High-Technology Clusters: Specialisation, Interaction and Transportation G. M. Peter Swann Manchester Business School University of Manchester, UK Workshop on High-Tech Business and Clusters Robinson College, Cambridge, 28th May 2003 A supporting paper is available from: pswann@man.mbs.ac.uk
Objectives To recognise the role of specialisation and interaction in the development of clusters. and therefore, to analyse the role of a cost-effective communication and transportation infrastructure in cluster development Illustrate with reference to personal computer manufacture.
What sort of Cluster are we discussing here? The term cluster has come to mean different things to different authors as Ron Martin has shown Here we discuss the highly specialised and narrowly defined clusters that emerge in some high-tech industries A value chain is broken into many different stages and companies (and indeed clusters) specialise in a small part of that value chain Illustrate this with reference to manufacture of personal computers:
Personal Computer (PC) Manufacture In PC manufacture, it is almost meaningless to ask in what country a PC is manufactured: Open up a personal computer and inspect the electronic chips: they read like the United Nations (Scott and Hayen, 1991) The various components are manufactured in many different countries, assembly may be done in more than one country, and the final "badge" may be added somewhere else again.
Origin of Components for Typical Personal Computer Brand: USA Final Dispatch: Ireland Main Box: Ireland Motherboard Chips: USA, Korea, Taiwan, Philippines Motherboard Battery: Philippines CD ROM Drive China (Japanese Parts) CD-R (consumables) Germany Hard Disk Drive: Singapore 3.5" Floppy Disk Drive: Philippines
Origin of Components. (continued) Modem Card: Netherlands (chips from USA, Korea, Taiwan) Graphics Card: China (chips from USA, Korea, Taiwan) Specialist Video Card: USA Monitor: UK (origin of components?) Keyboard: Mexico Mouse: Mexico Child's Mouse: Taiwan
Origin of Components. (continued) Loudspeakers: Malaysia Microphone: Mexico Inkjet Printer: Spain Zip Drive: Malaysia Scanner: Taiwan Webcam: China Power Supplies: Taiwan, China, Malaysia, Mexico Manuals: Scotland, Ireland, Wales, Germany Environmental Mark: Sweden
Internationalisation It is well recognised that Personal Computer industry is one of those in which internationalisation and intraindustry trade is furthest developed: vertical trading chains spanning a number of countries, each specialising in a particular stage of production (OECD, 2002) This splitting of production across countries may be most beneficial for complex manufactured products, involving many components (OECD, 2002) Krugman (1995) speaks of supertrading countries which depend on the slicing up of the value added chain
Internationalisation and Shocks This internationalisation of production means that a shock in any one country can have implications for a widely dispersed supply chain E.g. the major earthquake in Taiwan (September 1999) caused disruption to many parts of the high-tech industry: It will ripple through the whole supply chain (CNET News.com, 1999) This is an old idea, see for example, Jevons (1878): A second disadvantage of the division of labour is that trade becomes very complicated, and when deranged, the results are ruinous to some people.
Internationalisation and Clusters The processes of internationalisation and clustering are inter-related Particular companies and indeed particular clusters may become specialised in one (or a few) particular part(s) of the overall process In short, we have a process of specialisation and interaction which depends on transportation for its development
Simple Model: Key Concepts Cluster development proceeds along with division of labour and specialisation The cluster model of economic development requires interaction between different stages of supply chain, and hence a cost-effective communication and transportation structure This interaction may be the rich exchange of tacit knowledge of the (mythical?) Silicon Valley model... Or it may be the more mundane interaction required for completion of supply chain logistics
Figure 1 Geographical Maps of City, Towns and Villages
Figure 2.1 Dispersed: Production Co-located with Demand
Figure 2.2 8 Vertically Integrated Towns
Figure 2.3 4 Vertically Integrated Towns
Figure 2.4 Single City Cluster
Figure 2.5 4 Specialised Clusters
70 Figure 3 Transportation to Customers Frequency by Length of Journey 60 50 Frequency 40 30 20 10 0 0 2 4 6 8 10 12 0 2 4 6 8 10 12 0 2 4 6 8 10 12 0 2 4 6 8 10 12 0 2 4 6 8 10 12 4 Clusters 1 City 4 Towns 8 Towns Dispersed
Model of Specialisation and Interaction Two Parts: Model of Production Costs - with economies of scale, scope and agglomeration Model of Interaction Costs - between production stages and final customers
Model of Specialisation and Interaction: (1) Production Costs Any production stage enjoys economies of scale: { } PC = X exp α X PC ij ij j ij i N = j= 1 PC ij Any cluster enjoys economies of scope: N PC = PC exp β X i i ij j= 1
Model of Specialisation and Interaction: (1) Production Costs, continued. Any cluster enjoys economies of agglomeration, but with a risk of congestion costs in very large clusters PC PC.exp Z X Z X 2 N N i = i γ1 i + γ ij + 2 i + ij j= 1 j= 1 PC = K i= 1 PC i
Model of Specialisation and Interaction: (2) Interaction Costs Communication costs in a given supply chain: ( 1 θ ) [ ] h pq pq pq C C { } CC = Φ e + f.d i( p,h ),i( q,h ) Transportation costs in a given supply chain : h pq pq pq T T { [ ]} TC = θ Φ e + f.d i( p,h ),i( q,h )
Model of Specialisation and Interaction: (2) Interaction Costs, continued... Interaction costs are the sum of these two, summed over all nodes of the supply chain, and all supply chains IC = CC + TC h h h pq pq pq IC H N N = h= 1 p= 1 q= 1 IC h pq
Parameter Values and Model Outcomes (1) a simple chart of how the most efficient structure depends on the parameters (2) a radar chart of average parameter values amongst simulations with a particular outcome
Parameter Values and Model Outcomes: (1) Simple Chart a two-dimensional mapping from parameters to outcomes, under the following assumptions: Online communication costs are negligible (e C = f C = 0) Transportation costs within a cluster are negligible (e T = 0) Offline proportion (θ ) is fixed at 10% Congestion cost parameter (γ 2 ) is fixed at 0.00006 Scope economy parameter is half the value of the scale economy parameter (β = α /2) The agglomeration economy parameter is one tenth of the scale economy parameter (γ 1 = α /10)
Table 1 Total Communication/Transportation Distances, by Industrial Structure A B B C C D D FC 4 Specialised Clusters 512 512 512 388 1 City Cluster 0 0 0 264 4 Vertically Integrated Towns 0 0 0 168 8 Vertically Integrated Towns 0 0 0 104 Dispersed Production Colocated with Customers 0 0 0 0
Figure 4 Outcome by Parameter Values Transport Cost Parameter (x 10-1 ) 1300 590 260 120 52 23 10 4.6 8 Towns Dispersed (Production Co-located with Demand) 4 Towns 8 Towns 2.0 0.9 4 Clusters 1 City 0.4 0 1 2 3 4 5 6 7 8 9 10 Scale Economy Parameter (x 10-2 ) (also scope and agglomeration)
Table 2 Ranking of Different Industrial Structures by Efficiency, for certain parameter values Scale = 0 Scope = 0 Agglomeration = 0 Congestion = 0.00006 Scale = 0.05 Scope = 0.25 Agglomeration = 0.005 Congestion = 0.00006 Scale = 0.1 Scope = 0.05 Agglomeration = 0.01 Congestion = 0.00006 Offline = 10% Comm Cost = 0 Transp Cost = 26 8 Towns (100%) Dispersed (75%) 4 Towns (68%) 4 Clusters (11%) 1 City (0%) Dispersed (100%) 8 Towns (40%) 4 Towns (29%) 1City (19%) 4 Clusters (3%) Dispersed (100%) 8 Towns (25%) 4 Towns (16%) 1 City (10%) 4 Clusters (1%) Offline = 10% Comm Cost = 0 Transp Cost = 1 8 Towns (100%) 4 Towns (73%) 4 Clusters (53%) Dispersed (43%) 1 City (0%) 4 Towns (100%) 1 City (95%) 8 Towns (51%) Dispersed (31%) 4 Clusters (21%) 4 Towns (100%) 8 Towns (69%) 1 City (69%) Dispersed (26%) 4 Clusters (9%) Offline = 10% Comm Cost = 0 Transp Cost = 0.04 8 Towns (100%) 4 Towns (74%) 4 Clusters (73%) Dispersed (41%) 1 City (0%) 4 Clusters (100%) 1 City (53%) 4 Towns (39%) 8 Towns (14%) Dispersed (7%) 1 City (100%) 4 Towns (52%) 4 Clusters (14%) 8 Towns (7%) Dispersed (2%)
Parameter Values and Model Outcomes: (2) Radar Chart a 6-dimensional radar chart showing the average value for each parameter amongst all simulations with a particular outcome: compute model outcomes over a grid of 4 6 = 2 12 = 4096 different parameter combinations and outcomes for all simulations in which a particular structure is most efficient, we compute the average value of each parameter each axis runs from 0% to 100%, since each parameter average is expressed relative to the maximum value for that parameter
Table 3 Grid of Parameter Values Value 1 Value 2 Value 3 Value 4 Scale Economy 0.006 0.024 0.042 0.06 Scope Economy 0.0006 0.0024 0.0042 0.006 Agglomeration Economy 0.00006 0.00024 0.00042 0.0006 Congestion Costs 0.000006 0.000024 0.000042 0.00006 Offline % 10% 40% 70% 100% Transportation Costs 0.3 1.2 2.1 3.0
Figure 5 Mean Parameter Values for Simulations with Different Outcomes Scale Transport Costs Scope Offline % Agglomeration 4 Towns 8 Towns 1 City 4 Clusters Dispersed Congestion Costs
Conclusions The specialised cluster strategy requires cost effective transportation and communication Codification may increase the proportion of transactions online, but cannot displace the face-to-face If transportation is too expensive to support specialised / distributed clusters, then the single city cluster may be the the only viable clustered structure But is the single city cluster desirable? Congestion costs Unbalanced development Very relevant to UK
Conclusions, continued Distributed cluster structures may be the only way to provide the mix of economies of scale, scope and agglomeration required for international competitiveness while avoiding congestion costs of single city cluster. But if the social cost of transportation is higher than private cost, then distributed cluster structures may be environmentally unsustainable