The effect of the innovative environment on exit of entrepreneurial...

Strategic Management JournalStrat. Mgmt. J., 27: 519–539 (2006)

Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/smj.534

THE EFFECT OF THE INNOVATIVE ENVIRONMENTON EXIT OF ENTREPRENEURIAL FIRMS

M. B. SARKAR,1 RAJ ECHAMBADI,1 RAJSHREE AGARWAL2* andBISAKHA SEN3

1 College of Business Administration, University of Central Florida, Orlando, Florida,U.S.A.2 College of Commerce and Business Administration, University of Illinois atUrbana–Champaign, Champaign, Illinois, U.S.A.3 School of Public Health, University of Alabama at Birmingham, Birmingham,Alabama, U.S.A.

To foster ‘creative destruction,’ entrants must survive the turbulent conditions they face in theirfirst crucial years in the industry. We investigate how the external knowledge milieu of an entrant,conceptualized as its innovative environment, causes systematic variation in survival patterns.We test our model from 3,431 firms in 33 industries over 80 years. We depict the innovativeenvironment along two knowledge-related dimensions, namely technology regime and technologyintensity. While the aligned state of the innovative environment, where product innovation existsin tandem with abundant innovation opportunities, promotes entrant survival, we find that thisbeneficial effect is more pronounced for small entrants due to a possible mitigation of scaledisadvantages. Copyright 2006 John Wiley & Sons, Ltd.

The entry of new firms, the reorganization andrationalization of existing firms, and subsequentexits are all recognized as evolutionary forces thatshape competitive dynamics within markets. Suchforces are unleashed when entering firms intro-duce innovations that are based on new techno-logical knowledge or customer insights (Hayek,1945; Schumpeter, 1934). Knowledge, therefore,is the genesis of entrepreneurial entry (Venkatara-man, 1997). However, although contemporary per-spectives have typically focused on the strategicbenefits of knowledge that is internal to a firm, the

Keywords: entrepreneurship; technological intensity; in-dustry evolution; innovation*Correspondence to: Rajshree Agarwal, College of Commerceand Business Administration, University of Illinois at Urbana–Champaign, Champaign, 1206 South Sixth St, Champaign, IL61820-6980, U.S.A. E-mail: [email protected]

evolutionary and innovation literatures suggest theintriguing possibility that knowledge conditions inthe external environment of an entrant may alsocarry important implications for its survival. In thispaper, we draw attention to how knowledge shapesthe external milieu, or the innovative environmentof an entering firm, and examine the implicationof different types of such environments for entrantsurvival.

Studying the root causes of survival of industryentrants is fundamental to understanding whetherentering firms will survive in the focal marketlong enough to have a discernible effect on it.By influencing the rate at which economicallyvaluable knowledge is created (Arrow, 1962),entrepreneurial entry impacts economic welfareand growth (Acs et al., 2003). Thus, as agentsof change, entrants can profoundly influence the

Copyright 2006 John Wiley & Sons, Ltd. Received 25 March 2003Final revision received 7 November 2005

520 M. B. Sarkar et al.

social diffusion of innovations. However, the vul-nerability of a large number of entrants to liabili-ties of newness and smallness has been emphasized(Freeman, Carroll, and Hannan, 1983; Stinch-combe, 1965). In fact, empirical findings of endem-ic entrant failure (Huyghebaert and Van de Gucht,2004) vindicate Geroski’s (1995) claim that entryis often times easier than survival. Of particularconcern is whether the pattern of change in marketsresembles a ‘revolving door,’ where the majorityof entrants exit soon after they enter, rather than‘creative destruction,’ where entrants displace oldfirms that have lost their ability to create value.However, our understanding of the forces govern-ing change in an industry and the conditions underwhich small, new firms can survive is limited. AsShane (2001) wrote, there is a need to explorevariance in the exit rates of entrants to accuratelydetermine the level of entrepreneurial churn in anindustry.

Previous research suggests that it may makeparticular sense to study the confluence of entrytiming, demographics, and industry in order tofurther our understanding of the phenomena ofentrant survival. First, both organizational ecol-ogy and economics research indicates that early-year survival depends on an entering firm’s size(Audretsch and Mahmood, 1995; Freeman et al.,1983). Second, studies also indicate that entrantsurvival may be influenced by the technologyregime, or the evolutionary stage, of the indus-try at the time of entry (Agarwal, Sarkar, andEchambadi, 2002). In this vein, evolutionary schol-ars examine how the relevance of a particulartype of knowledge base to innovation changesas the industry progresses through its life cycle,and how such transformations impact competitiveadvantage that accrues to different types of firms(Gort and Klepper, 1982; Nelson and Winter, 1982;Suarez and Utterback, 1995; Tushman and Ander-son, 1986). Third, there are indications that thescientific nature of an industry, as captured byits technology intensity, may impact competitivedynamics within industries and thereby influenceentrant survival (Geroski, 1990; Klevorick et al.,1995; Zahra, 1996). For example, investments inknowledge-producing activities may push out thetechnological frontier of an industry, and createmore innovation opportunities for new entrants(Kamien and Schwartz, 1982; Mansfield, 1968;Silverberg, Dosi, and Orsenigo, 1988).

In other words, previous research suggests thatpatterns of entrant survival may vary across time(when), industry (where), and size (how ) of entry.Although there may arise intriguing possibili-ties when considered in conjunction, the interplayamong these temporal, spatial, and scale dimen-sions of entry and their joint implications onentrant survival have been largely unexplored. Inthis paper, we attempt to address this lacuna inresearch. Towards this end, we propose that twoof these dimensions—technology regime (whichrelates to timing of entry) and technology inten-sity (which relates to the industry where entryis occurring)—comprise interrelated knowledge-based dimensions of an entrant’s innovative envi-ronment. Technology regime is related to the com-parative importance of knowledge that is possessedby entrants to that of incumbents with respect toapplicability to innovation. Technology intensity,on the other hand, relates to cross-sectional dif-ferences in the ‘inventive potential’ of industries(Rosenberg, 1974) and innovation opportunitiesthat are made possible by knowledge investmentsin an industry (Romer, 1986). Within this frame-work of the innovative environment where entrytakes place, we investigate how entrant size shapessurvival patterns. We report tests of our hypothesesusing longitudinal data on 3,431 firms that entered33 industries between 1908 and 1991.

THEORETICAL FRAMEWORK

Knowledge and the innovative environment

According to the knowledge-based view of thefirm, the generation, combination or recombina-tion, and exploitation of knowledge (Conner andPrahalad, 1996; Kogut and Zander, 1992) underliesall firm organizations. While contemporary worktends to view knowledge predominantly throughthe lens of competitive advantage (Barney, 1991;Grant, 1996), the entrepreneurship literature haslong acknowledged knowledge to be the fountain-head of innovative entry of new firms (Venkatara-man, 1997). When entrepreneurs seek to reap theeconomic value of unique and idiosyncratic knowl-edge stemming from technological breakthroughsand/or customer insights, they enter a marketwith their innovations (Hayek, 1945; Schumpeter,1934). Typically, the emphasis in the strategy lit-erature is on knowledge that a firm possesses,

Copyright 2006 John Wiley & Sons, Ltd. Strat. Mgmt. J., 27: 519–539 (2006)

Innovative Environment and Exit of Entrepreneurial Firms 521

and which is thus internal to a firm. However,rich research streams on industry evolution and oninnovation have emphasized that knowledge thatis external to a firm may impact entry and exitpatterns.

Our interest lies in being able to predict pat-terns of firm survival, which we define as con-tinued operations in the focal industry. Whilethe equivalence between economic performanceand survival has been questioned (Gimeno et al.,1997), we justify our focus on survival, or on itscorollary of exit, on the basis of theory, prece-dence and context. First, it has been argued thatwell-performing firms survive, while poor per-formers exit the market (Penrose, 1952; Win-ter, 1984; Williamson, 1991). Accordingly, sem-inal studies in population ecology (Freeman et al.,1983), IO economics (Geroski, 1995), and strat-egy (Agarwal et al., 2002) have considered firmsurvival as a valid organizational outcome. More-over, in the present context where we wish tostudy entrepreneurial churn as a precursor to newentrants being able to fulfill their larger role aschange agents, continued survival during the ini-tial years of heightened mortality is a first-ordercondition which takes precedence over any otherparameter of performance.

The evolutionary view suggests that the typeof knowledge that underlies technological inno-vation undergoes transformation over time as anindustry matures (Gort and Klepper, 1982; Nelsonand Winter, 1982). Accordingly, competitive con-ditions and survival advantages change over timeas an industry evolves. Complementary researchon innovation and technological progress highlighthow spillovers from knowledge investments cre-ate differences in innovation opportunities acrossindustries (Arrow, 1962; Jaffe, 1986; Mansfield,1968; Romer, 1986; Rosenberg, 1974). These per-spectives on temporal and cross-sectional differ-ences in knowledge conditions suggest that theirroles as complementary factors together determinethe innovative environment facing an entrant. Thefirst, technology regime, relates to the potential rel-evance of entrant vs. incumbent knowledge, whilethe second, technology intensity, relates to inno-vation opportunities available to entrants in theindustry. We define and discuss each dimension ofthe innovative environment, and then draw conclu-sions about their joint impact on entrant survivaland exit.

Technology regime

Scholars in economics, organizational ecology, andstrategy argue that evolutionary processes shapethe type of knowledge that forms the founda-tion behind firms’ innovative actions in an indus-try (Gort and Klepper, 1982; Hannan and Free-man, 1989; Nelson and Winter, 1982; Tushmanand Anderson, 1986; Utterback and Abernathy,1975). While emphasizing these dynamics of inno-vation, Nelson and Winter (1982) further proposethe notion of ‘regimes of technological change’as a way to describe the knowledge bases under-lying firms’ innovation processes, as well as themodal properties of their learning processes andsources of knowledge (Dosi, 1982). FollowingSchumpeter’s (1934) identification of historicalphases of economic development, Nelson and Win-ter (1982) distinguish between ‘entrepreneurial’and ‘routinized’ technology regimes. During theinitial entrepreneurial regime of an industry, thestock of industry-specific knowledge is low, andknowledge that is critical to innovation lies out-side established incumbent routines, thus favoringentrants rather than incumbents (Gort and Klepper,1982). As the industry matures into a routinizedregime, innovation is increasingly determined byaccumulated stocks of non-transferable, internal-ized, market-based expertise. Therefore, while anentrepreneurial regime facilitates innovation bynew firms, the routinized, experience-based regimehelps innovation by industry incumbents (Dosi,1982; Gort and Klepper, 1982; Winter, 1984). Inother words, the knowledge-based advantage ofentrants over incumbents reverses over time, as doex ante prospects for making technical advances.

We note that technology regimes are analogousto Tushman and Anderson’s (1986) ‘era of fer-ment’ and ‘era of incremental change.’ The con-cept of regime is also consistent with work oncompetitive density, legitimation, and competitionin organizational ecology (Hannan and Freeman,1989). Further, Gort and Klepper (1982) describefive stages of a product or technology life cycle,and Klepper and Graddy (1990) divide this lifecycle into three stages. Integrating the different lit-erature streams, Agarwal et al. (2002) note thatthe entrepreneurial regime (which they call the‘growth phase’) and the routinized regime (the‘mature phase’) are separated by a major changein competitive conditions, which suppresses entry



into the observed industry. Regardless of differ-ences in the naming and numbering of stagesof industry evolution, the notion that knowledge-based advantage changes sides over time appearsconsistently in the relevant scholarly literature.Entrants have an innovation advantage over estab-lished firms in an initial period, but establishedfirms have the innovation advantage in a laterperiod. Here, we refer to periods in the knowledgehistory of industries as technology regimes.

Technology intensity

We conceptualize technology intensity as the de-gree to which the industry invests in creative activ-ities that increase the stock of scientific knowledgeand its use in new applications. Research sug-gests that technology intensity reflects not onlythe munificence of innovation opportunities withinthe industry, but also the ability of firms toappropriate economic returns from new devel-opments (Klevorick et al., 1995). Investments inknowledge-generating activities not only have a‘spillover’ effect on innovation opportunities(Arrow, 1962; Mansfield, 1968), but more impor-tantly, technology-intensive industries can sustainhigher rates of technological opportunities becauseeven though firms may be rapidly exploiting pre-vailing opportunities, new ones are being createdequally rapidly (Allen, 1977; Jaffe, 1986). Theinherent relationship of investments in knowledge-producing activities with opportunities for innova-tion has implications for entrepreneurial entry andthe pattern of change in an industry.

First, technological opportunities are a by-pro-duct of knowledge investments. Knowledge spill-overs from scientific discoveries stimulate tech-nological opportunities, which in turn create animpetus for innovation and entrepreneurial entry(Acs et al., 2003; Levin and Reiss, 1988). Invest-ments in knowledge-generating activities thus cre-ate positive externalities and growth opportunities(Arrow, 1962; Mansfield, 1968; Romer, 1986). Inother words, the public good aspect of knowl-edge and associated spillover effects in the indus-try implies that technology intensive environmentsare likely to have a greater magnitude of inno-vation opportunities that entrepreneurial entrantscan take advantage of (Acs et al., 2003). Forinstance, recent work indicates that, due to thepermeable nature of organization boundaries, tech-nological investments by incumbent firms create

entrepreneurial opportunities for their employees(Agarwal et al., 2004; Klepper, 2002).

Second, technology intensity increases the num-ber of ways in which an entrant’s offering can bedifferentiated from existing products (Blair, 1972;Comanor, 1967). Scientific and market break-throughs from knowledge investments tend toengender products constantly increasing in sophis-tication (Shaked and Sutton, 1987). Often, thesescientific breakthroughs result in opportunities tocreate new markets that are ignored by existingfirms but exploited by entrants (Christensen, 1997).According to Comanor, expenditures in research‘foster and promote a rapid rate of new productintroduction, which then serves to facilitate theachievement of differentiation’ (Comanor, 1967:646). Not surprisingly, therefore, it has been notedthat high-technology industries, or those character-ized by high investment in scientific know-how,experience more entrepreneurial activity and fos-ter faster new product and process innovation (Acsand Audretsch, 1988; Geroski, 1990; Zahra, 1996).Since rapid innovation tends to ‘deconcentrate’markets (Blair, 1972; Geroski and Pomroy, 1990),it seems that technology intensity may have impli-cations not only for the entry of innovators, butalso for their survival.

The innovative environment and entrantsurvival

We combine the two complementary dimensionsof the innovative environment discussed above,namely, technology regime and technology inten-sity, to define four states of the innovative envi-ronment facing an industry’s potential entrants.Figure 1 shows these four states. We name theentrepreneurial regime of high-technology-inten-sity industries the aligned environment (Cell I),since in this state entrants possess a knowledgeadvantage and innovation opportunities are abun-dant. In direct contrast, the nonaligned environ-ment (Cell IV), or the routinized regime of low-technology-intensity industries, represents condi-tions in which entrants lack a knowledge advan-tage and have fewer innovation opportunities. Theother two states are partially aligned, since entrantseither have either fewer innovation opportunities(Cell II) or a knowledge disadvantage (Cell III).

Entrants in the aligned (Cell I) environment arelikely to enjoy a better survival rate than thosein any other quadrant. Unlike entrants in either



Technology Intensity High Low

Entrepreneurial

I: Aligned Environment:

Entrant Knowledge Advantage

More Innovation Opportunities

II: Partially-aligned Environment

Entrant Knowledge Advantage

Fewer Innovation Opportunities

Technology

Regime

Routinized

III: Partially-aligned Environment

Entrant Knowledge Disadvantage

More Innovation Opportunities

IV: Nonaligned Environment

Entrant Knowledge Disadvantage

Fewer Innovation Opportunities

Figure 1. The innovative environment facing entrants

the nonaligned Cell IV or the partially alignedCell III, firms that enter in Cell I experience theentrepreneurial regime of an industry. As discussedabove, during this period, the stocks of industry-specific knowledge and routinized information thatwould favor incumbents are relatively low (Acsand Audretsch, 1988). In Cell I, the knowledgeadvantage the entrepreneurial regime confers isfurther reinforced by the presence of innovationopportunities conferred by the technology inten-sity of the industry, thus providing entrants inthe aligned environment an advantage over thefirms entering the partially aligned Cell II. Asdiscussed above, positive spillovers of technologyand increased scope for differentiation in indus-tries with high technology-intensity permit entrantsto seize innovation opportunities (Kessides, 1991)or occupy strategic niches (Porter, 1980), thusincreasing their probability of survival. In short,entrants into an aligned environment can benefitfrom the synergy of relevant knowledge and abun-dant opportunities.

On the other hand, the nonaligned Cell IV rep-resents a hostile environment for entering firms,because they not only face an unfavorable tech-nology regime, but are also starved of innovationopportunities. The routinized regime implies thatincumbents have the knowledge advantage vis-a-vis entrants, and the low technology intensityimplies that there are few opportunities for enjoy-ing spillover or differentiation benefits. As a result,

these entrants are disadvantaged with respect toboth dimensions of innovative environment.

In the partially aligned Cells II and III, the situa-tion is less clear. In Cell II, knowledge conditionsare favorable because the entrepreneurial regimeprevails, but technological opportunities are lack-ing because technology intensity is low. In Cell III,technological opportunities abound because tech-nology intensity is high, but knowledge condi-tions disadvantage entrants because the routinizedregime prevails. Clearly, these partially alignedenvironments are not as good as Cell I, but arepreferable to Cell IV. Differentiating between theentrant conditions in Cells II and III requires judg-ing the relative importance of the two dimen-sions. On the one hand, if appropriate knowledgeis a precursor to exploiting market opportunities,Cell II is likely to be better for entrants thanCell III. On the other hand, if abundant marketopportunities will allow entrants to occupy mar-ket niches left vacant by incumbent organizations,even though these entrants have knowledge disad-vantages, Cell III appears more favorable than CellII for entrants. Therefore, because there are strongtheoretical arguments in both directions, we treatCells II and III similarly in our hypothesis below,and leave the resolution as an empirical issue.

Hypothesis 1: New entrants into an alignedinnovative environment are less likely to exitthe market than new entrants into nonaligned



innovative environments. New entrants into par-tially aligned environments will have intermedi-ate likelihoods of exit.

Entry size and the innovativeenvironment–survival relationship

In the above discussion, we did not distinguishamong the firms that enter a new industry. Inessence, we treated entrants as homogeneous, sinceour focus was the innovative environment theyface. We now turn to a key entry characteristicof a firm—one that has received significant atten-tion—the size or scale of entry. We posit that sizemay moderate the relationship between innovativeenvironment and entrant survival.

Studies in organizational ecology (Freemanet al., 1983; Hannan and Freeman, 1984) andindustrial organization (Dunne, Roberts, andSamuelson, 1988; Sutton, 1997) indicate a posi-tive relationship between firm size and survival.Organizational ecologists propose that the liabilityof smallness stems from three factors: the selec-tion processes that favor the structural inertia oflarger organizations (Hannan and Freeman, 1984),their access to capital and trained manpower andtheir legitimacy with external stakeholders (Baumand Oliver, 1991). In economics, size advantageemanates from market power (Bain, 1956) andminimum efficient scale considerations (Jovanovic,1982; Mansfield, 1962). Thus, small entrants suffera liability of smallness because they are likely tolack production and procurement economies, insti-tutional support and linkages (Baum and Oliver,1991; Stinchcombe, 1965), refined routines, andthe ability to produce outputs of consistent quality(Hannan and Freeman, 1984).

Since small entrants lack the additional cush-ion of size that is available to large entrants thereceptivity of an environment becomes particularlyimportant, and an aligned innovative environmentmay be necessary for high levels of survival. Giventhe vulnerabilities of small entrants, an environ-ment that is not only rich in innovation opportu-nities, but also gives them a knowledge advantageis likely to facilitate their survival more than anyother type of environment. On the other hand, largeentrants may additionally find other states of theinnovative environment, particularly the partiallyaligned ones, favorable to their survival prospects,because they do not have the liability of small-ness. Their larger scale of entry and access to

deep pockets and resources may shield them inan environment that lacks perfect alignment ofthe dimensions that facilitate entry and entrantsurvival. While entering an aligned environmentwould certainly benefit large entrants, their scaleof entry may mitigate the disadvantage of eitherfew innovation opportunities or no entrant knowl-edge advantage. Our next hypothesis reflects thisdifference in the innovative environment–survivalrelationship for small and large firms.

Hypothesis 2: Entry size moderates the rela-tionship between innovative environment andlikelihood of exit such that the aligned innova-tive environment will be more crucial for smallentrants than for large entrants.

METHOD

Data

We used longitudinal data on firm entry and exitfrom 1908 to 1991 in 33 manufacturing indus-tries to test our hypotheses. The Thomas Registerof American Manufacturers, our source of infor-mation for firm-level data, is a buying guide forthe full range of products manufactured in theUnited States (Lavin, 1992: 129). Researchers ineconomics, strategy, and marketing have relied onthe guide to study issues related to the diffusionof innovations and market evolution (see Gort andKlepper, 1982; Klepper and Simons, 2000; Robin-son and Min, 2002). The publishers of the ThomasRegister seek to obtain the complete representationof domestic manufacturing activity by analyzing abroad range of industry newsletters as well as start-up ventures in university incubators. Moreover, theguide does not charge a fee for inclusion, whichfurther fosters its completeness and accuracy.

The 33 manufacturing industries included in ourstudy are drawn primarily from the Gort and Klep-per (1982) study of 46 industries that resulted fromproduct innovations. We designated the first yearthat an industry appeared in the Thomas Registeras the year of an innovation’s commercial intro-duction.1 The Gort–Klepper study tracked only the

1 Gort and Klepper (1982) supplemented the Thomas Registerwith other sources for some of their products. As a result,their year of commercialization sometimes preceded the yeara product was first listed in the Thomas Register. However,since the number of firms that entered in the early years for the



number of firms existing in these industries, andtheir data are censored at 1973. We developed thedata independently and extended the time periodthrough 1991. As a result, several of the indus-tries in the Gort–Klepper study could not be usedfor new data development. Some industries (e.g.,nylon, computers) experienced substantial changesin definition over the years for which we con-ducted the analysis. Other industries (e.g., DDT,cryogenic tanks) were omitted due to their discon-tinuation over the years for which the analysis was

few products for which Gort and Klepper (1982) had additionalinformation is very small, we do not expect our reliance on theThomas Register for the systematic identification of the year ofcommercialization to yield substantively different results.

extended. Still other categories (e.g., streptomycin,penicillin) were discarded in favor of the broaderindustry group of antibiotics. A few industrieswere not included in the analysis owing to timelimitations on the development of data. Finally,two new industries—contact lenses and video cas-sette recorders—were included since they gainedprominence after the Gort–Klepper study was pub-lished. The final set of 33 industries (see Table 1)compares favorably with the number of industriesinvestigated by other historical studies (cf., forexample, Sultan, Farley, and Lehman, 1990).

Two sets of research assistants independentlymade lists of the firms that entered and exited eachindustry from its first listed year through 1991. Inaddition to the names and addresses of firms, the

Table 1. List of industries in the sample

Industry name Year ofcommercialintroduction

Onset ofroutinizedregimea

Technologyintensity

Number ofentrants in

industry

Number of firms thatexited industry

within 5 years of entry

Antibiotics 1948 1968 High 66 17Artificial Xmas Trees 1938 1962 Low 56 19Ball-point Pens 1948 — Low 226 85Betaray Gauges 1956 1971 High 19 3Cathode Ray Tubes 1935 1967 High 122 41Combination Locks 1912 1942 Low 93 33Contact Lenses 1936 — High 73 15Electric Blankets 1916 1963 Low 47 11Electric Shavers 1937 1949 Low 58 31Electrocardiographs 1942 1975 High 41 8Freezers 1946 1966 Low 132 45Freon Compressors 1935 1980 Low 74 29Gas Turbines 1944 — High 138 35Guided Missiles 1951 1966 High 386 90Gyroscopes 1915 1978 High 121 36Heat Pumps 1954 — Low 117 38Jet Engines 1948 1965 High 79 14Microfilm Readers 1940 1977 High 94 31Nuclear Reactors 1955 1965 Low 82 24Outboard Motors 1913 1974 Low 119 45Oxygen Tents 1932 1965 High 48 16Paints 1934 1967 Low 221 36Phonograph Records 1908 1927 Low 237 88Photocopying Machines 1940 1971 High 94 34Piezoelectric Crystals 1940 1964 High 93 27Polariscopes 1928 1959 High 42 12Radar Antenna Assemblies 1952 1966 High 113 38Radiant Heating Baseboards 1947 1965 Low 46 6Radiation Meters 1949 1967 High 65 24Recording Tapes 1952 — Low 160 54Rocket Engines 1958 1966 High 38 4Styrene 1938 1980 Low 89 23Video Cassette Recorders 1972 — Low 44 12

a Six industries in the study did not exhibit the onset of entry barriers for the period under investigation.



following firm-specific data were collected: yearof entry, year of exit, asset size, and diversifica-tion index. The year of entry (exit) was designatedas the first year that the firm was listed (delisted) inthe Thomas Register in the industry records. Theasset size was obtained from the entry year list-ing. Information on diversification was obtained byconsulting the annual firm index volumes of theThomas Register for the year preceding a givenfirm’s industry entry to see if the firm had pro-duced in any other manufacturing category prior toits entry in the focal industry. We then comparedthe data from the two sets of assistants to rec-oncile discrepancies, rectify mistakes, and ensurethe accuracy of our records. For example, beforeclassifying an event as an entry or as an exit, weverified data from successive years. We comparedthe names, addresses, and other relevant informa-tion of firms to ensure that an actual entry/exit hadoccurred, and excluded the cases in which exist-ing firms had merely been renamed or relocated.Identifiable mergers were treated as the continu-ance of the larger firm and the exit of the smallerfirm (Mansfield, 1962).2 Additionally, we obtained

2 Coding firm exits in this manner, while consistent with otherstudies (Dunne et al., 1988; Robinson and Min, 2002), leadsto the fundamental question: Does every merger/acquisitionindicate a firm’s failure, or do some reflect only success?Although redefining risks of exits that allow this delineationmight prove extremely useful, data limitations did not allow us tomake this distinction. Moreover, a check revealed that less than3 percent of our exits were attributable to identifiable mergers,

information on the R&D intensity of the indus-tries from the Survey of Industrial Research andDevelopment conducted by the National ScienceFoundation.

Our data provide several advantages. First, theyrepresent a great number of industries that resultedfrom product innovations, and these industry histo-ries span almost the entire twentieth century. Thesefeatures lend a potential for generalizability lack-ing in single-industry studies. Second, our use ofobjective, historical data from industry directoriesensures that our study does not suffer from poten-tial self-report bias of survey-based data. Third,using data recorded at the time of event occurrenceenables us to better study time-related effects with-out the fear of survivor bias. Finally, future repli-cation and validation studies are feasible, since ourdata were compiled from secondary data sources.

Variable definitions

Entrants are defined as firms who are less than5 years old in the focal industry. The 5-year timespan has been used in earlier studies of entrantsurvival (e.g., Baldwin, 1995; Robinson and Min,2002), and is consistent with our focus on early-year survival patterns. Our final dataset consistsof 3,431 firms for a total of 14,173 firm–yearobservations. Table 2(a) concisely describes the

leading us to believe that our substantive results would not varyas a result of this limitation.

Table 2(a). Operationalization of key variables appearing in the model

Variable name Variable description

Firm exit = 1 if the firm exited; = 0 if the firm survivedTechnology regime = 1 if the firm entered in the entrepreneurial regime; = 0 if the firm entered in the routinized

regimeTechnology intensity = 1 if the firm entered a high-technology industry; = 0 if the firm entered a low-technology

industrySmall size at entry = 1 if the firm entry size is in the smallest two asset categories; 0 if the firm entry size is in

the 3rd, 4th, and 5th asset categoriesAge The number of years since the time of firm entry into the industryAge2 A squared term of age to account for nonlinear effects of ageDiversifying entrant = 1 if the firm existed in some other industry prior to entry in focal industry; = 0 if the firm is

a de novo entrantDensity The number of firms in the industry at the time of founding relative to the peak number of

firms in the industryConsumer good = 1 if the firm entered a consumer industry; = 0 otherwiseLagged entry rate The 1-year lagged values of entry rateLagged exit rate The 1-year lagged values of exit ratePost-World War II = 1 if the firm entered the industry post-World War II; = 0 otherwise



Table 2(b). Means, standard deviations, and correlations of key variables

Variable Mean S.D. 1 2 3 4 5 6 7 8 9 10 11

1 Firm exit 0.07 0.26 12 Technology regime 0.69 0.46 0.03 13 Technology intensity 0.44 0.50 0.02 0.07 14 Small size at entry 0.68 0.47 −0.03 −0.06 −0.04 15 Density 0.62 0.27 −0.02 −0.19 0.06 0.06 16 Diversified entrant 0.64 0.48 0.04 0.03 0.07 −0.21 0.04 17 Post-World War II 0.89 0.32 0.03 −0.16 0.17 −0.03 0.30 0.18 18 Consumer good 0.52 0.50 −0.02 0.12 −0.30 0.06 −0.05 −0.11 −0.12 19 Age 2.80 1.40 −0.01 0.05 0.01 −0.02 −0.04 0.02 0 −0.01 110 Age2 9.81 8.40 0 0.04 0.01 −0.02 −0.04 0.02 0 −0.01 0.98 111 Lagged entry rate 0.17 0.31 0 0.16 0.04 −0.03 −0.33 −0.01 −0.15 −0.03 −0.24 −0.21 112 Lagged exit rate 0.06 0.06 −0.05 −0.08 −0.03 0.02 0.09 −0.02 −0.09 0.03 −0.03 −0.03 0

n = 14,173 observations

measurement of the variables in the study, andTable 2(b) provides the corresponding descriptivestatistics and correlation matrix.

Dependent variable: firm exit

Our dependent variable, firm exit, is coded as1 if a firm exited the focal industry in a givenyear, and 0 otherwise. Since firms exiting thefocal industry may still be in existence in otherindustries, we refrain from using the term ‘firmfailure’ synonymously with firm exit. For exam-ple, for a firm that exited 3 years after entry,the dependent variable took the value of 0 forthe first 2 years and the value of 1 in the thirdyear. For a firm that survived for 5 years afterits entry, which is the survival period of inter-est in this study, the dependent variable is coded0 for all five yearly observations. As noted inTable 2(b), about 7 percent of firms exited in anygiven year.

Key explanatory variables

The three explanatory variables in our study aretechnology regime, technology intensity, and firmsize at time of entry in the focal industry. Whiletechnology intensity and size may also be mea-sured on a continuous scale, we use dichoto-mous category values for all three variables in ouranalysis. We are motivated to do so in light ofour category-based research framework, combinedwith a methodological need to interpret three-wayinteractions (Gruber, 1994), as detailed in the fol-lowing section.

Technology regime

Industry evolution studies have used patterns inthe number of firms, net entry, and gross entry,as depicted in Figure 2, to distinguish amongthe stages of the industry life cycle (Agarwalet al., 2002; Gort and Klepper, 1982; Klepper andGraddy, 1990). We base our measure of technologyregime on Agarwal et al.’s (2002) work becauseit differentiates the entrepreneurial and the rou-tinized regimes on the basis of gross entry rates,rather than net entry (gross entry less gross exit), oron the basis of number of firms (Gort and Klep-per, 1982; Klepper and Graddy, 1990). Agarwalet al.’s measure of gross entry satisfies the impor-tant condition that the operationalization of lifecycle regimes for a study investigating firm exitrates not be functionally related to the dependentvariable of firm exit.

Accordingly, we define the entrepreneurial re-gime as the period extending from the commercial-ization of an innovation to the significant reductionin entries (that is, the large hill in the gross entrypattern depicted in Figure 2), and the routinizedregime as the period after that point (that is, fromthe period of low/zero entry to the subsequentresurgence, as depicted in Figure 2). Statistically,to distinguish between two consecutive intervals,we determine the delineating year between the tworegimes using the generalized discriminant anal-ysis procedure first used by Gort and Klepper(1982). Briefly, this methodology allows us to dis-tinguish between any two consecutive intervals byexamining the data on annual gross entry rates foreach industry. To determine the delineating year,we first partition the entry rate series into three



Entrepreneurial Regime

Number of Firms

Number of Entrants

Routinized Regime

Time from Commercialization of Product Innovation

Figure 2. Stylized patterns of entry and number of firms over the industry life cycle

categories—the first and third categories containthe years where the gross entry rates clearly reflectthe entrepreneurial and routinized regimes. Peri-ods for the ‘in-between’ years are then optimallyclassified based on mean values. The Appendixdescribes in detail the application of this procedureto distinguish the two regimes.

Table 1 identifies the delineating year betweenthe two regimes for each industry. We conductednumerous sensitivity tests to ensure that our resultswere robust to changes in the year that separatedthe regimes. Also, since six industries did notexperience the onset of routinized regimes by theend of our sample period, we conducted robust-ness checks to ensure that the results were notsensitive to the inclusion or omission of theseindustries. Consistent with the patterns depictedin Figure 2, the descriptive statistics in Table 2(b)show that 69 percent of all firms entered duringthe entrepreneurial regime of the industries.

Technology intensity

Technology intensity captures the investment inknowledge-producing activities in an industry.Accordingly, we base our measure on total (com-pany, federal, and other) industrial R&D fundsallotted as a percentage of sales. These figures areprovided at the three-digit SIC (Standard Indus-trial Classification) level in the Survey of Indus-trial Research and Development. For the 1987–97period, the average R&D intensity for all man-ufacturing industries in the United States was

4.7 percent. We define those industries in our sam-ple that were above the average R&D intensityas high-technology-intensity industries and thosethat were below the average as low-technology-intensity industries. An examination of the R&Dintensity values of the industries in our samplerevealed a distinct clustering at the low and highends of the scale, suggesting that the use of thisdichotomous measure did not result in a loss ofsignificant information.

Since technology intensity is a critical inde-pendent variable in our analysis, we conductedrobustness checks to ensure the validity of ouroperationalization. In accordance with Robinsonand Min (2002), we experimented with a labor-based measure for R&D intensity, using Had-lock, Hecker, and Gannon’s (1991) classification,and obtained similar results. We also examinedwhether the categorization results changed whenthe dividing line was specified differently. Theuse of the median value of R&D intensity tosplit the sample into high- and low-technology-intensity industries resulted in identical catego-rization results. Finally, we compared the R&Dintensity of our sample to the R&D intensity ofall industries at the 3-digit SIC level, and usedthe sample mean and median as alternative divid-ing lines. Our categorization results were virtu-ally identical for alternate specifications, therebyindicating that our dichotomous measure of R&Dintensity possessed adequate face validity.

We note that our measure of technology inten-sity across industries does not vary over time,



since the temporal dimension is being capturedin the technology regime, or life cycle, measure.Although industries may undergo periods of peaksand troughs in the amount of effort and resourcethey devote to knowledge-generating activities,cross-industry variations in technology intensityseem to be relatively robust. As Klevorick et al.note, ‘[A] striking characteristic of industries thatare commonly thought to be rich in technologicalopportunities is that high R&D intensities and highrates of technical advance tend to be sustained overtime’ (Klevorick et al., 1995: 188). Accordingly,we believe that, although not without limitations,a time-invariant measure of technology intensity isadequately valid. As reported in Table 2(b), 45 per-cent of the firms entered high technology-intensityindustries.

Firm size

Firm size at the time of entry is measured by itsasset size as listed by the Thomas Register in theyear a firm entered a relevant market. Althoughfirm size has been otherwise measured in a varietyof ways, such as by employees, sales, asset value,the empirical results have so far demonstrated thatall of these definitions of firm size provide verysimilar results (Child, 1972; Chandy and Tellis,2000).

Each firm’s asset values are reported in nominal,or at current dollar value, asset categories in theThomas Register, and the data span over 80 years,so there was a need to control for inflation. Accord-ingly, we recorded the asset value of each firmin the year it entered the industry of interest, andconverted it to 1982 U.S. dollars to obtain a realdollar value. On the basis of their real adjusted val-ues, the firms were then classified into five assetcategories. Fifty percent of the firms had assets lessthan $2.8 million; 12 percent had assets between$2.8 to $5.5 million; 5 percent had assets between$5.5 to $8.3 million; 32 percent had assets between$8.3 and $11 million; and less than 1 percent ofthe firms had assets greater than $11 million. Thesize distribution based on asset size is consistentwith the distribution by employee size reported byChandy and Tellis (2000), and is also generallyrepresentative of manufacturing industries.

Given the bimodal distribution in the size offirms, and since our hypotheses concern the differ-ing impact of the innovative environment on firmsurvival for small and large firms, we classified the

firms in the two lowest asset categories, with realvalue less than $5.5 million, as small; and clas-sified all other firms as large. Robustness checksrevealed that the effect of size on survival was thesame whether we used the all five asset classes,or the dichotomous measure of small vs. large.The latter was chosen for ease of exposition andanalysis. Further, when we tested an alternativeoperationalization of large firms, dropping the thirdasset category ($5.5 million to $8.3 million) fromthe large firm category, our results did not change.To corroborate our empirical operationalization ofsize, we also experimented with an alternative clas-sification. Defining small firms as those below the60th percentile of the size distribution of the firmsin their decade of entry produced largely simi-lar results, thus demonstrating the robustness ofour operationalization. The descriptive statistics inTable 2 show that small entrants accounted forapproximately 68 percent of the firms operatingin any given year.

Control variables

In addition to the above variables of interest, weinclude several firm- and industry-level controlsto account for both fixed and time-varying effects.Our firm-level controls included the firm’s age, anda diversification dummy. A quadratic specificationfor age is used to allow for nonlinear effects ofage. We further include a diversification dummy(1 = diversified, 0 = otherwise) to control for dif-ferences between diversified and de novo entrants.Our industry-level controls included the measureof founding density, or the number of firms attime of entry, because the organizational ecol-ogy literature suggests that firm exit varies withthe number of firms in a given population. Giventhat our study encompassed many industries, therewere differences in numbers of entrants per indus-try, as revealed in Table 1. To accommodate thesedifferences and to enable cross-industry compar-isons, we computed a relative measure of densityby dividing the number of firms in an industry inany given year by the maximum number of firmsobserved in the industry, which is represented bythe peak of the number-of-firms curve depictedin Figure 2. Further, we used a dummy variableto distinguish between consumer goods and com-ponent or industrial goods, because the demandconditions facing firms in each type may differ.To control for temporal differences, we calculated



lagged entry and exit rates in an industry. Entry (orexit) rates are computed as the number of entrants(or exits) in a year divided by the total numberof firms that existed in the industry in the preced-ing year. We used 1-year lagged values rather thancontemporaneous values for entry and exit ratesto avoid concerns of endogeneity of populationlevel entry and exit rates with our dependent vari-able of firm-level exit. We also included a dummyvariable to differentiate between firms that enteredbefore and those that entered after World War II,because the war was a major economic event thatsignificantly altered the business environment inthe United States. For example, historical evidencesuggests that after World War II large firms insti-tuted organizational features that better supportedradical innovation (Chandler, 1956).

Model specification and estimation

We tested our hypotheses by examining entranthazard rate, which is the probability of a firmnot surviving another year after a particular age.Although several discrete- and continuous-timeanalysis techniques are available (Allison, 1995),we chose the random-effects complementary log-log model to control for unobserved heterogeneityamong industries that our control variables may nothave captured. As described by Allison (1995), thecomplementary log-log distribution is preferableto other distributions (e.g., logistic), because itaccommodates for the possibility that exits canoccur at any point during a given year, even thoughthe exits are recorded only at yearly intervals. Also,we use random effects to control for unobservedheterogeneity because fixed effects models arenot available in the existing statistical packages.Unconditional fixed effects estimates are biased,and a sufficient statistic allowing the fixed effectsto be conditioned out of the likelihood does notexist. To ensure robust results, we tested additionalmodel specifications, including the probit, logistic,and Cox proportional hazards, and our results wereconsistent across specifications.

The empirical analysis of our hypotheses re-quired testing contingencies or interactions alongthe following three distinct dimensions: technol-ogy regime, technology intensity, and size. Giventhe complexities of interpretation, and collinearityproblems associated with three-way interactionmodels, we employed the variant of a techniquedeveloped by Gruber (1994) in the economics

literature. The technique categorizes firms intoexclusive groups depending on the values taken byeach dichotomous independent variable of interest.These n groups are then represented by (n − 1)dummy variables, the last group being the con-trol group. This method nests traditional interac-tion models and enjoys several advantages, par-ticularly with respect to three-way interactions.First, these models did not suffer from potentialmulticollinearity because Gruber’s technique usesindependent variables to classify observations intomutually exclusive groups; moreover, in this study,these groups depended on size and innovative envi-ronment. Second, Gruber’s technique allowed theinvestigation of differences in firm exit in differ-ent environments, at different sizes, and permittedthe immediate comparison of the probability ofthe survival of each group relative to the controlgroup. Thus, the technique circumvents problemsin the interpretation of the coefficients of inter-action terms in traditional three-way interactionanalysis, and prevents the need for additional sub-group analysis. Third, Gruber’s technique providesa generalized and unrestricted model against whichthe validity of models that suggest the survivalprobabilities in any two of the groups to be thesame can be easily tested using likelihood ratiotests.

For the test of Hypothesis 1, we specify thefollowing model:

h(t) = γ1 T1 + γ2 T2 + δXit (1)

where:

T1 = 1 for entrants into the aligned innovativeenvironment (Cell I in Figure 1), and 0 other-wise;

T2 = 1 for entrants into partially aligned environ-ments (either Cell II or Cell III in Figure 1),and 0 otherwise;

and Xit represents the vector of control variablesfor firm i at time t .In Equation 1 the underlying control group con-sists of entrants into the nonaligned environment(Cell IV in Figure 1). The coefficients γ1 and γ2

can then be interpreted as showing the differencein firm exit of each group of firms from that of thecontrol group.

For the test of Hypothesis 2, we use two dif-ferent models. First, we conduct a subgroup anal-ysis and test the innovative environment–firm



exit relationship shown in Equation 1, for smalland large entrants separately. Second, we estimatethe full model for the three-variable interactionsin which eight groups are represented by sevendummy variables. Thus, the model is specified as:

h(t) = β1 ∗ G1 + β2 ∗ G2 + β3 ∗ G3 + β4 ∗ G4

+ β5 ∗ G5 + β6 ∗ G6 + β7 ∗ G7 + αXit (2)

where:

G1 = 1 for large entrants into the aligned innova-tive environment (Cell 1), and 0 otherwise;

G2 = 1 for large entrants into the entrepreneurialregimes of low-technology-intensity industries(Cell 2), and 0 otherwise;

G3 = 1 for large entrants into the routinized re-gimes of high-technology-intensity industries(Cell 3), and 0 otherwise;

G4 = 1 for large entrants into the nonaligned inno-vative environment (Cell 4), and 0 otherwise;

G5 = 1 for small entrants into the aligned innova-tive environment (Cell 1), and 0 otherwise;

G6 = 1 for small entrants into the entrepreneurialregimes of low-technology-intensity industries(Cell 2), and 0 otherwise;

G7 = 1 for small entrants into the routinizedregimes of high-technology-intensity industries(Cell 3), and 0 otherwise;

and Xit represents the vector of control variables.

In Equation 2, the control group consists of smallentrants into nonaligned environments (Cell 4).The coefficients β1 through β7 can then be inter-preted as showing the difference in firm exit ofeach group of firms from that of the control group.

RESULTS

Table 3 provides the results of our analysis forHypothesis. Model 1 in Table 3(a) reports the coef-ficients for the two groups defined in Equation 1relative to the control group. Our results reveal thatthe coefficient of the aligned environment is nega-tive and significant (γ = −0.13; p < 0.05), imply-ing that entrants into aligned environments (Cell Iin Figure 1) have higher survival rates than non-aligned environment entrants (Cell IV in Figure 1).The relative magnitudes of the coefficient esti-mates of the aligned and partially aligned envi-ronments indicate that survival rates in the alignedenvironment are higher than those in the partiallyaligned environments. Also, as posited, Model 1reveals a difference in the survival rates of entrantsinto partially aligned environments (Cells II andIII in Figure 1) and nonaligned environments (CellIV in Figure 1) (γ = −0.05; p < 0.05), implyingthat entrants into partially aligned environmentsalso have higher survival rates than nonalignedenvironment entrants. Together, these results indi-cate support for Hypothesis 1.

Table 3(a). Estimates of probability of firm exit in innovative environments

Model 1 Model 2

Intercept −1.17∗ (0.08) −1.16∗ (0.08)Aligned environment (Cell I) −0.13∗ (0.03) −0.13∗ (0.03)Partially aligned environment −0.05∗ (0.02) —Entrepreneurial, low-technology-intensity (Cell II) — −0.08∗ (0.03)Routinized, high-technology-intensity (Cell III) — 0.02 (0.04)Age 0.16∗ (0.04) 0.16∗ (0.04)Age2 −0.02∗ (0.007) −0.02∗ (0.01)Diversified entrant −0.11∗ (0.02) −0.11∗ (0.02)Density 0.19∗ (0.05) 0.18∗ (0.05)Consumer good 0.05∗∗ (0.03) 0.05∗∗ (0.03)Lagged entry rate 0.10∗ (0.04) 0.10∗ (0.04)Lagged exit rate 0.85∗ (0.18) 0.82∗ (0.19)Post-World War II −0.08∗ (0.04) −0.09∗ (0.04)

Log likelihood −3559.45 −3556.47Number of observations 14173 14173

Routinized, low-technology-intensity is used as the baseline group.Standard errors are given in parentheses.Significant at: ∗ 0.05 level; ∗∗ 0.10 level



Table 3(b). Results from the tests of restrictions of Table 3(a)

Chi-squarea Significance

Likelihood ratio test from pooling the aligned and partiallyaligned environments in Model 1

10.07∗ Significant

Likelihood ratio test from pooling the aligned andentrepreneurial low-technology-intensity environments (CellII) in Model 2

4.06∗ Significant

a computed as −2[difference between unrestricted and restricted log likelihood]∗ Significant at 0.05 level

Because we characterize two distinct types ofenvironments as partially aligned, we conductedadditional analysis to ascertain differences in sur-vival rates between the aligned and each of thetwo partially aligned environments. The resultsfrom Model 2 in Table 3(a) indicate that althoughsurvival rates are highest in the aligned environ-ment (Cell I) (γ = −0.13; p < 0.05), entrants intothe entrepreneurial, low-technology-intensity envi-ronment (Cell II) are better off (γ = −0.08; p <

0.05) than either nonaligned entrants (Cell IV) orentrants into routinized, high-technology-intensityenvironments (Cell III) (γ = 0.02; p < 0.10). Wenext verified that the survival probability in thealigned environment (Cell I) differed statisticallyfrom the survival probabilities in either partiallyaligned environment (Cells II and III). We testedfor the invariance of the survival coefficients,in two separate sets of comparisons, between

the (a) aligned environment (Cell I) and the par-tially aligned environments (Cells II and III); andbetween the (b) aligned environment (Cell I) andthe more advantageous of the two partially alignedenvironments, namely, the entrepreneurial regimewith low-technology-intensity industries (Cell II).The likelihood ratio tests reported in Table 3(b)illustrate that the aligned environment could notbe pooled with the partially aligned environments(χ 2 = 10.70; p < 0.05) or even with the moreadvantageous partially aligned environment (CellII) (χ 2 = 4.06; p < 0.05), thereby reconfirmingsupport for Hypothesis 1. Clearly, entrants intoaligned environments (Cell I) are advantaged overall other entrants.

Table 4(a) provides the results for Hypothe-sis 2, which we tested by conducting a sub-group analysis for small and large entrants (asdepicted in Models 1 and 2, respectively). Our

Table 4(a). Estimates of probability of firm exit in innovative environments for small and largeentrants

Variable Model 1(small entrants)

Model 2(large entrants)

Intercept −1.17∗ (0.10) −1.22∗ (0.15)Aligned environment (Cell I) −0.14∗ (0.04) −0.13∗ (0.05)Entrepreneurial, low-technology-intensity (Cell II) −0.06 (0.04) −0.12∗ (0.04)Routinized, high-technology-intensity (Cell III) 0.05 (0.05) −0.04 (0.08)Age 0.19∗ (0.05) 0.11 (0.08)Age2 −0.03∗ (0.01) −0.01 (0.01)Diversified entrant −0.12∗ (0.03) −0.02 (0.05)Density 0.17∗ (0.06) 0.20∗ (0.09)Consumer good 0.07∗ (0.03) −0.01 (0.05)Lagged entry rate 0.12∗ (0.05) 0.06 (0.08)Lagged exit rate 0.71∗ (0.23) 1.01∗ (0.34)Post-world war ii −0.09∗ (0.05) −0.08 (0.08)

Log likelihood −2521.43 −1026.249Number of observations 9681 4492

Routinized, low-technology-intensity is used as the baseline group.Standard errors are given in parentheses.∗ Significant at 0.05 level.



Table 4(b). Results from tests of restrictions for Models 1 and 2 in Table 5(a)

Chi-squarea Significance

Small firmsLikelihood ratio test from pooling firms in the

aligned and the entrepreneurial, low-technologyenvironments (Cell II)

5.31∗ Significant

Likelihood ratio test from pooling firms in thealigned and the routinized, high-technologyenvironments (Cell III)

13.94∗ Significant

Large firmsLikelihood ratio test from pooling firms in the

aligned and the entrepreneurial, low-technologyenvironments (Cell II)

0.05 Not significant

Likelihood ratio test from pooling firms in thealigned and the routinized, high-technologyenvironments (Cell III)

3.92∗ Significant

a computed as −2[difference between unrestricted and restricted log likelihood]∗ Significant at 0.05 level

results reveal that for small entrants only thealigned environment has a negative and signif-icant coefficient (γ = −0.14; p < 0.05), whileboth the partially aligned environments have sta-tistically insignificant negative coefficients. In con-trast, for large entrants, the coefficients are neg-ative and significant for both the aligned envi-ronment (γ = −0.13; p < 0.05) as well as theentrepreneurial, low-technology-intensity environ-ment (γ = −0.12; p < 0.05). The likelihood testsreported in Table 4(b) show that for small entrantsthe aligned environment cannot be pooled witheither partially aligned environment (χ 2 = 5.31and 13.94 for Cell II and Cell III, respectively;p < 0.05). However, for large entrants, not onlyare the coefficient values almost the same for thealigned and the entrepreneurial, low-technology-intensity environments; but furthermore, the like-lihood test fails to rule out the possibility thatthe two environments can be pooled (χ 2 = 0.05;p > 0.10).

To ensure that the results are not sensitive tothe assumptions inherent in the subgroup analysis,we analyzed the full model with three-way interac-tions, and Table 5 presents the results. The resultsare consistent with those reported in Table 4. Fur-ther, the analysis presented in Table 5 enables thetests of pooling small and large entrants withinan innovative environment. In particular, we con-ducted likelihood tests to confirm that in thealigned environment the exit rates of small entrantsare not significantly different from those of large

entrants (χ 2 = 0.09; p > 0.10). Thus, the alignedenvironment mitigates any scale disadvantages thatsmall entrants may experience. However, in theentrepreneurial, low-technology-intensity environ-ment (Cell II), the likelihood test disallows thepossibility that small and large entrants can bepooled together (χ 2 = 3.94; p < 0.05), becauselarge entrants have significantly lower exit rates.

To interpret the magnitude of the coefficients, wecalculated marginal effects associated with aboveanalysis. This enables us to compute the differ-ences in the estimated values of the exit ratesacross size and innovative environment. Using thenonaligned environment as the control group, wecalculated the percentage differences in the exitrates for small and large firms in each of the otherinnovative environments. We obtained the initialpredicted probabilities of exit, by holding the othercontrol variables at their means. Thereafter, wenormalized the predicted probability of the base-line group to 0, and adjusted the other proba-bilities for the other groups accordingly. Table 6gives these results. Small entrants experienced a32 percent decline in exit rates in the alignedenvironment relative to the nonaligned environ-ment; but in all other environments the differencesin exit rates are not statistically significant. Forlarge entrants, exit rates are statistically equal, thatis, 27 percent lower in the aligned environmentand 29 percent lower in the entrepreneurial, low-technology-intensity environment. Together, theresults show that entering the aligned environment



Table 5. Estimates of probability of firm exit in different environments across differentfirm sizes

Variable Model 3

Intercept −1.16∗ (0.08)G1: Large entrant in aligned environment (Cell I) −0.13∗ (0.05)G2: Large entrant in entrepreneurial, low-technology intensity (Cell II) −0.14∗ (0.05)G3: Large entrant in routinized, high-technology intensity (Cell III) −0.02 (0.07)G4: Large entrant in routinized, low-technology intensity (Cell IV) −0.01 (0.06)G5: Small entrant in entrepreneurial, high-technology intensity (Cell I) −0.14∗ (0.04)G6: Small entrant in entrepreneurial, low-technology intensity (Cell II) −0.06 (0.04)G7: Small entrant in routinized, high-technology intensity (Cell III) 0.04 (0.05)Age 0.16∗ (0.04)Age2 −0.02∗ (0.01)Diversified entrant −0.10∗ (0.03)Density 0.18∗ (0.05)Consumer good 0.04∗ (0.03)Lagged entry rate 0.10∗ (0.04)Lagged exit rate 0.81∗ (0.19)Post-World War II −0.09∗ (0.04)

Log likelihood −3554.32

Standard errors are given in parentheses.Significant at: ∗ 0.05 level; ∗∗ 0.10 level

Table 6. Differences in probabilities of exit across dif-ferent environments for small and large entrants

Innovative environment Smallentrants

Largeentrants

Aligned environment(Cell 1)

32% lower∗ 27% lower∗

Entrepreneurial, low-technology-intensity(Cell 2)

12% lower 29% lower∗

Routinized, high-technology-intensity(Cell 3)

11% higher 2% lower

The values in this table are computed based on the marginaleffects from the analysis in Tables 4 and 5. The nonalignedenvironment is the baseline group.∗ Differences significant at the 0.05 level.

is the only way that small entrants can enhancetheir odds of survival, while large entrants haveat least one other environment that increases theirprobability of survival. Thus, we find support forHypothesis 2.

It is important to note that our resultsindicate that for large entrants, entering duringthe entrepreneurial regime increases survival,irrespective of technology intensity. Alignmentbrings large entrants no added advantage over andabove the beneficial effect of the entrepreneurialregime. Their survival rates are regime-variant, not

technology-intensity-variant. However, for smallfirms, it is only in the aligned environment thattheir survival is significantly enhanced.

DISCUSSION AND CONCLUSION

Although new entrants have been celebrated asagents of innovation and change, recent empiricalfindings reaffirm their high levels of vulnerabilityto exit during the early years of their existence(Huyghebaert and Van de Gucht, 2004). Whilesuch findings resonate with existing theoreticalarguments related to liabilities of newness andsize (Stinchcombe, 1965), what is less well under-stood are factors that determine variance within theexit rates of new entrants (Shane, 2001). In otherwords, understanding how certain entry conditionsmay provide a protective umbrella to small newentrants, and therefore mitigate their early yearvulnerabilities seems salient since it will inform usabout entrepreneurial churn, a phenomenon regard-ing which we have limited knowledge.

Our theoretical framework and empirical find-ing focuses on this gap, and draws attentionto how the external knowledge environment ofentry influences a new entrant’s likelihood ofearly exit, or conversely survival. Our typologyof the innovation arena of an entrant is based on



two knowledge-related dimensions of an entrant’senvironment: technology regime and technologyintensity. While the former reflects the tempo-ral dimension of the external knowledge environ-ment, the latter mirrors its cross-sectional com-ponent. By complementing the temporal elementof an entrant’s entry environment, namely theevolutionary stage of the industry’s life cycle inwhich it enters, with a cross-sectional character-istic of the industry, namely its technology inten-sity, we intersect evolutionary literature with thaton technological progress. On one hand, evolu-tionary research seeks to explain ‘the movementof something over time, or to explain why thatsomething is what it is at a moment of time interms of how it got there’ (Dosi and Nelson,1994). Adopting a dynamic approach, evolution-ary scholars emphasize temporal changes in bothcompetitive conditions and the nature of knowl-edge that underlies innovation over the life cycleof an industry. On the other hand, research on tech-nological progress concerns itself with trying toexplain what causes cross-sectional differences inthe ‘inventive potential’ or the richness of inno-vation opportunities across industries (Rosenberg,1974; Klevorick et al., 1995). Adopting a knowl-edge spillover perspective (Arrow, 1962; Mans-field, 1968; Jaffe, 1986; Griliches, 1979), it is gen-erally argued that resources devoted to knowledge-generating activities in an industry endogenouslydetermine the technological opportunities availablewithin an industry (Romer, 1986; Lucas, 1988).

Integrating the two perspectives, we propose thatthe innovation setting of an entrant may be char-acterized by a typology of alignment (or the lackthereof) based on technology regime and technol-ogy intensity. We hypothesize that firms enteringaligned innovative environments, where entrantshave both a knowledge advantage and ample tech-nological opportunities, will enjoy a survival pre-mium denied to firms entering other environments(Hypothesis 1). Further, we expect the size ofentry to impact the relationship between innova-tive environment and survival relationship in sucha way that the aligned state of the innovative envi-ronment will be more critical for small than forlarge entrants (Hypothesis 2). We test our hypothe-ses with data on 3,431 entrants in 33 industriesover the years 1908–91. Using this longitudinaland comprehensive dataset allowed us to test thehypotheses in a manner precluded by the use of

cross-sectional data, which can have survivor biasand data-censoring problems.

We found support for both hypotheses. Entrantsinto aligned environments had significantly lowerexit rates than entrants into all other environments.Further, we found that entrant size conditioned theinnovative environment–survival relationship suchthat alignment of the two dimensions of innova-tive environment is a necessary condition for theenhanced survival levels of small entrants, but notfor large entrants. Our analysis allowed us to dis-cern the effect of each of the two dimensions ofinnovative environment on the survival rates ofsmall and large entrants. In the context of thefirst dimension, technology regime, our results arein support of the general arguments derived fromindustry evolution theories; that in general, firmsthat enter during the entrepreneurial regime havea distinct survival advantage as compared to thosethat enter during the routinized period of the indus-try’s life cycle (Agarwal and Gort, 1996; Suarezand Utterback, 1995). However, our findings indi-cate that this result is not unequivocally true forall types of entrants. In line with our theoreticalpredictions, we find that small entrants are bene-fited by entry into the entrepreneurial regime onlywhen the industry they enter is one that is tech-nology intensive and characterized by high levelsof R&D investment. Thus, our results demonstratehow the knowledge intensity of the industry andscale of entry together form boundary conditionson the effects of technology regimes on entrantsurvival.

Our study has several potential contributions.First, we shed light on a phenomenon that is cen-tral to our understanding of competitive evolution,and thus critical to scholarship in the domainsof entrepreneurship and strategic management. Totruly act as agents of change that reshape marketsand help diffuse innovations through society, newentrants need to be able to survive the initial yearswhen they are especially vulnerable to exit due toliabilities of newness. Otherwise, in sharp contrastto the utopian Schumpeterian vision of creativedestruction, entrepreneurial churn may resemble arevolving door since entrants are likely to exit soonafter they enter. Our study integrates insights fromevolutionary and innovation literatures to suggestthat a confluence of knowledge conditions in theexternal environment helps entrants by provid-ing them with an innovation advantage vis-a-visincumbent firms. We thus attempt to explain such



variations in entrant survival rates through ourframework of the firm’s innovative environment.

Second, while much of strategy literature is ded-icated to examining knowledge that is internal toa firm, we draw attention to the important roleplayed by the external knowledge environment onentrant performance. In developing our theoreticalrationale on how technological intensity relates toinnovative opportunities in an industry, we drawon insights from Kenneth Arrow’s pioneering workon knowledge spillovers and endogenous growththeory research conducted by Paul Romer and col-leagues. By merging considerations of how invest-ments in innovation create in turn more innova-tion into an evolutionary life cycle approach, ourstudy points to the increasing relevance of a co-evolutionary approach to strategy research.

Third, our findings contribute to the under-researched area of technology intensity. Quiteintriguingly, there has been little empirical inves-tigation of the relationship between technologyintensity and entrant survival. A notable exceptionis Audretsch and Mahmood (1995), who conceptu-alized an industry’s technology intensity as equiv-alent to entry barriers and reported a positive rela-tionship between exit rates and technology inten-sity. However, their research was cross-sectional innature, and temporal differences in the innovationenvironment were ignored. Our marked departureis on two fronts: one, in line with endogenousgrowth theory, our work proposes that technol-ogy intensity may be actually reflective of lowerbarriers to entry/survival due to higher innovationopportunities in the industry; and two, we demon-strate the importance of considering both evolu-tionary dynamics and scale issues in considerationsof how technological intensity may impact entrantexit.

Fourth, although our hypotheses and findingspoint to size as a key entrant characteristic, wefind that the distinction between start-up, or denovo entry, and diversifying entry as an importantfactor that influences exit. We conducted additionalanalyses to check for the effects of innovative envi-ronment on the subgroup of de novo entrants, andfound the results to be consistent with our overallresults. Small de novo entrants in particular ben-efited from aligned environments, enjoying highersurvival rates than small de novo entrants in allother environments. Large de novo entrants foundthe entrepreneurial regime conducive to their sur-vival. To the extent that entrepreneurial entry is

associated with size as well as start-up status, ourresults imply that small firms are benefited by inno-vative environments that do not penalize their scaledisadvantage.

From a practitioner perspective, our work indi-cates that the timing of entry decision seems con-tingent on the type of industry being consideredand the amount of resources the entrepreneur hasat her disposal. Our study explains conditionswhere small entrants may be able to benefit fromspillovers of technology investments, and situa-tions where small entrants may be more com-petitive. Also, firms that deal with a portfolio ofinvestments in start-ups may find our findings ofvalue in that our research points out to the inno-vative environment being a criterion along whichthey may consider delineating their portfolio.

Although we tested our hypotheses using datafrom several industries, thus increasing the gener-alizability of our results, several study limitationsstem from the data constraints of a multi-industrystudy. First, entrant size was a dichotomous mea-sure based on asset values. Given the key role ofentrant size, collecting the size data using an alter-native measure, such as number of employees orsales volume, would greatly enhance the validityof the results. We believe the similarity of sub-stantive results using alternative measures (Child,1972; Chandy and Tellis, 2000) partly mitigatesthis concern. Further, our data indicated a bimodaldistribution of size that allowed us to use ‘small’and ‘large’ as broad categories, obviating the needfor multiple measures of asset size. Still, the anal-ysis would have been stronger without this datalimitation.

Similarly, though we included important controlvariables capturing inter-firm and inter-industrydifferences (for instance, age, diversifying entrant,consumer good, and entry after World War II),and applied an estimation technique that accom-modates unobserved heterogeneity, including addi-tional firm- and industry-level control variableswould further increase the validity of our results.Third, in view of prior research findings, our mea-sure of technology intensity is time-invariant, inthat each industry is characterized as highly tech-nology intensive, or not, for its entire life cycle.Our technology regime variable is dynamic, andpartially captures differences in technology inten-sity over time; however, a time-variant measureof technology intensity would have enabled usto explicitly test the validity of the assumption



that industry technology intensity remains stableover time, and would have allowed us to cap-ture dynamics not attributable to industry evolutionalone.

The above limitations provide further avenuesfor potential future research. For instance, vari-ables such as market share and financial perfor-mance may be investigated, in addition to survival,as in this paper. Another fruitful area of researchwould be the analysis of how the two dimensionsof entrepreneurial entry (size and de novo status)affect entrant survival in conjunction with the twodimensions of innovative environment (technol-ogy regime and intensity). Finally, our data couldalso be used to test for chronological differencesin the size–survival relationship, while controllingfor temporal differences arising due to industryevolution. The data would need to be extendedto adequately represent industries that were in theroutinized regime in the early part of the 20th cen-tury and those in the entrepreneurial regime in thelater part of the 20th century, so that effects oftime that are not confounded by the selection ofindustries could be discerned.

In summary, our paper refocuses attention onknowledge that is external to firms as being animportant determinant of their performance, asis internal knowledge. Further, we integrate twostrands of research that have developed largelyparallel to each other, creating a new theoreticalconstruct relating to the innovative environmentfacing industry entrants. Our paper thus extendseach strand via an integrative model that showsthe importance to entrant survival of the alignmentof two aspects of innovative environment. Finally,by focusing on the differing impact of innovativeenvironment on small and large entrants, we pro-vide support for the importance of the fit betweenthe resource characteristics of firms and the envi-ronments that firms enter.

ACKNOWLEDGEMENTS

The authors have contributed equally. The researchwas supported in part by a grant from the E. Mar-ion Kauffman Foundation. The authors would liketo thank Associate Editor Will Mitchell and twoanonymous reviewers for their constructive feed-back. Inigo Arroniz, David Audretsch, Tom Dean,Malathi Echambadi, Constance Helfat, GlennHoetker, Quintus Jett, Andrew King, Margaret

Petaraf, Alva Taylor, and Mary Tripsas providedhelpful comments on prior versions of this arti-cle. We also appreciate the copy-editing servicesprovided by Persephone Doliner.

REFERENCES

Acs ZJ, Audretsch DB. 1988. Innovation in largeand small firms: an empirical analysis. AmericanEconomics Review 78: 678–690.

Acs Z, Audretsch DB, Braunerhjelm P, Carlsson B.2003. The missing link: the knowledge filter andendogenous growth. Working paper, presented at theDRUID’s Knowledge Conference, June 2003.

Agarwal R, Gort M. 1996. The evolution of markets andentry, exit and survival of firms. Review of Economicsand Statistics 78: 489–498.

Agarwal R, Echambadi R, Franco A, Sarkar MB. 2004.Knowledge transfer through inheritance: spinoutgeneration, development and survival. Academy ofManagement Journal 47(4): 501–522.

Agarwal R, Sarkar MB, Echambadi R. 2002. Theconditioning effect of time on firm survival: anindustry life-cycle approach. Academy of ManagementJournal 45(5): 971–994.

Allen T. 1977. Managing the Flow of Technology . MITPress: Cambridge, MA.

Allison PD. 1995. Survival Analysis Using the SASSystem: A Practical Guide. SAS Institute: Cary NC.

Arrow K. 1962. The economic implications of learningby doing. Review of Economic Studies 29: 155–173.

Audretsch DA, Mahmood T. 1995. New firm survival:new results using a hazard function. Review ofEconomics and Statistics 76: 97–103.

Bain JS. 1956. Barriers to New Competition . HarvardUniversity Press: Cambridge, MA.

Baldwin JR. 1995. The Dynamics of Industrial Competi-tion: A North-American Perspective. Cambridge Uni-versity Press: Cambridge, MA.

Barney JB. 1991. Firm resources and sustained com-petitive advantage. Journal of Management 17(1):99–120.

Baum JA, Oliver C. 1991. Institutional linkages and orga-nizational mortality. Administrative Science Quarterly36: 187–218.

Blair J. 1972. Economic Concentration . Harcourt, Brace& Jovanovich: New York.

Chandler AD. 1956. Management decentralization: anhistorical analysis. Business History Review 30:111–174.

Chandy RK, Tellis GJ. 2000. The incumbent’s curse?Incumbency, size and radical product innovation.Journal of Marketing 64: 1–17.

Child J. 1972. Organization structure, environment, andperformance: the role of strategic choice. SociologyReview 6: 1–22.

Christensen C. 1997. The Innovator’s Dilemma: WhenNew Technologies Cause Great Firms to Fail . HarvardBusiness School Press: Boston, MA.



Comanor WS. 1967. Market structure, product differen-tiation, and industrial research. Quarterly Journal ofEconomics 81(4): 639–657.

Conner KR, Prahalad CK. 1996. A resource-basedtheory of the firm: knowledge versus opportunism.Organization Science 7: 477–501.

Dosi G. 1982. Technological paradigms and technologicaltrajectories: a suggested interpretation of the deter-minants and directions of technical change. ResearchPolicy 11(3): 147–162.

Dosi G, Nelson RR. 1994. An introduction to evolution-ary theories in economics. Journal of EvolutionaryEconomics 4: 153–172.

Dunne T, Roberts MJ, Samuelson L. 1988. Patterns offirm entry and exit in U.S. manufacturing industries.RAND Journal of Economics 19(4): 495–515.

Freeman J, Carroll GR, Hannan MT. 1983. The liabilityof newness: age dependence in organizational deathrates. American Sociological Review 48: 692–710.

Geroski PA. 1990. The effect of entry on profit marginsin the short and long run. Anales d’Economie et deStatistique 15–16: 333–353.

Geroski PA. 1995. What do we know about entry?International Journal of Industrial Organization 13:450–456.

Geroski PA, Pomroy R. 1990. Innovation and theevolution of market structure. Journal of IndustrialEconomics 38(3): 299–314.

Gimeno J, Folta TB, Cooper AC, Woo CY. 1997.Survival of the fittest? Entrepreneurial humancapital and the persistence of underperforming firms.Administrative Science Quarterly 42(4): 750–783.

Griliches Z. 1979. Issues in assessing the contribution ofresearch and development to productivity growth. BellJournal of Econometrics 10(1): 92–116.

Gort M, Klepper S. 1982. Time paths in the diffusion ofproduct innovations. Economic Journal 92: 630–653.

Grant RM. 1996. Toward a knowledge-based theoryof the firm. Strategic Management Journal , WinterSpecial Issue 17: 109–122.

Gruber J. 1994. The incidence of mandated maternitybenefits. American Economic Review 84(3): 622–641.

Hadlock P, Hecker D, Gannon J. 1991. High technologyemployment: another view. Monthly Labor Review114(7): 26–30.

Hannan MT, Freeman J. 1984. Structural inertia andorganizational change. American Journal of Sociology49: 149–164.

Hannan MT, Freeman J. 1989. Organizational Ecology .Harvard University Press: Cambridge, MA.

Hayek FA. 1945. The use of knowledge in society.American Economic Review 35(September): 519–530.

Huyghebaert N, Van de Gucht L. 2004. Incumbentstrategic behavior in financial markets and the exitof entrepreneurial start-ups. Strategic ManagementJournal 25(7): 669–688.

Jaffe A. 1986. Technological opportunity and spilloversof R&D: evidence from firms’ patents, profits, andmarket value. American Economic Review 76(5):984–1001.

Jovanovic B. 1982. Selection and the evolution ofindustry. Econometrica 50(3): 649–670.

Kamien M, Schwartz N. 1982. Market Structure andInnovation . Cambridge University Press: New York.

Kessides IN. 1991. Entry and market contestability:the evidence from the United States. In Entry andMarket Contestability: An International Comparison ,Geroski PA, Schwalbach J (eds). Blackwell: Oxford;23–48.

Klepper S. 2002. The capabilities of new firms and theevolution of the U.S. automobile industry. Industrialand Corporate Change 11(4): 645–666.

Klepper S, Graddy E. 1990. The evolution of newindustries and the determinants of market structure.RAND Journal of Economics 21(1): 27–44.

Klepper S, Simons K. 2000. Dominance by birthright:entry of prior radio producers and competitive ramifi-cations in the U.S. television receiver industry. Strate-gic Management Journal , Special Issue 21(10/11):997–1016.

Klevorick A, Levin RC, Nelson RR, Winter SG. 1995.On the sources and significance of interindustrydifferences in technological opportunities. ResearchPolicy 24: 185–205.

Kogut B, Zander U. 1992. Knowledge of the firm,combinative capabilities, and the replication oftechnology. Organization Science 3: 383–397.

Lavin MR. 1992. Business Information: How to Find It,How to Use It (2nd edn). Oryx Press: Phoenix, AZ.

Levin RC, Reiss PC. 1988. Cost-reducing and demand-creating R&D with spillovers. RAND Journal ofEconomics 19(4): 538–556.

Lucas RE. 1988. On the mechanics of economicdevelopment. Journal of Monetary Economics 22:3–42.

Mansfield E. 1962. Entry, Gibrat’s law, innovation andthe growth of firms. American Economic Review52(4): 1023–1051.

Mansfield E. 1968. Industrial Research and Technolog-ical Innovation: An Econometric Analysis . NortonPress: New York.

Nelson RR, Winter SG. 1982. An Evolutionary Theoryof Economic Change. Harvard University Press:Cambridge, MA.

Penrose ET. 1952. Biological analogies in the theory ofthe firm. American Economic Review 42: 804–819.

Porter M. 1980. Competitive Strategy . Free Press: NewYork.

Robinson WT, Min S. 2002. Is the first to marketthe first to fail? Empirical evidence for industrialgoods businesses. Journal Marketing Research 39(1):120–128.

Romer P. 1986. Increasing returns and long-run growth.Journal of Political Economy 94(5): 1002–1037.

Rosenberg N. 1974. Science, invention and economicgrowth. Economic Journal 84(333): 90–108.

Schumpeter JA. 1934. The Theory of Economic Develop-ment . Harvard University Press: Cambridge, MA.

Shaked A, Sutton J. 1987. Product differentiation andindustrial structure. Journal of Industrial Economics36(2): 131–146.

Shane S. 2001. Technology regimes and new firmformation. Management Science 47(9): 1173–1190.



Silverberg G, Dosi G, Orsenigo L. 1988. Innovation,diversity and diffusion: a self-organization model.Economic Journal 98(393): 1032–1054.

Stinchcombe AL. 1965. Social structures and organiza-tions. In Handbook of Organizations , March JG (ed).Rand McNally: Chicago, IL; 142–193.

Suarez FF, Utterback JM. 1995. Dominant designs andthe survival of firms. Strategic Management Journal16(6): 415–430.

Sutton J. 1997. Gibrat’s legacy. Journal of EconomicLiterature 35(1): 40–59.

Sultan F, Farley JU, Lehman DR. 1990. A meta-analysisof applications of diffusion models. Journal ofMarketing Research 27(1): 70–77.

Tushman ML, Anderson P. 1986. Technological discon-tinuities and organizational environments. Administra-tive Science Quarterly 31: 439–465.

Utterback JM, Abernathy WJ. 1975. A dynamic model ofprocess and product innovation. OMEGA 3: 639–656.

Venkataraman S. 1997. The distinctive domain ofentrepreneurship research: an editor’s perspective. InAdvances in Entrepreneurship, Firm Emergence, andGrowth , Brockhaus JKR (ed). JAI Press: Greenwich,CT; 119–138.

Williamson OE. 1991. Strategizing economizing and eco-nomic organization. Strategic Management Journal ,Winter Special Issue 12: 75–94.

Winter SG. 1984. Schumpeterian competition in alter-native technological regimes. Journal of EconomicBehavior and Organization 5: 287–293.

Zahra SA. 1996. Governance, ownership, and corporateentrepreneurship: the moderating impact of industrytechnological opportunities. Academy of ManagementJournal 39(6): 1713–1735.

APPENDIX: GENERALIZEDDISCRIMINANT PROCEDURE USED TOIDENTIFY ENTREPRENEURIAL ANDROUTINIZED REGIMES

To distinguish between the entrepreneurial androutinized stages, we examined the dataset of theannual gross entry rates for each industry, whichwere typically characterized by a large ‘hill’ sep-arated from a later period of little or no entry.To determine the break year for each industry,in which gross entry rates slowed, we first par-titioned the series into three categories. CategoriesA and B contain the years when the gross entry rateclearly reflects the entrepreneurial and routinizedstages, respectively. The series of the T consec-

utive in-between years of the third category arethen labeled x1, x2, . . . , xT . The problem wasthen to choose an optimal dividing year j suchthat observations x1, x2, . . . , xj would be classi-fied in the period before entries become numerous,and xj+1, xj+2, . . . , xT would be classified in theperiod after entries become numerous. This can beaccomplished using the following three-step pro-cedure:

For eachj = 1, 2, . . . , T , we computed

d1(j) =j∑

i=1

xi/j (1)

d2(j) =T∑

i=j+1

xi/(T − j)

The choice of the dividing year was limited tothose values of j for which

|d1(j) − µ1| ≤ |(µ1 − µ2)/2| (2)

|d2(j) − µ2| ≤ |(µ1 − µ2)/2|

where µ1 and µ2 represent the mean rates of grossentry for the entrepreneurial and routinized cat-egories. If there were no values of j satisfy-ing Equation 2, then all observations were classi-fied in the entrepreneurial stage if |d1(T ) − µ1| <

|d1(T ) − µ2| and in the routinized stage. The ratio-nale behind this step was that the mean of theobservations classified into each of the two stagesis closer to the sample mean of the observationsinitially classified into its respective stage, ratherthan to the mean of those placed into the otherstage.

If multiple values of j satisfied Equation 2, thenwe selected the value of j from this set thatmaximized |d1(j) − d2(j)|.

This step ensured that, among the classificationsthat would satisfy Step 2, the classification thatwas chosen maximized the difference betweenthe means of the points classified into the twoalternative stages.


The effect of the innovative environment on exit of entrepreneurial...

Documents

Transcript of The effect of the innovative environment on exit of entrepreneurial...