Information Storage Industry Center

David McKendrick

Report 98-03
November, 1998

Abstract
      The disk drive industry is both internationally dispersed and regionally concentrated. U.S. companies held 80% of the market in 1995, but they assembled 70% of their disk drives in S.E. Asia; and while only 20% of the people employed in the HDD industry worked in the U.S., R&D employees were overwhelmingly in the United States, centered on Northern California. The purpose of this paper is to explain how the industry came to have this pattern of dispersed concentration. It traces the location of each plant that assembled disk drives and each facility engaged in product development since the industry's establishment in 1956. The paper makes two specific points. First, the spatial pattern of the industry reflects the emergence of two kinds of clusters: operational and technological. Each cluster is differentiated by function and motivated by its own logic, with technological clusters "stickier" than operational ones. Second, industry location is dynamic: as an industry evolves, so does its location. The evolving location of industry results from the interaction of four locational determinants: market access, factor costs, public policy and agglomeration economies. In addition, the relative importance of each factor to the attractiveness of a specific location changes with time.

      Sociologists and economic geographers who study industrial districts and economic clustering emphasize their formation and stability.¹ Continuity and social embeddedness have been staples of the industrial districts literature. For sociologists, industrial districts are historic and continuous. They owe their success to the presence of small, innovative firms that are embedded within a regionally cooperative system of relations (Sabel and Piore, 1984; Sforzi, 1989; Saxenian, 1994; Best, 1990). Economic geographers give more emphasis to the processes by which industrial clustering occurs. Locations "selected" early on are likely to be "locked in" (Arthur, 1986). Both share the view that these agglomerations, once formed, are stable and durable. They are remarkably "sticky" places (Markusen, 1996) in an internationally mobile world.
      The hard disk drive industry offers an interesting lens through which to view the clustering of economic activity. Its spatial patterns are quite different from those associated with the industrial districts and economic geography literatures. The disk drive industry is both internationally dispersed and regionally concentrated. Economic activity in HDD stretches from the U.S., across the Pacific to Japan and S.E. Asia, increasingly China, and, continues across the Atlantic to Ireland, Germany and Hungary, with a very small presence reaching south to Mexico. But it is also geographically concentrated. While U.S. companies held 80% of the market in 1995, they assembled 70% of their disk drives in S.E. Asia, predominantly Singapore; and only 20% of the people employed in the HDD industry worked in the U.S., but R&D employees were overwhelmingly in the United States, centered on Northern California (Gourevitch, et al., 1997).
      The purpose of this paper is to move beyond a snapshot to consider how the industry came to have this pattern of dispersed concentration. It provides a sense of the sequence in which today's spatial division of labor took shape, offering evidence that clustering is not only a regional phenomenon but an outgrowth of global economic, regulatory and technological processes. The paper makes two points. First, there is more than one kind of industry cluster. As Audretsch and Feldman (1996: 630) point out, "[t]he emerging importance of location as a unit of observation argues for examining both production and innovation within a geographic context." I identify two kinds of clusters that emerged in this industry: operational and technological. Each is differentiated by function and motivated by its own logic; research can benefit from unbundling the two. Second, industry location is dynamic. The geographic pattern observed today is due to the evolving interaction of four locational determinants: market access, factor costs, public policy and agglomeration economies, each playing a relatively more or less important role during different times. Most empirical research on industry location does not capture this temporal and interactive character of location.
      The paper begins with a description of the history of the disk drive industry's location since the introduction of the first hard disk drive in 1956. It shows how the industry's innovative center became increasingly concentrated in California, while assembly dispersed overseas but became concentrated in S.E. Asia. It then explains why these two distinct types of functional clusters emerged and why they clustered where they did. The discussion is limited to U.S. firms, which account for over 60% of all firms that ever shipped a disk drive. The data are distilled from event histories of every U.S. company that made a hard disk drive. In addition to entry and exit information, I identified the location of each plant that assembled disk drives since 1956 and each facility engaged in product development. No one has previously traced the location of both manufacturing and product development in a single industry since its founding.² Unlike studies that examine industry location in one or two points in time, covering the entire history of an industry's location provides an opportunity for understanding the factors underlying both continuity and change in industry location.

Industry Origins and the Diversity of Location: The First 25 Years
      A variety of captive and independent firms entered the disk drive business in the United States between 1956 and 1980. Most were computer system manufacturers and diversified industrial enterprises, some were specialists in peripherals or other types of magnetic storage, but few were de novo start-ups. They can be classified into three groups: computer systems makers, independent disk drive suppliers, and "second generation" firms that mostly copied and improved upon IBM products.
      The first group consisted of computer systems manufacturers. In the early 1950s, IBM's San Jose lab was searching for a capacious storage device. IBM was lagging behind Remington Rand, which in 1954 had announced a magnetic drum-based storage device for its Univac File Computer, and Remington Rand was an experienced drum manufacturer. (Magnetic drums were early competitors to the disk drive.) IBM responded with the first prototype disk drive, which was delivered to a customer site in June, 1956. The following June, 1957, the first production unit came off the line (Bashe, et al., 1986). IBM's drive was an immediate commercial hit, and other systems manufacturers judged that they needed to make disk drives in order to be competitive. By the early 1970s, General Electric, Control Data, NEC, Hitachi, Fujitsu, Honeywell, Burroughs, NCR and Toshiba had all begun to make drives to bundle with own mainframe systems. Their data storage activity usually co-located with their systems development: GE in Arizona, CDC in Minnesota, Burroughs and NCR in Southern California, and Digital Equipment in Massachusetts. IBM and Honeywell were exceptions: IBM's San Jose lab was set up in 1952 in order to tap the engineering skills in California, far from the company's mainframe operations; Honeywell's storage operations were initially in Massachusetts rather than co-located with its Minnesota systems groups.
      The second group included machinery and computer peripheral companies that entered to serve computer system manufacturers that had not yet made their own disk drives. These firms placed a wide variety of different technological bets, experimenting with various disk and recording head technologies. They were also geographically dispersed. Bryant Computer Products (Michigan) and Data Products (initially Minnesota, later moving to Southern California) were the first two companies to enter the industry after IBM. While IBM's first disk drive used disk 24 inches in diameter, Bryant's were 39 inches and Data Products' were 31 inches. Other early entrants among this group came from Northern California (Friden) and Massachusetts (Anelex).
      The third group consisted of "second generation" producers of disk drives. Engineers and managers from IBM left to start their own disk drive companies, others would reverse engineer IBM drives, and some firms relied on technology transfer agreements with existing players.³ ISS and Memorex, two significant IBM spinoffs in Northern California, made drives completely plug compatible with IBM's, and even improved upon the performance of IBM's products. Another set of firms came from the electronics and defense industries and for the most part reverse engineered IBM disk drives. Many of these second generation firms were in Southern California, which after the end of World War II became the home to technology companies, especially aerospace and defense firms. The western part of the San Fernando Valley into southern Ventura County was often thought of as a smaller version of what was becoming known as Silicon Valley, but disk drive activity also emerged in Orange County, to the south of Los Angeles. Century Data Systems, Marshall Industries, Pertec, and Data Products were four prominent disk drive companies with headquarters and product development in Southern California.
      Although Northern California and Southern California were the two locations with the most HDD production, a great many other places in the U.S. had some disk drive activity. The two most successful entrants outside of California were Control Data and Digital Equipment Corporation. Control Data was for many years the leader in OEM production and the source of Minnesota's present day capabilities in magnetic recording. DEC was one of the most successful of the captive drive makers, for many years second only to IBM, with operations in Massachusetts, later centered on Colorado. General Electric invested in storage technology in Arizona, moving it to Oklahoma City in the late 1960s.⁴ Potter Instrument, of Melville, New York, leveraged its know-how in tape drives to enter the disk drive business. Although a leader in magnetic drum storage, Univac had a brief and unhappy experience designing disk drives for its mainframes at its Blue Bell, Pennsylvania, facility.⁵ Storage Technology, an IBM spinoff that originally made tape drives, helped to establish Colorado as a center for data storage technology.
      Figures 1a through 1e show the headquarters location of all American firms producing disk drives between 1960 and 1980 (in five-year increments).⁶ Activity was quite dispersed geographically, reflecting the variation in states of origin. The leading firms in the industry were also geographically dispersed. Figures 2a through 2e show the location of the five largest HDD producers between 1960 and 1980. Led by IBM, the captive disk drive companies came to dominate the industry by 1980, when they earned 77% of worldwide revenue. IBM alone accounting for 33%, followed by Burroughs, Control Data (which led in the OEM market as well), and DEC. Whether measured by number of firms or size, disk drive activity was considerably spread across the United States.

New Markets, New Firms, New Technologies, New Locations
      The emergence of new companies to make shoe box-sized 5.25-inch disk drives -- "little trifles," as one IBM engineer called them -- changed the logic of location. The introduction of the IBM PC in the early 1980s and the subsequent explosion of the PC market ultimately encouraged expansion and mass production of HDDs. In 1982 Seagate was making 12,000 drives a month. By the beginning of 1984, volume manufacturing at some companies reached 300,000 units a month. In 1988, the industry shipped almost 9 million 5.25-inch drives. Today, smaller 3.5-inch drives have supplanted the 5.25-inch drive, and the largest plants are capable of assembling 1.5 million units per month.
      Changing markets encouraged the development of new technology (smaller drives), and both had implications for firm location.⁹ Figures 3a through 3d show the location of all U.S. HDD firms between 1985 and 1997. Between 1980 and 1985, 40 companies entered the industry. Many of the companies that pioneered the new desktop market were spinoffs of incumbent firms that made larger disk drives (especially IBM), and these new firms set up business in the same region. However, very few new entrants were spinoffs from Control Data (Minnesota) or DEC (Massachusetts). Throughout the 1980s, new entry occurred primarily in Northern and Southern California, with some additional activity in Colorado. So gradually, disk drive development and manufacturing became more concentrated in California. Interestingly, these rapidly growing new companies were largely independents and began to displace the large, captive drive manufacturers. Figures 4a through 4d shows that by 1990, four of the five largest companies were headquartered in Northern California. In 1991, all five were from Northern California: IBM, Seagate Technology, Quantum, Conner Peripherals, and Maxtor.

Technological clusters
      By technological cluster I mean the co-location of activities that lead to the recognition of new market opportunities, the development of new technologies, and the design of new products.¹¹ But the term is meant to emphasize places where "innovation" occurs. The centripetal forces of technological clustering are driven by new firm entry and repeated intra- and inter-firm coordination in product design.
      New firm entry can replenish the competitive strength of a location through the introduction new technologies and is thus an important contributor to the longevity of a cluster. As in other industries, many of the founders of start-ups are engineers with an idea that their employers were unwilling to support with resources. The founders of many initial 5.25-inch and 3.5-inch entrants had worked in product development and marketing for other disk drive firms located in Northern California. These new companies included Seagate, Conner Peripherals, Rotating Memory Systems, International Memories, Atasi, La Pine, Quantum, Maxtor, and Areal Technology. While fewer in number, most early entrants in Southern California -- Tandon, Computer Memories, Peripheral Technology, Microcomputer Memories -- were founded by engineers with experience at other Southern California disk drive firms, including floppy disk drive companies.¹² Figures 7a through 7c show that the pioneers of smaller disk drives came primarily from these two regions.

Operational clusters
      Disk drive firms acquire a wide variety of inputs: recording heads and head subassemblies, disks, motors, electronics, plus various lower value-added components such as baseplates, magnets, clamps, etc. Today, the locations for these assembly and manufacturing activities are strongly concentrated in Singapore, Malaysia and Thailand, with a small cluster of activity in China and Ireland. These are "operational clusters" based on the economies of proximity in input-output relations: speed of throughput, product changeovers, increasingly specialized process engineering and assembly labor. Operational clusters may on occasion be sources of new product ideas, but their principal goal is to achieve operational efficiencies, and any new technologies they create are meant to improve production processes or supply chain management.
      In one sense, it is surprising to see manufacturing activity move from its original location. Jaffe et al. (1993) observe that one reason innovative activity clusters is because production does. Early in an industry's life, products and processes are uncertain and undergo a lot of experimentation. Products are made in ways that are often quite craft-like: Some of the early disks were coated with iron oxide particles poured by hand from a cup while a disk was spun. Producing more units provides more experience in the manufacturing process and greater understanding of how to produce additional units even more cheaply -- the learning-by-doing phenomenon. Such experience can also enable a firm to produce new products incorporating similar or related technologies at the same plant. Moreover, firms develop backward linkages with suppliers and forward linkages with customers, making a specific location increasingly more attractive (Krugman, 1991). Both these plant-specific skills and inter-organizational linkages serve as strong centripetal forces to concentrate production.
      Why would a firm, faced with such an attractive operational cluster, decide to locate production elsewhere, seemingly foregoing the benefits of agglomeration? One explanation turns on industry maturity and plant-level learning. As the industry matures and production increases, production processes become more uniform, and the emergence of standard kinds of equipment gives all competitors access to similar production and test equipment. Production costs begin to play a greater role in firm survival. Product specifications for manufacturing also become more routinized, component changes between new product introductions are kept to a minimum, and so changes to the manufacturing process become relatively straightforward and routine. Improvements in plant-level understanding of manufacturing processes also gradually allow the firm (and industry) to replicate a plant's operations elsewhere, including abroad. Manufacturing thus no longer needs to be in the initial high technology core, but can move to more distant locations, as Vernon (1966) and students of technology transfer (Teece, 1976) have explained. Thus, at some point manufacturing proximity to product development becomes less important than access to lower cost inputs, including labor.
      A second, complementary explanation, of course, is that political and technological changes have enabled firms to exploit the latent benefits associated with spatially separating stages of production. Several rounds of trade agreements, liberalization of foreign investment rules, improvements in transportation and communications have made possible the dispersion of manufacturing to many more locations around the world, allowing geographic advantage to play a more important role than in the past. These enabling technologies and policies suggest operational dispersion rather than clustering, as does the standardization and routinization of manufacturing processes along the lines suggested by the product life cycle. However, as explained below, production once again tends to concentrate. In the case of the disk drive industry, the operational clusters today do not result from processes emphasized by product cycle theory or the technology transfer literature, or from explanations offered by comparative advantage. Assembly in S.E. Asia does not occur after new products are first introduced and assembled in the United States. Instead, new products are now launched from plants overseas, with only prototype assembly occurring in the U.S. Clustering is encouraged in order to benefit from common pools of labor with experience in the industry, space, materials and service -- from the public sector as well as the private. Clustering in operations allows specialization and economies of scale, by which ancillary firms can spread their output across large local customers, and risk pooling in which a concentration of suppliers constitutes a sort of "depot" of specialized inputs, whether components, services or labor, serving different clients in more stable, standardized industries. Assemblers will thus have lower costs if they operate at a location with a larger number of suppliers of differentiated inputs (Klimenko, 1998). Through second-sourcing strategies, assemblers can drive prices down and obtain supplies in time to meet changing demand.
      Table 2 represents the different kinds of benefits derived from participating in a technological and operational cluster. The list should not be viewed as comprehensive; it is rather an attempt to list what are among the most important reasons for clustering in practice. Once co location of innovation and manufacturing is unnecessary from a technological point of view, these factors become relatively more important in shaping the production location decision (as well as the location of technical activities). It is easy to see that the logic underlying each kind of cluster differs, and it is analytically useful to distinguish between the two.

The Dynamics of Location: Market Access, Factor Costs, Public Policy, and Agglomeration Economies
      The separation of activities associated with product development from those of manufacturing, of course, implies an evolutionary process. Some strong forces are at work behind the international spread of the disk drive industry. Researchers have identified a number of factors that shape location behavior. For purposes of this paper, I focus on market access, factor costs, public policy, and agglomeration economies. The dynamic and interactive quality of these factors comprises one of the major elements in the spatial evolution of economic activity.
Market access
      Market access is one of the most compelling reasons to locate production (Dunning, 1993). Regional economists like Allen Pred (1966) and Raymond Vernon (1966) long ago emphasized that a large market makes it profitable to produce goods in a region that had previously been an importer. Overcoming host government import controls has been an especially important factor influencing the decision to produce abroad, but the need to adapt or customize products for the local market and the saturation of the domestic market have also been reasons for firms to move into new foreign markets.
Factor costs
      Resource endowments in standard accounts of international trade and investment would predict rather widespread dispersion of economic activity according to straightforward examination of factors of production: land, labor, capital and all the ancillary items --- utilities, transportation, and communication. Since factors of production between countries are not very mobile, the search for lower factor costs, especially for labor, could be expected to compel investors to disperse economic activity fairly widely.
Public policy
      Public policy shapes costs to manufacturing by structuring markets and market incentives. Regulations influence taxes, rebates, training, purchasing, local content, import duties, and other important items that affect the bottom line directly or indirectly.
Agglomeration economies
      Agglomeration economies explain that firms cluster not because the market is large, but because by locating near other similar firms they benefit from technological spillovers, specialized labor, and intermediate inputs (Arthur, 1986; Krugman, 1991; Head, et al., 1995). They are "economies" to the extent that the net benefits to being in a location with other firms in the value chain increase with the number of firms in the location (Arthur, 1986). In investing abroad, the benefits may include the flow of experienced-based knowledge on how to operate efficiently in a given location, including such mundane things as how to meet government regulations or adapt to the local labor force. We would expect specific parts of the production process to be clustered in individual countries if information spillovers occur largely through worker mobility across firms. Note also that clustering is not evidence for agglomeration economies. Similar firms may merely have the same needs, such as cheap labor or adequate water supplies and from which they benefit regardless of the presence of other firms in the value chain.

Temporal and interactive determinants of location
      Many discussions of the determinants of location refer to one or another of these factors but seldom in combination or in consideration of their timing and sequence. None of these factors alone explain the geographic distribution of activity in the disk drive industry. The industry is too concentrated to fit the factor cost explanation easily. Many low-cost areas around the world are not used, and too much activity remains in high-cost areas, notably California and also, currently, Singapore. Nor do firms assemble disk drives in Singapore in order to serve the Southeast Asian market or merely to get tax subsidies from the Malaysian and Singaporean governments. Agglomeration economies are clearly important in explaining today's pattern, but industrial clustering in 1997 looks very different than it did in 1975.
      The evidence from the history of the disk drive industry suggests the determinants of location are dynamic and interactive. That is, the inducements of market access, lower factor costs, supportive public policy and agglomeration economies interacted over time, and location decisions emerged from this four-way interaction to generate the industry's geographic structure. Certainly IBM's development of the disk drive in San Jose could be treated analytically as the result of chance and historical accident, which in turn accounts for a large part of the early location of the industry. At the same time, the interactions among these four factors offer strong explanations for the centrifugal forces dispersing assembly away from the U.S. The relative power of these forces changed over time, with factor costs diminishing in importance and agglomeration economies becoming more salient; public policy, however, has played a consistently important role.
      During the 1960s and 1970s, the primary benefits from locating disk drive assembly abroad was to enhance market access. IBM, Burroughs and Control Data preferred to have some assembly activity in or near to their largest markets, primarily Europe. For these companies, the primary customers for their computer systems -- governments, banks and insurance companies - were also likely to look favorably upon firms committed to local assembly of systems and peripherals.
      By the 1980s, that motivating logic had changed. Driven by computer makers' demand for cheaper components, initial HDD overseas investments by American companies making desktop drives were motivated by an effort to lower costs by tapping pools of cheap labor. But regulation can skew the geographical division of labor away from a pure factor cost model: A country with low wages can use public policy to make it stand apart from other low wage countries in the eyes of investors. Singapore stood out in the early 1980s, offering not only generous government incentives but a supportive policy environment. HDD companies required the ability to import needed components and ship final product. The necessary conditions for meeting these needs included liberal rules governing foreign investment, a free trade regime, high quality physical infrastructure, and facilitative "trade infrastructure" (transport, customs, etc.). Singapore met each of these requirements. Subsequently, similar motivations drove later investments (and some relocation) of HDD assembly to Malaysia and Thailand, and into the Philippines and China as well.
      Gradually, the production system, with Singapore at its center, became more complex (Doner and Haggard, 1998). The initial wave of drive assemblers served to attract a second wave of assemblers to Singapore. Among them were Conner Peripherals and Western Digital, whose S.E. Asian assembly operations contributed to their becoming among the industry's largest companies. There was also temporal clustering of investments in linked segments. Not only have drive assemblers followed drive assemblers, but both independent media and heads producers, as well as a number of second-tier suppliers in products such as spindle motors and actuators, also appear to follow their customers (Doner and Haggard, 1998). These investments began to bid up wage rates, but they also began to generate modest agglomeration economies, with ancillary firms reaching scale economies and helping to lower input costs for HDD firms. Singapore and Malaysia introduced additional policies in an attempt to hold activities in these increasingly high cost locations. These included not only targeted tax breaks and incentives for in-house and collective training but also a number of public or quasi-public institutions, such as a Local Industries Upgrading Program that involves extensive participation by foreign disk drive firms.
      Although some of the simpler, labor intensive activities now go to Thailand and China, HDD firms and the governments of Singapore and Malaysia continue to invest in activities that increase agglomeration economies. These efforts are designed to create and to exploit more knowledge- and skill-intensive capabilities through the managerial and technical workforce and the local supply base (Doner and Haggard, 1998). One is devoted to the data storage industry itself. In cooperation with the National University of Singapore, the Data Storage Institute offers undergraduate course in magnetic media technology, masters' level course in areas ranging from heads and media design to tribology and servo engineering, and advanced seminars by industry and outside academic experts. HDD firms are also involved in a research consortium on PCBs at Gintic Institute. HDD firms continue to invest in HDD assembly in Singapore even as its factor costs have escalated. IBM selected Singapore for its first S.E. Asian facility in 1995 because of its ability to ramp up production more quickly than anywhere else in the world. The government prepared the multistory facility specifically for drive and heads production, and IBM was able to hire a lot of local engineers and managers with previous disk drive experience to facilitate a fast start. Seagate even restructured its Thai HDD assembly operations and relocated its drive production to Singapore in 1998. The benefits of clustering in Singapore, and to a lesser extent Malaysia, have increased to the point where HDD firms can add value in ways consistent with the industry's ever more demanding product requirements and time pressures. Reaching volume production quickly requires rapid tooling and retooling, quick response in the supply of components, which often involve design modifications, and the ability to quickly "debug" a modified production process (Doner and Haggard, 1998).
      Table 3 summarizes this locational dynamic in a stylized way, showing the changing reasons firms invested in one country, Singapore. For firms producing the smaller disk drives, the initial incentive to invest in Singapore in the early 1980s was to exploit low wage rates, government subsidies, and a liberal trade and investment regime. Market access was unimportant, and agglomeration economies were absent. Gradually, companies that invested in Singapore benefitted in additional ways. As more firms located in S.E. Asia the supporting industries emerged in the region so that by 1990 three-fourths of the parts needed to produce a disk drive could be purchased in Asia (LAT, 1990; BT, 1993). These additional investments began to generate modest agglomeration economies. At the same time, public policy continued to support the industry, even offering targeted benefits. Today, the reasons for continuing to invest in disk drive assembly in Southeast Asia have in large part changed. Factor costs are no longer central to the economic motivation. Instead, even as the cost of wages and land increase, Singapore and Malaysia continue to attract investment because of the benefits of locating near others in the value chain. The agglomeration economies in HDD assembly in Singapore have become quite important to HDD companies. The industry continues its search for factor cost savings -- the recent HDD investments in China are so motivated. But it will be some time, if ever, before China offers disk drive companies the low operational costs they enjoy in Singapore or its supportive public policies and agglomeration economies.

Conclusion
      The U.S. disk drive industry is globally dispersed yet regionally concentrated: assembly is largely in S.E. Asia and R&D almost entirely in the United States. This pattern grew out of the interaction of four locational determinants: factor costs, market access, public policy and agglomeration economies. The importance of each factor changed over time, with firms initially motivated by a search for cheaper inputs and subsidies. Over time, as more firms invested in Singapore and Malaysia, the disk drive industry began to benefit from agglomeration economies.
      The kind of activities carried out in S.E Asia are quite different from those in the United States, suggesting it might be useful to distinguish among two types of industrial clusters: operational and technological. Once assembly and production become standardized, firms are able to focus more on lowering costs by shifting their location to places offering cheaper inputs; gradually, one or more of these locations develops an operational cluster, wherein firm proximity generates economies that include scale of production and speed of ramp-up. By contrast, technological clusters are relatively "stickier" due to the need to engage in non-routine problem solving in product design, often in collaboration with other firms. This requirement reduces opportunities to shift so-called "knowledge work" to distant locations.
      It is unclear whether similar locational patterns and underlying processes are visible in other industries. Most research on industrial clustering or agglomeration seek to explain the growth or success of cities or regions and use aggregate data covering multiple industries. The research summarized here focuses on a different level of analysis -- the geography of only one industry -- and disk drives may show a higher degree of technological clustering and greater international production than many other industries. It would be useful to conduct a similarly fine-grained study of the evolution of location in other industries.

Endnotes

• ¹I would like to thank Steph Haggard, Allen Hicken, Rich Carney and Roger Bohn for their contributions to this paper.
  
  
• ²I also have complete data for Japanese, European, Latin American and other Asian firms. For greater detail on the geographic pattern of these firms, see McKendrick and Hicken (1997).
  
  
• ³Internationally, disk drive design and production were mainly diffused through technology transfer arrangements with U.S. firms. NEC, Hitachi and Hokushin had technical tie-ups with Honeywell, Bryant Computer Products and Diablo Systems, respectively. In Europe, Compagnie Internationale de lÕInformatique had a license to make Control Data drives, and BASF acquired technology from Century Data Systems. In the 1980s, Brazilian firms entered through licenses from Seagate and Hewlett Packard, among others.
  
  
• ⁴Perhaps surprisingly, Oklahoma City has among the longest histories in the disk drive industry. When GE sold its computer business to Honeywell in 1970, its storage business was included. In 1974 Honeywell put all of its disk drive assets, including its Oklahoma City facility, into a joint venture controlled by Control Data. After Control Data assumed sole ownership of the joint venture, Oklahoma City became an important HDD center. Seagate bought Control DataÕs disk drive business in 1989, and Oklahoma City remains an important location for the company. It is the only U.S. location that does any HDD assembly other than prototypes and today is responsible for development of 3.5-inch disk drives intended for use in minicomputers, supermicrocomputers, workstations and file servers.
  
  
• ⁵As one Univac computer engineer who was on the receiving end of these drives wrote to me, "The 8414 [disk drive] became the only computer storage peripheral known for developing frequent oil leaks. There were a number of other mechanical unreliabilites which, combined with logic errors in early versions of the fantastically complicated controller which interfaced the byte-oriented discs to the 36-bit word mainframe and operating system driver bugs, caused the early adopters of these units an enormous amount of pain."
  
  
• ⁶By headquarters I mean the center of a firmÕs disk drive operations, not necessarily its corporate headquarters. For example, the core of IBM's disk drive operations are in San Jose, CA, not Armonk, New York; Hewlett-Packard's disk drive headquarters were in Boise, Idaho, not Cupertino, CA.
  
  
• ⁷IBM and International Memories both shipped an 8-inch drive in 1979, but it never became as dominant as the bigger disks or the smaller disks that followed.
  
  
• ⁸Europe was a larger market for Control Data than the United States during the late 1960s and early 1970s. Its highest volume during those years went to ICL, English Electric and Siemens.
  
  
• ⁹This is not to deny endogeneity in the technology development process, only that the commercialization of the smaller disk drives followed the introduction of the desktop computer.
  
  
• ¹⁰The bulk of the jump in share of unit shipments in the other category is due to Quantum's reliance on its Japanese manufacturing partner, MKE, which assembles a large share of Quantum's drives in Japan. Note also that shipments from Korea and China are included in the S.E. Asian share, but their unit shipments were quite small and so do not distort the overall picture.
  
  
• ¹¹This idea behind a technological cluster is similar to what Storper (1992) calls a technological district except that the unit of analysis is the industry rather than multiple industries that comprise the district.
  
  
• ¹²In fact, this part of Southern California was called "Floppy Valley" and became the center of America's floppy disk drive industry.