing industrial R&D expenditures and input± output relations, provide new oppor-tunities to the comparative analysis of technology transfers. An analysis byPapaconstantinou et al. (1996), as well as a follow-up analysis by Sakurai et al.(1996), used the STAN input± output matrices and ANBERD data on R&Dexpenses as the foundation of their analysis of the diþ usion of R&D and industrialperformance (productivity) in the manufacturing industry in ten OECD-countries(see also Sakurai et al., 1997).

The present paper does not deal with the productivity issue, but will concentrateon identifying the patterns of technological interdependencies within the four major

Eco

nom

ic S

yste

ms

Res

earc

h 20

00.1

2:37

7-39

9.

Industrial Interdependence in National Systems of Innovation 381

OECD countries, applying the same data sets as Papaconstantinou et al. (1996)and Sakurai et al. (1996). The main feature distinguishing the present paper frommost previous studies of interdependence is the focus on the industrializationpathway as a factor contributing to the understanding of the present day structureof R&D interdependence. Thus, the R&D-structure of each national system isrelated to the national history of industrial development.

The model applied is a graph theoretical model (see Appendix B for thetechnical description of the model) which transforms the input± output system to aminimal ¯ ow system where only ¯ ows exceeding a preset value will be included.

R&D expenditures are weighted by input± output coeý cients expressing theeconomic interdependence between industries, and are used as expressions ofembodied technology ¯ ows between industries. That is, it is assumed that theembodied technology ¯ ows are proportional to the R&D expenditures of R&Dperforming industries, as well as the quantitative extent of the ¯ ows of intermediategoods between the user and producer industries. The advantage of this method isthat it captures the combined eþ ect of R&D activities and the structure of theproduction system in which these activities are transported through intermediatecommodity ¯ ows from producers to users.

An analysis of embodied knowledge ¯ ows, applying input± output data, impliesthat the sectors of utilization of knowledge carried out by an industry are propor-tionally the same as the sectors of utilization of goods and services in input± outputtables. As pointed out by Archibugi (1988, p. 273) there is no certainty that theR&D ¯ ows of an industry have the same direction as the industry’s products.However, as Archibugi also states, even though the analysis of knowledge ¯ owsmight only produce indications, these indications are still valuable in the absenceof a de® nite proof regarding ¯ ows between sectors.

The graphs are constructed for Germany, the United Kingdom, Japan and theUnited States for the year 1990, which is the most recent year available in theSTAN input± output database. All values are calculated in US dollars, but the ® ltervalues are scaled according to the total business sector R&D expenses in eachcountry, i.e. the ® lter value for the United States is approximately 1.5 times aslarge as the ® lter value for Japan, since the R&D expenditures in the United Statesare approximately 1.5 times as large as the Japanese R&D expenditures. Diþ erent® lter values have been calculated in order to test the stability of the relations, aswell as to achieve structures that represent a balanced consideration of twocon¯ icting interests: reducing the complexity of the graphs (achieved throughsetting a high ® lter value), and including as many relations as possible (achievedthrough setting a low ® lter value). Thus, the graphs presented in Figures 1 through4 are the result of a trial and error process of applying diþ erent ® lter values. In the® gures, ¯ ows go from left to right (except when they are marked with an arrow).Bold lines express bilateral relations.

4. Exploring DiVerences in Structures of Industrial Interdependence in

National Systems of Innovation

The four countries dealt with in this paper might, super® cially, look similar froman overall industrial point of view. Looking at the main economic indicators of thefour countries in Appendix A, one ® nds that food, motor vehicles and non-electricalmachinery are among the ® ve most important industries from a production volumepoint of view in all countries. Food and non-electrical machinery are among the

Eco

nom

ic S

yste

ms

Res

earc

h 20

00.1

2:37

7-39

9.

382 I. Drejer

® ve largest employment industries in all countries. Concerning R&D eþ orts,communication equipment and motor vehicles are among the ® ve most importantindustries in all countries. The export specialization indicator is the only one wherethe similarities between countries are not too obvious, although some commonfeatures also appear here. The United States distinguishes itself from the otherthree countries by being most strongly export specialized in high-tech industriesonly: aerospace, oý ce machines and computers, instruments, communicationequipment and semiconductors as well as industrial chemicals.

The industries spending most on R&D among the ® ve industries in whichGermany is most strongly export specialized are motor vehicles, industrial chemicalsand non-electrical machinery; the two other industries in the `export specializationTop 5’ are fabricated metal products, and electrical machinery. The UK is moststrongly export specialized in the R&D-intensive aerospace industry, followed byother manufacturing industries. In the following places we ® nd another research-intensive industryÐ pharmaceuticalsÐ as well as oý ce machines and computers,and instruments. Japan is most strongly export specialized in transport industries(shipbuilding and other transport) followed by communication equipment, oý cemachines and instruments.

Despite the similarities in economic indicators, the four countries have somedistinct diþ erences in structures of interdependence, which are illustrated in Figures1 through 4.

The most distinct diþ erence between the four systems is that while two separateclusters emerge in the cases of the UK, Japan and the US, the German systemseems more integrated, with industrial chemicals being a general source of R&Dfor a range of industries. In the UK case, one cluster draws on industrial chemicalsand pharmaceuticals as R&D sources, while another cluster draws on electricalmachinery and communication equipment as R&D sources. The same two typesof clusters are present in Japan, but here the communication equipment andelectrical machinery cluster is more widespread than the other cluster, whichconsists of industrial chemicals as its only source industry, and food, textiles andrubber and plastics as the receiver industries (i.e. pharmaceuticals are not includedin the cluster). In the case of the United States, the two main clusters aresupplemented by two minor, one-to-one clusters consisting of pharmaceuticals andfood and non-electrical machinery and motor vehicles respectively.

These diþ erences might partly be explained by the fact that the four countrieshave gone through quite diþ erent development processes. The UK was the domi-nant industrial nation in the 19th century, but has since lost its World leadership.The United States took over that leadership in the 20th century and, in particular,in the post World War II-period, the American economy ¯ ourished. Both theGerman and Japanese economic systems were severely struck by their defeat in theSecond World War, but they both managed to rebuild their nations very fast tobecome major economic powers. Their foundations for rebuilding were quitediþ erent, although they were both subject to American restrictions and support.The role of the American protection and support played, in particular, an importantrole in restructuring the two economies after the Second World War.

The paper thus presents four systems that diþ er considerably in the formal andinformal institutions guiding economic behaviour, in the economic structuresdetermining technological change, and in the historical development leading to thecurrent structures, i.e. they represent four very diþ erent national systems of

Eco

nom

ic S

yste

ms

Res

earc

h 20

00.1

2:37

7-39

9.

Industrial Interdependence in National Systems of Innovation 383

innovation. The aim of this section is to explore whether these diþ erences can berelated to the structure of interindustrial technological linkages.

4.1. Germany

Porter (1990, p. 356) sums up the remarkable success of the postwar Germanindustrial strategy by stating that no country in the world, including Japan, exhibitsthe breadth and depth of industries with strong international positions as Germany.

Chemicals played a dominating role in the process of industrialization inGermany. The foundation for this can be found in the 19th century. Many of theGerman competitive positions were created by the turn of the century,1 whenGermany was characterized by a close connection between universities, TechnischeHochschulen and industrial ® rms. With the universities and the Technische Hoch-schulen, Germany had established a sophisticated system for education in scienti® c,technical and commercial matters, reaching from elementary school to doctorallevel (Keck, 1993, p. 122). This system has had a signi® cant in¯ uence on thestructure of the German system as we know it today.

The ® rst major science-based industry in Germany was the beet-sugar industry,which became a major exporter in the late 19th century. In addition to chemicalresearch, the industry had a base in agricultural research. Important lines ofbusiness in the chemical industry also supplied inputs to the textile industry (inparticular bleaching and dyeing). Germany’s largest chemical companies, BASF,Hoechts and Bayer, were all founded in the 1860s, and were, at the time, the mainproducers of synthetic dyes (Keck, 1993, pp. 125± 126). Porter actually perceivesthe pressure from factor disadvantages as a major factor behind the success of theGerman chemical industry, as the lack of available raw materials stimulatedbreakthroughs in synthetic materials (Porter, 1990, p. 371).

A strong feature in the German chemical industry is the tendency to integratebackwards into the production of basic chemicals and intermediates. In general,German ® rms have prospered from taking advantage of the economics of scope(but not disregarding economies of scale though), which also meant that theprimary focus was on process innovation rather than product (Murmann & Landau,1998, p. 31).

The dominance of the chemical industry in Germany was so strong that,during the interwar years, German industrial power became synonymous with IGFarbenindustrie Actiegesellschaft, which was a result of the merger of the majorGerman chemical companies in 1925 (IG included contemporary giants such asHoechts, BASF and Bayer) (Murmann & Landau, 1998, pp. 49± 50).

The German chemical industry did not suþ er as much as could have beenexpected after the end of the Second World War, even though the Allies con® scatedknow-how, trademarks and patents from German industry as a part of a policy ofmaking public all information from the enemy. Germany’s luck’ was that the endof the Second World War coincided with a shift from coal-based technology inchemicals to petrochemicals, thus most of the information obtained by the Alliessoon became obsolete as the technological frontier in chemicals moved in adirection new to all parties. Moreover, restrictions on chemical production andresearch were soon removed, as it became obvious that a reversal of the Germaneconomic decline required allowing German chemical ® rms again to produce theirproductsÐ chemicals were perceived as the lifeblood’ of a modern economy(Murmann & Landau, 1998, pp. 60 ± 61). Germany has since experienced a decline

Eco

nom

ic S

yste

ms

Res

earc

h 20

00.1

2:37

7-39

9.

384 I. Drejer

Figure 1. R&D linkages in Germany, 1990.

Figure 2. R&D linkages in the UK, 1990.

Eco

nom

ic S

yste

ms

Res

earc

h 20

00.1

2:37

7-39

9.

Industrial Interdependence in National Systems of Innovation 385

Figure 3. R&D linkages in Japan, 1990.

Figure 4. R&D linkages in the US, 1990.

Eco

nom

ic S

yste

ms

Res

earc

h 20

00.1

2:37

7-39

9.

386 I. Drejer

in competitive advantage, and the dominance of the chemical industry in Germanyis nowadays not very pronounced when looking at the economic indicators ofAppendix A.

Turning to the non-chemical part of German industry, Porter (1990) pointsout metals, metalworking and associated machinery, as well as construction ofmetallurgical plants, as other major ® elds in Germany. The mining and metalprocessing industries have their origins in the mining schools that trained genera-tions of administrators and managers in the 18th and 19th centuries, leading to aneþ ective transfer of technology from abroad as well as to graduates pioneering innew processes (Keck, 1993, p. 127).

Turning to the structures of Figure 1, it becomes apparent that industrialchemicals still have a central position as a technology driver’ in Germany. Industrialchemicals, together with rubber and plastics, are the sources of embodied R&Dfor a broad range of industries including paper, food, textiles, electrical and non-electrical machinery as well as motor vehicles. In other words, the main bulk ofembodied R&D ¯ ows run from industries in a main group related to chemistrytowards industries related to electronics and metal processing. Thus, industrialchemicals is a dominant industry that appears to be a general technology sourcefor the entire system, including industries in the transport and machinery cluster’ .Furthermore, it is worth noticing the bilateral relations in the `transport andmachinery cluster’ , which consist of non-electrical machinery, electrical machineryand motor vehicles.2

The metal processing industries are closely related to transportation equipment.The motor vehicles industry, electrical machinery and non-electrical machineryindustries are all related through bilateral relations, i.e. these industries are bothsources to, and receivers of, technology from each other. The highly integratedmachinery± motor vehicles cluster is a notable characteristic of the German systemof innovation, illustrating the other important position of relative strength inGermany besides the industrial chemical industry. Judged from Figure 1, themachinery and transportation (motor vehicles) industries are, in fact, also closelyrelated to the industrial chemicals industry. As will be illustrated below, Germanyis the only one of the analysed national systems where the chemical industry isfound to be so deeply integrated in the web of R&D interdependence in thenational system.

Germany also shows a lack of industrial strength in some areas. The serviceindustries, as well as electronic products, computers and semiconductors etc., areweak spots in the German system (Porter, 1990, p. 367), despite the fact thatGerman ® rms devote a major part of their R&D expenses to some of these ® elds(see Appendix A).

Summing up, skills and technology have played an important role for thepresent German system of innovation, which is particularly characterized by thestrong position in the R&D-intensive chemicals industry. Thus, the in¯ uence ofthe institutional set-up most notably expressed by the educational system cannotbe neglected. However, Germany seems to be better at improving performanceand staying ahead in its traditional areas of specialization than at moving to new® elds of growth (such as computers). This is underlined in Figure 1 by the factthat Germany is the only one of the four countries analysed that has not one singleelectronics, communications or computer related industry as a source of embodiedR&D ¯ ows. A proposed explanation for this is that while Germany’s strength hasbeen built on upgrading advantages by raising the quality of human and technical

Eco

nom

ic S

yste

ms

Res

earc

h 20

00.1

2:37

7-39

9.

Industrial Interdependence in National Systems of Innovation 387

resources, these advantages in human skills, through an outstanding quality ofeducation, seem to have been eroded in the past decades. Thus, the Germansystem now appears to be structured mainly around `old’ positions of strength,which might prove to be a vulnerable position in the long run.

4.2. The United Kingdom

As opposed to the German system, Britain has experienced a long period of relativedecline in economic power from a 19th century position of dominance. While theGerman industries are praised for their depth, British clusters are described asshallow’ , i.e. the vertical strength that is seen in, for example, the German chemicaland related industries is missing in Britain’s major industries (Porter, 1990,pp. 484± 494).

The United Kingdom has performed quite successfully in the chemicals andpharmaceuticals industries, which can be explained partly by a heavy investmentin R&D (see Appendix A), and partly by the establishment of strong linkages touniversities (Porter, 1990, p. 498; Walker, 1993, p. 180).

The UK started out by leading in the electrical and mechanical areas, while theposition of strength in chemicals and pharmaceuticals came later. For most of the18th and 19th centuries the British innovation system was generating revolutionarychanges in techniques of energy and material transformation (the coal, iron andsteam nexus), in the organization of production (the factory system), and intransportation (railways and the steam ship). These industries have all declined inimportance and have now lost their world dominance (Walker, 1993, pp. 187 ±188). The relatively poor performance of engineering related industries has,amongst others, been blamed on defence procurement, which has absorbed a largeproportion of high technology engineering resources, a position which is alsosupported by Porter (1990, p. 498). The eþ ect of defence procurement ontechnological development does not always have to be negative though, as isillustrated in the cases of Japan and the United States below. The proposed reasonsfor the negative eþ ects of the involvement in defence markets in the UK are a smallspinoþ into the civil sector, and that the defence involvement has in¯ uenced thestyle’ of technological activity (product rather than process innovation) (Walker,1993, p. 177). Another straightforward explanation for the decline in engineering-related industries is the lack of an educational strategy in engineering, as the UKhas a comparatively poor quality of (especially) vocational training and secondarylevel education (Mason et al., 1992).

Figure 2Ð the main R&D linkages in the British system of innovationÐ showsa system that is split into two distinct clusters: (1) an electronics related cluster,with electrical machinery and communication equipment being sources for the(from a R&D perspective) major industries, oý ce machinery and aerospace, aswell as non-electrical machinery and motor vehicles; and (2) a chemicals relatedcluster with industrial chemicals and pharmaceuticals as sources of knowledge fora range of low-tech industries.

The development of a `cluster’ around industrial chemicals and pharmaceuticalscannot be dated as far back as the `engineering cluster’ . Food, which is among thereceiver industries in this cluster, has played a considerable role in the world marketas Britain is home to some of the world’s largest food, drink and tobacco companiesthat have origins in the 18th and 19th century (Walker, 1993, p. 161). Thechemical and pharmaceutical industries appear to be somewhat newer. However,

Eco

nom

ic S

yste

ms

Res

earc

h 20

00.1

2:37

7-39

9.

388 I. Drejer

as a consequence of the demand created by the industrial revolution (by industriessuch as textiles, glass, steel etc.) Britain was, in fact, the home of the largestchemical industry in the world in the middle of the 19th century. The Germanchemical industry outperformed the British around the turn of the century throughGerman investments in both manufacturing, marketing, and management, allowingit to reap cost advantages from both economies of scale and economies of scope.In addition, the large German investments in R&D, which resulted in a continuousstream of new or superior products and processes, were causing Britain to fallbehind (Murmann & Landau, 1988, pp. 28± 30). The above-mentioned Germaneducation system was one of the factors allowing these investments.3 However,during the First World War, British ® rms were very close to catching up withGerman ® rms, since chemicals played an important role in the war (poisonousgases), and the British government induced the creation of an infrastructure forinteraction among science, industry and government, which had been missingpreviously, but which turned out to be very bene® cial for industry. In addition, thetransfer of know-how from Germany to Britain during and after the First WorldWar played a crucial role for the technological catch-up in chemicals (Murmann& Landau, 1998, pp. 46 ± 47). The British chemical industry never reached thesame overwhelming importance for the national industrial society as was the casein Germany, but in recent years an eþ ort in the area of higher education has madethe British level of science world class (Murmann & Landau, 1998, pp. 63 ± 64).

A reason for the lack of linkages between the two main clusters in Britishindustry could be the above-mentioned lack of synchrony in the development ofthe clusters. Britain now possesses some of the world’s leading chemicals andpharmaceuticals ® rms, probably due to these industries being closely linked toscience’ , a ® eld in which the UK has a considerably stronger position than inelectronics and related industries (Walker, 1993, p. 180). Thus, the structureillustrated in Figure 2 is in accordance with the uneven development of the twomain areas of British industry.

In summary, it is a characteristic of Britain that its major role in the worldeconomy is maintained in manufacturing industries that are typically science-based’ (cf. Pavitt, 1984). The success has been less in engineering-based industrieswhere R&D needs to be much more `development’ than research’ (von Tunzel-mann, 1995, p. 186). Against the background of the institutional set-up, a reasonfor this can partly be found in the relatively low priority in the British educationsystem given to engineering related education. The chemicals related industrieshave bene® ted from linkages to the science world, and the educational strategy hasbeen inspired by the successful German system. Thus, the two main clusters inthe British innovation system are the result of two diþ erent strategies: in thechemical cluster, a deliberate strategy inspired by the German success story; andin the engineering cluster, a lack of suý cient initiatives strengthening the basiccompetencies needed to keep pace with the world leaders.

4.3. Japan

As mentioned in the introduction to this section, Japan shares its impressive successafter the destruction of the Second World War with Germany. However, it is claimedthat, unlike Germany, Japan did not have a strong historical position in areas suchas chemicals and machinery on which to rebuild (Porter, 1990, p. 384). The con-siderable competitive strength gained by Japan in consumer electronics, oý ce

Eco

nom

ic S

yste

ms

Res

earc

h 20

00.1

2:37

7-39

9.

Industrial Interdependence in National Systems of Innovation 389

machines, electronic components and computing equipment, transport equipmentand related machinery, as well as steel and fabricated metal products, is based onthe existence of a strong heavy industry dating back to the 19th century though.

The ® ve industries in which Japan is most strongly export specialized are: othertransport, shipbuilding, communication equipment and semiconductors, oý cemachines and computers, and instruments (see Appendix A). The dominance ofelectronics and transport related industries in Japan can partly be explained by therole of the Japanese military in the previous century. In terms of industrialcomposition, food processing and textiles were the largest industries in the late19th century, but then metal, machinery, chemicals and other heavy industriesbegan to grow fast. Even in the Meiji era (1868 ± 1912), the military and thegovernment played, in general, an important role for those industries that make upthe backbone of the engineering industries.

Two years after emperor Meiji came to power, in 1870, a Ministry of Industry(Koà bushoà , sometimes also translated as Ministry of Engineering or Ministry ofConstruction) was established. The Ministry was abolished in 1885, but in its 15years of existence it played a crucial role in the process of industrialization fromabove’ , where industrialization was forced through an ambitious programme ofimporting relatively advanced western technology. The emphasis was put on hiredforeigners passing on their knowledge to their Japanese counterparts and thengoing home as soon as possible. This strategy was used in relation to railwayconstruction, the creation of a nationwide telegraph network, mining, as well asiron works. The experts hired were mainly British. Historians generally agreethat the Meiji policies, with their bias towards importing relatively sophisticatedtechnologies, were commercial failures when viewed from the point of view of stateenterprises, but the policies were crucial to Japan’s technological development.Technology was transferred from government to private ® rms from 1881 onwardsthrough selling government enterprises to a selected group of private buyers atvery low prices. The entrepreneurs who bought the mines and factories not onlygained cheap machinery and equipment, but also a ready trained source of technicalexpertise, as well as established technical links with western ® rms. Mitsubishi isone example of a major company growing out of this process (Morris-Suzuki,1994, pp. 73 ± 79).

The focus on education during the Meiji era should also be mentioned. In1871, compulsory primary education, with a considerable emphasis on scienti® cenquiry, was introduced. Regarding higher level education, the Imperial College ofEngineering was established in 1873, with a strong electrical engineering facultywhich, as far as is known, appointed the ® rst Professor of Electrical Engineeringin the World. These colleges also, to a large degree, depended on foreign importedknowledge (Morris-Suzuki, 1994, pp. 80± 82).

The military arsenals and navy dockyards also played a crucial role in thedevelopment of Japan as they used relatively sophisticated imported techniquesand were sending their leading technicians abroad for training in major westernarmament ® rms. The military expansion around the turn of the century hadimportant spin-oþ s for civilian industry. For example, government arsenals pro-duced a wide range of industrial machinery, which was sold to private enterprises,and workers trained in arsenals often moved to civilian industries, taking theirknowledge of imported production techniques with them. In addition, militarydemand provided a market for many of Japan’s more technically advanced industriesin their early stage of development. Toshiba is one example of a major company,

Eco

nom

ic S

yste

ms

Res

earc

h 20

00.1

2:37

7-39

9.

390 I. Drejer

with its roots in industrial machinery, which bene® ted from military demand(Morris-Suzuki, 1994, p. 79).

The Japanese economy took oþ around the First World War with increasedproduction, especially in steel, machinery and other heavy industry (Odagiri &Goto, 1993). This was the time of the second industrial revolution’ in the West,with a widespread diþ usion of electrical power, the introduction of the automobileand aeroplane, and the techniques of mass production. To the Japanese governmentthe technologies of the second industrial revolution’ were not so much a source ofgrowth, rather they were perceived as a challenge to Japanese security. The fear fora total war’ of the 20th century implied a blurring of the distinction betweenmilitary and non-military industries. Thus, the military played a central role in thedevelopment of both the automobile industry as well as the aircraft industryÐ thetechnologies of war and peace were interrelated (Morris-Suzuki, 1994, p. 107 andpp. 124± 145).

The Second World War caused Japanese production facilities to suþ er severely,but still more than two thirds of the production capacity in the heavy industrieswas left intact after the war (Odagiri & Goto, 1993, pp. 83± 85). Thus, there wasa considerable foundation for rebuilding the economy. In addition, somewhatparadoxically, the unsuccessful attempt to win what was perceived as the war ofscience and technology’ left Japanese institutions, human skills and public attitudeswell prepared for the new massive import of western technology in the years thatfollowed after the surrender to the Allies in 1945 (Morris-Suzuki, 1994, p. 157).In addition, the combination of military procurement and protectionism, throughrestrictions on imports and foreign investment until the early 1970s, providedexcellent conditions for the development of Japanese heavy industry, in particularfor the motor vehicles industry (Odagiri & Goto, 1993).

Thus, the government has played a signi® cant role in the development of theelectronics and transport related industries. However, the success with chemicalsrelated industries was much more moderate. Internationally, Japan does not seemto have any particular advantages in chemicals-related industries, with maybetextiles as one exception as Japanese ® rms are strong in synthetic textile ® bres(Porter, 1990, p. 403). Despite this, Japan does have, in fact, several large chemical® rms, but Hikino et al. (1998, p. 103) point to the puzzle that the Japanesechemical industry, despite an impressive growth and a large size, basically remainsinvisible on the international economic scene. One explanation might be that thestrengths of Japanese chemical companies have been their ability to learn quicklyand their incremental process innovation capabilities, while they have not beenvery successful in developing great technological competencies in radical productor process innovation. Japanese chemical companies have been biased towardcapital and resource intensive areas, and away from the knowledge intensive partsof the chemical industry, including pharmaceuticals (Hikino et al., 1998, pp. 107 ±108). Education, or lack of education, is also proposed as part of the problem,since little eþ ort has been put into high level education in chemical engineering.What the industry needed in order to absorb and operate foreign technology wasprimarily a large supply of plant-level engineers, not trained researchers (Hikinoet al., 1998, pp. 118± 119). Finally, the process of acquisition of foreign knowledgein chemicals diþ ered from the one characterizing the electrical machinery industry.A large degree of the technology ¯ ows from the west in electrical machinery duringthe turn of the century was facilitated through partnerships between western andJapanese ® rms, as a way for western ® rms to gain access to the Japanese market,

Eco

nom

ic S

yste

ms

Res

earc

h 20

00.1

2:37

7-39

9.

Industrial Interdependence in National Systems of Innovation 391

which resulted in a combination of patent licences, technical assistance andinvestment. The western chemical companies relied on their own ability to domi-nate overseas markets, and thus were not interested in engaging in partnershipswith Japanese ® rms. As a result of this, the imports of know-how in chemicalstended to be limited to single patent± licencing arrangements (Morris-Suzuki, 1994,pp. 113± 114).

The dominance of the engineering-related industries over the chemicals-relatedindustries is supported in Figure 3. The electronics related cluster includes moreindustries and is more dominating compared with the chemical cluster.4 Thus, thedominance of electronics and transport related industries is illustrated by thedensely connected engineering cluster. Figure 3 supports Porter’s ® ndings ofsemiconductors (included here in communication equipment) and electronics asuniting a number of clusters/industries (Porter, 1990, p. 394). Industrial chemicalson the other hand are related to traditional low-tech manufacturing industries asreceivers: food, textiles and rubber and plastics. Among these, food is the mainindustry from the perspective of production and employment. Industrial chemicalsis an important industry in relation to R&D spending, but the industry is relatedto low-tech traditional manufacturing industries that might be important from aproduction and employment point of view, but none of these industries areimportant in an international context. In addition, Porter describes chemicals asan area of continuing weakness in Japan (Porter, 1990, p. 420). Figure 3 clearlyillustrates the segmentation of Japanese industry and, as opposed to Figure 2 forBritain where it was not possible from the ® gure alone to determine which areawas the most prosperous, there is no doubt that the engineering area is dominatingin the case of Japan.

4.4. The United States

Before the Civil War (1861 ± 1865) the American economy was characterized by aregional specialization in functions, but the war resulted in a political and ideologicalframework that permitted structural change and regional integration (von Tunzel-mann, 1995, pp. 189± 190).

Just like Japan, the United States’ process of industrialization started oþ byborrowing and copying technologies from abroad. However, the technologies wereadapted to diþ erent supply and demand conditions. On the supply side, a majordiþ erence from BritainÐ which was the major technology sourceÐ was the abun-dant supply of land, which in¯ uenced both the mechanization of agriculture(allowing a limited labour force to work more land) and the transport industry(providing mass transportation over large distances). On the demand side, thedemand structure, dominated by a large number of rural households with a strongpreference for moderately priced consumer goods, allowed for standardization andmass production, which was the key to American industrialization (von Tunzel-mann, 1995, pp. 192± 196).

Formal R&D was ® rst developed for comparatively simple purposes (primarilyin the major advancing industries like metallurgy, food processing and construc-tion), and the emphasis in the 19th century was, especially in chemistry, on `old’sciences. A greater use of experimental science was seen around the turn of thecentury, and the nature of the science base shifted from a concentration ofchemistry-based research toward more physics based research in areas such aselectricity, transportation and instruments (von Tunzelmann, 1995, pp. 201 ± 202).

Eco

nom

ic S

yste

ms

Res

earc

h 20

00.1

2:37

7-39

9.

392 I. Drejer

Rosenberg (1976) ascribes a large strategic role in the industrialization processto the machine tool industry that emerged in the last half of the 19th century. Inthe earliest stages, machine producing establishments were part of, or related to,factories specializing in the production of a ® nal product, most notably textiles,but eventually they developed into independent ® rms. The skills acquired in theproduction of one type of machine were transmitted to other types of machines,e.g. locomotive works grew out of the cotton textile industry. In general, heavy,general-purpose machinery grew out of the textile machine shops, while lighter,more specialized high-speed machine tools grew out of the production requirementsof arms makers. The technological developments of the 19th century were largelydependent on the convergence of functional processes throughout the machineryand metal-using sectors, which contributed to the simultaneous growth of several,technologically related industries. Thus, the machine tool industry can be perceivedas the holder of a pool of skills and technical knowledge that could be used in theentire machine-using sectors of the economy. For example, the automobile industrywas built on basic skills and knowledge already existing in the machine tool industry(Rosenberg, 1976, pp. 10 ± 26).

The role of public funding of R&DÐ in particular related to defence, atomicenergy and aeronauticsÐ cannot be ignored. The Second World War had a crucialin¯ uence on the industrial success of the United States. The war spurred electron-ics, chemicals and pharmaceuticals to new heights (Porter, 1990, pp. 294± 296).Huge expansions of Federal research funding during and after the Second WorldWar moved universities and colleges into the lead in high-tech industries, and thewar also resulted in a large in¯ ow of top scienti® c talent. The military has beenstrong in areas such as aircraft, computers and semiconductor technology, andmuch of the early exploratory research on computers was done under governmentcontracts (Nelson, 1982, p. 452). Signi® cant federal money has also gone intopharmaceuticals, but into the ® eld of cancer and `orphan drugs’ which haverelatively small commercial markets (Nelson, 1982).

The chemicals industries had already been ¯ ourishing before the war. The early20th century was dominated by the chemicals industry and related industries, andthe chemicals, glass, rubber and petroleum industries accounted for almost 40%of the number of laboratories founded during the period 1899± 1946 (Mowery &Rosenberg, 1993, pp. 32± 33). The strong position in chemicals was based ® rstlyon natural resources, but also on technological capabilities (Arora & Rosenberg,1998, p. 76). While the United States’ raw materials endowment was crucial tothe growth of the chemical industry in the early phase, the size of the home marketas well as research-based technological advances in both products and processestook over in importance. The national oil and natural gas stocks thus played a rolein giving the United States a ® rst mover advantage in petrochemical technologies,where the United States now takes a leading position (Arora & Rosenberg, 1998,p. 98). There is also a close linkage between the chemicals and petroleum industries(illustrated in Figure 4). The American story of chemicals points to the importanceof complementarities, as technology, market size, resource endowments, and supplyof entrepreneurial capital were all important factors behind a successful Americanchemical industry (Arora & Rosenberg, 1998, p. 99).

The postwar expansion resulted in an American world dominance in innovation,but the primary focus was shifted from the industries that had represented theadvanced technologies in the ® rst half of the 20th century, i.e. chemicals, metalworking, synthetics and plastics, toward a dominance in new sectors primarily

Eco

nom

ic S

yste

ms

Res

earc

h 20

00.1

2:37

7-39

9.

Industrial Interdependence in National Systems of Innovation 393

related to electronics (von Tunzelmann, 1995, p. 228). The development of theelectronics-related industries was, as in Japan, largely related to the military, alsoafter the war. A huge defence programme provided a market for advanced goodssuch as aircraft and electronics (Porter, 1990, p. 284). Moreover, in the 1950s and1960s, the US military market provided an important springboard for start-up® rms in microelectronics and computers, with new ® rms playing a large role incommercializing product technologies within the ® elds of semiconductors andcomputers as well as biotechnology. Pro® ts and overheads from military procure-ment contracts supported company funded R&D and this might have generatedmore civilian spillovers than R&D that were directly funded by the military.Additionally, in the 1950s, defence procurement lowered the marketing barriers toentry, which allowed small ® rms to direct their development eþ orts to meeting theperformance and design requirements of a single large customer (Mowery &Rosenberg, 1993, pp. 48 ± 54). However, as the military’s needs have become morespecialized, defence demand is no longer an undisputed strength, and it is claimedthat the huge defence market in recent years has distracted American ® rms frommore important areas (Porter, 1990, p. 526). However, the American positions in,for example, aircraft and computers are still related through the role of governmentspending (Porter, 1990, p. 511).

The separate development of chemicals and engineering related industries is inaccordance with the two separate main clusters that are illustrated in Figure 4.5

The Engineering cluster consists of four of the ® ve most R&D spending industries inthe United States (see Appendix A): oý ce machinery, communication equipment,instruments, and aerospace. The ® fth industry in the engineering cluster is electricalmachinery. The four R&D heavy industries are also among the industries in whichthe United States is most strongly export specialized. These engineering-relatedindustries account for the strongest manufacturing positions of the Americansystem of innovation.

Two other small `one-to-one’ clusters appear in the American system: onebetween pharmaceuticals and food, and one between non-electrical machinery andmotor vehicles.6 Regarding the ® rst `one-to-one’ cluster, agricultural products havebeen important from an export perspective for a long time, and their R&Ddependence on the pharmaceuticals industry may indicate a relatively techno-logically sophisticated food industry. Motor vehicles and non-electrical machineryis another mini-cluster, which in this case is isolated from the rest of the engineeringcluster. The history of industrialization does not give any strong indications as towhy this is the case.

Brie¯ y concluding on the characteristics of the American system of innovation,similarities are found with Japan in the role of technology import in the 19thcentury, but with very diþ erent needs for adaptation to technology. In the Americancase, a very large home market enabled adjusting technologies to mass production,which suited the emergent culture of mass consumption. The system is largeenough to function quite eþ ectively with separate clusters, which are not techno-logically related in any crucial way, but both have dominant positions in the worldmarket, also from a technological point of view. However, the major strength hasshifted from the chemicals to the electronics-related `paradigm’ .

4.5. Summing up

Despite some common characteristics in the economic features of the four countriesanalyzed here, both with regard to economic indicators and with regard to the

Eco

nom

ic S

yste

ms

Res

earc

h 20

00.1

2:37

7-39

9.

394 I. Drejer

main technology sources being related to chemicals and/or electronic machineryetc., this paper has presented four distinct systems of R&D interdependence.Diþ erent factors, institutional as well as `exogenous’ (e.g. engagement in wars)have shaped the systems.

The technology ¯ ow maps reveal two main patterns: in the case of the UK,Japan and the US, at least two separate clusters appear. One cluster is centredaround chemicals-related industries, while another is centred around communi-cation equipment, electronic and/or transport industries. At the pre-set ® lter values,no embodied R&D ¯ ows connect the two clusters. Germany, on the other hand,reveals a more interrelated system with less clear tendencies of separate clustering,due to the crucial role played by the German chemical industry, as well as therelatively weak position in electronics and related industries. However, even if theUK, Japan and the US share a main structure, considerable diþ erences betweenthe three countries are also revealed in the patterns of relations.

It has been argued (DuÈ ring & Schnabl, 2000) that the production patterns andR&D eþ orts become more and more similar in a global economy, and that nationalR&D systems therefore should become much more similar over time. On the otherhand, it can be argued that diþ erences in both formal and informal institutions, aswell as the role played by history, are of such fundamental importance thatdiþ erences between systems will persist, and these diþ erences will continue tostand out in, for example, graphical representations focusing on particular `corners’of the systems, as in this case illustrated by embodied R&D linkages.

5. Conclusions

This paper has dealt with the extent to which the diþ erences in structure ofinterdependence can be understood by relating it to the underlying characteristicsof each individual innovation system. The `history matters’ assumption has beenthe major guiding point.

The claim was that a relatively simple input± output based graph theoreticalmodel would be able to capture some underlying diþ erences in the basic featuresof each individual system. There has been a focus on the industrial developmentof each country, and the institutional factors in¯ uencing this development, relatingthis to the structure of interdependence of the system. It is argued that it waspossible to illustrate that the structure of technological interdependence at theindustry level can be related to the historical process of industrialization.

A general conclusion that can be deduced from the analysis is that, if two majorindustrial ® elds develop at a diþ erent pace, with diþ erent strategies, and withdiþ erent starting points, these areas tend to remain largely separated in a techno-logical sense. It is claimed that this is why separate chemicals and electronicsclusters are found in three of the four countries. In the German case, the chemicalindustry was the forerunner of the industrialization process, which could helpexplain why industrial chemicals is still today a generic source of embodiedtechnology, while the electronics cluster is relatively weak.

Thus, the distinctiveness of technology bases might not so much be determinedby the inherent characteristics of technologies, but rather by whether diþ erenttechnological areas have developed in isolation’ from each other, or whether theyhave developed in an integrated process.

Eco

nom

ic S

yste

ms

Res

earc

h 20

00.1

2:37

7-39

9.

Industrial Interdependence in National Systems of Innovation 395

Notes

1. Historical continuity is very strong in GermanyÐ despite the destruction during the World WarsÐalso at the ® rm level. This is illustrated by the fact that 19 of the 25 largest ® rms in 1989 werefounded before 1913. These ® rms were mainly to be found in the ® elds of chemicals, vehicles,electricity, energy, steel and machinery (Keck, 1993, p. 136).

2. Figure 1 con® rms all the relations found for Germany in 1990 by DuÈ ring & Schnabl (2000), butmore relations are included in Figure 1 than in the DuÈ ring± Schnabl graphs. We ascribe this to alarger ® lter value applied by DuÈ ring & Schnabl.

3. The German education system actually produced too many highly quali® ed scientists, which allowedfor a British import of talented chemical engineers from Germany.

4. DuÈ ring & Schnabl (2000) ® nd chemicals to have a much more central role as a technology sourcein Japan than is the case in the present analysisÐ in addition to industries in the electrical±transportation cluster. One explanation for this diþ erence could be that DuÈ ring & Schnabl includeimported goods in their analysis, based on the assumption that the R&D structure of the importedgoods corresponds to the structure of the importing country.

5. The defence connection is, of course, not the only factor that has in¯ uenced the prosperousengineering cluster in the US. The ® nancial system with a well-developed market for venture capital,and the `entrepreneurial spirit’ cannot be ignored. However, in terms of proposing explanations forthe separate development of the engineering and chemicals clusters, the defence connection mighthave played an important role.

6. Just like for Japan, DuÈ ring & Schnabl (2000) ascribe a more central role to chemicals than is foundhere, but the major relations from chemicals to petroleum and plastics and rubber respectively arecon® rmed. Electrical apparatus is a central technology source in DuÈ ring & Schnabl’s analysis, whilethis is not the case here. However, we ® nd communication equipment, which is included in electricalapparatus in DuÈ ring & Schnabl’s analysis, to be a central source in the electronics cluster. Finally,DuÈ ring & Schnabl ® nd motor vehicles to be connected to electrical machinery/apparatus, which isnot the case here.

References

Archibugi, D. (1988) In search of a useful measure of technological innovation (to make economistshappy without discintenting technologists), Technology Forecasting and Social Change, 6, pp. 253± 277.

Arora, A. & Rosenberg, N. (1998) Chemicals: a U.S. success story, in: A. Arora, R. Landau & N.Rosenberg (eds) Chemicals and Long-Term Economic Growth (New York, Wiley).

DeBresson, C., Hu, X., Drejer, I. & Lundvall, B.-AÊ . (1998) Innovative Activity in the Learning EconomyÐ

a comparison of systems in 10 OECD countries, Draft report.Drejer, I. (1999) Technological change and interindustrial linkagesÐ introducing knowledge ¯ ows in

input± output analysis. PhD thesis (Department of Business Studies, Aalborg University).Drejer, I., Kristensen, E. S. & Laursen, K. (1999) Studies of clusters as a basis for industrial and

technology policy in the Danish economy, in: Boosting InnovationÐ The Cluster Approach (Paris,OECD).

DuÈ ring, A. & Schnabl, H. (2000) Imputed interindustry technology ¯ owsÐ a comparative SMFAanalysis, Economic Systems Research, 12, pp. 363 ± 375.

Edquist, C. & Lundvall, B.-AÊ . (1993) Comparing Danish and Swedish systems of innovation, in: R.Nelson (ed.) National Innovation SystemsÐ A Comparative Analysis (New York, Oxford UniversityPress).

Harary, F., Norman, R. Z. & Cartwright, D. (1965) Structural Models. An Introduction to the Theory of

Directed Graphs (New York, Wiley).Hikino, T., Harada, T., Tokuhisa, Y. & Yoshida, J. (1998) The Japanese puzzle, in: A. Arora, R. Landau

& N. Rosenberg (eds) Chemicals and Long-Term Economic Growth (New York, Wiley).Keck, O. (1993) The national system for technical innovation in Germany, in: R R. Nelson (ed.)

National Innovation SystemsÐ A Comparative Analysis (New York, Oxford University Press).Lundvall, B.-AÊ . (1998) Why study national systems and national styles of innovation?, Technology

Analysis & Strategic Management, 10, pp. 407 ± 421.Mason, G., Prais, S. J. & van Ark, B. (1992) Vocational education and productivity in the Netherlands

and Britain, National Institute Economic Review, pp. 45± 63.Morris-Suzuki, T. (1994) The Technologica l Transformation of Japan (Hong Kong, Cambridge Univer-

sity Press).

Eco

nom

ic S

yste

ms

Res

earc

h 20

00.1

2:37

7-39

9.

396 I. Drejer

Mowery, D. C. & Rosenberg, N. (1993) The U.S. national innovation system, in: R. R. Nelson (ed.)National Innovation SystemsÐ A Comparative Analysis (New York, Oxford University Press).

Murmann, J. P. & Landau, R. (1998) On the making of competitive advantage: the development of thechemical industries of Britain and Germany since 1850, in: A. Arora, R. Landau & N. Rosenberg(eds) Chemicals and Long-Term Economic Growth (New York, Wiley).

Nelson, R. R. (1982) Government and Technical ProgressÐ A Cross-Industry Analysis (New York, Perga-mon Press).

Odagiri, H. & Goto, A. (1993) The Japanese system of innovation: past, present and future, in: R. R.Nelson (ed.) National Innovation SystemsÐ A Comparative Analysis (New York, Oxford UniversityPress).

Papaconstantinou, G., Sakurai, N. & Wyckoþ , A. (1996) Embodied technology diþ usion: an empiricalanalysis for 10 OECD countries, STI Working Paper 1996/1 (Paris, OECD).

Patel, P. & Pavitt, K. (1993) National systems of innovation: why they are important and how theymight be de® ned, compared and assessed. Document de Travail No. 93/05, Sophia Antipolis,Universite de Nice, LATAPSES.

Pavitt, K. (1984) Sectoral patterns of technical change: towards a taxonomy and a theory, Research

Policy, 13, pp. 343± 373.Polanyi, M. (1966) The Tacit Dimension (London, Routledge & Kegan).Porter, M. E. (1990) The Competitive Advantage of Nations (New York, Free Press).Rosenberg, N. (1976) Technological change in the machine tool industry, 1940± 1910, in: N. Rosenberg

(ed.) Perspectives on Technology (Cambridge, Cambridge University Press), pp. 141 ± 210.Sakurai, N., Ioannidis, E. & Papaconstantinou, G. (1996) The impact of R&D and technology diþ usion

on productivity growth: evidence for 10 OECD countries in the 1970s and 1980s, STI Working

Papers, 1996/2 (Paris, OECD).Sakurai, N., Papaconstantinou, G. & Ioannidis, E. (1997) Impact of R&D and technology diþ usion on

productivity growth: empirical evidence for 10 OECD countries, Economic Systems Research, 9,pp. 81± 109.

Saviotti, P. P. (1997) Innovation systems and evolutionary theories, in: C. Edquist (ed.) Systems of

Innovation. Technologies, Institutions and Organizations (London, Pinter), pp. 180± 199.Schnabl, H. (1994) The evolution of production structures, analyzed by a multi-layer procedure,

Economic Systems Research, 6, pp. 51± 68.Schnabl, H. (1995) The subsystem-MFA: a qualitative method for analyzing national innovation

systemsÐ the case of Germany, Economic Systems Research, 7, pp. 383± 396.von Tunzelmann, G. N. (1995) Technology and Industrial ProgressÐ The Foundations of Economic Growth

(Aldershot, Elgar).Walker, W. (1993) National innovation systems: Britain, in: R. R. Nelson (ed.) National Innovation

SystemsÐ A Comparative Analysis (New York, Oxford University Press).

Eco

nom

ic S

yste

ms

Res

earc

h 20

00.1

2:37

7-39

9.

Industrial Interdependence in National Systems of Innovation 397

Ap

pen

dix

A:

Mai

nE

con

om

icIn

dic

ato

rsfo

rF

ou

rO

EC

DC

ou

ntr

ies,

199

0

Pro

du

ctio

n*

Em

plo

ymen

t*R

&D

*S

RC

A(e

xport

spec

.)*

Ger

many

1.

Moto

rve

hic

les

(12

.8%

)1.

Non

-ele

ctri

calm

ach

iner

y(1

1.7

%)

1.

Co

mm

.eq

uip

.an

dse

mic

on

du

cto

rs(1

9.1

%)

1.

No

n-e

lect

rica

lmac

hin

ery

(0.1

5)

2.

Fo

od

,d

rin

kan

dto

bac

co(1

2.2

%)

2.

Fab

rica

ted

met

alp

rod

uct

s(1

0.7

%)

2.

Mo

tor

veh

icle

s(1

7.7

%)

2.

Mo

tor

veh

icle

s(0

.13

)

3.

Non

-ele

ctri

calm

ach

iner

y(1

1.7

%)

3.

Fo

od

,d

rin

kan

dto

bac

co(9

.4%

)3

.In

du

stri

alch

emic

als

(15

.7%

)3

.F

abri

cate

dm

etal

pro

du

cts

(0.1

2)

4.

Ind

ust

rial

chem

ical

s(8

.9%

)4.

Oth

ertr

ansp

ort

(9.1

%)

4.

No

n-e

lect

rica

lmac

hin

ery

(10

.8%

)4

.E

lect

rica

lmac

hin

ery

chem

ical

s(0

.11

)

5.

Fab

rica

ted

met

alp

rod

uct

s(6

.6%

)5.

Ele

ctri

calm

ach

iner

y(7

.6%

)5

.A

ero

spac

e(8

.7%

)5

.In

du

stri

alch

emic

als

(0.8

)

Gre

atB

rita

in

1.

Fo

od

,d

rin

kan

dto

bac

co(1

7.2

%)

1.

Non

-ele

ctri

calm

ach

iner

y(1

3.1

%)

1.

Ph

arm

aceu

tica

ls(1

8.0

%)

1.

Aer

osp

ace

(0.3

4)

2.

Ind

ust

rial

chem

ical

s(8

.6%

)2.

Fo

od

,d

rin

kan

dto

bac

co(1

0.7

%)

2.

Aer

osp

ace

(17

.1%

)2

.O

ther

man

ufa

ctu

rin

gin

du

stri

es(0

.32

)

3.

Non

-ele

ctri

calm

ach

iner

y(8

.4%

)3.

Tex

tile

san

dle

ath

er(9

.9%

)3

.C

om

m.eq

uip

.an

dse

mic

on

du

cto

rs(1

3.5

%)

3.

Ph

arm

aceu

tica

ls(0

.29

)

4.

Pap

eran

dp

rin

tin

g(8

.4%

)4.

Pap

eran

dp

rin

tin

g(9

.6%

)4

.In

du

stri

alch

emic

als

(12

.8)

4.

Oý

cem

ach

ines

and

com

pu

ters

(0.2

2)

5.

Moto

rve

hic

les

(7.0

%)

5.

Fab

rica

ted

met

alp

rod

uct

s(6

.7%

)5

.M

oto

rve

hic

les

(8.2

%)

5.

Inst

rum

ents

(0.0

9)

Japan

1.

Moto

rve

hic

les

(11

.2%

)1.

Fo

od

,d

rin

kan

dto

bac

co(1

1.2

%)

1.

Co

mm

.eq

uip

.an

dse

mic

on

du

cto

rs(1

6.3

%)

1.

Oth

ertr

ansp

ort

(0.4

6)

2.

Fo

od

,d

rin

kan

dto

bac

co(1

0.1

%)

2.

Tex

tile

san

dle

ath

er(1

0.9

%)

2.

Mo

tor

veh

icle

s(1

4.4

%)

2.

Sh

ipb

uil

din

g(0

.43

)

3.

Fer

rou

sm

etal

s(8

.5%

)3.

Non

-ele

ctri

calm

ach

iner

y(9

.9%

)3

.E

lect

rica

lmac

hin

ery

(11

.2%

)3

.C

om

m.eq

uip

.an

dse

mic

on

du

cto

rs(0

.43

)

4.

Non

-ele

ctri

calm

ach

iner

y(8

.5%

)4.

Com

m.e

qu

ip.an

dse

mic

on

du

ctors

(9.1

%)

4.

Ind

ust

rial

chem

ical

s(1

0.1

%)

4.

Oý

cem

ach

ines

and

com

pu

ters

(0.2

8)

5.

Com

m.e

qu

ip.an

dse

mi-

con

du

ctors

5.

Fab

rica

ted

met

alp

rod

uct

s(7

.9%

)5

.O

ýce

mac

hin

esan

dco

mp

ute

rs(1

0.1

%)

5.

Inst

rum

ents

(0.2

6)

(8.4

%)

Unit

edS

tate

s

1.

Fo

od

,d

rin

kan

dto

bac

co(1

4.3

%)

1.

Pap

eran

dp

rin

tin

g(1

1.9

%)

1.

Aer

osp

ace

(23

.2%

)1

.A

ero

spac

e(0

.48

)

2.

Pap

eran

dp

rin

tin

g(1

0.2

%)

2.

Tex

tile

san

dle

ath

er(9

.7%

)2

.C

om

m.eq

uip

.an

dse

mic

on

du

cto

rs(1

3.9

%)

2.

Oý

cem

ach

ines

and

com

pu

ters

(0.2

8)

3.

Ind

ust

rial

chem

ical

s(8

.3%

)3.

Non

-ele

ctri

calm

ach

iner

y(9

.2%

)3

.O

ýce

mac

hin

esan

dco

mp

ute

rs(1

3.2

%)

3.

Inst

rum

ents

(0.1

6)

4.

Moto

rve

hic

les

(7.5

%)

4.

Fo

od

,d

rin

kan

dto

bac

co(8

.7%

)4

.M

oto

rve

hic

les

(11

.5%

)4

.C

om

m.eq

uip

.an

dse

mic

on

du

cto

rs(0

.14

)

5.

Non

-ele

ctri

calm

ach

iner

y(6

.7%

)5.

Wo

od

and

furn

itu

re(6

.9%

)5

.In

stru

men

ts(7

.9%

)5

.In

du

stri

alch

emic

als

(0.0

5)

*P

erce

nta

ges

are

mea

sure

din

rela

tion

toto

tal

pro

du

ctio

nvo

lum

e,em

plo

ymen

tan

dR

&D

exp

end

itu

rein

the

man

ufa

ctu

rin

gin

du

stri

esre

spec

tive

ly.

SR

CA

isth

esy

mm

etri

cre

veal

ed

com

par

ativ

ead

van

tage

(exp

ort

spec

iali

zati

on

),ra

ngin

gfr

om

21

(de-

spec

ialize

d)

to1

(sp

ecia

lize

d).

Ava

lue

of0

isn

eutr

al(n

eith

ersp

ecia

lize

do

rd

e-sp

ecia

lize

d).

Th

eex

port

spec

ializa

tio

nis

calc

ula

ted

rela

tive

toth

e2

0O

EC

Dco

un

trie

sfo

rw

hic

hd

ata

are

avai

lab

le:

Au

stra

lia,

Au

stri

a,B

elgiu

m,

Can

ada,

Den

mar

k,

Fin

lan

d,

Fra

nce

,G

erm

any,

Gre

ece,

Icel

and

,It

aly,

Jap

an,

the

Net

her

lan

ds,

New

Zea

lan

d,N

orw

ay,P

ort

uga

l,S

pai

n,S

wed

en,U

nit

edK

ingd

om

and

the

Un

ited

Sta

tes.

Eco

nom

ic S

yste

ms

Res

earc

h 20

00.1

2:37

7-39

9.

398 I. Drejer

Appendix B: The Model for the Graph Theoretical Minimal FlowAnalysis

The following is a modi® ed version of the model presented in Schnabl (1994,1995). The model starts with a Leontief system, where total production equals thedirect and indirect intermediate ¯ ows of goods and services (as expressed in theLeontief inverse) multiplied by ® nal demand. This expresses the total productionrequirement for producing for the actual ® nal demand:

X 5 (I 2 A) 2 1 á y ñ

á ñ expresses the diagonalization of a vector. This system is `normalized’ by pre-multiplying X by the inverse of the diagonalized vector of outputs, thus making allrows sum to 1. Thus, we now have relative requirements:

S 5 á x ñ 2 1(I 2 A) 2 1á yñ

Knowledge or technology is now introduced through the diagonalized vector á k ñ .This step weights the production requirements by the technology levels in thedelivering industries:

Xk 5 á k ñ á x ñ 2 1(I 2 A) 2 1á y ñ

Since (I 2 A) 2 1 by de® nition equals I + A + A2 + A3 . . . , the equation for Xk canbe expressed by the following set of equations:

X1,k 5 á k ñ á x ñ 2 1A á yñ , X2,k 5 á k ñ á x ñ 2 1A2 á y ñ , etc

In order to make the system binary, and thus allowing for the use of graph theoreticmethods, the values of the X1,k matrix are ® ltered’ through a preset minimal value,thus making cells with a value less than the minimal value equal to 0, while cellswith a value equal to or larger than the minimal value are given the value 1. (Theprocedure of only using X1,k and not the whole range of matrices Xn,k is based onHarary et al., 1965.) A new matrix W1,k with cells having the value 0 or 1 is created.The W1,k matrix (from now on written as W) is used for calculating the `dependence’or reachability’ matrix D:

D 5 #(W + W 2 + W 3 + W4 + . . . )

where # expresses Boolean summation. D is used in calculating a `connection’matrix, C:

ci j 5 di j + [di j dj i] + ki j

where ki j 5 1 if there is a relation, regardless of direction, between the industries iand j, or else ki j 5 0. K is calculated as

K 5 #[(I + I ¢ ) + (W + W ¢ ) + (W + W ¢ )2 + (W + W ¢ )3 + (W + W ¢ )4 + . . . ]

where the summation of the transposed W matrix (i.e. W ¢ ) and W `dissolves’ thedirection in the relation between industries i and j by making the sum matrix(W + W ¢ ) symmetric. The elements of C can take the values 0, 1, 2 or 3 (see forexample Harary et al., 1965):

ci j 5 0: no relation between i and j.ci j 5 1: there is a weak relation between i and j, in the sense that i and j are both

connected to a third industry, but there are no ¯ ows, neither direct norindirect, between i and j.

Eco

nom

ic S

yste

ms

Res

earc

h 20

00.1

2:37

7-39

9.

Industrial Interdependence in National Systems of Innovation 399

ci j 5 2: there is a one-way-relation from i to j. The direction form i to j is theresult of the multiplication di j dj i.

ci j 5 3: a bilateral relation between i and j exists, i.e. the relation is both from ito j and from j to i.

The matrix C is used for calculating centrality coeý cients, de® ned as the sum ofthe rows divided by the sum of the columns. The centrality coeý cients revealwhether the industry in question is a technology source (more out¯ ows than in¯ ows)or a technology receiver (more in¯ ows than out¯ ows). Using the coeý cients todecide the position of the industries in the graph and the values of the cells todecide whether there is a relation between two industries, and if so, whether thisrelation is one-way or bilateral, a directed graph of the embodied technology ¯ owsbetween industries is constructed.

Appendix C: Industries included in the analysis

ISIC, rev 23100 Food, drink and tobacco3200 Textiles, footwear and leather3300 Wood, cork and furniture3400 Paper and printing3510 + 3520 ( 2 3522) Industrial chemicals3522 Pharmaceuticals3530 + 3540 Petroleum re® neries3550 + 3560 Rubber and plastics3600 Stone, clay and glass3710 Ferrous metals3720 Non-ferrous metals3810 Fabricated metal products3820 ( 2 3825) Non-electrical machinery3825 Oý ce machines and computers3830 ( 2 3832) Electrical machinery3832 Communication equipment and semiconductors3841 Shipbuilding3842 + 3844 + 3849 Other transport3843 Motor vehicles3845 Aerospace3850 Instruments3900 Other manufacturing industries

Eco

nom

ic S

yste

ms

Res

earc

h 20

00.1

2:37

7-39

9.