You are on page 1of 18

Journal of Urban Economics 67 (2010) 15–32

Contents lists available at ScienceDirect

Journal of Urban Economics
www.elsevier.com/locate/jue

The origin and growth of industry clusters: The making of Silicon Valley and Detroit
Steven Klepper *
Department of Social and Decision Sciences, Carnegie Mellon University, Pittsburgh, PA, United States

a r t i c l e i n f o a b s t r a c t

Article history: Data for all producers of automobiles and integrated circuits on their origins, base location, and perfor-
Received 7 May 2009 mance are used to analyze the factors behind the historical clustering of the two industries in Detroit and
Revised 10 September 2009 Silicon Valley, respectively. Key ideas concerning organizational reproduction and heredity are elabo-
Available online 17 September 2009
rated and used to explain how spinoffs from incumbent firms in the same industry can lead to clustering.
Findings concerning the spawning of spinoffs, entry by firms in related industries, and firm performance
JEL classification: suggest that organizational reproduction and heredity were the primary forces underlying the clustering
L2
of the two industries.
L6
R1
Ó 2009 Elsevier Inc. All rights reserved.

Keywords:
Clusters
Spinoffs
Competence

1. Introduction written about the rise of Silicon Valley, including the well known
book by Saxenian (1994) concerning the triumph of Silicon Valley
Arguably the two most impressive industry clusters in the his- over Route 128 and the recent history of the semiconductor industry
tory of the United States are the semiconductor industry in Silicon in Silicon Valley by Lecuyer (2006). They lay out a theory that reso-
Valley and the automobile industry in Detroit. Silicon Valley got its nates with modern theories of geography. Once semiconductor firms
name from the semiconductor industry and Detroit’s moniker as began to congregate in Silicon Valley after the emergence of Fairchild
the Motor City was derived from the automobile industry. At the Semiconductor as a leader of the industry, labor pooling, technolog-
start of the semiconductor industry in 1950, the population of San- ical spillovers, and a rich supplier industry stimulated further firm
ta Clara County, the heart of Silicon Valley, was .3 million people. In growth and entry of semiconductor firms in the Valley. The evidence
the next 30 years, nearly 100 semiconductor firms entered in Sili- that has been compiled about clusters is broadly consistent with the
con Valley, including five of the industry’s top 10 firms, and the importance of such agglomeration economies (Rosenthal and
population of Silicon Valley more than quadrupled to 1.3 million. Strange, 2004). While this was not the story told historically about
In its heyday, Detroit’s growth was even more impressive. During Detroit and the automobile industry (cf. May, 1975; Rae, 1980), no
the first 30 years of the automobile industry, over 100 automobile consensus has emerged around these historical accounts, which
firms entered in the Detroit area, including over half of the indus- has left the door open for explanations based on agglomeration
try’s leaders, and the population of Wayne County, the home of economies (Tsai, 1997).
Detroit, swelled from .3 to 1.8 million people. The main purpose of this paper is to bring together data col-
Such extreme industry clusters are rare (Ellison and Glaeser, lected and analyzed by Klepper (2007, 2008) for US automobile en-
1997) and call out for explanation, particularly when there is no trants and Klepper (2009) and Klepper et al. (2009) for US
obvious regional natural advantage underlying the clustering. Yet semiconductor entrants to compare the factors behind the geo-
there has been little systematic empirical analysis of the forces that graphic clustering of the two industries. These data include infor-
caused the semiconductor industry to be so concentrated in Silicon mation about the origins of the entrants, including whether they
Valley,1 and until recently the same could be said about the automo- produced other products prior to entry, and for new firms whether
bile industry and Detroit. Numerous articles and books have been they were spinoffs, defined as firms whose founders previously
worked for another firm in the same industry. Spinoffs have been
celebrated in the semiconductor industry (Lindgren, 1971; Saxe-
* Fax: +1 412 268 6938.
nian, 1994; Sporck, 2001; Lécuyer, 2006) and implicated by indus-
E-mail address: sk3f@andrew.cmu.edu
1
See Scott and Angel (1987), Fallick et al. (2006), and Ketelhöhn (2006) for three try insiders as key to the clustering of the industry in Silicon Valley
relevant analyses. (Sporck, 2001; Moore and Davis, 2004). Klepper (2007) argued that

0094-1190/$ - see front matter Ó 2009 Elsevier Inc. All rights reserved.
doi:10.1016/j.jue.2009.09.004

annual Electronics Buyer’s Guide (EBG). 7 Active module producers were first listed in 1962. Inexplicably some firms that produced ICs continuously based on other sources biles manufactured in the United States from 1895 to 1966. ture of conventional components and semiconductor substrates). electronics components. hopefully the questions will help frame future investiga. their founding histories in Kimes and Clark (1996) and whether tion of the Detroit automobile and Silicon Valley semiconductor they were listed in Smith as producing other products prior to clusters. and were not listed in some years as producers. are analyzed. it was also deter- analyze each industry is discussed. The EBG also listed transistor pro- ing on employee mobility? While definitive answers are hardly ducers (from 1949 to 1987). Silicon Valley firms was due to the disproportionate number that Data on US semiconductor producers were compiled from the descended from the leading firms and the greater propensity of Sil. grated circuits (ICs). entry and performance of auto and semicon. and if the main founder worked at a prior firm conductor industries. including the treatment of acquisitions (firms acquired by non-automobile producers were treated as continuing producers as were automobile producers that The analysis of both industries begins with their commercial acquired other automobile producers.4 former that got the spinoff process going in their regions. was used to determine the initial cap- firms in Detroit and Silicon Valley. were used to determine the producers of each product and their first The paper is organized as follows. and prior products of producers of all makes of automo- time. Consistent with the theory. that the spinoff process operated more intensively in De. Previously an attempt was made to separate these are raised and considered. startups. which lists the producers icon Valley firms to enter at the technological frontier. also provided for many different types of ICs (the types changed over ence of regional conditions on entry. Klepper / Journal of Urban Economics 67 (2010) 15–32 spinoffs were also key to the clustering of the automobile industry production of all makes a firm produced in Smith (1968). available. which 2 Buenstorf and Klepper (2009a) also feature the role spinoffs played in the appears to encompass producers of all kinds of circuits of assembled discrete clustering of the US tire industry historically around Akron. in the evolu. and active module7 producers (from 1962 to 1987). All other entrants were lumped into a residual category labeled offs. 2. of ICs from 1965 to 1987. Diversifiers are firms that added automobiles to their Key ideas concerning organizational reproduction and heredity product line or firms that were founded by individuals that previ- based on a theoretical model in Klepper (2008) are elaborated and ously headed pre-existing firms. and in semiconductors the superior performance of makes are listed in any given year). With spinoffs not venturing of American Manufacturers. In Section 7. 3 years before ICs. The main product each diversifier produced prior to automobiles was determined from Kimes and Clark (1996). with the acquired automobile producers inception. For each IC producer. The prior employer of the main founder is designated that are used to analyze the evolution of the automobile and semi. in autos. do the findings for semi. and were classified as diversifiers. why do spinoffs occur? Furthermore. the findings firms organized by the head of the pre-existing firm with the similar name or pre- are discussed and various theoretical and policy-related questions existing firms that modified their names to reflect an expanded product line when they diversified into automobiles. This was used to determine the first and however. this led to a buildup of superior been published since 1905.3 Spinoffs are firms with one or used to explain how spinoffs can lead to clustering. Most aggregated into three broad categories: monolithic ICs (all compo- fundamentally. Entrants around Detroit. including a few prominent ones. raising the specter of some kind of influ. spinoffs. 5 See Klepper (2008) for how the annual lists of automobile producers in Thomas’ Data on US automobile producers were compiled primarily Register was matched to Smith’s (1968) list. The evidence also indicates. as the spinoff’s parent. Firm entry producing ICs toward the end of the data period (1965–1987). two types of firms but they performed similarly (Klepper. the broad evolu- tion of the two industries and their clusters is reviewed. Some firms. better firms reproduced at a higher rate and their ducers and on the output of the largest producers. It was often difficult to tell whether they were new ductor firms. that firm is designated as a secondary parent of the spin- in turn influenced the rate at which its employees left to form spin. which provides a brief description of the available information. this superiority italization of producers. In Section 2. and the EBG listed producers of various types tance were key to their clustering. of each type of IC was recorded.6 Separate lists were troit and Silicon Valley. 3 Many firms entered the automobile industry with very similar but not exactly the In Sections 5 and 6. Data were also collected on the initial sizes of automobile pro- Subsequently. and film ICs (composed of layers of film on top of semiconductor conductors and automobiles pertain to other industries as well. which lists the names. In both regions. their main location was determined based on where their production was concentrated over and city).16 S. the superior perfor. The different types of ICs were are raised by the importance of spinoffs in the two clusters.5 The output of the leading producers of automobiles each year mance of firms in the Detroit area was due to the disproportionate was determined from Bailey (1971). semiconductors. off. and for simplicity they are combined here. and these were corrected based on the Kimes and Clark (1996). . which is dated as 1895 for automobiles and 1949 for treated as censored exits). The semiconductor industry clustered in the era of inte- industries suggests that organizational reproduction and inheri. The theory has more founders that previously worked at another automobile firm a number of implications regarding entry and firm performance on Smith’s list. Detroit and Silicon Valley each had an early exemplary per. of electronics products through 1987. the theoretical framework used to analyze the data is presented. after which it was no longer The role played by the spinoff process in the clustering of both published.2 The analysis focuses on the role of spinoffs. Ohio. 2007) for detailed procedures that were followed in compiling the data. and for each firm the first and last year of production and is the formation of spinoffs influenced by public policies bear. last year of production of every IC producer. 6 from Smith (1968). same names as pre-existing firms. which lists the annual number number that descended from the leading firms and entered at the of cars produced of the leading makes of automobiles (up to 20 largest sizes. In Section 4. Numerous other questions time as new ICs were introduced). respectively. and these lists tions concerning the emergence and growth of industry clusters. Data 4 See Klepper (2002. and some of the early producers of ICs were listed initially as active module producers. diode producers (from 1952 to 1987). size. an annual marketing directory that has far from their geographic origins. Thomas’ Register offspring were superior performers. the data used to and last year of production. substrates). or startups according to more broadly organizational competence and heredity. In Section 3. automobiles. hybrid ICs (a mix- performance of spinoffs and their ‘‘parents” related. which was used as a measure of entry manifested itself at the time of entry. base location (state Some IC producers had multiple locations listed in some years. 2007). a firm’s pre-entry on Smith’s list or a secondary founder worked at another firm on experience critically shaped its performance and its performance Smith’s list. that began founding conditions of every automobile producer. why is the nents made on doped semiconductor substrates). offs contribute to the growth of regions. were not listed as IC and exit dates of producers are based on the first and last year of producers and thus are not included in the analysis. how do spin.

Entry 1910 seven of the top 10 producers of automobiles were located in into the industry was concentrated in its first 15 years. simpler electronics products. 25% in the mid 1920s and then over 50% by 1941. to approximately 15 firms per year from 1911–1922. Nearly classified into (only) one of the four product categories based on this hierarchy. and number of automobile firms. Firms 200 State Total entry Diversifiers Startups Spinoffs NY 98 35 48 15 150 PA 52 13 28 11 100 IL 70 25 39 6 OH 89 35 38 16 50 MO 27 8 17 2 MA 55 15 36 4 0 IN 69 23 30 16 1895 1915 1935 1955 1975 MI 135 30 46 59 Fig. Evolution of the automobile and semiconductor industries area10 through 1941. followed Motors in 1908) it spawned the most spinoffs of any firm in by diodes. Eleven of the Semiconductor. Fig. This information was used to identify (the largest) ber of automobile entrants. and they were classified as in the firm that brought in Charles Sporck. 2. One other the leading makes of automobiles and data from the FTC (1939) on the total annual firm. active modules. and Klepper (2009) traced the backgrounds 1941 despite enormous growth in the industry’s output. MOS Technology. Klepper / Journal of Urban Economics 67 (2010) 15–32 17 300 Entry Table 1 Automobile entry by state and background for the leading eight states. This was compiled from annual data reported in Bailey (1971) on the output of Following general practice. increased in 1907.9 entrants were dispersed throughout the Northeast and Midwest. 20 a genealogy was compiled by the organization SEMI listing the foun. web searches. ordered by 250 Exit population.8 firms per year from 1901 to 1905 and then peaked at 82 firms area firms. None of the initial 69 entrants from 1895 to 1900 entered in the Detroit area. 2 plots the annual percentage of firms located in the Detroit 3. Entry remained high for the next 3 years and then dropped their dominance of the industry. Nearly all of these firms were spinoffs. General Motors. Automobiles years greatly understates the clustering of the automobile industry there. had a similar history to National and was classified as a spinoff production of automobiles. compiled annually the sales of mer- chant semiconductor producers whose sales exceeded a minimum only 15 firms entered through 1966. The 100-mile distance was chosen to reflect movement and Fairchild Camera and Instrument. which was a successful en- 8 gine producer.5 firms per year. 50 diodes. General Motors and its constituents had as many spinoffs over a longer period.8 It was not possible to trace comprehensively the backgrounds of all of the other IC producers. 10 ders of every semiconductor entrant in Silicon Valley between 1955 and 1986. From 1895 the Detroit area. which lists the market shares of the leading automo- The annual number of automobile entrants. dropping to 9 by by 1986 were identified. A total of 101 firms that entered at 272 in 1909. to region if they spent the majority of their years producing there. for example. A private consulting company. was the first great firm in the industry and in its The 5-year rule was used to exclude from diversifiers firms that entered with the intent of producing ICs but first produced other. defined as the prior semiconductor employer Fig. Percentage of automobile firms in the Detroit area. or other electronics products at least 40 5 years before they were listed as IC producers were classified as 30 diversifiers into ICs. and the genealogy 0 1895 1905 1915 1925 1935 1945 was used to identify the IC entrants in Silicon Valley that were spin- offs and their parents. spinoffs outside of Silicon Valley and their parents. Olds Motor Works. with Detroit area firms having a combined market to 1900.12 including three of the industry’s leaders. . which increased to share of 65%. led by Ford. Subsequently it fell sharply. exits. and then other electronics products. entering in 1901. While Michigan was the leading state. active modules. of the spinoff’s primary founder. the broad outlines of the evolution of area. 1895–1966. after which Much of the growth of the industry around Detroit was attrib- utable to spinoffs. of 92 of these firms using the Silicon Valley genealogy. which was located in Connecticut.1. and diversifiers were the industry. 1. Table 1 lists the leading eight states in terms of the total num- and other sources. 12 of Motorola (see Klepper (2009)). 1. the head of manufacturing at Fairchild. 1895–1941. 11 reconstitute its efforts in Silicon Valley. Firms that were listed in the EBG as producers of transistors. 60 The pre-entry backgrounds of the IC producers were also traced. reaching around industry around Silicon Valley are described. Entry. The short life as an independent firm (it was acquired by General closest product to ICs in terms of technology and market was the transistor. with the first entrant in the Detroit Following Klepper (2009). Table 2. The share of automobile firms in the Detroit area in the early 3. National was classified as a spinoff of Fairchild. Fairchild. entry averaged 11. Subsequently the percent- the automobile industry around Detroit and the semiconductor age of producers in the Detroit area rose steadily. all the rest of the later entrants in the Detroit area that became 9 A few of the firms classified as spinoffs were financed by non-semiconductor firms (and sometimes organized as subsidiaries) or involved a reconstitution of an 10 existing semiconductor firm in which the new ‘‘founders” were given an ownership Firms were classified in the Detroit area if they located in Michigan within interest.11 indicates that by from 1895 to 1966 based on Smith (1968) is plotted in Fig. effectively giving birth to a new firm. Olds Motor Works. National branching of firms within approximately a 100-mile distance of Detroit. exit. Inte- grated Circuit Engineering (ICE). The number of firms peaked threshold for the period 1974–2002. This share rose further after 1910 as the leading Detroit 36. a Long Island military contractor. % Det Firms mined when they first produced any electronics product listed in the 70 EBG. was an example of a reconstituted entrants moved in or out of the 100-mile region. For Silicon Valley. and later Chrysler. and producers bile producers every 5 years from 1900 to 1925. was financed by and later became a subsidiary of 100 miles of Detroit. S.

MI 2 1 2 1 Hupp 1909 Detroit. MI 3 1 2 7 Paige-Detroit 1909 Detroit. from 1953 to 1973. Un. MI 4 1 Brush 1907 Detroit. Entry. which firms entered into the production of transistors. and startups. illustrates the importance of spinoffs in its. and producers from 1949 to 1987 based on the EBG. Early entrants Entry year Entry location 1900 1905 1910 1915 1920 1925 Pope 1895 Hartford. 3. CT 18 Knox 1900 Springfield.3 Packard 1900 Warren. Fig. CT 36 Stanley 1896 Watertown. MI 4 4 2 Stoddard 1904 Dayton. 1949–1987. NY 3 Detroit-area firms 0 58 65 83 52 85 leaders of the industry (see Table 2) were spinoffs. IN 2 Ford 1903 Detroit. as reflected in breaks down the entrants in the leading states into diversifiers. 3. ex- spinoffs. WI 16 2 3 Later entrants Studebaker 1902 South Bend. IN 9 10 6 6 Reo 1904 Lansing.2. 4 reports the fraction of transistor producers in four consol- idated metropolitan statistical areas: Boston.02 4 Olds/GM 1901 Detroit/Lansing. The number of producers grew steadily to 90 by and Indiana with 16 each. MI 5 7 5 Dort 1915 Flint. respectively. exit. Entry Exit Firms 120 100 80 60 40 20 0 1949 1954 1959 1964 1969 1974 1979 1984 1989 Fig. Los Angeles. Silicon Valley had no producers before 1955 and no more tors only for its own use and the government market. MI 6 Oakland/GM 1907 Pontiac. Franklin 1900 Syracuse. IN 8 5 3 4 Anderson/Union 1902 Anderson. entry was fairly steady. MI 2 2 1 H. NY/Detroit. After the Michigan versus the other leading states. MI 1 6 12 Saxon 1913 Detroit. and then it dropped to 7. In contrast.8 firms per year from the next closest states in terms of number of spinoffs were Ohio 1974 to 1987. NY 4 White Sewing Machine 1901 Cleveland. Semiconductors York. three cities. which presents the annual number of transistor entrants. Fig. Thomas-Detroit/Chrysler 1906 Detroit. MI 1 Durant 1921 New York. MI 26 1 2 1 Cadillac/GM 1902 Detroit. OH/Detroit. Numerous than 8% of the producers through 1960. Klepper / Journal of Urban Economics 67 (2010) 15–32 Table 2 Market shares of leading US automobile firms. OH 1 E. entrants. MI 3 1 1 3 Hudson 1909 Detroit. Producers concentrated early around the first scientists and effectively started the semiconductor industry. OH 2 Dodge Brothers/Chrysler 1914 Detroit. . MI 3 17 5 6 5 Willys 1903 Terre Haute. MA 2 Locomobile 1899 Bridgeport. Michigan had a total of first few years. Table 1. AT&T liberally disseminated its know-how ducers and Boston and LA around 15% by the latter half of the and licensed its transistor patents and agreed to produce transis. MI 2 Chandler 1913 Cleveland. averaging 11 firms per year 59 spinoffs that constituted 44% of all of its entrants. 1900–1925. MI 7 18 56 22 44 Maxwell Briscoe/Maxwell/Chrysler 1903 Tarrytown. MI 3 6 5 2 4 Buick/GM 1903 Flint. with New York accounting for around 40% of the pro- der antitrust pressure. and San Francisco. and number of transistor firms. MA 0.R. where the latter region is primarily com- posed of Silicon Valley firms (and hereafter is referred to as the Sil- The transistor was invented in 1947 by three Bell Labs (AT&T) icon Valley area).H. OH 0. 1950s. MI 1 Chevrolet/GM 1911 Flint. New 3. MI 16 6 2 1 1 Jeffery/Nash 1902 Kenosha.18 S. which constituted 18% and 23% of their 1975 and then leveled off.

Subse- to 1987 based on the EBG. 5 presents IC producers were concentrated in New York. 5. S. which increased further to over 23% by 1987. %BOS %LAX 40 %NYC 35 %SFO 30 25 20 15 10 5 0 1965 1970 1975 1980 1985 1990 Fig. steadily rose and by 1979 Silicon Valley was the leading area with quently entry dropped to an average of 20. 6. New York. San Francisco. driven by the co-production of transistors and ICs by IC entrants. Klepper / Journal of Urban Economics 67 (2010) 15–32 19 %BOS %LAX 100 %NYC 80 %SFO 60 40 20 0 1949 1954 1959 1964 1969 1974 1979 1984 1989 Fig. Boston. which was first commercially produced in Los Angeles. Along with Texas 1980. 6. San Francisco. exit. each of which contained around 20% of the producers. . reaching a high of 210 in 1987. Los Angeles. Fig. The first notable semiconductor producer in Silicon Valley was 1987 and the number of firms leveled off until it grew again after Fairchild Semiconductor. Percentage of transistor firms in Boston.13 Subse. Fig. and San Francisco areas. and producers from 1965 ton. 1965–1987. Los Angeles. which reports the share of IC producers in the New York. Percentage of IC firms in Boston. Entry. 4 indicates that the share of transistor producers 13 in the Silicon Valley area also grew after the advent of ICs. Los Angeles. it pioneered the silicon transistor and then the inte. exits.7 quently the percentage of producers in the Silicon Valley area firms per year and the number of producers grew to 154.9 firms per year through around 20% of the IC producers. Fig. and Bos- the annual number of IC entrants. largely The sharp drop in the number of firms in 1970 corresponds to a change in the categories of ICs listed. which entered in 1957. Entry Exit Firms 250 200 150 100 50 0 1965 1970 1975 1980 1985 1990 Fig. indicates that at first 1961 and eventually took over much of the industry. New York. grated circuit (IC). Instruments. and number of integrated circuit firms. From 1965 to 1973 entry averaged 39. 4.

and the ing semiconductor producers from 1957 to 1990. TX 2 6 A Micron Technology 1978 Boise. NJ 6 7 5 7 4 3 2 Raytheon 1951 Boston. as reflected in the Silicon Valley genealogy. CA 11 5 – – C C C C General Instrument 1955 Long Island. ules. 1960. 1980. 61% in LA. 61% in Boston. MA 12 9 3 3 0.5 2 VLSI Technology 1979 San Jose. c Includes Raytheon. and 1966 and the ICE sales data for subsequent years. as will be discussed further below. entrants in New York. a Dates for receiving tube firms and early leaders based on Tilton (1971). CA 2 5 7 6 Mostek 1969 Dallas. remaining IC producers. 1985. CA 10 11 10 9 Harris 1967 Melbourne. p. 1960.20 S. CA – – 4 – C C C C Hughes 1955 Los Angeles. 1963. Receiving tube firms Entry yeara Metropolitan location 57 60 63 66 75 80 85 90 General Electric 1951 Syracuse. CA – 1 2 LSI Logic 1980 Milpitas. through its own cludes all the firms that were determined to be spinoffs and the growth but even more importantly as the source of many of the sub. or other electronics products versus 57% of the IC entrants offs. This re- flects both the paucity of prior electronics producers in Silicon Val- 14 This was compiled from market share data reported in Tilton (1971.14 By 1975 five of rest of the US are broken down according to whether they pro- the top 10 semiconductor producers were located in Silicon Valley duced transistors. Transistors Diodes Active modules Electronics Other firms Total Boston 5 6 9 11 48 79 Los Angeles 7 1 8 24 62 102 New York 8 6 12 27 71 124 San Francisco 4 2 1 10 68 85 Other 16 8 21 58 130 233 Total 40 23 51 130 379 623 Similar to Detroit. and 1966 market share data. active mod- Fairchild was responsible for an extraordinary number of spin. NY 9 8 8 8 C C C C RCA 1951 Camden. Table 3 lists the periodic market shares of the lead. which increased to 48% five years later. PA 3 6 4 3 Other early leaders Texas Instruments 1953 Dallas. MA – 1 1 2 2 AMI 1966 Santa Clara.5 Delco Radio (GM) 1956 Kokomo. MA 5 4 – – 1 1 1 0. Much of this residual category labeled ‘‘other firms. active modules. TX 20 20 18 17 20 19 18 15 Transitron 1953 Boston. and 1990 market shares are based on annual compilations of ICE. with the rest of the IC producers placed in a of the industry. 1963. Sources: see Tilton (1971) for sources for 1957. ID – 0. 1957–1990. diodes. CA – – 5 6 5 Analog Devices 1965 Boston. the 1975. Among the other four leading Silicon offs but whose background could not be determined. IN – – – 4 C C C C Fairchild 1957 Mountain View. or other electronics and collectively Silicon Valley firms accounted for 43% of the output products before ICs. three were spinoffs from Fairchild and the veys a clear message: 80% of the IC entrants in the Silicon Valley fourth was a second generation descendant of Fairchild. the share of transistor and IC firms in Silicon the backgrounds of IC entrants in the different regions in the US to Valley greatly understates the clustering of the semiconductor understand the effect spinoffs had on Silicon Valley. FL 2 3 3 4 Intel 1968 Santa Clara. C: Captive producer in the listing of Integrated Circuit Engineering (ICE). CA – 5 9 13 9 7 5 A Motorola 1958b Phoenix.5 Sylvania 1953 Boston. Table 4 Backgrounds of IC entrants by region. PA 2 6 4 5 C C C C Philco-Ford 1954 Philadelphia. Los Angeles. CA 7 10 10 17 AMD 1969 Sunnyvale. which was based in Silicon Valley as of 1975. CA – 2 3 Silicon Valley Share Leading firmsc 0 5 9 13 38 42 42 38 Leaders + other ICE firmsc 43 48 49 47 –: Firm was producer.” The latter category in- growth was driven by Fairchild Semiconductor. In Table 4. but no market share data reported. area were not prior producers of transistors. It is instructive to consider in New York. IC industry there. A: Acquired by a semiconductor producer. San Francisco. MA 4 3 – – Westinghouse 1953 Pittsburgh. process there. Motorola used semiconductors only for its own purposes before 1958. Table 4 con- Valley firms in 1975. b According to Tilton (1971). NY – – – 4 3 2 1 0. CA – 4 2 1 1 National 1967 Santa Clara. Klepper / Journal of Urban Economics 67 (2010) 15–32 Table 3 Market shares of leading US semiconductor producers. 66) for the ley before the advent of ICs and also the richness of the spinoff years 1957. . many of which may also have been spin- sequent leaders of the industry. diodes. Boston. AZ – 5 10 12 8 11 13 17 Later leaders Signetics 1961 Sunnyvale.5 TRW 1954 Los Angeles. and 56% elsewhere.

with spinoffs playing a key role in the Entrants have a home region. it follows that: as being an H firm in the new industry depends on the firm’s ability Proposition 2. the average fined over the interval [1/2. (L) competence. for which pd is greater). The brief accounts of the evolution of the automobile and semi. It is assumed that than L firms and only diversifiers that are H firms in their original for a diversifier to be an H firm in the new industry it must be an H industry and spinoffs of H incumbents can be H firms in the new firm in its own industry. their lack of organizational and industry experience. For simplicity. It is assumed that the more relevant a diversi- fier’s industry to the new industry. and (d) spinoffs from H versus L incumbents. For sim- by studies of spinoffs in a number of industries (Agarwal et al. The ideas are also used to derive various predictions that will edge about their home region. in every period a smaller percentage of H than L firms are at risk of centrally involve a firm’s ability to manage technological change. with q(0) > 0 and related an industry is to the new one (i. Then t peri- high competence depends on the ability of the spinoff founder to ods after entry. and (other) startups. (b) diversifiers that are H versus L firms in their own otherwise. Among contemporaneous entrants. competence would Intuitively. To explain differences in the length of firm survival in the new Spinoffs can exploit knowledge about the new industry that industry. let the fraction of these firms of type k that survive to the 2004. H potential entrants have a higher probability factor that is assumed to be drawn from a uniform distribution de. reflecting conductor industries indicate that the composition of entrants var. entry experience are distinguished: diversifiers. von Rhein. For diversifiers this is where they clustering of both industries.. Hence at the time of entry. 2006. with probabilities p. beginning of period t equal its expected value of at. 1/2]. Klepper / Journal of Urban Economics 67 (2010) 15–32 21 4. 2009a). Potential entrants enter if their prof. where their parent firm was located). Various E(lt) = 0. denoted as q(). firms experience a permanent additive shock lt ent” firm. Analogously. where it is assumed that gt < PL + 1/2. startups. though. which is also the same for firms of either type. and PL and PH are normalized and maximum entry size is greater for H than L firms. and (1  2p) respectively. Theory The last group of entrants. worked (i. For simplicity. plicity. It is assumed that their founders gained while working in the industry at their ‘‘par. 2006. greater expected number of spinoff entrants than L firms in the new Three types of entrants into a new industry in terms of their pre- industry. it is assumed that for same for firms of either type. g > 0. for spinoffs it is where their founders ing organizational competence based on a model of industry evo. 2006. which implies that the hazard of exit is probability that a spinoff of an H incumbent firm will itself be an greater for L than H firms. it is assumed that serve as a basis for testing the theory. the larger pd) then the q0 > 0.e. the output of entrants of type k is greater the probability that firms in the industry enter the new uniformly distributed over the interval [q(0). the new industry. assumed that entrants have valuable economic and social knowl- ters. Klepper and Thompson. of type k = L or H. They are all assumed to be L firms in the new industry. There is assumed to be a 1–1 mapping between the size firms in the same industry to enter the new industry.. It is prominent shared features of the Detroit and Silicon Valley clus. is composed of new firms founded by individuals without experience in the new industry. q(Pk + 1/2)]. industry. sifiers that are H firms in their original industry and spinoffs of H its are nonnegative. Accordingly.e. industry. Cassiman and firms at risk of exit are those that had profits at the time of entry Ueda. and spinoffs of H incumbents enter at a greater probability that an H firm in another industry will be an H firm in maximum and average size than spinoffs of L incumbents. (c) diversifiers from more related industries (i. where e is a idiosyncratic framework. its parent (in the new industry) must survivors with profits less than or equal to g equal its expected va- be an H firm. exit. g. to their profits and exit if their profits fall below 0. Only diver- such that 1/2 < PL < PH < 1/2. At the time of entry. Spinoffs are expected to inherit traits industry being either diversifiers that were H firms in their original from their parents. a mechanism to induce exit is needed. . in every period t. Let s denote the probability that firms in the industry or spinoffs of H incumbent firms. the profits of potential entrants A number of results about entrants follow directly from this with competence k = L or H equal Pk + e. industry. it follows that: new industry have some particular trait. Let pd denote the greater than startups. ied greatly across regions. 16 17 In high-tech industries like semiconductors and automobiles. or of such employees that can found spinoffs and each has the same 0. as being lue of bt. Diversifiers are assumed to be either high or low Since the average and maximum entry size of H firms is greater competence producers in their original industry. This implies that the probability of entry of po. In this section a few key ideas regard. on average the probability of a spinoff having the trait is greater than s if its parent hazard of exit in each period is lower for: (a) diversifiers and spinoffs had the trait (when the spinoff was founded) and less than s than startups. Klepper. potential entrants into mance—for example. incumbent firms can be H firms in the new industry. 2009).e. A key component of the theory is that firms differ innately in Otherwise the location of firms has no effect on their perfor- terms of their competence. it tential entrants of type k is Pk + 1/2 and the profits at entry of follows that (see the Appendix for proofs of all propositions): entrants of type k are uniformly distributed over the interval Proposition 1. Pk + 1/2]. though. let the fraction of these a spinoff to be an H firm. produced in their industry.. ceteris paribus.17 Coupled with H entrants in a new H firm in the new industry. (b) The more of firms at entry and their profits. the greater the value of pd. spinoffs. and for startups lution in Klepper (2008) are laid out and used to explain it is where their founders previously worked and/or resided. which is the Buenstorf and Klepper. Franco and Filson. 15 Buenstorf and Klepper (2009a) used a similar approach to analyze the historical clustering of the US tire industry. First. 2008. and (c) H firms in the new industry spawn a The competence of firms is based on their pre-entry experience. This is only a necessary condition. Therefore. Consider the hazard of exit t periods after entry of firms theories of spinoffs predict that more competent firms spawn bet. The only ter-performing spinoffs (Franco and Filson. high (H) and low number or market share of firms in its home region. which is supported less than or equal to gt and that are still in the industry. The maximum and average entry size of spinoffs is to transfer its experience into the new industry.16 For simplicity. (a) H firms in another industry are more likely than L [0. it is assumed that every firm has the same number it is assumed that lt can take on three possible values. For simplicity. 2007. S.15 all entrants locate in their home region to exploit this knowledge. Spinoffs are typically formed by high level employees. It is assumed that the Proposition 3. so that probability of leaving to form a spinoff in any given period. the hazard of exit of firms of type k equals patbttg/ exploit his or her experience at the parent firm. a firm’s profitability is not affected by the a new industry are assumed to come in two types.. Let ps denote the [Pk + 1/2  (1  at)tg]. This is only a necessary condition. of entry than L potential entrants and among entrants. p.

In other regions. over dominates the distribution for spinoffs outside of the Detroit time the percentage of firms and industry output accounted for area and for startups. but DLj > 0. tested using the data that were collected for the automobile and Aj > 0 for j – 1 and diversifiers and startups (all of which are entrants: low competence) enter in other regions. Furthermore. Consider first the predictions concerning spinoffs. so the fraction of entrants that are firms spawn spinoffs is no different in Detroit than spinoffs is greater in region 1 than elsewhere (feature (3)). suppose DL1 = A1 = 0. Fujita and Thisse. the proposed theory implies that tionate number of spinoff entrants. Let there be j = 1. and the H spinoffs in turn offs in the Detroit area having parents located there. the maximum profits of firms of number of which became leaders of the industry. . 1991. should also hold: (6) Firms in the Detroit area survive longer than firms else- where. and engine industries survive Detroit and Silicon Valley clusters. the share of the industry’s firms and output and this would only be because on average they had more compe- accounted for by both regions increased. In contrast. spinoffs. this dominance should be confined to the spin- ture (5)). er hazards of exit. suppose that high competence is such an advantage that after a certain point in the (1) The leading firms. the rate at which ersifiers. as featured ing automobile make (through 1924) are listed along with whether . Duranton and Puga. tent parents. Nearly all the (2) In each period. the data that were collected. (3) After controlling for the quality of firms. gions shared five notable features: 2002. carriages and wagons. entrants are a mix of div. a for all types of firms. Krugman. Furthermore. To explain these five patterns. and let Dkj denote the number of firms of type k = L or H in re. . regions. which produced engines for a firm of type k = L or H. For each firm. how the theory can explain the most notable aspects of the evolu. with the spin- spawns both H and L spinoffs in region 1. 2. bicycle. These re. try rates and lower hazards of exit if they were located in a cluster.22 S. DHj > 0. The three proposi. spawn H and L spinoffs. Suppose DH1  0 but by chance this are carriages and wagons (65 firms). Suppose pd is very low and only one diversifier attains high of 224 diversifiers. the following two predictions longer than other diversifiers. leading both regions to have a dispropor. bicycles. Krugman clustering that characterized Detroit and Silicon Valley. otherwise. the industry evolves as follows. and more nar- rowly only for the spinoffs of the leading firms. There are too many parents to list them all. Let M denote the firm’s additional profits if it is located in (2) Both had an initial entrant that became an early leader of the a cluster. where the top competence in the new industry. . the total Alternatively. Subsequently. and no gines (22 firms). with the longer survival confined to spinoffs in the (1) The size distribution of entrants in Detroit and Silicon Valley Detroit area and in particular to the spinoffs descended from should dominate the size distribution of entrants in other the leading firms that entered at the largest sizes. trated in the Detroit area. where region 5. and as will type k would be Pk + 1/2 + M if they were located in a cluster and be seen a number of their spinoffs in turn were fertile Pk + 1/2 otherwise. The predictions of the theory that can be tested are dictated by try. 2004). first. Suppose for simplicity there is only one industry that supplies diversifying entrants to the new indus. They are composed there. Belleflamme et al. only spinoffs and not diversifiers or startups would have higher en- (4) Both regions had firms of above average size. and startups. offs in the Detroit area that descended from the leading This is of course an exaggerated account to illustrate simply firms. and Venables. Further. Klepper / Journal of Urban Economics 67 (2010) 15–32 The theory can now be used to provide a simple account of the in modern theories of geography (cf. with 714 entering by startup entrants in region j based on the level of economic activity 1925. it entrants in the Detroit area than elsewhere. J regions.. (5) Diversifiers and spinoffs survive longer than startups. Then industry—Olds Motor Works. and Fairchild Semiconductor. and en- one firm enters in region 1. where it is assumed that 1/2 < PL < PH < 1/2  M. have the highest value of pd). in the other regions are L firms. suppose that being located in a cluster provides number of its spinoffs and the number that ever produced a lead- firms with an advantage that increases their profits. 2000. the probability of entry would equal before automobiles. A total of 725 firms entered the auto- gion j in that industry. Last. which Pk + 1/2 + M if the firm was located in a cluster versus Pk + 1/2 produced transistors before ICs. Let Aj denote the number of potential mobile industry between 1895 and 1966. whether in or outside the Detroit area. A total of 96 firms have lower hazards of exit than firms elsewhere. with this holding only for spinoffs. then diversifiers from the car- When these propositions are applied to the explanation for the riage and wagon. the firms in Detroit and Silicon Valley should spinoffs entered in the period 1899–1924. there are no firms in region 1 (feature (1)). and 348 startups. following Klepper (2009) the 28 that spawned two or more spin- offs in 1899–1924 are listed in Table 5. with 68 spawning will hold only for spinoffs. and if tion of the Detroit and Silicon Valley clusters. and this firm does not enter the three products produced by the diversifiers prior to automobiles new industry when it begins. (5) Over time. ward. by the firms in region 1 rises and the industry clusters there (fea. 142 spinoffs. but this spawned one or more spinoffs in this period. Hence all else equal. . so the firms in region 1 on average (4) The distribution of entry sizes for spinoffs in the Detroit area are larger than the firms in other regions (feature (4)). a stylized account based on the theory is employed. At highest rate. The first H firm (2) The fraction of entrants that are spinoffs is greater for in the industry locates in region 1 (feature (2)). which are disproportionately concen- industry’s evolution only H firms survive. so firms of all types in clusters would have low- sources of spinoffs. The implications of this in the context of the simple framework laid out above are straightfor- (1) Neither region had many entrants at first. All firms elsewhere. but spinoffs of the leading firms that enter at the largest sizes. Last. Automobiles 1 is Detroit and Silicon Valley. and more narrowly only for the only one spinoff. and engines are considered tions summarize more generally patterns that should be the three most related industries to automobiles (and thus expected if the theory underlying the stylized account is correct. All tests are confined to these 714 firms. bicycles (26 firms). entry would be greater in clusters (3) Both Olds and Fairchild were the source of many spinoffs. 1995. The following predictions of the theory can be other diversifier or startup enters in region 1. spawn spinoffs at the Under these assumptions.

C. so the period examined is ended at 1924. The coefficient estimates of age and age squared the firm itself ever produced a leading automobile make. Klepper / Journal of Urban Economics 67 (2010) 15–32 23 Table 5 duction began or 1899. diction.H. not only did it have the most spinoffs along greater around when firms were acquired by either auto or non- with Buick/GM (in a shorter time interval). respectively. as exemplified by the spinoffs in the Detroit implies (roughly) a 2. at the . ing the top two makes had the highest spinoff rate (reflected in the sum of the two coefficient estimates) followed by firms producing any leading make.77 quantifies how much greater the odds ratio is for firms in by General Motors. initially financed Buick (see Klepper (2007)). Ransom Olds.20 Consistent with the second pre.23 It has been found Northern 1902–1910 3 1 that firm age and whether a firm was recently acquired affect the Hupp Motor Car 1909–1924 3 Yes firm spinoff rate (Klepper and Sleeper. which had little effect on the estimates. Packard 1900–1924 2 Yes age.6. played a key role in the success of Cadillac and Ford Motor Co. Since the annual probability of spawning a spinoff is quite low. who had initially worked for Olds The coefficient estimate is the derivative with respect to being located in Detroit before moving to Ford. 18 22 No spinoff occurred in 1925. respectively.. They imply a maximum spinoff rate at age 17. and another one of included in the analysis. although alternative explanations for this Following Klepper (2007).19 Olds Motor Works was within the sample range. Eight 20 spinoffs were founded more than 5 years after the exit of its parent and thus were not Its two main subcontractors. indicating that firms produc- a Classified in Detroit area if majority of years of production there. Firm Years (through # # Leading Leading The dependent variable equals 1 in a year in which a firm has one or 1924) Spinoffs spinoffs firm more spinoffs22 and 0 otherwise. the top five parents and seven of the at the . the coefficient estimates of the two Ideal 1911–1924 2 variables pertaining to producing a leading automobile make are Biddle 1915–1922 2 both positive and significant at the . The probability of a spinoff is significantly especially influential. the Detroit area. 1916. after Olds were all related to Olds. this leading one observer of the industry to describe its leader. the coefficient estimate area versus 15% of the entrants elsewhere. but the top six parents auto firms at the . automobile firms spawned spinoffs is used to formally analyze the spinoff process. Hupp was founded by Robert Hupp of Ford.24 A 1–0 dummy vari- Duryea 1896–1907 2 Yes able equal to 1 for years after a firm exited is included to test F. 2005). after production ceased (through 1924). as might be expected.25 This is inconsistent with the third prediction. Blomstrom 1903–1909 2 both are based on the total number of years of production). with the former significant tent with the first prediction. Stearns 1898–1924 2 Yes whether the spinoff rate declined after exit. To test the effect of Co.01 levels. Chevrolet 1911–1916 2 Yes two 1–0 dummies for acquisitions by automobile and non-automo- Saxon 1913–1922 2 Yes Hupp Corp. the number of years a firm produced automobiles and its square Jackson 1902–1918 2 were included as explanatory variables (for years after a firm exited. 1911–1916 2 bile firms are included based on data in Smith (1968) on ownership changes. Standard errors are corrected by clustering the observations for Lozier 1904–1915 2 1 each firm. and continuing 5 years Spinoffs of automobile producers. and a 1–0 Maxwell Briscoe/ 1904–1924 4 Yes dummy equal to 1 if the firm had produced the number one or Maxwell two make in the current or preceding 5 years. 44). 25 Last. who also Forty-six firms exited by being acquired by another automobile firm and there co-founded Maxwell-Briscoe. including six with two 19 See Klepper (2007) for a genealogical tree encompassing the spinoffs of all of spinoffs and one with three spinoffs. It is consistent there and all but 10 of the 59 spinoffs with parents in the Detroit with clustering increasing firm profitability (as featured in modern area located there. which is the basis for the interval considered. Last. There were 126 firm-years with one or more spinoffs. Consistent with the first prediction and the pat- Single Center 1907–1909 2 terns reflected in Table 5. To test Imperial 1909–1917 2 whether spinoff rates are higher around the time of acquisitions.01 level and the latter falling just short of significance at the top eight all produced leading automobile makes and the top seven .” (Doolittle.10 level. the observations with multiple spinoffs. To test whether better firms had Detroit-area producersa higher spinoff rates.21 All firm-years are pooled. are positive and negative. exp{1. a 1–0 dummy equal to one for firms located in the Chocolate Detroit area is included to test if clustering affected the firm spinoff Stoddard 1903–1910 2 Yes rate. Olds Motor Works had a great impact on the industry. A number of other well known individuals in the industry also of the log of the odds of a firm spawning a spinoff relative to not spawning a spinoff. p. 21 A number of spinoffs were founded after the parent firm exited (generally within 5 years of its exit date) and two were formed before its parent began production. theories of geography). as the translates roughly into Detroit firms having a 2. Berg 1902–1906 2 Year dummies are included to reflect variations in entry conditions Jeffery 1902–1924 2 Yes Metz/American 1903–1923 2 Yes over time. The number of top makes to include in the latter variable was chosen based on fit. a logit analysis of the rate at which pattern are also considered later.77 higher probability of spawning a ‘‘schoolmaster of motordom. All told. Last. making Olds a secondary parent of Maxwell-Briscoe.01 level (for any leading make) Barley 1916–1924 2 and . spinoff than firms elsewhere. 23 Olds’ subcontractors.02} = 2. Leland and Faulconer and the Dodge Brothers. spinoffs accounted for 48% of the 112 entrants in the Detroit in the 5 years after a firm has exited.05 and .05 level (for the top two makes). S.B. 24 Northern was a spinoff of Olds that was co-founded by Jonathan Maxwell. the probability of a spinoff is significantly lower. . which is parents were all located in the Detroit area.10 level.77 greater spinoff rate for firms in the Detroit area—49 of the 54 spinoffs in the Detroit area had parents located area.18 Consis. worked for Olds during its brief life as an independent producer before being acquired Therefore. were 120 acquisitions by non-automobile firms. Spinoffs generally located of the Detroit dummy is positive and significant at the . two explanatory variables are constructed based Olds 1901–1908 7 3 Yes on the annual list of leading automobile makes reported in Bailey Buick/GM 1903–1924 7 2 Yes (1971): a 1–0 dummy equal to 1 if the firm had produced a leading Cadillac 1902–1908 4 3 Yes Ford 1903–1924 4 2 Yes automobile make in the current or preceding 5 years. Each equals 1 in the year a firm was acquired and in the Non-Detroit area producersa Haynes Apperson 1895–1924 2 year before and 2 years after the acquisition.01 level and close to their parents. Not surprisingly. York 1905–1917 2 The coefficient estimates of all but the year dummies are re- Palmer and 1907–1914 2 Springer ported in Table 6. Each firm’s lifetime as an automobile producer is broken into annual intervals starting with the year before pro. whichever is later. Benjamin Briscoe. An ordered logit was estimated to accommodate these firms.

17. These The coefficient estimates for this model are reported in Table 8 percentages are reported separately for startups in the Detroit under the column headed Model 2.23)* troit dummy.02 (0.” and an additional cat. Consis- Acquired by auto firm 0. and c and b are coefficient vectors.05 level.10 level. and dummies Register. suggesting that if anything the opposite was true. lower annual hazards than the omitted group of start- size categories. bicycle. the top three size categories).061.38)** Acquired by non-auto firm 0. Among for the diversifiers from the carriage and wagon. with standard errors in parentheses.88 (0.01 level. ard of exit model over the period 1895–1966. Klepper (2008). The coefficient estimate for the Detroit spinoffs is spinoffs with parents that ever produced a leading automobile also less than one and significant at the . The comparison is similar if extended to the of entry variables are less than one and significant (at the . so that numbers Variable Coefficient estimate below (above) one indicate a reduction (increase) in the hazard rel- Leading make 1. Additionally. a much up entrants. Klepper (2002. The second fect the hazard proportionally at all ages.001. affects the hazard.. entered at the highest three size categories. entered in these lower still if they came from the carriage and wagon.38)** An initial version of the model was estimated with a single var- Years of production 0. ceteris paribus.42 (0. respectively.01 level. denoted as Largest Top Spin- hit ¼ expðfc0 þ c0 zit gageit Þ expðb0 þ b0 xit Þ.6% of the Detroit gine industries.03 (0. the spinoffs not in the Detroit area. The estimates imply that diversifiers and spinoffs had 37% and 34%. a 1–0 dummy equal to 1 for firms located in the Detroit Years of production squared 0. was added in xit for diversifiers that previously produced carriages and wagons. are respectively 0. . The two-tailed p-values for the respective compar. as might be expected if equilibrating forces diminished any advantage conferred by being located in the Detroit area. Detroit area was confined to the spinoffs located there. with a still greater 26. 1905–1909.003) area. 30 the Detroit area and elsewhere. All of the coefficient estimates area and elsewhere. This was not due to the introduction of controls for the time of entry and firm isons are . . bicycle.90 (0. 2008) Last. lowered the hazard only at older ages.05 level. offs. 29 used to test whether the probability of entrants starting with a capitalization of There are no a priori predictions about the functional form of the relationship $300. and the annual hazards of the diversifiers were 30% higher percentage of the Detroit spinoffs. . backgrounds—when these variables were included without the dummy for the 27 This is consistent with the model of industry evolution developed in Klepper Detroit spinoffs.56 (0.01 and .4% of the non-Detroit spinoffs entered in these three respectively. equals 1 for the terms. Top 1 or 2 make 0. coefficient estimates. ups. ageit. the coefficient estimate of the Detroit the ones that descended from the leaders. Model 1. for the first two cohorts were included in both xit and zit. Next dummies were introduced in xit for diversifiers and spin- ** Significant at the .17 (0.26 levels) only in the interaction with age.3%.28 * Significant at the .72 izations below $300. bicycles. Furthermore. is less than one and significant at the .24)*** tent with the sixth prediction. or engines to test if they had lower hazards than other diversifiers.23)*** labeled Model 1.034. The first variable. entrants were broken into three cohorts of roughly off entrants that entered in each of the size categories in Thomas’ equal size: 1895–1904. c0 and b0 are constant variable.18 area.85 (0. Following ported in Table 7 on the percentage of various startup and spin. Similarly. The 26 Assuming firms in the unknown and unobserved categories had initial capital.05 top five size categories.” for entrants not found in Thomas’. ‘‘unobserved. confirming the make entering in these three size categories. coefficient estimates are reported as hazard ratios. spinoffs in the Detroit area and elsewhere. which is reported in Table 8 under the column Detroit 1. 3. and the startups in well.01 level. implying that earlier entry The last two predictions are tested by estimating an annual haz.26). reflecting The Detroit dummy was arbitrarily divided into two dummies covering the that their influence varied according to the age of firms. gine industries is also less than one and significant at the . and en- the startups. or en- three top size categories.01 level and the coefficient estimate the entry year of the spinoff or the preceding 5 years. two 1–0 dummy variables for the spinoffs of leading firms found that a Gompertz model fit the data well for the automobile that entered at the largest sizes are added as explanatory variables industry: in xit to test if the greater longevity of the Detroit spinoffs was con- fined to these firms. equals 1 for the nine spinoffs that entered at the highest three where zit is a vector of variables that condition how the age of firm i size categories and had a parent that produced a leading automo- in year t. dummy equals one and is insignificant.0001. including a category ‘‘unknown. the distribution of en. Fisher’s exact test was (0.6% and 5. conform with the predictions of the theory. denoted as Next Largest Top Spinoffs. The spinoff dummy is also interacted with the Detroit dummy to test if the lower hazard of the firms in the The fourth prediction can be tested using the information re. Klepper / Journal of Urban Economics 67 (2010) 15–32 Table 6 exited by being acquired by another firm or that were still producing Coefficient estimates of the automobile spinoff logit model (standard errors in at the end of the data period in 1966 were treated as censored.27 Firms that periods 1895–1920 and 1921–1966 to test if the lower hazard of the Detroit firms was confined primarily to the earlier period.e. bile make in its entry year or the preceding 5 years. All parentheses).000 or greater was larger for spinoffs in the Detroit area with a leading parent between time of entry and the hazard. dummy vari- 28 ables for entry cohorts were entered in both xit and zit. Following Klepper 18 spinoffs that entered in the next two highest size categories (2008). and 1910–1924. The cohort division is arbitrary but fits the data than the other Detroit spinoffs. and .4% of the non-Detroit and Detroit start.005 (0.29 egory. the coefficient estimate for the Detroit dummy hardly changed from (2002).24 S. suggesting that it was only try sizes for the other Detroit spinoffs is similar to the distribu. xit is a vector of variables that af. the spinoffs in the Detroit area and not the other firms located tions for the non-Detroit spinoffs and the startups in the Detroit there that had lower hazards. to test if these firms survived longer. offs to test if they had lower hazards. The coefficient esti- and spinoffs in the Detroit area broken down according to mates of the diversifier and spinoff dummies are both less than whether their parent produced a leading automobile make in one and significant at the . and an additional dummy *** Significant at the . the coefficient estimate of the De- Not producing 0. 4. The estimates are robust to different cohort divisions.06)*** iable in xit. all variables are entered in the vector xit to allow them to af- fect the hazard proportionally at all ages.13) and 0. # Firm-year observations 7762 implying a 32% lower annual hazard for firms located in the Detroit Log likelihood 551. the lower hazard of spinoffs located in the Detroit area.32)*** ative to the omitted group. Consistent with the fourth prediction. Consistent greater size of the spinoffs in the Detroit area is confined to with the sixth prediction.000 (i.30 The coefficient estimates of the time area and elsewhere.

with all other areas serving as the omitted automobiles.02)*** 0.7 Not observed 51.92 (0. .0 0.12)*** 0.1 0.14)** 0.5 $50–100 10. can be used to test the assumption that spinoffs inherit traits from The estimates for this specification are reported in Table 8 un.70 (0. production of ICs and the performance of IC entrants according to A series of models are estimated.15) 1.77 (0. which ceding 5 years.9 8. Size category % of 311 non.5 20.14) 0.0 $300–500 1. Semiconductor industry through 1987 are treated as censored.68 (0.0 4. and active ables are less than one and significant at the . the analysis focuses on entry into the the opposite.0 0.97 (0.4 19.0 0.01) 1.0 4.0 1.13)*** Entry 1895–1904 219 0. All coefficient estimates are reported in ratio form.8 3.00 (0. diode.0 0. *** Significant at the .12)*** C&W.5 6. can be used to test how their experience in these industries prior 32 All tests failed to reject the null of proportionality. Data were also col- lected on transistor.27) Largest top spinoffs 9 0.84 (0.01) 1.000s Detroit startups startups Detroit spinoffs spinoffsa leading parentsa without leading parentsa $1000+ 1.92 (0. and New York and a 1–0 dummy for location otherwise the same tests could be done for semiconductors as in Silicon Valley.13) 0.6 3.5 $2.0 Unknown 8.0 0.4 11. 1–0 dummies for location in the three early electronics clusters of Data on the entry sizes of the IC entrants were not available. the coeffi.7 10. Consider first entry into ICs by transistor.0 0.7 13. der the column Model 3. Accordingly.3 4.01 and .0 1.3 0. Table 8 Coefficient estimates of the automobile hazard of exit models (standard errors in parentheses).9 13.6 13.0 0.1 40.9 23.9 0.6 2.5 $500–1000 1.7 5.3 7.63 (0.1 20.14) Entry 1905–1909 271 0.32 Firms that never entered 6.0 1.3 $25–50 6.00 (0.14)** Spinoffs 142 0.31 of the theory.05 levels.01 level.7 22.5 0. these 28 firms were cient estimate of the Detroit spinoff dummy rises and is closer to excluded from the analysis and a model of the hazard of entry into one and no longer significant.9 3.28)** Number of firms 714 714 714 714 Log likelihood 1908.5 34.0 27.3 0.57 (0.5–5 2. but Boston.63 (0. their parents.96 (0. The coefficient estimates for both vari. consistent with the sixth prediction ICs from 1965 to 1987 was estimated for the other 360 firms.3 a At the time of entry. Los Angeles.5 $5–10 1.71 (0.0 0.0 $100–300 5.1 7. try experience (see Footnote 8). diode.2 13.6 2.6 0.1 7.3 5.3 27.95 (0.014)*** Detroit 112 0. Of these. albeit less comprehensively given the limited number category.1 5.99 (0. 28 subsequently entered ICs spinoffs of leading firms that entered at the largest sizes and a with less than 5 years of experience in their product and thus did 43% lower annual hazard for the other group of spinoffs of leading not qualify as a diversifier based on the required 5 years of pre-en- firms that entered at the larger sizes. % of 52 Detroit % of 30 Detroit spinoffs with % of 22 Detroit spinoffs 1.02)*** Entry 1905–1909  age 271 0.5 $10–25 10. which 31 The estimates were similar when the model was estimated with all 388 firms.10 level. Data were also collected on the type of IC produced at entry. so that estimates above one indicate variables The semiconductor industry clustered in Silicon Valley after ICs that increased the hazard of entry and estimates below one indicate were developed.15) Diversifiers 224 0. % of 37 Detroit % of 90 non. and active module producers.3 0.05 level.45 (0. engine diversifiers 113 0.83 * Significant at the . and had a parent that produced a leading automobile make in its to the advent of ICs influenced whether they entered ICs and entry year or the preceding 5 years and the six spinoffs that en.13)*** 0. Variable # Firms Model 1 Model 2 Model 3 *** *** Age 714 0.83 (0.06 (. The estimates imply an 87% lower annual hazard for producing these products.13 (0.5 25. The theory predicts that firms with greater tered at any of the five highest size categories and had a secondary pre-entry experience are more likely to enter and survive longer.8 8. bike.12)*** 1.68 (0. Accordingly.06 (0.7 1.4 3. as of 1964 388 firms were respectively.38 1845. with a Cox proportional hazard model used to obviate having to specify a functional form for the hazard.7 10.66 (0. Based on the EBG.6 18.7 10.20 1835. how long they survived.97 (0. S. ** Significant at the .0 4.1 13.0 0. Klepper / Journal of Urban Economics 67 (2010) 15–32 25 Table 7 Entry size distribution of automobile startups and spinoffs in the Detroit area and elsewhere.54)*** Next largest top spinoffs 24 0.13) Entry 1895–1904  age 219 0. parent that produced a leading make in its entry year or the pre.0 4.0 $1–2. The first model contains three their heritage and location.02)** 0. The next model adds two 1–0 dummies for production of IC entrants whose origins could be traced.0 0. module producers.26)*** 0.02)* Detroit spinoffs 54 0. More important.

as exemplified by the Silicon Valley spinoffs— more likely to enter than diode producers.32 (0.66 (4.35) 1. Boston. consistent with the con Valley firm and the founder of the fourth had previously theory the coefficients are all greater than 1 and significant at vari. Proposition 1a implies that more experienced firms off of one its spinoffs.06)* Diode exp. Alternatively.18 (0. which is the only group a 1–0 dummy equal to 1 for a firm with one or more spinoffs in a gi- whose origins could be comprehensively traced. worked at Fairchild). Assuming years starting another firm first). the most prolific spawners in Silicon Valley than elsewhere.74)*** Diode 74 2. which had little effect on the estimates. which in turn are expected to be more spinoffs that made it onto the ICE lists and 24 of the 91 total spinoffs likely to enter than active module producers. or having a founder that at one point worked and larger transistor producers should have higher hazards of entry.38) New York 102 0.10 level. cated in New York. Intel. Silicon Valley has icon Valley. a spin- competence.73 (0. accounted for the greatest number of spinoffs. which had the greatest number of transistor and The top six Silicon Valley firms in Table 10 were all top 20 diode producers in 1964. It alone accounted for 14 of the 53 enter than diode producers. and the number of firms elsewhere with that firms in clusters should be no more likely to enter than other spinoffs.67 (0. Thus. Consequently.01 and . the next years of production (of the respective product) through 1964 for three Silicon Valley firms with the most spinoffs.05 (0.16 * Significant at the . transistor producers are expected to be more likely to spinoff process in Silicon Valley. 34 There were 45 firm-years with one or more spinoffs. Intersil. if clusters raise firm profitability then firms lo. the leading firms in Table 9 reports coefficient estimates for the various models. The coefficient estimates of the regional dummies are Analogous to automobiles.70 (2.67 (1. including six with two icon Valley firms. The product most related to ICs was transistors. The contrast between the number of firms in Silicon Once the backgrounds of firms are controlled.99) 1. 22. and of the four with non-Silicon Valley parents. This of entry through its date of exit (or 1986 if it was still in the industry was done for the larger firms that made it onto the listings compiled in 1986)33 and all firm-years were pooled.81)*** Number of firms 360 360 360 360 Log likelihood 482.53) 1.85) 2. Also the largest coefficient estimate. respectively.01 level.05 levels.01 (4.08)* Active module exp. Consistent with firms (42%). as is the greater number of spinoffs of firms. are all spinoffs of Fairchild. which in turn were more nearly every spinoff with a Silicon Valley parent located there likely to enter than (the omitted group of) active module producers.03) Technology Transistor 61 5.26)*** Leading transistor 11 7. 74 1. *** Significant at the .96 (0. Table 10 reflects the extraordinary influence of Fairchild on the osition 1b. Variable # Firms Model 1 Model 2 Model 3 Geography Boston 46 1.61 (0.58 439.26 S. a logit model of the spinoff process all insignificant. so the analysis is the Silicon Valley genealogy but not on the ICE listings. vides a comprehensive estimate of the number of spinoffs for the Sil. of transistors and diodes. Moreover.15 (0.34 A firm’s competence is each firm the number of its spinoffs on the ICE listings. and the years the firm pro- omitted group.05. Furthermore. Klepper / Journal of Urban Economics 67 (2010) 15–32 Table 9 Coefficient estimates of the IC hazard of entry models (standard errors in parentheses). and the next Silicon Valley leading transistor producers in 1957 and 1960 based on market firm. ings whose backgrounds could be traced. would be more likely to enter. Valley. Table 10 lists for ven year that made it onto the ICE listings.25) Silicon Valley 13 2. Most of the other Silicon Valley firms are of experience and being a larger transistor producer are proxies for connected to Fairchild as well. ous levels.14 (0. all four had Sil- When years of experience for each product and the dummy for the icon Valley roots (three had a non-primary founder from a Sili- leading transistor firms are added in Model 3. duced semiconductors. the number of the observations with multiple spinoffs. consistent with the theory.27)** Experience Transistor exp. 2.27) 0. consistent with the theory.77 (0. . In the industry. Also reported is whether the firm reached the top spinoffs and one with three spinoffs.38) Los Angeles 60 1.42) 1. based on Prop. which were disproportionately concentrated in Sil- Model 1 with only the four regional dummies. 225 1.05 level. but none of the coefficient esti. suggesting that being in a cluster did not increase was estimated for the US merchant ICE firms (96 in total) in which the probability of entry. 53 of the 56 Silicon uct dummies for transistor and diode producers are greater than one Valley firms (95%) were spinoffs versus only 15 of the other 36 and significant at the . Model 3 adds the when the other Silicon Valley spinoffs are added. ** Significant at the . consistent with the theory.05 (0.01 (0. The column labeled # spinoffs adds for each firm the spinoffs that appeared on 33 No firm had a spinoff before it began producing or after it exited. fol. lowed by diodes and active modules. followed by firms as were all the firms that had spinoffs outside of Silicon firms in Los Angeles. transistor producers were from their roots. spinoffs did not generally stray far the ordering of their relevance to ICs. among the 92 firms on the ICE list- mates is significant. The dependent variable is by Integrated Circuit Engineering (ICE). for Fairchild.59)** 3. either as a spinoff of Fairchild.93 461.10 (1.22) 0.35) 1. and Silicon Valley. and producers of each of the three products and also a dummy for the Signetics. with producers of active modules the its spinoffs that ever attained this status. which pro- confined to the years each firm produced.11 (0. five. was founded by one of the founders of Fairchild (after share data for those years reported in Tilton (1971).05 (0. 61 1. An ordered logit was estimated to accommodate 20 producers in sales in any of the ICE annual listings. National. In Model 2. the coefficient estimates of the prod. each firm’s history was broken into annual intervals from its date Next the rate at which firms spawned spinoffs is analyzed. the theory predicts Valley with spinoffs.44)*** 8.87 (0. is also striking.

1–0 dum. With controls for year and age.05 level and implies a (roughly) 5.16 to 2. and both are significant at the that a larger fraction of IC entrants in Silicon Valley than elsewhere .5% if their 1974 ICE market share was less than 0. and 35 Some firms entered before the first ICE listing in 1974. With nearly all Silicon Val- impression from Table 10 that better firms spawned spinoffs at a ley firms descended from Fairchild.34. The first model includes the four regional dummies. be expected given the small sample of acquired firms (eight by 53% in Los Angeles.16 higher spinoff rate given that the probability of functional form. which located in Silicon Valley. New York. Monolithic ICs eventually dominated ICs and could be consid- The coefficient estimates of all but the year dummies are re. The coefficients of expectation. To test whether the estimates might duced semiconductors and the number of years squared. two dummies for entry between 1965– 1969 and 1970–1974 to test if early entry was advantageous. In a few cases of firms with no market share data for any year through in the Silicon Valley percentage and the percentages in Boston.36 This is not consistent with mated38 over the period 1965–1987.01 level. tern will be considered later. inclusion of the time of entry variables has little effect on the other estimates.0% or 0. 52% in New York. Consistent with this at age 15.16 times greater Successive Cox proportional annual hazard of exit models are esti- spinoff rate for firms in Silicon Valley. The coeffi. they were assigned a market share equal to one half of the lowest market share and elsewhere are all less than .39 The second model adds dummies for transistor.64} = 5. and electronics producers. year dummies. formance is measured by longevity (years of production of ICs). they were assigned a market share of 1. 1986. diode. ity. the two-tailed p-values corresponding to the difference those years. the performance of IC entrants is considered. it would be expected positive and negative. the coefficients were re-estimated with all observations for firms that take the value 1 in the year prior to and (up to) 2 years after Fairchild deleted. which is within the sample range. the not spawning a spinoff in any given year was close to 1. the observations for each firm. While there is no a priori basis for the choice of entry cohorts. if a firm lacked market share data for 37 the last year or two of its existence. respectively. Los Angeles.5%. the coefficient estimate of market time to produce hybrid and film ICs.1. 36 39 The odds ratio of a spinoff is exp{1. which was a clear out- mies for firms acquired by semiconductor and non-semiconductor lier. was reported for earlier years in Tilton (1971). confirming the the leading innovator of monolithic ICs. and a 1–0 dummy equal to 1 for firms for the implied greater spinoff rate for Silicon Valley firms. it was assigned its last recorded market share for Using Fisher’s exact test.37 semiconductor firms. As expected.01 level. the effect of time of entry is identified via which translates into roughly a 5. This had little effect on the estimates except the acquisition. have been unduly influenced by Fairchild. If their market share the hazard.001. with transistor firms years. nificant at the . and 44% elsewhere.16 higher for firms in Silicon Valley. Their 1974 ICE market year dummies to accommodate a downward trend over time in share was used as their market share for the years 1969–1973. ered the technological frontier. Klepper / Journal of Urban Economics 67 (2010) 15–32 27 Table 10 Spinoffs of semiconductor producers. as might entry was 92% for firms in Silicon Valley versus 39% in Boston. it was used as their market share for years before 1969. the percentage of firms producing a monolithic IC at both acquisition variables are positive but not significant. The coefficient estimates of age and age squared are their spinoffs as conjectured in the theory. All tests failed to reject the null of proportionality.35 Similar to the logit model for automobiles. They imply that the firm spinoff rate reached a maximum produced monolithic ICs when they entered. S. although as noted earlier alternative explanations for this pat- other independent variables include the number of years a firm pro. Standard errors are corrected by clustering declined from 5. the theory but is consistent with clustering raising firm profitabil- ings and Tilton (1971). if firms passed down traits to greater rate. Firm Years (through 1986) # Spinoffs # ICE spinoffs # Top 20 spinoffs Top 20 firm Silicon Valley producers Fairchild 1957–1986 24 14 7 Yes National 1967–1986 9 4 1 Yes Intel 1968–1986 6 6 2 Yes Signetics 1961–1975 5 2 1 Yes Intersil 1967–1981 4 2 0 Yes Synertek 1973–1985 4 3 1 Yes Semi Processes 1975–1985 4 1 0 AMI 1966–1986 3 2 0 Yes AMCC 1979–1986 3 2 1 Seeq 1981–1986 3 3 1 Amelco 1961–1986 2 0 0 Yes Micro Power 1971–1986 2 1 0 Raytheon/Rheem 1961–1986 1 0 0 Yes Siliconix 1963–1986 1 0 0 Yes Avantek 1965–1986 1 0 0 AMD 1969–1986 1 1 1 Yes Exar 1971–1986 1 1 0 Cal-tex 1971–1975 1 0 0 Nitron 1972–1985 1 0 0 Zilog 1974–1986 1 1 1 Yes Supertex 1976–1986 1 0 0 Exel 1983–1986 1 0 0 Non-Silicon Valley producers General Instrument 1960–1986 4 2 0 Yes Texas Instruments 1952–1986 3 3 2 Yes Motorola 1958–1986 2 2 1 Yes Mostek 1969–1985 2 2 2 Yes RCA 1950–1986 1 1 0 Yes measured by its market share in the current year based on the ICE list. . At first per- cient estimate for being located in Silicon Valley is positive and sig. 16 by non-semiconductor firms). Fairchild developed and was share is positive and significant at the . but many firms continued over ported in Table 11. Last. If their market share was not listed in Tilton (1971) for the earlier active module. For years after 1974. 38 of any firm on the ICE list in the respective year.

causing the coefficient estimates to be biased toward one. In all the models. More important. and longevity Acquired by semiconductor firm 0.01 level. respectively) lower hazards.10 level.and post-diversifiers. reflecting the greater percentage of entrants descended Coefficient estimates of the semiconductor spinoff logit model (standard errors in from the leading firms in Silicon Valley than elsewhere (22 in Sili- parentheses). Klepper / Journal of Urban Economics 67 (2010) 15–32 Table 11 nearly one.01 level.01 level. When the dummy for being a spinoff of a ordered as expected.35 (0. The coefficient estimate for the Silicon icant and the coefficient estimate for the second cohort significant Valley dummy is about 30% lower than in Model 2. it has a large. firms the much greater fraction of entrants in Silicon Valley that pro- in Silicon Valley and Boston had significantly (at the . and for brevity the esti- mates for the logit of attaining the top 20 producers are presented. alternative measures were created based on the ICE sales Log likelihood 128. the coefficient further analysis was done of the probability of producing a mono- estimate for the dummy for spinoffs of top 20 parents is .24 (0. The years of experience variables were also 1964 (denoted by the prefix post-) under the expectation that firms tried but had little predictive power and these results are not re- that did not directly enter into ICs might actually have been less ported. Logit analyses were estimated for whether firms ever at- * Significant at the . which effec- a higher hazard. tively pares down the sample to the 329 firms that produced a with the coefficient estimate for the first entry cohort always signif. entry and the probability of making it into the top rank of firms. largely confined to the Silicon firms that were spinoffs of the unobserved attributes that compensated for their lack of experi. the coefficient estimates of the post-dummies is significant.78 (0. and the dummy The coefficient estimates are reported in Table 13.16 and sig. which may Silicon Valley firms reaching the top echelons of the industry is reflect that less experienced diversifiers that entered ICs had other. Silicon Valley 1. frontier. duced a monolithic IC at entry. The models estimated were similar to those for the hazard except ule. while the coefficient estimates for the other Years of production 0.60) over such a period may have its limits as a performance measure. which are by definition larger firms. Also in. the greater likelihood of rience variables are all close to one and insignificant. with the regional and 40 Parental heritage could only be comprehensively measured for firms on the ICE technological dummies included as explanatory variables. Separate dummies are included as a continuous measure (year of entry) under the expectation that for firms that produced these products by 1964. nological frontier at entry is controlled. and significant at the . with three of the four significant.28 S. In contrast. reflecting in some of the models. The coefficient esti- Variable Coefficient estimate mate for the Boston region is less than one and significant (at the Market share 0. One differ- for being a leading transistor firm.73)** To probe the robustness of the estimates to the measure of perfor- # Firm-year observations 1194 mance. active mod. this is the appropriate variable to test the theory. and electronics producers given the technological and market the year dummies were not relevant and time of entry was entered proximity of each product to ICs.012 (0. the coefficient estimate will be biased downward to reflect a lower hazard producing a monolithic IC ever produced a monolithic IC later. earlier entry lowered the hazard. To the extent these were the spinoffs coefficient estimate that was significant was for prior transistor that entered at the largest sizes. as in Model 1.and post-dummies and the dummy for diction of the theory that firms with superior heritage have lower the leading transistor producers. the coefficient estimates of the expe. the 20 largest in any year. This tests the pre. con Valley versus seven in all other regions). The last model adds a 1–0 dummy ence from the hazard of exit estimates is that in Model 1 without equal to 1 for firms producing a monolithic IC at entry to test if being firm controls. tained the ten largest in any year. which and significant coefficient estimate and the coefficient estimate is consistent with firms that chose to produce another electronics for Silicon Valley declines by approximately 50% and becomes product before ICs being no more competent than those that di. 2002). Years of production squared 0. insignificant. Consistent with the theory. hazard of exit estimates is that producing a monolithic IC at en- iod in 1987 were treated as censored. Coefficient estimates that were not identified because they competent. consistent with the was more likely for firms located in clusters.11)*** two regions are close to one and not significant.01 level. and so it should still be is consistent with heritage rather than location being the primary possible to test if any greater longevity of firms in clusters is attributable to their determinant of whether a firm ever reached the technological heritage.40 Firms that were still producing at the end of the data per. expected to have the lowest hazard followed by diode.55 data. positive. the year before the later a firm entered then the less time it would have to attain ICs were first listed in the EBG (denoted by the prefix pre-). a try is close to one and is not significant. entered separately for the pre. This of exit assuming larger firms survive longer. were perfect predictors of failure also could not be reported. try has a strong effect on performance. no coefficient can be estimated for this var- form. indicating that these firms had an 84% lower ducing a monolithic IC. Furthermore. ply were large enough to be on the ICE list in any year (through *** Significant at the . In Model 3. reflecting some kind theory the coefficient estimate for the Silicon Valley dummy rises to of technological spillover (that is not captured in the proposed the- ory).10 level). and after the top 20 producers. They all yielded similar estimates. The only listings. ence. as is commonly found. In In light of the relationship between producing a monolithic IC at Model 4.04)*** . reflecting the fact that The coefficient estimates of the models (except for the year every firm that got into the top 20 produced a monolithic IC at dummies) are reported in Table 12. monolithic IC at entry. or sim- ** Significant at the . In Model 1 without any firm controls. This persists in Model 2 with the spinoff’s entry year or the preceding 5 years.05 lev.05 level. None of top 20 parent is included in Model 4. once being at the tech- rectly entered ICs. the Silicon Valley effect is much more pronounced at the technological frontier at entry lowered the hazard. the inclusion of the pre.99) Acquired by non-semiconductor firm 0. . Another difference from the hazards. In Model 2 the pre-dummies are cant at the . only 13% of the firms that entered not Otherwise. Of particular interest was whether this hazard than other IC entrants. lithic IC in a later year for those 294 firms that did not enter pro- nificant at the .004)*** The data on IC production span only 23 years. production. They are expressed in ratio entry.88 (0. leading firms (and attained the ICE listing). though.10 and . but it can be entered as a control in Model 3. Consequently. Even if it is biased. A Cox proportional hazards model of producing a monolithic IC in a later year after entry was estimated. but it is still sizable and signifi- els. The next model adds the years of experience variables. the coefficient estimate of producing a monolithic IC at en. so values below one indicate a lower hazard and above one iable.64 (0. it should be comparably biased for firms in all regions. This reflects the disproportionate cluded is a 1–0 dummy equal to 1 for the 29 spinoffs on the ICE list number of Silicon Valley firms that became leaders of the indus- whose parents’ sales were among the top 20 firms on the ICE list in try (20 versus 15 in all other areas).

14) (0.17) (0. 42 ** If so.13) (0.13) (0.62 41 The evidence is more discriminating for automobiles whereas in semiconductors a there were not many non-spinoff IC entrants in Silicon Valley to compare to the Not identified. Consistent with a broader process of organizational (0. Indeed. S. Discussion the Northeastern corner of the state.16)** (0. valuable local knowledge based on their pre-entry experience that can substitute for vative firm entered in both regions.98 mies related to labor pooling.51 ductors. are striking.17)** of agglomeration economies and limited the performance of non- Entry 70–74 184 0.60 0.13)*** (0. through spinoffs and in Olds Motor Works’ case close to their geographic roots.84) (0. (0.11)*** (0.17)* and better spinoffs but firms in more closely related industries Pre-active 35 0. this led to (0. * spinoffs and data on firm origins were available only for a subset of the IC entrants.39) analysis was not directed toward evaluating the effects of Post-electronics 32 0.09 trolling for various factors. suggesting that agglomeration econ- Pre-active 35 0.98) ized knowledge spillovers played a similar role.18)* (0.05 level. yet Leading transistor 11 0. but Buenstorf and Klepper (2009b) found that all types of entrants located close to their geographic origins and this was no greater for The parallels between the Silicon Valley and Detroit clusters spinoffs (or other entrants) that originated in Northeastern Ohio than elsewhere.46) (0. While the Post-active 16 1.10 the superior performance was largely restricted to the spinoffs (0. (0. but New York 124 1.17 While organizational reproduction and heredity seem to have (0.58 had a major influence on the emergence and growth of both (0.14) (0.77 0.47 0.13)* (0. Helsley and Strange.08) sessed at the time of entry.13) firms in both regions were on average superior performers. contributing Firms to a disproportionate number of spinoff entrants in both regions. (0.12)* (0. Once the spinoff process got going in both regions. the theory sketched out in the paper is not adequate to analyze such a possibility. (0.43 0.48 0. perhaps the right heritage is more Parent top firm 29 0.06) appears to have been based on innate characteristics they pos- Pre-diode exp.20) (0. which resonate with those of Figueiredo et al. to test whether entrants were more likely to locate where they originated if they originated in the clusters than elsewhere.95 were more likely to diversify into both industries and perform module (0.47 0.02) Post-transistor 8 0.72 0. 2003).93 hence probability of entry of indigenous potential entrants exp.16)* (0.94 0.67 0.00 0.69 0. such as the Entry 65–69 195 0. But data would be needed on the geographic origins of all Significant at the .13) (0.63 1. Pre-electronics 98 1.10 1. Significant at the .45) (0.05 2445.20) better than other entrants.15)** Pre-diode 20 0. Alternatively. (0. and with Boston 79 0. and local- (1.00) (1. but then an outstanding inno.15) (Rosenthal and Strange. In both instances.94 (0.41) (0.14) (0.90 1.90 3. including spinoffs elsewhere.10 level.56 2448.16) Post-transistor 8 1.99 rior spinoffs in both industries were distinguished at birth.25) spinoffs descended from the leaders that entered at the largest Entry cohort sizes.70 0.10 1.49 bidding up of wages and prices in the clusters. (2002) for modern Portuguese startups.05) superiority.68 0.13)* a buildup of successful firms in Detroit and Silicon Valley.10) it did not compete away the advantages realized by the spinoffs Firm heritage in the clusters. (0.56) (0. better firms not only spawned more (0. But if agglomeration economies were module exp.08 0.17)* (0.01 level. this firm the benefits associated with agglomeration economies and induce them to locate contributed.62) clusters.16 valuable in clusters due to the greater presence of specialized in- (0.42 Year dummies 623 Yes Yes Yes Yes Number of firms 623 623 623 623 623 Log likelihood 2460.26) (0. which could explain why the firm Post-diode exp. .10 1. Alternatively. firms nearby. suggest that entrants of all types and geographic origins have in either Silicon Valley or Detroit.16)** (0.12)** (0.70 (0.75) (5.60 1. Los Angeles 102 1. agglomeration economies might have Pre-electronics 98 0.00 clusters were largely indigenous entrants and their superiority exp.42) (0.18) flected in their initial capitalization in automobiles and their Technology propensity to produce at the technological frontier in semicon- Pre-transistor 32 0.08) firms would have been superior performers in the clusters. 20 1.40) (0. to the creation of many other leading Coefficient estimates of IC hazard of exit models (standard errors in parentheses).85 agglomeration economies on the evolution of the two clusters.96 such a process was operative it would have to be explained why entry (0. 3 –a spinoff rate was so much higher in the clusters even after con- Post-active 16 1.55 reproduction and heredity.26) (0.14) located there that were descended from the leading firms. it could help explain why spinoffs of firms in the two clusters generally Significant at the . Ohio in 7.07) strongly at work it might have been expected that all kinds of Post-electronics 32 0.72 0. offset the benefits (0.14) (0. Klepper / Journal of Urban Economics 67 (2010) 15–32 29 Table 12 subcontracting as well. Clearly.02 (0.66 0. it is less clear whether traditional agglomeration econo- Post-diode 3 0. The industry clustered around Akron. Variable # Model 1 Model 2 Model 3 Model 4 it operated much more intensively than elsewhere.25 1.76 0.47 the superiority was largely restricted to spinoffs and in particular (0. 2002). Supe- Silicon Valley 85 0.44 0.28 module (0.99 0.33 0.43) (0.12)** (0.75 spinoffs not straying far from their geographic roots.54 0. entrants.99 fueled entry in the clusters by enhancing the profitability and exp.99 0.53 1. (0.17) spinoff entrants in the clusters.99 1.82 2433.12)** (0.00) (0.94 0.07)*** puts to innovation (cf.15 0. no firms were initially located Their findings.39 0. In both industries. as re- (0. Buenstorf and Klepper (2009b) collected the requisite data for historical entrants in the US tire industry in the state of Ohio. *** located in the clusters. proximity to suppliers. Geography Spinoffs of leading firms performed especially well.41 Perhaps some kind of equilibrating process. The superior entrants in the Pre-transistor 32 1.99 omies did not pull entrants to the clusters nor nurture their module exp.97 exp. But if Technological frontier Monolithic at 329 0.16)* (0.25) two observations seem pertinent.

49)*** 1.03)** 0. In con- wouldn’t the spinoff process be a zero-sum game in which the trast. the two clusters. nearly all the successful entrants could be traced to go from the theory to questions about policy regarding clusters.61)*** # Observations 623 623 623 623 623 Log likelihood 109. firms as the industry evolved. Variable # Observations Model 1 Model 2 Model 3 Model 4 Geography Boston 79 0.12) 0. But firms did not en. with Fairchild close behind. spinoff rate? The leading spinoffs in automobiles and semiconduc- ually at the forefront of the industry and its market share was tors tended to be formed by very high level employees.30 S.94 (1.75 (0. which features spinoffs inheriting ther region was a likely place for their industries to cluster. For one. to the concentration of the industry in Silicon Valley rather than A related question about learning involves from whom do spin- Dallas.25 (0. 2000)..96 (1.48 97.04)*** 0.34 (1.10)* 1. TI preneurship (Sorensen.07 (0. this was not done in semiconductors.87) 1. the the- ter in either Detroit or Silicon Valley early on.53 (0.85) New York 124 1.85) 0. whom were even founders of their firms (Klepper. Clearly. which was instrumental across firms of different size. to form spinoffs? In both industries. though. ket along with Fairchild’s 7% whereas on its own Fairchild’s market tions that in retrospect favored the development of each industry share never exceeded 13%.60 (0. But is the to- and Fairchild were credited with co-inventing the integrated cir.13 (0. TI’s base location. Tsai.26) 2.42 (1.23)* 1.01 (0. does not tell such a story. do leading But that only seems part of the story.15) 0. ory sketched out in the paper.26) 1.22) 0.65 (0.44)*** 2.84) 0.74) 0.83) 0.01 level.58) Technology Pre-transistor 32 0. and perhaps even a minor firms have more spinoffs because there is more to learn in such part judging from numerous regions that are blessed by an early firms or merely because they are larger and have more candidates innovator that never develop into an outstanding industry cluster.36 (0. degree done things differently from their parents.19) Pre-active module 35 –a –a –a Pre-electronics 98 –a –a –a Post-transistor 8 –a –a –a Post-diode 3 –a –a –a Post-active module 16 0.26) Entry Entry year 623 0. Somehow spinoffs must have to some in its cluster (cf. 2005)? Was Silicon Valley a more hospitable area founder. back in one way or another to Olds Motor Works and Fairchild.63 (1.09 (1. this was operationalized entrepreneurial culture that encouraged the formation of spinoffs empirically by linking spinoffs to the prior employer of their main (Gompers et al. will require such a story. Sturgeon.69 (0.08 (0.79) 0.28 (1. in part because of the gains of the spinoff came at the expense of its parent (or other absence of the requisite data but also because it would have made .06) 1.61 (1.96) Leading transistor 11 2.28)* Post-electronics 32 0. and the Fairchild had many more spinoffs than TI and many more that at. 1999)? which seems to have been particularly true for semiconductor If agglomeration economies were not strongly at work in shap. traits from their parents.20 (1.50)*** 0. In 1980. 2007. it would seem to be the chance entry there of these two firms The theory itself and some of the findings raise a number of that was the key impetus for the two clusters.47 84. and AMD. *** Significant at the .10) 0.05 (1.82) Los Angeles 102 0. Klepper / Journal of Urban Economics 67 (2010) 15–32 Table 13 Coefficient estimates of IC logit of attaining top 20 in sales. a key question is how could spinoffs have to take into account secondary founders and earlier auto employ- fueled the growth of both clusters? Phrased differently. some of consistently greater than Fairchild’s.33 (1.10 level. Any attempt Furthermore. suggesting that nei. Numerous questions are raised by the role of spinoffs and more firms in the region)? Fairchild demonstrates dramatically that this broadly organizational reproduction and heredity in the growth of was not the case in semiconductors. possibly due to its prohibition on the sometimes work for more than one firm in the same industry.13) 1.89 (0.68) 0.14 a Not identified. which accords with the findings of general studies of entre- pioneered the silicon transistor. tal number of employees the right denominator to compute the cuit and both were among the first producers of ICs. Instruments (TI) and Dallas.35 70.08 (0.05)*** Technological frontier Monolithic at entry 329 –a –a Firmheritage Parent top firm 29 3.50 (1.45 (1.15) Pre-diode 20 0.83 (1.10 (0. Intel. 2008). 2009).86 (0. Thus.05 level. TI was contin. TX in the semiconductor industry.12)* Silicon Valley 85 2. 1997.09) 1. accounted for 32% of the mar- Silicon Valley? Various attempts have been made to identify condi.33 (0.40 (0.27 (0. additional questions about the spinoff process. TI 2009). Why? Did Fairchild have a uniquely off founders learn? In both industries. Yet as reflected in Table 10. National. the number of spinoffs per Indeed.04)** 0.81 (0. enforcement of employee non-compete covenants (Gilson.31 (0. First.34 (1. four of its spinoffs. why ers of the main founder in the analysis of firm performance.. why did the clusters form in Detroit and Signetics.79) 0. ** Significant at the .62 (1. Elfenbein et al. * Significant at the . one does not need to go far to find such an example—Texas employee seems to have been lower in the larger firms (Klepper. In automobiles an attempt was made ing the two clusters. number of such candidates to form spinoffs may not vary greatly tained the ranks of the industry’s leaders. But many spinoffs involve multiple founders and founders for spinoffs than Dallas.

2009). which forms the basis of a disagreement theory of spinoffs that Klepper Proof. but National was a pre-existing Connect.e. Valley. Proof.. and (c) H firms in the new industry spawn a It is tempting to close by not trying to address the various greater expected number of spinoff entrants than L firms in the new questions raised. hence the maximum and average entry size of minimum amount of time to qualify as diversifiers? How should firms be treated that spinoffs of H incumbents is greater than that of spinoffs of L are started by individuals that previously headed another. as featured in Klepper and Thompson (2009). Similarly. as appears Appendix A to be true generally (Dumais et al. and the extent to the geographic structure of other industries besides automobiles which they can change their competences over time. settings. referees. where they derive such competences. ties that have been accumulating concerning spinoffs in a num- even defining spinoffs can run into some of the same issues. h shunned but turned out to be successful (Klepper. Much remains to be spinoffs and clusters. A key question is learned about what it means for organizations to have compe- whether a similar process involving spinoffs operated to shape tences. Should it be classified as a spinoff with Fairchild as its par. an industry is to a new industry then the greater the probability come less agglomerated over time. might it have had something to other industries may yield new insights into the emergence do with peer effects. this could help explain Proposition 2. two gence and growth of their clusters and the base location of en. which is greater than the probability of entry of a spinoff from edge of their industries. 2002. and hence firms in the industry over- Olds Motor Works and Fairchild seem instructive. (a) H firms in another industry are more likely than L tinued vibrancy of Silicon Valley (Zhang. only spinoffs of H incumbents can themselves Similar questions arise concerning firms with experience producing other products prior to entry. but over time activity in both industries shifted away from their clusters via branching and related actions. of integrated circuits. Marx et al. 2003) compared to the firms in the same industry to enter the new industry. 2). Propositions and proofs clusters? While attention focused on the early evolution of the Detroit and Silicon Valley clusters. which clearly will require a lot more study. Ohio. ceteris paribus. also played a key role in the historical agglomera. might help to explain how spinoffs are not merely a zero-sum Fairchild’s head of production. that H firms in the industry. the larger pd) then the eration economies or in time will Silicon Valley inevitably de- greater the probability that firms in the industry enter the new cline like Detroit? industry.1. Regarding the geographic influence of spinoffs in other industries. duction and heredity may be important forces operating in many criminating data. it is hard to overlook the con- Proposition 1. greater than startups. the probability that an L firm in the same industry enters the tion of the US tire industry around Akron. In the case of the semiconductor industry. employees to do the same (cf. Serguey Braguinsky. (PL + 1/2). Stuart and Soren. questions abound about how they operate and influence It is not hard to come up with yet further questions about the formation and growth of clusters. including autos and semiconductors. are limited in their ability to evaluate promising ideas that arise child. 2009). and/or might it have reflected a kind of localized external economy asso- ciated with the organizing and finance of startups (Kenney and Acknowledgments Florida. some quite fundamental. If firms example. Last. Should they be required to produce these products for some be H firms. 1999. that digging into the origins and performance of entrants in son. to reconstitute the firm in Silicon game but fuel the growth of regions. The probability that an H firm in another industry enters a gue that spinoffs. Klepper (2008) implicates the lack of successful spinoffs (1  pd)(PL + 1/2) is an increasing function of pd. Could this have something to do with agglom- related an industry is to the new one (i.. industry. Klepper / Journal of Urban Economics 67 (2010) 15–32 31 it more difficult to sort out the distinctive heritage of spinoffs in and Thompson (2009) use to explain various statistical regulari- Silicon Valley given the pervasive influence of Fairchild. which certainly will be a challenge. The maximum and average entry size of spinoffs is why it became concentrated in Silicon Valley rather than Dallas. (b) The more decline of Detroit. the more related as a key factor causing the US television receiver industry to be. As to why spinoffs occur. Indeed. Detroit and the semiconductor industry in Silicon Valley experi- troit and Silicon Valley? Did it have something to do with the law enced such striking parallels in the way they evolved suggests on employee non-competes (cf. related firm? incumbents. S. 2000)? Why did spinoffs generally locate close to home and why does this appear to be generally true of new firms I thank Rosemarie Ziedonis for sharing longitudinal data she (Figueiredo et al. What can be learned from these patterns about the forces governing the evolution of A. trants. Some spinoffs are H firms whereas all startups are L firms. h . Nanda and Sorensen. and spinoffs of H incumbents enter at a greater Similar forces seem to have been at work in other automobile maximum and average size than spinoffs of L incumbents. Dahl and Sorenson. 2009). 2009b)? ski for all his help in collecting and analyzing the data on producers The analysis of both industries focused only on the emer. this icut semiconductor company before it hired Charles Sporck. hence spinoffs enter at a greater maximum and average size than 43 startups. and more broadly organizational reproduction new industry is pd(PH + 1/2) + (1  pd)(PL + 1/2). Both were all. But theorizing and evidence related to a number of the questions offers some relevant insights. The fact and semiconductors? Another key question that was alluded to that two such celebrated clusters as the automobile industry in earlier was why the firm spinoff rate was so much higher in De. While these observations suggest that organizational repro- ent?43 Addressing questions like these may require yet more dis. and the editor provided helpful comments. and semiconductor firms and in other industries as well. Buenstorf and Klepper (2009a) ar. Guido Buenstorf. At the same new industry. 2003. Therefore. H firms spawn a greater to leave both firms to exploit ideas that their parent firm expected number of spinoffs than L firms (part (c)). which exceeds and heredity. Buenstorf compiled on the sales of semiconductor producers and Jon Kowal- and Klepper. 2008. 2002).. Gilson. from within. This led their founders and many others an L incumbent. enter the new industry (part (b)). the probability of entry innovative firms that experienced considerable internal turmoil of a spinoff from an H incumbent is ps(PH + 1/2) + (1  ps)(PL + 1/ associated with financial control by outsiders with little knowl. National Semiconductor was defined as a spinoff of Fair. For ber of industries. with (successful) spinoffs encouraging other and growth of industry clusters. (PL + 1/2) (part (a)). Since pd(PH + 1/2) + time.. which is common.

Journal of Urban Economics 48. High-Tech Start-ups and Industry Dynamics in Silicon Valley. h Klepper. Making Silicon Valley. 2009. Anthony J. 2006. Nanda. pp. In: Kenney. Geographic concentration as Arbor. Knowledge Lécuyer. economies. 1996. Ramana. Hamilton. and survival. Lindgren. Optimal project rejection and new firm start. Kevin. mimeo. New York. Gompers.. 2004. Francisco. George S.. Home-field Stuart. Jacques-Francois. Charles. Glenn. 1997. Steven. Josh. transfer through inheritance: spinout generation. Saranac Publishing. 2004. DC. mobility. Jacques-Francois (Eds. Thisse. spinoffs. Economics and Statistics 88. hazard of exit in each period is lower for: (a) diversifiers and spinoffs Kenney. pp. San Francisco. Handbook of regional and urban economics. Journal of Public Policy Institute of California. stars. Buenstorf. Finance 60. Ueda. Bruno. William B. Christopher. Management Science 55. 98–123.. A. and the Michigan experiment. Thisse. Masako. Working Paper. Toby E. Charles E. Saxenian. Sorenson. pp. 37–61. NY. (c) diversifiers from more related industries (i. In: Vernon Henderson. Steven. Eerdmans Publishing Company. UK. 79–115. Wheels within Wheels. . tire cluster revisited. ups.. Steven. Economic Journal 119. Glaeser. Olav. 36–51. The Romance of the Automobile Industry.. NJ.. Rae. 2007. Peter. Stanford. The Handbook of Urban and Regional Economics. Klepper. Gilles. mimeo. 387–412... Doolittle. 501–522. Tilton. Picard. 2007. 2009b. Steven. Thompson. Diego. Sleeper. 2063–2117. Klepper. Guimaraes. Venture capital in Silicon Valley: fueling than startups.. Disagreements. 25–45. Evidence from the US Tire Industry. Jesper. New Marx. References Krugman. Semiconductors. Steven. Thisse. Glenn. Masahisa. Klepper. hence on average they have lower hazards Klepper. 2006. Klebold Press. Gambardella. Firm survival and the evolution of oligopoly. Louis. Cambridge University Press. 1987. 577–614. Florida. Cambridge. Strumsky. Learning the Silicon Valley way. Niels W.J. Beverly R. 575–629. 841–860.. The Entrepreneurial North Holland. In: Dealing with Academy of Management Journal 47. 2002. Krause Publications. Ketelhöhn. 193–204. 08-051. 2008. Journal of Urban distribution of entrepreneurial activity. 175–201.). 679–699. Krugman. Kimes. Sarkar. Sorenson. Paulo. third ed. Journal of Political Economy 99. Michael. The US semiconductor industry: a locational analysis. 2006. Paul R. Puga. Davis. Geographic concentration in US Scott. Venables. Daniel L.. Understanding Silicon Valley. Henry A. Zoltan. Lee.. Filson. Rosenthal. Knowledge diffusion through employee Sturgeon. 2001.. Bart.. St. New York von Rhein. Edward L.). Building High-Tech Clusters: Silicon Cassiman. Timothy. Daniel. Amsterdam. Among contemporaneous entrants. The role of clusters as sources of dynamic externalities in the US semiconductor industry... Fallick.. Alfonso (Eds. Micro-foundations of urban agglomeration and agglomeration. Steven. James. CA. England.. Washington. vol. James R.. Angel.). 875–912. 2009a. John E. 2003.B.. 2004. J. Agarwal. Management Science 52. 2003. and Public (b)). 341–361. Harvard University Press. 1968. Globalization and the inequality of nations. Cambridge. Entry by spinoffs. (b) diversifiers that are H versus L firms in their own new firm formation. Journal of Political Economy International Journal of Urban and Regional Research 19.. Grand Rapids. Klepper. Paul R. Duranton. 483–499. The Geography of Organizational Knowledge. 2005. Klepper. hence on average spinoffs from H incumbents have lower hazards mimeo. exit (part (c)). 1991. Zenger. Darren. Route 128. on average the Helsley. whereas all startups are L firms. and covenants not to compete. mimeo. International Diffusion of Technology: The Case of Fujita. 2002. Automobile Spinoffs 1886-1939. Pierre. The Spatial Aggregation of Automobile Manufacturing Activities Gilson. Saranac Lake. 472–481. 158–184. The splintering of the solid state industry. Todd. Raj.. 2004. 859–880.e. The University of Michigan Press. 1971. Guido. Elsevier. 2002. MI. Thisse. 377–393. 2006. Strange. Stuart S. John B. pd is greater). Review of entrepreneurial entry. David. Smith. William C. CA. Administrative Science Quarterly 48. Administrative Science Quarterly 52. 1–8.). 15– Federal Trade Commission. Quarterly Journal of Economics 60. Jon. 1986–1999.P. Philip H. US 47. Octavio. 1975. Franco. Matt. Kristina. Technological Spillovers and the Agglomeration of the Semiconductor Industry in Silicon Valley. industry. Debra. mimeo. Laurence (Eds. in the new industry. In: Acs. Washington.. Funk and Wagnalls. 4. Fleischman. Sporck. April M. and (d) spinoffs from H versus L incumbents. 262–275. Cambridge. of exit than diversifiers that are L firms in their industry (part Audretsch. Silicon Valley – a chip off the old Detroit bloc.. Paul. Why Michigan? In: Lewis. Rosenthal. Ann Dumais. MIT Press. 2008. Jacques-Francois Amsterdam. Goldstein.. pp. Regional Advantage. 616–631. Entrepreneurship. Standard Catalog of American Cars. Cambridge. Steven. 1971. vol. a dynamic process. Valley and Beyond. Steven. 1994. Rajshree.. Stuart S. Strom.. Journal of Economic Geography 6.. public corporations and the genesis of new ventures.. 889–927. Richard. 875–889. Peer Effects and Entrepreneurship. In: Kenney. Franco.32 S. Bruce. Cambridge. Timothy J. Spinoff. Martin (Ed. Cambridge University Press. Clark Jr. for which Stanford University Press. (Eds. 2009. hence the lower their hazard of Klepper. 4. 1997. The legal infrastructure of high technology industrial in the American Midwest. Olav. Washington University. Heritage and agglomeration: the Akron American Automobile Industry. DC. of agglomeration economies. Cities and Geography. Heritage and Firm Survival-An Analysis of German University Law Review 74. Spawning of Scientists and Engineers: Stars. 2003. Lucia B. 705–733. mimeo. Report on the Motor Vehicle Industry. A Most Unique Machine: The Michigan Origins of the Buenstorf. pp. Entrepreneurial spawning: Zhang. Sorensen. Martin (Ed. MA. M. Business School Entrepreneurial Management Working Paper No. New York. L. Kowalski. a related industry that are H firms. Veloso. William C. Noncompetes and inventor York. The greater pd then the greater the fraction of diversifiers from Policy. Management Science Belleflamme. 2005. Economics of Agglomeration. RAND Journal of Economics 33.. Increasing returns and economic geography. D. Why Does Entry Cluster Geographically? Moore. Guy. Robert W. Economics Bulletin 12. 2002. Job hopping in Silicon Sorensen. 2006. Harvard Dahl. districts: Silicon Valley. Douglas. hence on average diversifiers Klepper. An economic 55.). development. Scharfstein. MI. May. Steven. theory of regional clusters. Ellison. Stanford. Woodward. Some fraction of diversifiers and spinoffs are H firms 1890–1942. 105. Rebitzer. 2119–2171. Robert (Eds. Review of Economics and Statistics 85. Echambadi. 1–9. and spinoffs have lower hazards of exit than startups (part (a)). Sally. Stanford University Press.). Princeton.. Jesper B. and the evolution of Detroit as the Only diversifiers that are H firms in their industry can be H firms capital of the US automobile industry. Journal of Urban Economics 51. and the Small Firm Effect. Bureaucracy and entrepreneurship: workplace effects on Valley: the microfoundations of a high tech industrial cluster. Cambridge University Press. Fleming. 2008.. 2002. Guido. Innovation and input sharing. Disagreements and Intra-Industry Spinoffs.. Paul.. Martin. Scott.. 2000. 2008. Economics 52. 1995. Strange. pp. Tsai. 1999. How Silicon Valley came to be. Ellison. The Social Attachment to Place. 2009. Edward L. Technological Change. 2009. Only spinoffs from H incumbents can be H firms.). Gordon. Junfu. Glaeser. Geography. AnnaLee. William C. The Brookings Institution. pp. 1291–1306. MA. The American Car Since 1775. Ronald J. of exit than spinoffs of L incumbents (part (d)). In: Bresnahan. Figueiredo. Lerner. Auerbach Publishers. 2000. RAND Journal of Economics 37. Jacques-Francois. Elfenbein. 2000. 2009. 1939. WI. Automobile Quarterly Inc. Growth. April M. Review of Economics and Statistics 84. Proof. In: Vernon Henderson. Government Printing Office. Understanding Silicon Valley. Management Science 53. Evidence on the nature and sources J. 1916. Klepper / Journal of Urban Economics 67 (2010) 15–32 Proposition 3. pp. Slugs. 7–39.. 1971.. David B. 1980. The Automobile and American Culture. manufacturing industries: a dartboard approach. Liquidity events and the geographic advantage: location decisions of Portuguese entrepreneurs. industrial organization. Bailey. Nilo. Strange. mobility: specialists. Iola.