papers

...more papers

The Limits of Autarky:
Regional Networks and Industrial Adaptation
in Silicon Valley and Route 128

AnnaLee Saxenian
Department of City and Regional Planning
University of California at Berkeley
Berkeley, CA 94720-1850
anno@socrates.berkeley.edu

Prepared for HUD Roundtable on Regionalism sponsored by the Social Science Research Council, Dec 8-9, 1994. This paper summarizes some of the themes of Regional Advantage: Culture and Competition in Silicon Valley and Route 128 (Harvard University Press, 1994).

The competitive advantages of regional clusters have become the focus of scholarly and policy attention. Once only the province of economic geographers and regional scientists, the work of Paul Krugman (1991) and Michael Porter (1990) has spurred widespread interest in regions and regional development. These newcomers have ignored an already extensive and sophisticated literature on the dynamics of industrial localization (see, for example, Storper, 1989; Scott, 1988a, 1988b; Vernon, 1960). Yet, like their predecessors, they share a reliance on external economies to explain the advantages that are derived from the spatial of clustering of economic activity.

In this essay I compare California's Silicon Valley and Route 128 in Massachusetts to suggest the limits of the concept of external economies and propose an alternative, network approach to analyzing regional economies. The common notion of external economies is based on an assumption that the firm is an atomistic unit of production with clearly defined boundaries. By drawing a sharp distinction between what occurs inside and what occurs outside the firm, scholars overlook the complex and historically-evolved relations between the internal organization of firms and their connections to one another and the social structures and institutions of a particular locality. The network perspective helps explain the divergent performance of apparently comparable regional clusters, such as Silicon Valley and Route 128, and provides important insights into the local sources of competitive advantage.

THE LIMITS OF EXTERNAL ECONOMIES

Alfred Marshall (1920) developed the notion of "external economies of scale" to refer to the sources of productivity increase that lie outside of individual firms. In the classic view, producers derive external benefits by sharing the fixed costs of common resources, such as infrastructure and services, skilled labor pools and specialized suppliers, and a common knowledge base. In addition, some theorists distinguish external economies that depend on the size of the market, including such factors as a labor pool and specialized supplier base (pecuniary external economies) from those that involve spillovers of knowledge between firms (technological external economies). When these factors of production are geographically concentrated, firms gain the additional benefits of spatial proximity, or "agglomeration economies." Once established in a locality, such an advantage becomes self-reinforcing through a dynamic process of increasing returns (Arthur, 1990; Krugman, 1991; Scott, 1988b; Storper, 1989).

Students of regional development typically treat Silicon Valley and Route 128 as classic examples of the external economies that are derived from industrial localization. They are seen as cumulatively self-reinforcing agglomerations of technical skill, venture capital, specialized input suppliers and services, infrastructure, and spillovers of knowledge associated with proximity to universities and informal information flows (see, for example, Castells, 1989; Hall & Markusen, 1985; Krugman, 1991; Porter, 1990; Scott, 1988b). Some researchers have compared them to the 19th century industrial districts described by Alfred Marshall (Piore & Sabel, 1984).

Yet this approach cannot account for the divergent performance of the two regional economies. In spite of their common origins in postwar military spending and university-based research, Silicon Valley and Route 128 have responded differently to intensified international competition. Both regions faced downturns in the 1980s. Although Silicon Valley recovered quickly from the crisis of its leading semiconductor producers, Route 128 shows few signs of reversing a decline that began in the early 1980s. The rapid growth of a new wave of start ups and the renewed dynamism of established companies such as Intel and Hewlett-Packard were evidence that Silicon Valley had regained its former vitality. Along Route 128, by contrast, start ups failed to compensate for continuing layoffs at the Digital Equipment Corporation and other minicomputer companies. By the end of the 1980s, Route 128 producers had ceded their longstanding dominance in computer production to Silicon Valley.

Regional data underscore this divergence. Between 1975 and 1990, Silicon Valley firms generated some 150,000 new technology jobs--triple the number created along Route 128, even though they enjoyed roughly the same employment levels in 1975.

In 1990, Silicon Valley-based producers exported more than $11 billion in electronics products, almost one third of the nation's total, compared to Route 128's $4.6 billion (Electronic Business, 1992). Finally, Silicon Valley was the home of 39 of the nation's 100 fastest-growing electronics companies, whereas Route 128 claimed only 4. By 1990, both Southern California and Texas had surpassed Route 128 as locations of fast-growing electronics firms.

These rankings are based on five-year sales growth rates, but the list is not limited to small firms. Multi-billion dollar companies such as Sun Microsystems, Apple Computers, Intel Semiconductor, and Hewlett-Packard all ranked among the fastest- growing enterprises in 1990.

The concepts of agglomeration and external economies alone cannot explain why clusters of specialized technical skills, suppliers, and information produced a virtuous and self-reinforcing dynamic of increasing industrial advances in Silicon Valley, while producing relative decline in Route 128. These theories account for regional stagnation or decline through imprecise references to "diseconomies" of agglomeration or the accumulation of negative externalities. Yet if such diseconomies are related to the overall size of a regional cluster, the degree of congestion, or the costs of production, growth should have slowed in the more densely populated Silicon Valley long before Route 128. The simple fact of spatial proximity evidently reveals little about the ability of firms to respond to the fast-changing markets and technologies that now characterize international competition.

The distinction between internal and external economies is based on the assumption that the firm is an atomistic unit of production with clearly defined boundaries. Treating regions as collections of autonomous firms has even led some observers to conclude that Silicon Valley suffers from excessive, even pathological, fragmentation (Florida & Kenney, 1990). Proponents of this argument overlook the complex of institutional and social relationships that connect the producers in its fragmented industrial structure. Researchers who adopt the broadest interpretations of technological external economies recognize that firms learn from each other through flows of information, ideas, and know-how (Storper, 1989), but they do so only by denying the theoretical distinction between internal and external economies, between what is inside and outside the firm.

A NETWORK APPROACH TO REGIONS

Far from being isolated from what lies outside them, firms are embedded in networks of social and institutional relationships that shape, and are shaped by, their strategies and structures (Granovetter, 1985). The network perspective helps illuminate the historically-evolved relationships between the internal organization of firms and their connections to one another and to the social structures and institutions of their particular localities (Nohria & Eccles, 1992b; Powell, 1987).

A network approach can be used to argue that, despite similar origins and technologies, Silicon Valley and Route 128 evolved distinct industrial systems in the postwar period. The differences in productive organization have been overlooked by economic analysts or treated simply as superficial differences between "laid back" California and the more "buttoned-down" East coast. Far from superficial, these variations demonstrate the importance of the local social and institutional determinants of industrial adaptation. In particular, they help explain why these two regions have responded so differently to the same external forces, from the lowering of global trade barriers and the intensification of international competition to cuts in the domestic military budget.

Silicon Valley has a regional network-based industrial system that promotes learning and mutual adjustment among specialist producers of a complex of related technologies. The region's dense social networks and open labor markets encourage entrepreneurship and experimentation. Companies compete intensely while at the same time learning from each other about changing markets and technologies through informal communications and collaborative practices. Loosely linked team structures encourage horizontal communication among firm divisions and with outside suppliers and customers. The functional boundaries within firms are porous in the network-based system, as are the boundaries between firms and between firms and local institutions such as trade associations and universities.

The Route 128 region, in contrast, is dominated by autarkic (self-sufficient) corporations that internalize a wide range of productive activities. Practices of secrecy and corporate loyalty govern relations between firms and their customers, suppliers, and competitors, reinforcing a regional culture that encourages stability and self-reliance. Corporate hierarchies ensure that authority remains centralized and information tends to flow vertically. Social and technical networks are largely internal to the firm, and the boundaries between firms and between firms and local institutions remain far more distinct in this independent firm-based system.

Regional Networks and Industrial Adaptation

Understanding regional economies as networks of relationships rather than as clusters of atomistic producers, and thinking of the regions as examples of two models of industrial systems--the regional network-based system and the independent firm-based system--helps illuminate the divergent trajectories of the Silicon Valley and Route 128 economies during the 1980s. For example, Silicon Valley's superior performance cannot be attributed to differentials in real estate costs, wages, or tax levels. Land and office space were significantly more costly in most of Silicon Valley than in the Route 128 region during the 1980s; the wages and salaries of production workers, engineers, and managers were higher (Sherwood-Call, 1992), and there were no significant differences in tax rates between California and Massachusetts (Tannewald, 1987).

Nor can the differences in regional performance be traced to patterns of defense spending. Route 128 has historically relied more heavily on military spending than has Silicon Valley, and hence is more vulnerable to defense cutbacks; however, the downturn in the Massachusetts electronics industry began in 1984, when the value of prime contracts to the region was still increasing. Although defense spending cannot account for the timing of the downturn in the region's technology industry, the military spending cutbacks that began in the late 1980s exacerbated the difficulties of an already troubled regional economy.

Finally, while it may be tempting to attribute Silicon Valley's prosperity to the ability of local firms to shift low wage jobs elsewhere, this cannot account for the differential performance of the two regions. Technology firms from both Silicon Valley and Route 128 have, since the 1960s, moved their routine manufacturing operations to lower-wage regions of the U.S. and the Third World (Scott, 1988b; Saxenian, 1985).

Route 128's difficulties lie in the rigidities of its local industrial system. The independent firm-based system flourished in an environment of market stability and slowly-changing technologies because extensive integration offered the advantages of scale economies and market control (Chandler, 1977). It has been overwhelmed, however, by changing competitive conditions. Corporations that invest in dedicated equipment and specialized worker skills find themselves locked in to obsolete technologies and markets, and their self-sufficient structures limit their ability to adapt in a timely fashion. The surrounding regional economy in turn is deprived of resources for self-regeneration because large firms tend to internalize most local supplies of skill and technology.

Regional network-based industrial systems like that of Silicon Valley, in contrast, are well-suited to conditions of technical and market uncertainty. Producers in these systems deepen their capabilities by specializing while engaging in close, but not exclusive, relations with other specialists. Network relations promote a process of reciprocal innovation that reduces the distinctions between large and small firms and between industries and sectors (DeBresson & Walker, 1991). Evidence from the industrial districts of Europe suggests that the localization of know-how and information encourages the pursuit of diverse technical and market opportunities through spontaneous regroupings of skill, technology, and capital. The region, if not all the firms in the region, is organized to innovate continuously (Best, 1990; Sabel, 1988).

The competitive advantages of network organizational forms are reflected in the experience of Japanese industry as well. Japanese producers of electronics and autos, for example, rely on extensive networks of small and medium-sized suppliers, to which they are linked through ties of trust and partial ownership. Although Japan's large firms may have often exploited suppliers in the past, many firms increasingly collaborate with them, encouraging them to expand their technological capabilities and organizational autonomy (Nishiguchi, 1989). Like their Silicon Valley counterparts, these producers tend to be geographically clustered and depend heavily on informal information exchange as well as more formal forms of cooperation (Friedman, 1988; Imai, 1989).

As the case of Japan suggests, there are large as well as small firm variants of network-based systems (Dyer, 1993; Fruin, 1992, 1993; Herrigel, 1993). Large corporations can integrate into regional networks through a process of internal decentralization. As independent business units are forced by competition to achieve the technical and productive standards of outsiders, they often rely on external institutions that facilitate knowledge sharing and collaboration with suppliers and customers.

Of course, all economic activity does not cluster within a single regional economy. Firms in network systems serve global markets and collaborate extensively with distant customers, suppliers, and competitors. Technology firms, in particular, are highly international. However, the most strategic relationships are often local because of the importance of timeliness and face-to-face communications in complex, uncertain, and fast changing industries (Nohria & Eccles, 1992a).

REGIONAL NETWORK V. FIRM-BASED SYSTEMS

In the rest of this chapter I use a set of paired comparisons to illustrate the differences in the organization and adaptive capacities of Silicon Valley's regional network and Route 128's independent firm-based industrial systems. The comparison of Apollo Computers and Sun Microsystems--both 1980s' generation start ups competing in the emerging workstation market--demonstrates how small firms benefit from the open flows of information, technology, and know-how in a network system. The comparison of the Digital Equipment Corporation (DEC) and Hewlett-Packard Co.(HP)--the leading computer systems producers in the two regions--in turn shows how regional networks can facilitate the reorganization of large firms.

Clearly, these cases alone cannot encompass the experience of two complex regional economies. For an extended treatment of the origins and evolution of the two regional economies, see Saxenian (1994). Nor can the focus on individual firms fully portray the myriad decentralized relationships in a regional network-based system. Indeed, the resilience of Silicon Valley's network system lies precisely in the fact that it does not depend upon the success of any individual firm. However, these comparisons illustrate the social and institutional dimensions of productive organization that are overlooked in the concept of external economies and the competitive advantages of regional networks in the current economic conditions.

Start Ups: Apollo Computer and Sun Microsystems

The largest wave of start ups in Silicon Valley's history began in the late 1970s and accelerated during the 1980s. The region was the home to scores of new ventures that specialized in everything from workstations and semi-custom semiconductors to disk drives, networking hardware and software, and computer-aided engineering and design. These start ups contributed to the diversification of the regional economy away from its original concentration in semiconductors and into a complex of computer-related specialists.

In contrast with the upsurge of entrepreneurial activity in Silicon Valley, the pace of start ups along Route 128 slowed during the 1980s. Massachusetts experienced lower rates of new high-tech firm formation between 1976 and 1986 than either New England or the United States as a whole (Kirchoff & McAuliffe, 1988). Also, the performance of companies founded during the 1980s was disappointing. Nothing in the Route 128 experience matched the spectacular successes of the 1980s' generation of Silicon Valley start ups such as Sun Microsystems, Conner Peripherals, and Silicon Graphics. By the end of the decade, public companies that were started in Silicon Valley during the 1980s collectively accounted for more than $22 billion in sales, whereas their Route 128 counterparts had generated only $2 billion (Standard & Poor's, 1992).

Investment decisions reflected this divergence. Annual venture capital investments in Northern California during the 1980s were double or triple those in Massachusetts. Over the course of the decade, Massachusetts-based companies received some $3 billion in venture capital, or 75% of the total raised in the region, whereas firms in Northern California received $9 billion, or 130% of the total capital raised locally. Silicon Valley companies were consistently awarded at least one third of the nation's total venture capital pool.

By 1992, 113 technology enterprises located in Silicon Valley reported revenues exceeding $100 million, compared to 74 companies in Route 128. Moreover, the great majority of Silicon Valley's $100 million enterprises were started during the 1970s and 1980s, whereas those in Route 128 were overwhelmingly started prior to 1970 (CorpTech, 1993).

The comparison of Apollo Computer and Sun Microsystems demonstrates how the autarkic structures and practices of Route 128's independent firm-based system created disadvantages for start ups in a technologically fast-paced industry. Apollo pioneered the engineering workstation in 1980 and initially was enormously successful. By most accounts, the firm had a product that was superior to that of its Silicon Valley counterpart, Sun Microsystems (which was started two years after Apollo, in 1982). The two firms competed neck and neck during the mid-1980s, but in 1987 Apollo fell behind the faster moving, more responsive Sun, and never regained its lead. By the time Apollo was purchased by Hewlett-Packard in 1989, it had fallen to fourth place in the industry, whereas Sun led the industry with more than $3 billion in sales (Bell & Corliss, 1989).

Apollo's founder, 46-year-old William Poduska, one of Route 128's few repeat entrepreneurs, had worked for Honeywell and helped to found Prime Computer before starting Apollo. Not only was Poduska himself well steeped in the culture and organizational practices of the region's established minicomputer firms, but the entire Apollo management team moved with him from Prime. This history contrasts with the that of the typical Silicon Valley start up, in which talent was typically drawn from a variety of different firms, and even industries, representing a mix of corporate and technical experience.

Not surprisingly, Apollo's initial strategy and structure reflected the model of corporate self-sufficiency of the region's large minicomputer companies. In spite of its pioneering workstation design, for example, the firm adopted proprietary standards and chose to design and fabricate its own central processor and specialized integrated circuits. Although it sourced components such as disk drives, monitors, and power supplies, Apollo began with a proprietary operating system and architecture that made its products incompatible with other machines.

Sun, in contrast, pioneered open systems. The firm's founders, all in their twenties, adopted the UNIX operating system because they felt that the market would never accept a workstation custom designed by four graduate students. By making the specifications for its systems widely available to suppliers and competitors, Sun challenged the proprietary and highly profitable approach of industry leaders IBM, DEC, and HP, which locked customers in to a single vendor of hardware and software. This strategy allowed Sun to focus on designing the hardware and software for its workstations and to limit manufacturing to prototypes, final assembly, and testing. Unlike the traditional vertically integrated computer manufacturers, Sun purchased virtually all of its components off the shelf from external vendors and subcontracted the manufacture and assembly of their printed circuit boards. (In the late 1980s, Sun began assembling some of its most advanced printed circuit boards internally.) The firm even relied on outside partners for the design and manufacture of the reduced instruction set computing (RISC)-based microprocessor at the heart of its workstations and encouraged its vendors to market the chip to its competitors.

Although specialization is often an economic necessity for start ups, Sun did not abandon this strategy even as the firm grew into a multi-billion dollar company. Why, asked Sun's Vice President of Manufacturing Jim Bean in the late 1980s, should Sun vertically integrate when hundreds of Silicon Valley companies invest heavily in staying at the leading edge in the design and manufacture of integrated circuits, disk drives, and most other computer components and subsystems? Relying on outside suppliers greatly reduced Sun's overhead and ensured that the firm's workstations contained state-of-the art hardware.

This focus also allowed Sun rapidly to introduce complex new products and continuously alter their product mix. According to Bean: "If we were making a stable set of products, I could make a solid case for vertical integration" (Whiting, 1987). Relying on external suppliers allowed Sun to introduce an unprecedented four major new product generations during its first five years of operations and to double the price-performance ratio each successive year. Sun eluded clone-makers through its sheer pace of new product introduction. By the time a competitor could reverse engineer a Sun workstation and develop the manufacturing capability to imitate it, Sun had introduced a successive generation.

As a result, the Sun workstations, although vulnerable to imitation by competitors, were also significantly cheaper to produce and sold for half the price of the proprietary Apollo systems (Bulkeley, 1987). Sun founder and CEO Scott McNealy described the advantage for customers: "We were totally open with them and said, 'We won't lock you into anything. You can build it yourself if we fail,' whereas our competition was too locked up in this very East coast minicomputer world, which has always been proprietary, so that encouraging cloning or giving someone access to your source code was considered like letting the corporate jewels out or something. But customers want it" (Sheff, 1989).

It quickly became apparent that customers preferred the cheaper, non-proprietary Sun workstations. However, Apollo, like the Route 128 minicomputer producers, was slow to abandon its proprietary operating system and hardware. As late as 1985, the firm's management refused to acknowledge the growing demand for open standards and even turned down the offer of a state-of-the-art RISC microprocessor from Silicon Valley-based MIPS Computers. Apollo finally committed 30 percent of its research and development budget to RISC development in 1986, but the effort became an economic burden, and the chip they ultimately developed internally was no faster than the chip they could have bought two years earlier from MIPS.

Sun's innovative computing strategy was inseparable from the firm's location in the sophisticated and diversified technical infrastructure of Silicon Valley. Apollo, in contrast, responded sluggishly to industry changes in part because of a more limited regional infrastructure. According to Jeffrey Kalb, an engineer who worked for DEC in Route 128 for many years before moving to Silicon Valley to start the MasPar Computer Corp.:

It's hard for a small company to start in Route 128 because you can't get stuff like IC's and disk drives fast. Route 128 is dominated by large, vertically integrated firms that do everything themselves. In Silicon Valley, you can get anything you want on the market.

You can get all those things in Route 128 sooner or later, but the decisions are much faster if you're in Silicon Valley. From the East coast, interacting with the West coast is only possible for 3-4 hours a day because of the time difference, and you spend lots of time on the phone. It's no one thing, but if you get a 20-30% time to market advantage by being in Silicon Valley, that's really significant. (Kalb, 1991)

Apollo's other major misstep was in its 1984 choice of a President and CEO to replace Poduska. Following the tradition of the large Route 128 companies, they hired a long-time East coast corporate executive who had worked his way up the ranks at General Electric and then become the President of GTE Corporation. The 53-year-old Thomas Vanderslice was asked to bring "big-company organizational skills" to fast growing Apollo and help the firm to "grow up." He couldn't have had a more different background than the twenty-something graduate students and computer whizzes who had founded Sun Microsystems two years earlier (Beam & Frons, 1985).

The media played up the superficial differences between Apollo and Sun: the buttoned down, conservative Apollo executives alongside the casually attired, laid-back founders of Sun. It made for great journalism: Vanderslice enforced a dress code and discouraged beards and moustaches at Apollo, and Sun threw monthly beer bashes and employees showed up on Halloween in gorilla suits. Whereas Vanderslice was chauffeured to work daily in a limousine, an April Fool's Day prank at Sun involved placing founder Bill Joy's Ferrari in the middle of the company's decorative pond.

However, the important differences between the two firms lay in their management styles and organization: Vanderslice brought in a traditional, risk averse management team who focused on imposing financial and quality controls, cutting costs, and diversifying the firm's customer base. Former Apollo employees describe him as an archetypical "bean counter" who established formal decision-making procedures and systems in the firm at a time when flexibility and innovation were most needed.

This commitment to formality, hierarchy, and long-term stability--which typified most large Route 128 companies--could not have offered a greater contrast with the "controlled chaos" that characterized Sun (Weiss & Delbecq, 1987). Like many Silicon Valley companies, Sun developed decentralized organizational forms in its efforts to preserve the flexibility and enthusiasm of a start up even as it grew. Corporate strategy was generated by discussions among representatives of autonomous divisions rather than dictated by a central committee and Sun's culture encouraged informal communications, participation, and individual initiative (Levine, 1988).

In the late 1980s, when Sun surpassed Apollo in both sales and profitability, more than a dozen Apollo managers defected to their West-coast rival. They joined other experienced and ambitious engineers at ailing Route 128 companies who recognized that opportunities to join or start technologically exciting new ventures lay not in New England, but along the increasingly crowded freeways of Northern California. As skilled engineers moved west, the advantages of Silicon Valley's network-based industrial system multiplied.

Large Firms: Digital Equipment and Hewlett-Packard

The successes of the 1980s' generation start ups were the most visible sign that Silicon Valley was adapting faster than Route 128, but changes within the regions' largest firms were equally important. Established producers in Silicon Valley began to decentralize their operations, creating interfirm production networks that built on the region's social and technical interdependencies and strengthened its industrial system. By institutionalizing longstanding practices of informal cooperation and exchange, they formalized the process of collective learning in the region. Local firms redefined themselves by participating in local production networks, and the region as a whole organized to create new markets and sectors.

Adaptation in the Route 128 economy, by contrast, was constrained by the autarkic organization and practices of its leading producers. Focused inward and lacking dynamic start ups from which to draw innovative technologies or organizational models, the region's large minicomputer firms adjusted very slowly to the new market conditions. By the end of the decade, they were struggling to survive in a computer industry that they had once dominated.

Although it is very difficult to develop accurate and useful measures of vertical integration, one indication of the greater reliance of Route 128 firms on internal production is the lower sales per employee figures shown in Table 1 for the leading Route 128 firms and their Silicon Valley counterparts.

Table 1. 1990 Sales Per Employee: Silicon Valley and Route 128 ($ thousands)

Source: "The Electronic Business 200" Electronic Business (22 July 1991) 43-49; Annual 10K Company Reports

The comparison of DEC and HP during the 1980s highlights the differing relationship of large firms to the region in network and firm-based industrial systems. By 1990 both were $13 billion companies and the largest and oldest civilian employers in their respective regions. (Lockheed Missile and Space and Raytheon Corporation were the largest private employers in Silicon Valley and Route 128, respectively. But both were military contractors with limited commercial business.) Both DEC and HP were vertically integrated producers of proprietary minicomputers with shared origins in an earlier era of computing. Yet the companies responded differently to comparable competitive challenges. HP gradually opened up by building a network of local alliances and subcontracting relationships while strengthening its global reach. DEC, in spite of its formal commitment to decentralization, retained a substantially more autarkic organizational structure and corporate mindset.

The transformations in the computer industry during the 1980s placed a premium on speed and focus. Computer makers were forced to develop and bring new products to market faster than ever before, often in a matter of months. HP Vice President of Corporate Manufacturing Harold Edmondson claimed in 1988 that half of the firm's orders in any year came from products introduced in the preceding three years (Edmondson, 1988). At the same time, the cost of developing new products increased as they became more technologically complex. Innovation in all segments of the industry--from microprocessors and logic chips to system and applications software to disk drives, screens, input-output devices, and networking devices--meant that it was more and more difficult for a single firm to produce all of these components, let alone remain at the forefront of the underlying technologies. This increasingly competitive environment posed a challenge for established computer makers like DEC and HP. By 1990, however, HP had successfully managed the transition from minicomputers to workstations with open systems, whereas DEC remained dependent on its proprietary VAX line of minicomputers. As a result, even though both enjoyed 1990 revenues from electronics products of $13 billion, HP earned $771 million, and DEC lost $95 million.

Variations in corporate performance always have multiple causes, but the firms' organizational structures and their relationships to their respective regions help explain these differences. DEC maintained clear boundaries between itself and other companies or institutions in the region. This was, in part, a result of extensive vertical integration: the firm designed and manufactured internally virtually all software and hardware components for its computers internally. Moreover, DEC's corporate culture rewarded secrecy and corporate loyalty; departed employees were typically treated like pariahs and cut off from the corporate "family" (Rifkin & Harrar, 1990). As a result, the technical and social networks that mattered were all internal, and there were few opportunities for collaboration, learning, and exchange with other local firms.

HP was both less dominant in Silicon Valley and more open to the surrounding economy. DEC dominated the Route 128 region in a way that no firm did in Silicon Valley. With more than 30,000 Massachusetts employees in 1990, DEC accounted for almost 20 percent of regional high-technology employment, whereas HP's 20,000 Silicon Valley employees were only 8 percent of the regional total. HP benefitted from a long history of participation in the region's rich associational life and fluid labor markets. Continuous and open exchange about everything from the latest start ups to technical breakthroughs allowed local engineers to stay at the leading edge of new computing technologies and market trends (Vedoe, 1990).

HP's decentralized divisional structure also offered an ideal training ground for general managers. Former HP executive were responsible for starting more than 18 firms in Silicon Valley between 1974 and 1984, including such notable successes such as Rolm, Tandem, and Pyramid Technology (Mitchell, 1989). A 16-year veteran of DEC who now works for HP described how the firm's autonomous divisions preserve opportunities for entrepreneurship:

Running a business at the division level, you get a chance to be a general manager. You get a chance to learn . . . to be creative . . . There are a lot of new divisions springing up [within HP], new ideas springing up, brand new businesses, and old divisions that couldn't make it anymore transform themselves into new businesses. (P. Porter, 1993)

In contrast, DEC's matrix organization--which represented only a partial break from traditional functional corporate hierarchies--stifled the development of managerial skill and initiative in the Route 128 region. The matrix demanded continuous negotiations to reach consensus, and despite the addition of cross-functional relations among product groups, final authority remained highly centralized (Schein, 1985). As a result, aside from Data General, it is difficult to identify successful spin-offs from DEC.

Both DEC and HP began the 1980s with the bureaucracy and internal conflicts typical of large firms. Both missed opportunities and made false starts in workstation and RISC markets, and both had difficulty keeping up with newer, more agile competitors. Yet HP quickly became the leading producer in the fastest growing segments of the market. By 1990, HP controlled 31 percent of the $8 billion RISC computer systems market--a market in which DEC still had no presence. HP also boasted a 21 percent share of the $7.2 billion workstation market and 13 percent of the $33 million UNIX computer systems market, compared to DEC's 16 percent and 8 percent respectively. In addition, HP controlled 66 percent of the market for desktop laser printers and 70 percent of the market for ink jet printers (Nee, 1991).

Hewlett Packard reinvented itself by investing heavily in RISC microprocessor technology and the UNIX operating system well before most established computer companies recognized the importance of open standards. By betting the future of the computer division (which accounted for 53 percent of HP revenues) on RISC systems in 1985 and by undertaking internal reorganizations that unified and rationalized the firm's disparate computer divisions and component technologies, HP positioned itself advantageously for emerging markets (Yoder, 1991). In 1990, for example, the firm created an independent team to develop a RISC workstation. The ultimate product, the Series 700 workstations, was far ahead of the rest of the industry and allowed HP quickly to become one of the world's biggest sellers of UNIX systems. A financial analyst for Salomon Brothers assessed the situation: "they [HP] have done an excellent job of identifying trends in the computer market such as Unix, RISC, and PCs. No other major computer company has done a better job of positioning . . . They are the one company I can count on surviving. HP has a better base today than IBM or DEC" (Greene, 1990).

HP's ability to identify market trends early reflected the firm's openness to external changes and a Silicon Valley location that gave it easy access to state-of-the-art information markets and technologies. This flexibility contrasts sharply with DEC's prolonged denial of the growing demand for personal computers and UNIX-based systems. In the words of a former DEC marketing manager: "DEC had its head in the sand. They didn't believe that the world would really change . . . They got focused on the internal evolution of the company rather than on the customer or markets" (Vedoe, 1990). As late as 1985, DEC CEO Olsen referred to personal computers as "snake oil" (Harrar & Rifkin, 1990).

DEC was plagued by ongoing internal conflicts and a series of costly course reversals in its efforts to enter the workstation and open systems markets. The firm's strategy remained confused and inconsistent even after the defection of large customers such as GE and AT&T forced Olsen to authorize a shift to open systems and away from the vision of a single proprietary VMS operating system and VAX architecture for all DEC systems (DeNucci, 1990).

DEC's research lab in Silicon Valley developed state-of-the-art RISC and UNIX technologies in the early 1980s, but its discoveries were virtually ignored by headquarters, which continued to favor the highly profitable VAX-VMS system (Comerford, 1992). Insiders claim that DEC's Palo Alto lab contributed more to other Silicon Valley firms such as Sun and MIPS than it did to DEC because their findings quickly diffused to other Silicon Valley firms through technical papers and local industry forums (Basche, 1991; Furlong, 1991).

DEC finally decided to build its own RISC-based workstation in 1986, following conventional wisdom within the firm that the RISC microprocessor should be designed and built in house. It was not until 1992, however, after a series of costly reversals, that the firm finally introduced its own RISC processor, Alpha (Comerford, 1992). By this time, DEC controlled only 13 percent of the workstation market (McWilliams, 1992).

The contrast between DEC's Palo Alto Lab and its East coast operations is instructive. Engineers who worked at both emphasize how different the two were: DEC East was internally focused, whereas DEC Palo Alto was well integrated into Silicon Valley's social and technical networks. According to Joe DeNucci, a former employee:

DEC definitely relates differently to the regional economy in Silicon Valley than in Route 128. DEC is the largest employer in Route 128 and you come to think that the center of the universe is North of the Mass Pike and East of Route 128. The thinking is totally DEC-centric: all the adversaries are within the company. Even the non-DEC guys compete only with DEC.

DEC Palo Alto is a completely different world. DEC is just another face in the crowd in Silicon Valley; the adversaries are external, firms like Intel and Sun. It forces a far more aggressive and "prove-it" mind set. (DeNucci, 1991)

He described his years with the DEC engineering and development group in Palo Alto:

We had an immense amount of autonomy, and we cherished the distance from home base, from the 'puzzle palace,' and from the 'corridor warriors' and all the endless meetings. It was an idyllic situation, a group of exceptionally talented people who were well connected to Stanford and to the Silicon Valley networks. People would come out from Maynard and say 'this feels like a different company.' The longer they stayed, the more astounded they were. (DeNucci, 1991)

Tom Furlong, who headed a DEC workstation division in Maynard before moving west in 1985, described the newly formed Workstation Group in Palo Alto as a typical Silicon Valley start up. The group's autonomy from headquarters allowed members to take full advantage of the local knowledge available within the regional economy. At the same time, the group benefitted from the financial backing and reputation of a large, well-established corporation. By 1990, Furlong was the manager of a 275-person group. He compared his experience working in the two locations:

It would be very difficult for me to do what I'm doing here within DEC on the East Coast. I'm a fairly autonomous business manager out here, with all the functions necessary to success reporting to me and the freedom to use outside suppliers. Back East, I would have to rely on DEC's internal suppliers and functional groups for everything.

We're like a start-up organization here. We're not really significant to DEC, we're only contributing $.5 billion to them, but we have the advantages of their resources and name. (Furlong, 1991)

He explained the consequences of these organizational differences for new product development:

The same job of bringing a new workstation to market takes two times as long in the East coast and many more people than it does here. In Maynard, I had to do everything inside the company. Here I can rely on the other companies in Silicon Valley. Its easier and cheaper for me to rely on the little companies in Silicon Valley to take care of the things I need, and it forces them to compete and be more efficient. At DEC, the commitment to internal supply and the familial environment means that bad people don't get cut off. I had to depend on all sorts of inefficient people back at DEC East. (Furlong, 1991)

The Workstation Group did not achieve this independent position without resistance: "It was a huge embarrassment to them that we had to rely on external suppliers such as MIPS. DEC takes great pride in being vertically integrated, in having control over its entire system" (Furlong, 1991).

DEC was ultimately unable to assimilate the lessons of its geographically distant Palo Alto group, in spite of their technical advances, and in 1992 transferred it back to Maynard headquarters. Furlong and other members of the workstation team left DEC to work for Silicon Valley companies.

HP began the decade with a comparable level of vertical integration to DEC, but soon recognized that it could not continue to produce everything in house. In the late 1980s, HP began outsourcing most of the sheet metal fabrication, plastics, and machining for its computer systems. The firm also consolidated the management of some 50 disparate circuit technology units into two autonomous divisions, Integrated Circuit Fabrication and Printed Circuit Board Fabrication. These divisions were organized as internal subcontractors for the company's computer systems and instrument divisions. They were forced to compete with external vendors for HP's business and expected to remain competitive in technology, service, and cost to sell successfully to outside customers.

HP also built alliances with local companies that offered complementary technologies. During the 1980s, the firm created partnerships with Octel Communications for voice-data integration, 3Com for local area network-manager servers, and Weitek for semiconductor design. An HP manager explained the acquisition of a 10 percent stake in Octel: "In the business and office processing environment, no one company can develop everything on its own, so we're increasingly looking at forming alliances to meet our customers' needs" (Tuller, 1988).

The partnership between HP and semiconductor design specialist Weitek illustrates how a large firm benefitted from Silicon Valley's networks. Tiny Weitek, which lacked manufacturing capacity of its own, was the leading designer of ultra-high speed "number crunching" chips for complex engineering problems. In 1987, HP opened its state-of-the-art fabrication facility to Weitek for use as a foundry, hoping to improve the performance of the Weitek chips in its workstations. Realizing that the manufacturing process at the foundry Weitek used slowed down the chips, the HP engineers suggested fully optimizing the Weitek designs by manufacturing them with HP's more advanced fabrication process. This culminated in a three-year agreement that allowed the firms to benefit directly from each other's technical expertise.

The arrangement assured HP of a steady supply of Weitek's chips and allowed them to introduce their new workstation faster than if they had designed the chip in house. It provided Weitek with a market, the legitimacy of a close association with HP, as well as access to a state-of-the-art foundry. Moreover, the final product represented a significant advance over what either firm could have produced independently. This partnership allowed each to draw on the other's distinctive and complementary expertise to devise novel solutions to shared problems.

HP opened itself to outside influences during the 1980s, creating a model of a large firm that is internally decentralized and horizontally linked to networks of other specialists. DEC's dominant and isolated position in Route 128, by contrast, hindered its efforts to shift to new technologies or a new corporate form. Saddled with an autarkic organizational structure and located in a region that offered little social or technical support for a more flexible business model, DEC's difficulties worsened.

In 1992, DEC CEO and founder Ken Olsen was forced to resign after the company reported a $2.8 billion quarterly loss--the biggest in computer industry history. One year later, HP surpassed DEC in sales to claim the position as the nation's second largest computer company, after IBM. As a final irony, in 1993 DEC moved a design team for its new Alpha microprocessor from the East coast to Palo Alto to immerse Alpha engineers in the Silicon Valley semiconductor community. According to industry analyst Ronald Bowen of Dataquest: "Digital is finding the support network of other companies is very, very limited back East. In effect, what's been happening is the people who work on the East coast spend a lot of time flying to San Jose anyway" (Nash & Hayes, 1993).

CONCLUSION

This comparison of Silicon Valley and Route 128 highlights the analytical leverage gained by treating regions as networks of relationships rather than as collections of atomistic firms. By transcending the theoretical distinction between what lies inside and outside of the firm, this approach offers important insights into the structure and dynamics of regional economies. It directs attention to the complex networks of social relationships within and between firms and between firms and local institutions.

The Silicon Valley experience also suggests that network forms of organization flourish in regional agglomerations. Proximity facilitates the repeated, face-to-face interaction that fosters the mix of competition and collaboration required in today's fast-paced technology industries. Yet the case of Route 128 demonstrates that geographic clustering alone does not ensure the emergence of regional networks. Competitive advantage derives as much from the way that skill and technology are organized as from their presence in a regional environment.

REFERENCES

Arthur, B. (1990). Positive feedbacks in the economy. Scientific American, 262 (2), 92-99.

Basche, T. (1991). Vice President, Sparcstation Group, Sun Microsystems, personal communication.

Beam, A. & Frons M. (25 March 1985). How Tom Vanderslice is forcing Apollo Computer to grow up. Business Week, 96-98.

Bell, A. & Corliss E. (24 April 1989). Apollo falls to the west. Mass High Tech, 1, 9.

Best, M. (1990). The new competition: Institutions of industrial restructuring. Cambridge, MA: Harvard University Press.

Bulkeley, W. M. (6 July 1987). Culture shock: two computer firms with clashing styles fight for market niche. Wall Street Journal, A1.

Castells, M. (1989). The informational city: Information technology, economic restructuring, and urban-regional process. Oxford, UK: Basil Blackwell.

Chandler, A. D. (1977). The visible hand: The managerial revolution in American business. Cambridge, MA: Belknap.

Comerford, R. (July 1992). How DEC developed alpha. IEEE Spectrum, 26-31.

CorpTech. (1993). Technology company information: Regional disks. Woburn, MA: Corporate Technology Information Services.

DeBresson, C. & Walker, R. eds. (1991). Special issue on networks of innovators. Research Policy. 20 (5).

DeNucci, J. (1990). Vice President Entry Systems Group, MIPS Computer Systems, personal communication.

Dyer, J. (1993). The Japanese vertical keiretsu as a source of competitive advantage. Paper presented to the Vancouver Network Conference.

Edmondson, H. (1988). Vice President of Corporate Manufacturing, Hewlett Packard Corporation, personal communication.

Electronic Business (16 March 1992). The top 100 exporters. Electronic Business, 4-42.

Florida, R. & Kenney M. (1990). Silicon Valley and Route 128 won't save us. California Management Review 33 (1), 68-88.

Friedman, D. (1988). The misunderstood miracle: Industrial development and political change in Japan. Ithaca, NY: Cornell University Press.

Fruin, M. (1993). The visible hand and invisible assets: Network organization and supplier relations in the electronics industry in Japan. Paper presented to the Vancouver Network Conference.

Fruin, M. (1992). The Japanese enterprise system. Oxford, UK: Oxford University Press.

Furlong, T. (1991). RISC Workstation Manager, DEC Palo Alto, personal communication.

Granovetter, M. (1985). Economic action and social structure: the problem of 'embeddedness.' American Journal of Sociology 91 (3), 481-510.

Greene, T. (22 January 1990). Can HP find the right direction for the '90s? Electronic Business, 26-29

Hall, P. & Markusen A. (1985). Silicon Landscapes. Boston, MA: Allen & Unwin.

Herrigel, G. (1993). Large firms, small firms, and the governance of flexible specialization: The case of Baden-Wurtemmberg and socialized risk. In B. Kogut, ed. Country Competitiveness. New York: Oxford University Press.

Imai, K. (1989). Evolution of Japan's corporate and industrial networks. In B. Carlsson, ed. Industrial Dynamics Boston, MA: Kluwer.

Kalb, J. (1991) President, MasPar Computer Corporation, personal communication.

Kirchoff, B. and McAuliffe, R. (1988). Economic redevelopment of mature industrial areas. Report prepared for Technical Assistance and Research Division, Economic Development Administration, U.S Department of Commerce.

Krugman, P. (1991). Geography and Trade. Cambridge, MA: MIT Press.

Levine, J. B. (18 July 1988). Sun Microsystems turns on the afterburners. Business Week, 114-118.

Marshall, A. (1920). Industry and Trade. London: Macmillan.

McWilliams, G. (4 May 1992). Crunch time at DEC. Business Week, 30-33.

Mitchell, J. (9 Jan 1989). HP sets the tone for business in the valley. San Jose Mercury News,

Nash J. & Hayes, M. (19 July 1993). Key DEC project moving to Palo Alto. Business Journal (San Jose and Silicon Valley), 1, 17.

Nee, E. (1991). Back to basics at Hewlett-Packard. Upside, June/July, 38-78.

Nishiguchi, T. (1989). Strategic dualism: An alternative in industrial societies. Ph.D. dissertation, Nuffield College, Oxford University.

Nohria, N. & Eccles, R. (1992a). Face-to-face: Making network organizations work. In Nohria & Eccles, eds. Networks and organizations: Structure, form, and action. Boston, MA: Harvard Business School Press.

Nohria, N. & Eccles, R., eds. (1992b). Networks and organizations: Structure, form, and action. Boston, MA: Harvard Business School Press.

Piore, M. & Sabel, C. (1984). The second industrial divide: Possibilities for prosperity. New York: Basic Books.

Porter, M. (1990). The competitive advantage of nations New York: The Free Press.

Porter, P. (23 Aug 1993). Executive interview: HP's Gary Eichorn tackles enterprise computing. Mass High Tech, 3.

Powell, W. (1987). Neither market nor hierarchy: Network forms of organization. In B. Staw, ed. Research in Organizational Behavior. Greenwich, CT: JAI Press.

Rifkin, G. & Harrar, G. (1990). The ultimate entrepreneur: The story of Ken Olsen and Digital Equipment Corporation. Rocklin, CA: Prima Publishing.

Sabel, C. (1988). Flexible specialization and the reemergence of regional economies. In Hirst, P. and Zeitlin, J., eds. Reversing industrial decline?: Industrial structure and policy in Britain and her competitors. Oxford, UK: Berg.

Saxenian, A. (1994) Regional advantage: Culture and competition in Silicon Valley and Route 128. Cambridge, MA: Harvard University Press.

Saxenian, A. (1985) Silicon Valley and Route 128: Regional prototypes or historic exceptions? In Castells, M. ed. High technology, space and society. Beverly Hills: Sage.

Schein, E. (1985). Organizational culture and leadership. San Francisco, CA: Jossey-Bass.

Scott, A. (1988a). Metropolis: From the division of labor to urban form. Berkeley, CA: University of California Press.

Scott, A. (1988b). New industrial spaces: Flexible production organization and regional development in North America and Western Europe. London: Pion.

Sheff, D. (1989). A new ballgame for Sun's Scott McNealy. Upside, November/December, 46-54.

Sherwood-Call, C. (1992). Changing geographic patterns of electronic components activity. Economic Review (Federal Reserve Board of San Francisco). 2, 25-35.

Standard & Poor's (1992). Compustat PC+ database.

Storper, M. (1989). The transition to flexible specialization in the U.S. film industry: External economies, the division of labor, and the crossing of industrial divides. Cambridge Journal of Economics. 13, 273-305.

Tannenwald, R. (1987). Rating Massachusetts' tax competitiveness. New England Economic Review. November/December, 33-45.

Tuller, D. (12 August 1988). HP plans to buy 10% stake in Octel. San Francisco Chronicle.

Vedoe, C. (1990). Manager Workstation Marketing, Sun Microsystems, personal communication.

Vernon, R. (1960). Metropolis 1985. Cambridge, MA: Harvard University Press.

Weiss, J. & Delbecq, A. (1987). High technology cultures and management: Silicon Valley and Route 128. Group and Organization Studies. 12 (1), 39-54.

Whiting, C. (15 March 1987) For flexible, quality manufacturing don't do it yourself. Electronic Business, 46-47.

Yoder, S. (22 July 1991). A 1990 reorganization at Hewlett-Packard already is paying off. Wall Street Journal, A1.

...more papers