1 / 67

Franco Malerba, Richard Nelson, Luigi Orsenigo and Sidney Winter

INDUSTRIAL POLICIES IN A HISTORY-FRIENDLY MODEL OF THE CO-EVOLUTION OF THE COMPUTER AND SEMICONDUCTORS INDUSTRIES. Franco Malerba, Richard Nelson, Luigi Orsenigo and Sidney Winter.

hanzila
Download Presentation

Franco Malerba, Richard Nelson, Luigi Orsenigo and Sidney Winter

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. INDUSTRIAL POLICIES IN A HISTORY-FRIENDLY MODEL OF THE CO-EVOLUTION OF THE COMPUTER AND SEMICONDUCTORS INDUSTRIES Franco Malerba, Richard Nelson, Luigi Orsenigo and Sidney Winter L. Orsenigo, Pecs, july 2009

  2. The analysis of the effects of industrial policies is often based on static, equilibrium models, with agents characterized by complete rationality. Moreover, such analyses consider in most cases, one specific industry, without fully taking into account the effects that public intervention might bear on related industries L. Orsenigo, Pecs, july 2009

  3. growing literature on industrial dynamics, with heterogeneous agents, increasing returns and path-dependency even within this literature, policy implications have been by and large neglected, particularly as formal models are concerned. L. Orsenigo, Pecs, july 2009

  4. Antecedents The Schumpeterian trade-off Network externalities Blind Giants, Narrow Windows and Angry Orphans L. Orsenigo, Pecs, july 2009

  5. Microsoft case competition is not within the market, but for the market: temporary monopolies, provided that the entry of new potential monopolists offering potentially superior new technologies is not blockaded. Thus, antitrust policies should consider not only the immediate effects of the existence of a monopolist and of its actions but also the predictions of alternative possible futures (which might be very hard to predict indeed) L. Orsenigo, Pecs, july 2009

  6. efficacy of alternative policy measures designed to combat monopoly (if deemed appropriate on efficiency and political grounds), to promote industry growth and technological change. In general, little is known about the efficacy of industrial policies in dynamically related, co-evolving industries. L. Orsenigo, Pecs, july 2009

  7. Aims • Exploring the effects of alternative policies on the dynamics of two vertically-related industries in changing and uncertain technological and market environments. • “make or buy” decisions • dynamics of market concentration in contexts characterized by periods of technological revolutions punctuating periods of relative technological stability and smooth technical progress. L. Orsenigo, Pecs, july 2009

  8. History-Friendly Models • CLOSER RELATIONSHIP WITH HISTORICAL AND EMPIRICAL ANALYSIS • INDUSTRY-SPECIFICITIES • PUT MORE RESTRICTIONS ON MODELS • DERIVE TIME-PATHS, NOT “SIMPLY” LIMIT PROPERTIES • FORMALIZE AN APPRECIATIVE ARGUMENT • REPRODUCE A QUALITATIVE HISTORY: PROBING A THEORY, NOT REPRODUCING REAL TIME SERIES • BASIS FOR GENERALIZATIONS • TOOL FOR THEORETICAL, “HISTORY-FREE” INVESTIGATION L. Orsenigo, Pecs, july 2009

  9. VERTICAL INTEGRATION AND SPECIALIZATION: THE CONCEPTUAL APPROACH • Differences in firms’ capabilities and role played by the nature of the knowledge base in an industry. • Through learning firms accumulate over time capabilities in specific technological, productive and market domains. • Such competencies take time to be developed and are then typically sticky and local. • Established capabilities may adapt only slowly to changes in technology and demand and new capabilities more in tune with the new demand or technology may be necessary. • When products are systems with various components and subsystems, the capabilities of coordination and integration of these capabilities may be relevant. This may provide an advantage to integrated firms. • The growth and dynamics of competencies in each one of two vertically related industries influences the evolution of the other sector and shapes the dynamics of vertical integration and specialization L. Orsenigo, Pecs, july 2009

  10. Competencies and selection • The time dimension: • ompetencies take time to be developed and the previous history of the processes of construction of those capabilities often is important in determining what firms can and cannot do. • decisions to specialize and to vertically integrate are not symmetrical: • if a firm decides to discontinue the development and production of certain components, it might find it difficult to resume such activities later on and in any case time and efforts are required. Thus, these decisions are not entirely flexible as time goes by. • distribution of capabilities: the type and competences of all the other industry participants play a major role in affecting the decision to integrate or specialize. • market selection amplifies the impact of capabilities on the vertical scope of firms. • If specialized firms have superior capabilities, selection will push for grater specialization and vice-versa • market selection decreases heterogeneity among firms. • The process of capability development over time depends on the vertical scope of an industry: • Specialized firms that compete with other specialized firms accumulate knowledge and competence differently from vertically integrated firms. • In turn, the process of capability development affects the rooster of potential entrants. . L. Orsenigo, Pecs, july 2009

  11. The history • 3 ERAS • 1. late 1940s and early 1950s: the era of mainframe computers and transistors. • Early 1950s: entry of already existing firms: IBM, the Bunch (Burroughs, Univac Rand, NCR, Control Data, Honeywell), GE and RCA. In Europe and Japan, Philips, and Siemens. • IBM emerges as leader L. Orsenigo, Pecs, july 2009

  12. Vertical integration and specialization in the transistor era • At the very beginning of the industry, most computer producers were not integrated. • second half of the 1950s and the early 1960s : introduction of transistors and beginning of the semiconductor industry: • The largest firms (IBM, RCA and GE) were totally or at least partially vertically integrated. • The smaller firms purchased components on the market. L. Orsenigo, Pecs, july 2009

  13. Integrated Circuits • 2. The invention and development of the integrated circuits enabled further improvements in mainframe computers and reduced barriers to entry in the mainframe industry, stimulating the entry of new competitors to IBM. • ICs opened the possibility of designing minicomputers • Governmental anti-trust suit carried on for 13 years by the American Justice Department against IBM: • June 1982: IBM “ unbundles” its software and peripherals L. Orsenigo, Pecs, july 2009

  14. Integrated Circuits (ctd) • With the introduction of integrated circuits IBM became fully vertically integrated into semiconductors. • As a vertically integrated company, IBM produced the new system 360. By the end of the 1960s IBM enjoyed a market predominance of 70% in the world’s general service computer market. • Other mainframe producers also partially integrated into integrated circuits. • Three basic reasons for vertical integration: • integrated circuits embedded system elements and thus required close co-ordination between the system and the component producer in the design and development of both components and systems. • semiconductor designs became more and more “strategic” and key for system development, and therefore their design, development and production was kept in-house for fears of leakage of strategic information. • the rapid growth of the mainframe market and later on of the minicomputer market (1960s and 1970s) generated fears of shortages of various key semiconductor components among some of the largest computer producers. L. Orsenigo, Pecs, july 2009

  15. Microprocessors • Microprocessors enabled significant improvements in mainframes. • They made it possible to design personal computers. • Personal computers opened up a new demand which had not been touched by mainframes: small firms and personal users. • The great availability of low-priced high power computer components, led to the beginning of the microcomputer industry. • Apple Computer, Radio Shack and Commodore were all non-vertically integrated specialized in microcomputers. • In 1978 the whole personal computer market was practically ruled by those three firms which enjoyed together the 72% of the worldwide market. • However already by 1980, new start-ups were entering into the market with an increase in competition and an intensification of selection. • In 1980 IBM decided to enter the production of microcomputer. L. Orsenigo, Pecs, july 2009

  16. IBM, Intel and Microsoft • IBM strategy was to establish a common standard in the market through the production of a successful microcomputer (the PC) as the company did with the launch of the 360 system in the mainframe market. • IBM decided to buy its own components, peripherals and software from outside suppliers instead to build them internally. • IBM needed to speed up microcomputer production and did not have advanced internal capabilities in this respect. Moreover, also the software had to be developed independently from the hardware. Only the assembly of the minicomputers parts was supposed to be undertaken at IBM • IBM decided to choose Intel’s 8-bit older chip rather then the state of the art chips of Motorola or its clones which were much more powerful (and used by the most of IBM competitors) • IBM required Intel to sign a standard nondisclosure agreement and, in addition, stated that Intel should licence the chip out so that the IBM productive plant in Florida could be sure of a second alternative source. • IBM also turned to Microsoft for the standard operating software: MS-DOS. IBM agreed to let Microsoft licence its software products to others, because IBM aimed to lock the emerging market to its operating software. • In this way, however, Microsoft and Intel were able to conquer the respective software and microprocessor markets in few years. L. Orsenigo, Pecs, july 2009

  17. The PC Market • The huge unanticipated demand for microcomputers quickly transformed the microcomputer industry: established and start-ups companies swarmed into the minicomputer market because it was relatively easy developing or cloning the PC. • In the late 1984, as output began to catch up with demand, an industry shake-out occurred. • With time, IBM role of coordinator of decentralized technical progress by various suppliers weakened, because a shared technological leadership emerged with Intel and Microsoft. L. Orsenigo, Pecs, july 2009

  18. Vertical Disintegration Vertically integrated computer producers-- including IBM-- exited from large scale production of semiconductor components. Dis-integration took place because the new demand for semiconductors coming from personal computer producers had grown greatly; in response, a variety of highly advanced components were introduced by several merchant microelectronics firms. A key firm -Intel- emerged as the industry leader for microprocessors, thus determining a de facto standard in the semiconductor industry to which computer producers, out of necessity, complied. L. Orsenigo, Pecs, july 2009

  19. The Model: An Overview • At the beginning of the simulation, a given number of firms enters the market and begin to design and sell computers. • Computers are defined as a mix of characteristics, i.e. cheapness and performance. • Computers makers are born specialised and buy components on the marketplace from specialized component suppliers. The design of semiconductors is based on the available component technology, i.e. transistors. • Component firms sell their products to computer producers (as a function of the quality of their semiconductors) and to an external market, which is not explicitly modeled. • Computers are sold to heterogeneous groups of consumers as a function of their achieved merit of design. At the beginning, component technology makes it possible to design computers – mainframes - the characteristics of which appeal to consumers relatively more interested in performance rather than in their price. • Computer firms’ sales are also influenced by phenomena of inertia and brand-loyalty By investing profits in R&D firms improve the quality of their products. Some firms grow, others lose market shares and eventually exit. L. Orsenigo, Pecs, july 2009

  20. Integrated Circuits • Technological discontinuities in component technology. • First, integrated circuits- become exogenously available. • This new technology allows for the entry of new semiconductor firms. • As they invest in R&D and the new technology improves, they will gradually become more efficient than competitors belonging to the older generation, eventually displacing them. L. Orsenigo, Pecs, july 2009

  21. Microprocessors • After some more time microprocessors are introduced and again new component firms enter the market. • Microprocessors make it possible not only to design better mainframes but also to design a new typology of computers which appeal to new groups of customers relatively more interested in the cheapness rather than in the performance of the machines. • A new generation of computer firms enters the industry, opening up a new market – personal computers. L. Orsenigo, Pecs, july 2009

  22. Vertical integration • The decision to produce component in-house should in principle be driven by considerations related to the relative achievable quality of the components designed in-house as compared to those offered by the specialist suppliers. • However, computer firms can only conjecture about the quality of the components they might end up designing. • The decision to vertically integrate is led (probabilistically) by the relative size of computer firms vis-à-vis (the largest) component producer. • If computer producers are larger enough as compared to extant suppliers, they can fund a much larger flow of R&D expenditures and achieve better quality. • fears of supply shortages may induce vertical integration. Again, this is likely to be the case if semiconductor firms are small. • The decision to vertically integrate depends probabilistically on the age of the component technology: • Early stages: risks of getting stuck in an inferior trajectory • Established trajectories: the probability that new, superior generations of components may be frequently invented by component suppliers is lower. L. Orsenigo, Pecs, july 2009

  23. Specialization • Vertical disintegration is driven by a comparison between the merit of design of the components produced in-house vis-à-vis those made available by specialised semiconductors producers. • This is likely to happen in the early stages of the development of a new component technology and as semiconductor producers grow big enough to sustain a high level of R&D expenditures. • After signing a contract, the computer producer is tied to the component firm for a given number of periods. L. Orsenigo, Pecs, july 2009

  24. The Model • Computers • Computers are defined as a point in the space of characteristics, cheapness (i.e. the inverse of price) and performance. • As a consequence of firms’ R&D investment, the characteristics of computers of a given type improve over time. • The position of a particular computer design at any time defines its “merit of design” (Mod) or quality. In turn, computers result from the combination of two main inputs, systems and components. • The level of the merit of design, Mod, is given by a CES function: • (1) • with A > 1, 0 < t < 1 and r > -1. The elasticity of substitution is: = 1/ (1 + r). • Mainframes PCs. • No diversification L. Orsenigo, Pecs, july 2009

  25. Demand for Computers • Customers of computers are characterized by their preferences about the two attributes that define a computer design - performance and cheapness. • There are two customer groups: “big firms” and “small users” • Each customer group consists of a large number of heterogeneous subgroups. Within a particular sub group customers buy computers valuing its "merit", compared to other products. In addition, however, markets are characterized by frictions of various sorts, including imperfect information and sheer inertia in consumers behaviour, brand-loyalty (or lock in) effects • (2) • The propensity Li of computer i to be sold to a group of customers at time t is defined as: • (3) • where Sit-1 is the market share; the probability Pri of the computer i to be sold to a group of customers at time t is given by: • (4) • From this firm – if selected - the submarket buys a number of computers equal to Mit. L. Orsenigo, Pecs, july 2009

  26. The market for components • Systems are designed in-house by computer firms, while components may be also bought by specialized producers of semiconductors. • three different component technologies transistors, integrated circuits and then microprocessors • At the beginning of the simulation and at the time of each technological discontinuity a new cohort of firms (12 in this version of the model) enters the market, producing components with the latest available technology. • The demand for components, faced by component specialized firms, comes from two sources: • External demand, not modeled explicitly • demand for components from computer firms which have decided to outsource component production (specialized firms). • When a computer firm decides to outsource components production, it starts to scan the market for potential suppliers. Suppliers are chosen by computer firms on the basis of a ranking of the merit of design of the components produced by each supplier. Given uncertainty and imperfect information, this choice is partly stochastic. L. Orsenigo, Pecs, july 2009

  27. Market for components • Specifically, a specialized computer producer will “sign” a contract with a component producer selected by using a probability function that considers the technical quality of the components: the higher the quality of the component , the higher the probability of signing a contract with a computer producer. (5) • where LitCOMP is the propensity of component producer i to be selected and Pri,t is the probability of a supplier to be selected. • A component firm which signs a contract sells a number of components which is related to the proportion to which components and systems combine in order to build a computer (in the current parametrization, the proportion is one to one). After signing the contract the computer firm is tied to the component supplier for a certain number of periods. When this period expires, a new supplier might be selected, using the same procedure, if the firm still decides to buy component on the open market. L. Orsenigo, Pecs, july 2009

  28. Firms’ behaviour and technical progress • Firms start with a given (randomly drawn) mod and they start to sell make profits and invest in R&D spending. • Price is obtained by adding a mark-up, m , to costs which in turn are derived from the merit of design achieved by a computer. The price of components charged by component suppliers is determined symmetrically by adding a fixed mark-up to unit production costs. • R&D expenditures are calculated as . a constant fraction of profits • Technical progress: double draw scheme”. In each period firms draw the value of their Mod from a normal distribution. The number of draws that any one firm can take is set proportional to its R&D spending • In each period, the values of the Mod obtained through the firms’ draws are compared with the current Mod, and the higher among these values is kept. Thus, more draws increase the likelihood to get a higher Mod for both systems and components. L. Orsenigo, Pecs, july 2009

  29. Public knowledge • The extent to which technical progress is possible for each firm, given their R&D investment depends in turn on two variables: the level of publicly available knowledge and the value of the Mod achieved by the firm in the previous period • Public knowledge is specific to each basic component technology and it grows exogenously over time. When a new technology is introduced, its corresponding level of public knowledge is lower than that reached by current technology, but then it grows faster and at a certain time it overtakes the public knowledge of the older technology. The rate of growth of public knowledge starts to slow down as time goes by. An integrated computer firm decides to adopt the new technology when the mean of its own distribution becomes inferior to the level of the public knowledge of the new technology. • The mean of the normal distribution from which the values of the merit of design (Mod) of system or component are taken, is a linear combination of the Mod at time t-1 of firm i and of the level of publicly available knowledge, PK, at time t: • And • t>tmcK, lim and nu are parameters and tmckis the date of introduction of the new component technology. L. Orsenigo, Pecs, july 2009

  30. R&D • Integrated producers enjoy some coordination advantages as compared to specialized producers As a consequence, the productivity of their R&D efforts on components is enhanced by a spillover effect • cCOMP*m is the difference between the price of component in the open market and its actual cost for the producer; it represents savings gained by self-production. An integrated computer firm allocate these resources to component R&D. • Specialized computer producers invest all their R&D on systems and obviously do not enjoy the coordination advantages. • Component suppliers spend all their R&D on the development of components. L. Orsenigo, Pecs, july 2009

  31. Vertical Integration Probability of integration: Let: • where: • AgeOfTechK ( K =TR,IC,MP) = t – (Starting time of Technology K); Qit is the sales of the computer producer; biggestQt COMP is the sales of the largest component producers and w is a parameter Then: • where B is a parameter. If the probability of integration is bigger than a number drawn from a uniform distribution (0-1), integration occurs. L. Orsenigo, Pecs, july 2009

  32. Specialization • The probability of specialization for each firm is: • where maxModCOMP is the higher component Mod available on the market. • Then: • A is a parameter and if Prob(Specialize) is bigger than a number randomly drawn by a uniform distribution, specialization will occur. • A specialized computer firm may also decide to change its supplier, if a better producer has emerged in the market. The procedure for changing supplier follows the same rule for the specialization process. L. Orsenigo, Pecs, july 2009

  33. Exit • Computer firms: the variable • Eit = (1-e)*lshr + e*shareit • is computed, where lshr is the inverse of the number of firms active in the market at the beginning of the simulation (i.e. the market share that would have been held by n equal firms), shareit is the market share of firm i at time t and 0<e<1 is a parameter. Then, if Eit < E, where E is a constant threshold, the firm exits. • The rule governing the exit of the semiconductor producers is different and simpler. The probability of exiting of any one firm is an increasing function of the number of consecutive periods in which it doesn’t sell to a computer producer. L. Orsenigo, Pecs, july 2009

  34. Standard simulation • Assumptions: • the size of the external market is relatively small in the case of transistors and integrated circuits and significantly higher for microprocessors; • lock-in effects in demand are very important for mainframes and much less so for both PCs and components • the introduction of microprocessors allows much higher improvements in component designs as compared to the older technology: this technological discontinuity is much sharper than the previous one. L. Orsenigo, Pecs, july 2009

  35. The Standard Simulation: Results • A dominant firm emerges quickly in the mainframe industry and becomes vertically integrated. • In the semiconductor industry, concentration first rises as demand from computer producers exert strong selective pressures and firms leave the market. The decrease of the number of mainframe producers gradually softens competition and the Herfindahl index declines in the component market. Concentration begins to grow again as a vertically integrated monopolist comes to dominate the computer market: component suppliers are left with no demand from the mainframe firms and exit continues. • At the time of the introduction of integrated circuits, new semiconductor companies enter the market and concentration drops sharply. • The dominant mainframe firm remains vertically integrated, because the external market is not large enough to sustain a significant growth of the new entrants and of the quality of their components. The absence of a demand from the mainframe producer induces a shakeout and concentration gradually increases in the semiconductor market L. Orsenigo, Pecs, july 2009

  36. The age of microprocessors • Microprocessors constitute a major technological advance as compared to integrated circuit and a large external market supports a significant improvement in the quality of the new components. • the PC market opens up, generating a substantial new demand and fuelling further advances in the merit of the components. • The computer leader decides to specialize • Competition in components: large external market • The establishment of a monopoly in the supply of components contributes however to maintaining competition in the PC market, since all firms get their microprocessors from the same source. • In the last periods of the simulation, as the microprocessors technology matures, the incentives towards specialization become slightly less compelling and, in some simulations, the mainframe firm and some PC producers decide to vertically integrate L. Orsenigo, Pecs, july 2009

  37. History-Friendly Simulation: Results L. Orsenigo, Pecs, july 2009

  38. Effects of public policies in a dynamic setting: key variables a) the strength of increasing returns b) the timing of the arrival of the technological discontinuities and their magnitude c) the degree of heterogeneity of agents, in particular as it concerns consumers; d) the interactions between the upstream and downstream industries. Policy ineffectiveness Indirect and unintended effects L. Orsenigo, Pecs, july 2009

  39. Supply side policies Fostering the diffusion of knowledge support for public basic research Antitrust support the entry of new actors L. Orsenigo, Pecs, july 2009

  40. Demand side policies support open standards so that lock –ins are avoided; use public procurement in a selective way, so that the most advanced technologies are supported; use public procurement as an additional market, in order to provide larger markets and opportunities for firms’ growth and innovation. L. Orsenigo, Pecs, july 2009

  41. “policy ineffectiveness” cumulative nature of the market. Small initial advantages tend to grow bigger over time and catching up is almost impossible. Leaders do not only have a “static” advantage: they run faster than laggards. From “leveling the playing field” to “positive discrimination”? policies should make competitors able to run faster than the monopolist, and not just remove static disadvantages. L. Orsenigo, Pecs, july 2009

  42. Investment in basic research • Increase in the rate of growth of public knowledge • Faster integration in transistors and integrated circuits; increase in integration in the microprocessors era: better computers induce higher demand and faster firms’ growth • Great increase in the rate of technical change L. Orsenigo, Pecs, july 2009

  43. Favouring circulation knowledge • Firms can draw from existing techniques • Little effect on concentration in mainframes (high bandwagon) • Concentration decreases in PCs and components • Increase in the rate of technical change L. Orsenigo, Pecs, july 2009

  44. Antitrust • Antitrust intervenes when a firm has a market share > 90% and breaks the monopolist in two companies. • No effect: given strong lock-in effects, a new monopolist emerges very quickly • Antitrust intervenes also by reducing lock-in effects • Effects: • competition in mainframes • Slower vertical integration • Concentration grows in components market in the age of transistors and integrated circuits • Competition and specialization in all markets in the microprocessor era • No effects on the rate of technical change L. Orsenigo, Pecs, july 2009

  45. timing of antitrust policy: “Early” intervention has almost no effect because a new monopolist emerges very soon. “Late” intervention generates a duopoly, because increasing returns on the supply side were fading away by that time. small and transitory effects. L. Orsenigo, Pecs, july 2009

  46. cumulative nature of the market. Small initial advantages tend to grow bigger over time and catching up is almost impossible. Leaders do not only have a “static” advantage: they run faster than laggards. From “leveling the playing field” to “positive discrimination”? policies should make competitors able to run faster than the monopolist, and not just remove static disadvantages. L. Orsenigo, Pecs, july 2009

  47. Support to entry of new firms doubling the number of firms entering after each discontinuity in the computer and in the component market: No effect twelve new firms enter each market every forty period. In this latter case, new firms enter having the average merit of design present in the industry. L. Orsenigo, Pecs, july 2009

  48. periodic entry no changes in mainframes. In microprocessors, periodic entry does not change the leadership of the largest firm, but reduces industry concentration by greatly increasing the number of firms that are active and survive in the industry. bandwagon effects are weaker than in the mainframes market; new entrants in the component markets compete with a leader which is still building its dominant position and they are therefore able to survive. The later cohorts of entrants, though, find it increasingly difficult to compete. L. Orsenigo, Pecs, july 2009

  49. Timing Why can the entry of new firms – as contrasted with an increase in the initial number of firms - limit the tendencies towards monopoly even in the long run? If increasing returns are sufficiently weak to allow for a gradual process of concentration, the early new cohorts of firms enter with a level of the Mod which is higher than that attained by initial laggards and sufficiently high as compared to what has been achieved by the emerging leader, to provide them with a chance of surviving. L. Orsenigo, Pecs, july 2009

  50. Moreover the market size for microprocessors is rather large, because it is composed by the demand coming from the specialized mainframe producers and PC producers, and by the external market, so that the survival and growth of successful new firms is possible. L. Orsenigo, Pecs, july 2009

More Related