Canadian Biotechnology Start-Ups, 1991–1997: The Role of Incumbents' Patents and Strategic Alliances in Controlling Competition

Canadian Biotechnology Start-Ups, 1991–1997:The Role of Incumbents’ Patents and Strategic Alliances

in Controlling Competition

Tony Calabrese and Joel A. C. Baum

University of Toronto, Toronto, Ontario, Canada

and

Brian S. Silverman

Harvard University

Fligstein (1996) contends that organizations act to exploit the institutional context inwhich they are embedded so as to stabilize the competition they face. Drawing onFligstein’s theoretical analysis, we conceptualize incumbent biotechnology firms’ patent-ing and alliance-building activities as attempts to stabilize and control potential compe-tition and analyze how these activities shape rates of founding in the Canadian biotech-nology industry. We find that increases in the level and concentration of incumbents’patenting discourage founding, particularly in human application sectors of the industrywhere development and approval processes are more costly and time consuming. Incum-bents’ horizontal alliances depress start-ups; vertical alliances stimulate start-ups. Ourfindings highlight how technology appropriation and strategic alliances structure thecompetitive dynamics and evolution of high-technology, knowledge-intensive industries.© 2000 Academic Press

Key Words:patents; strategic alliances; organizational founding; control of competi-tion; biotechnology; Canada.

We are grateful to Fred Haynes (Contact International) and Denys Cooper (NRC) for permittingus access to their data. We are also grateful to three anonymousSocial Science Researchreviewersfor helping us to improve the article. Silverman acknowledges financial support from the Division ofResearch at Harvard Business School. We also thank Whitney Berta, Jack Crane, and Igor Kotlyarfor their help with data collection and coding.

Address correspondence and reprint requests to Tony Calabrese, Department of Sociology,University of Toronto, 725 Spadina Ave., Toronto, ON, M5S 2J4, Canada. E-mail: [email protected].

Social Science Research29, 503–534 (2000)doi:10.1006/ssre.2000.0679, available online at http://www.idealibrary.com on

503

0049-089X/00 $35.00Copyright © 2000 by Academic Press

All rights of reproduction in any form reserved.

The founding of new organizations involves the mobilization of resources inpursuit of perceived environmental opportunities. When entrepreneurs foundorganizations they must strategically select an organizational domain—definedby resource requirements and organizational capabilities—that targets a specificmarket and set of customers (Aldrich, 1990). The effect of these choices on thesuccess of new organizations depends upon the attributes of the socioeconomicsetting into which organizations are founded (Stinchcombe, 1965)—notably onthe resources available for potential new organizations and the level of compe-tition that such organizations will face.

In broad terms, the extant literature on organizational founding has followedtwo themes. Ecological research on organizational founding emphasizes howcompetition (for scarce common resources) and mutualism (based on comple-mentary differences) lead to differential selection among founding attempts.Ecological research on organizational founding has sought to determine theeffects of population density, prior foundings and prior failures on organizationalfounding rates. Ecological studies have demonstrated that incumbent firms’distribution among organizational domains shapes patterns of organizationalentry (Baum and Singh, 1994; Delacroix and Carroll, 1983; Hannan and Carroll,1992). Complementing ecological studies of organizational founding, institu-tional research has examined how legitimacy, social support, and approval fromexternal constituents increase the founding rate of new organizations by facili-tating resource acquisition (Baum and Oliver, 1992, 1996; Carroll and Huo,1986; Singh, Tucker, and Meinhard, 1991). Recent institutional research hasfurther explored how government policies affect founding rates by influencingthe degree of competition among organizations (Barnett and Carroll, 1993;Dobbin and Dowd, 1997). Although different in their specific themes, ecologicaland institutional research on organizational founding share a common focus onhow the collective arrangement of organizational environments shapes entrepre-neurial activity. This directs explanation away from the traditional trait-relatedconcerns of conventional theories of entrepreneurship (Gartner, 1989), stressinginstead that comprehension of organizational founding processes lies in graspinghow incumbent firms’ features and relationships constrain or enhance the prob-ability of new foundings into a population of organizations.

Lacking from both the institutional and ecological approaches to organiza-tional founding, however, is a conceptualization of how incumbents and entre-preneurs act to exploit the institutional context in which they find themselves.How exactly do incumbents’ features and relationships shape perceptions ofentrepreneurial opportunity in a given market? Fligstein (1996, p. 659) interpretsincumbents’ market actions as attempts to impose stability on markets. Giventhat market actors live in worlds characterized by uncertainty and causal ambi-guity, actors direct their action “toward the creation of stable worlds.”

The institutional context in which a market is embedded—the set of laws andnorms that establish the “rules of the game” concerning property rights, ex-change, and interorganizational interaction—influences the locally specific na-

504 CALABRESE, BAUM, AND SILVERMAN

ture of the “problem of competition” that incumbents must resolve to success-fully restrain potential competition (Fligstein, 1996, p. 666). Given the relevantinstitutions, then, firms act to create and sustain a “conception of control” or“locally viable, stable solution to the problem of competition.”1 Solutions typi-cally reflect market-specific agreements among firms on principles of organiza-tion, tactics for cooperation and competition, and the status ordering of firms inthe market. In addition, conceptions of control also function to structure actors’perceptions and understanding of how a market operates, permitting them tointerpret their world and act, if necessary, to control potentially destabilizingcompetitive situations.

In this article, we incorporate Fligstein’s insights on markets to build upon thecurrent institutional and ecological perspectives on organizational founding. Intechnology-based industries such as biotechnology, which is our focus here,knowledge development and appropriation are critical (e.g., Powell, Koput, andSmith-Doerr, 1996). So too, however, are capabilities and resources required forestablishing competitive advantage once an innovation is commercially viable(Pisano, 1990). In such markets, incumbents’ patents and alliances will, wecontend, be utilized to restrain and structure competition among incumbent firmsand in the process also influence rates of organizational foundings.

The prior literature on the biotechnology industry is consistent with ouremphasis on the important role played by firms’ patenting and alliance activitiesin structuring the industry’s competitive landscape. For example, patent lawsaffect competition by clearly defining property rights to new compounds (Barley,Freeman, and Hybels, 1992; Powell and Brantley, 1992; Fligstein, 1996). Patentlaws ensure that firm-specific investments in innovation are captured by organi-zations making the investments by allowing them the opportunity to extractmonopoly rents and thus avoid competition for a period of time. In such anenvironment interfirm alliances may emerge to take advantage of patent laws,playing both a crucial resource acquisition and industry integration role. Alli-ances connect firms to information and capabilities and deep pockets necessaryto support a firm through costly patent races and time-consuming productdevelopment and testing. As well, alliances can provide access to production,distribution, and marketing capabilities for product commercialization. Norms ofscience that characterize the university-based scientist community may alsofoster interfirm collaboration in the creation and development of patentable

1 At the same time that firms act to exploit the opportunities afforded by existing institutions, theyare likely to attempt to influence the institutional context through lobbying and other political activity(Fligstein, 1996). It is worth noting that Fligstein’s conceptualization of the interaction betweeninstitutions and firm behavior is consistent with much of the new institutional economics (e.g., Davisand North, 1971; North, 1991; Williamson, 1994) as well as the political economy of regulatorycapture (e.g., Stigler, 1971). It is also worth noting that Fligstein’s conceptualization of how firmsstrive to “control competition,” which is essentially through restraint of price competition (1996, p.659), is consistent with much of the empirical industrial organization literature (e.g., Bain, 1956;Caves and Porter, 1977).

505CANADIAN BIOTECHNOLOGY START-UPS

compounds (Argyres and Liebeskind, 1998). Finally, by increasing the requiredtime and resources necessary for product development and commercialization,governmental regulations, particularly (but not exclusively) in the “human ap-plications” segment of the biotechnology industry, contribute to incumbents’ useof patents and alliances (Fligstein, 1996, p. 666).

Overall, then, this literature implies that incumbent biotechnology firms op-erate within markets where conceptions of control prescribe outright competitionfor patents while, at the same time, encouraging interfirm collaboration for thedevelopment and sharing of knowledge, and its commercialization. In view of allthis, we propose that biotechnology incumbents’ intellectual property accumu-lation and strategic alliances are usefully interpreted as organizational actionsdesigned to stabilize and control the competitive environment.

In this study, we examine how incumbents’ patenting and alliance formationbehaviors’ affect opportunities for organizational founding in the Canadianbiotechnology industry. In other words, we conceptualize an incumbent biotech-nology firm’s patenting (an index of property rights) and alliance building (botha firm-level resource acquisition and industry-level integration mechanism) asattempts to deflect competition and analyze how these activities influence rates offounding in the Canadian biotechnology industry. By connecting conceptions ofcontrol to the institutional and ecological research on organizational founding,we show how incumbents’ solutions to the problem of competition define thebasis for ecological and institutional environmental processes shaping marketand organizational evolution and entrepreneurial opportunity structures. Thus, byemploying Fligstein’s (1996) theoretical ideas to conceptualize how incumbents’actions affect entrepreneurial opportunities we are able to move from generalnotions of market processes (e.g., density dependence and market concentration)to more precise, locally relevant means of competition.

Below, we explore the effects of incumbents’ patenting and alliance-buildingactivities more fully and report an analysis of 151 Canadian biotechnologystart-ups between January 1991 and December 1997. Canada’s biotechnologyindustry is the world’s second-largest national biotechnology industry, encom-passing more than 600 firms between 1991 and 1997. We disaggregate theindustry into 16 distinct “sectors,” examining effects of patenting and strategicalliances within each sector.2 Although somewhat crude, this disaggregation ismuch more fine-grained than is typical in research on organizational founding ingeneral (and in studies of biotechnology in particular) and permits us to distin-guish among firms more or less likely to be potential competitors. We considerthe entire patenting history of all incumbent Canadian biotechnology firms (BFs)between 1975 and 1998. We also consider several classes of alliances—Hori-

2 The 16 sectors comprising the Canadian biotechnology industry are (1) agriculture; (2) aquac-ulture; (3) horticulture; (4) forestry; (5) engineering; (6) environmental; (7) food, beverage, andfermentation; (8) veterinary; (9) energy; (10) human diagnostics; (11) human therapeutics; (12)human vaccines; (13) contract research; (14) biomaterials; (15) cosmetics, and (16) mining.


zontal: (1) with other BFs operating in the same sector(s) as the focal firm;Vertical [downstream]: (1) with pharmaceutical firms, (2) with chemical firmsand (3) with marketing firms; andVertical [upstream]: (1) with universities, (2)with research institutes, (3) with government labs, (4) with hospitals, and (5) withindustry associations.

PROPERTY RIGHTS, INCUMBENTS’ PATENTING, ANDORGANIZATIONAL FOUNDING

In science-based, technologically sophisticated industries, an organization’sinnovative capability provides it with a competitive advantage that is, at leastinitially, not easily imitated by competitors. Zucker, Darby, and Brewer (1998,pp. 290–291) suggest that the commercialization of biotechnology is “inter-twined with the development of the underlying science” that supports it, notingfurther that in the immediate aftermath of breakthrough invention, an organiza-tion may capitalize on “the natural excludability [that] arises from the complexityor tacitness of the information required to practice the innovation.” Similarly,Nonaka and Takeuchi (1995) argue that tacitness and complexity can lead to suchexcludability in a variety of industries including information technology andtelecommunications (see also Nelson and Winter, 1982; Rosenberg, 1982).

In some organizational environments, this natural excludability is comple-mented by an intellectual property rights system that regulates proprietaryownership of innovation. Intellectual property protection in the United States andCanada—both of which provide 20-year monopolies for patented innova-tions—is rated among the strongest in the world (Kondo, 1995). Organizationscan use such protection to deflect competition by facilitating an innovator’sability to appropriate the rents associated with its innovation (Pisano, 1990). Yeteven with such a patenting system, not all innovations enjoy an equally strong“appropriability regime” (Teece, 1986). For example, patented process innova-tions are notoriously difficult to enforce, given the inability to monitor otherorganizations’ internal production methods. In addition, different technologiesare characterized by different rates of obsolescence and different susceptibilitiesto being “invented around” (Levin, Klevorick, Nelson, and Winter, 1987).

As many scholars and industry observers have noted, the appropriabilityregime surrounding biotechnology patents is unusually strong because the pat-ented compounds are difficult to circumvent (Lerner, 1995). This appropriabilityregime ensures that a BF with a patent is in a favorable negotiating position toobtain complementary assets and skills (Pisano, 1990) and is more likely toobtain additional financing and willing partners to support commercializationactivities (Kenney, 1986). Prior research has demonstrated both positive effectsof a BF’s own patenting activity and adverse competitive effects of its rivals’patenting activity. Austin (1993) shows that publicly owned BFs experiencesubstantial increases in their market valuations upon announcement that theyhave been granted new patents. Complementing this result, Baum and Silverman


(1998) find a positive relationship between the number of pending patent appli-cations that a BF possesses and its survival chances. Of particular note, they alsofind that the number of pending and recent patents ofrival firms reduces a BF’ssurvival chances.

Given the organizational context described above, we propose that incumbentpatenting patterns may offer a mixture of inducements and discouragement topotential entrants. We attend to two features of incumbents’ patenting activitythat prior theory and research indicate are likely to be influential in shapingentrepreneurs’ perceptions of environmental opportunities and thus the rate oforganizational founding: (1) incumbents’ recent patenting rate and (2) the trendin concentration of recent patenting among incumbents.

Incumbents’ Recent Patenting Rate

To an extent, the issuance of a patent offers two contradictory signals topotential entrepreneurs: (1) the technological (and potentially commercial) via-bility of a particular technological avenue and (2) some degree of preemption orcontrol of that technological avenue. Absent a strong appropriability regime, wemight expect high rates of incumbent patenting activity to attract potentialentrants. However, as the strength of the appropriability regime increases, thepreemption or control effect should come to dominate the attraction effect.

Preemption and control can occur through several channels. First, Amburgey,Dacin, and Singh (1996) find that BFs that win a patent race in a particulartechnological area are more likely to win subsequent patent races in the samearea. Thus, a high rate of current patenting may indicate that incumbents have alead over existing or potential rivals infuture patent races and consequentlyshould reduce entry (Powell and Brantley, 1992, p. 388). Second, Lerner (1995)finds that small BFs tend to avoid patenting in areas where well-capitalized orpreviously litigious incumbents have been granted prior patents, a finding that heattributes to fears of “sham” litigation from “low-litigation-cost” firms. Thus,even if current patenting does not confer advantage in future patent races, to theextent that incumbents’ patents foreshadow costly litigation—a strategy partic-ularly valuable to well-capitalized firms in markets where property rights areemployed to constrain competition—incumbents’ patents may indicate preemp-tion or control of nearby technological areas, thus reducing entry. The thirdchannel is less direct. As described above, research indicates that patents, bysignaling innovative capabilities, help firms possessing them acquire additionalresources (Austin, 1993; Baum and Silverman, 1998). If incumbents’ patentingsuccess provides such a resource acquisition advantage—or potential foundersperceive that it does—this should reduce the rate of entry. Taken together, thesearguments and evidence lead us to expect that as the number of recent patentsgranted in a given industry sector increases, the subsequent founding rate of BFsshould decrease:


H1: Founding rates of new BFs in a given industry sector decrease as the number of recentpatents granted to incumbent BFs in that industry sector increase.

Concentration of Recent Patenting among Incumbents

In addition to the aggregate level of patenting activity in a biotechnologysector, theconcentrationof patenting activity among incumbents may influenceorganizational founding. An increasing concentration of innovative capabili-ties—and the rents that come with the acquisition of patents—within a fewincumbents is likely to discourage foundings in at least two ways. First, astechnical know-how becomes increasingly concentrated in, and controlled by, afew, innovative technological leaders, it may become concomitantly more diffi-cult for potential start-ups to catch dominant industry leaders (Suarez andUtterback, 1995; Utterback and Suarez, 1993). Second, the threat of litigationincreases with the dominance of a few firms. Litigation threats are driven in partby asymmetries in the capitalization of incumbents and entrants, and theseasymmetries are likely to be greater when a sector’s patenting activity is moreconcentrated. Further, since a reputation for litigiousness discourages entrantsfrom patenting nearby, an incumbent competing in multiple technological raceshas a greater incentive to litigate aggressively against any single threat to developsuch a reputation. Consequently, we expect the founding rate to decline as anindustry sector becomes increasingly characterized by a concentration of patentsgranted to a limited number of firms. Therefore, we predict that:

H2: Founding rates of new BFs in a given industry sector decrease as the concentration ofpatents among incumbents in that sector increases.

Technological Maturity

Changes in the rate of technological innovation over time play a further crucialrole in shaping resource and market opportunities for technology-based ventures(Utterback and Suarez, 1993). Technologically mature industry segments, char-acterized by lower rates of innovation, typically exhibit reduced levels oforganizational founding because of the constrained resource environment andlimited opportunities for innovation that characterize them (Eisenhardt andSchoonhoven, 1990). To the extent that technologically mature environments aretypified by increased stability and effective control by incumbents, existingorganizational routines and relationships further preclude entry. Finally, in in-dustries with strong appropriability regimes, the patent protection and enhancedresource access that incumbents’ older patents continue to provide may, byfurther stabilizing competition and reinforcing incumbents’ control, also contrib-ute to slowing the entry rate into technologically mature sectors.

Controlling for the overall level of recent patenting activity by incumbentfirms (H1) and for the level of concentration in incumbents’ patenting success(H2), we propose that a decreasing number of recent patents relative to older


patents will gauge the technological maturity of an industry sector. H1 and H2are premised on the idea that incumbents’ current property rights serve resourceprovisioning and competition control functions and that these rights will discour-age potential organizing attempts. In contrast, adecreasein the number ofrecentpatents relative toolder patents signals a maturing technology, which discour-ages foundings because of the lack of overall environmental munificence andopportunity, and extent of incumbents’ continuing control of competition. To thedegree that potential entrepreneurs interpret such a slowing of patenting activityas a gauge of the overall resource/opportunity availability (or lack thereof), anindustry sector’s technological maturity should discourage firm foundings.Therefore, we hypothesize that:

H3: Founding rates of new BFs in a given industry sector decrease as the ratio of olderpatents/current patents granted to incumbent BFs in that sector increases.

THE EFFECT OF REGULATORY POLICY ON PATENTING–FOUNDING RELATIONSHIPS

The foregoing hypotheses assume that incumbents’ patenting activity exerts asimilar influence across all 16 sectors of the biotechnology industry (see footnote3 for the list of sectors). However, there is good reason to expect variations acrosssubsets of these sectors. For example, scholars have noted the unique effect ofFDA regulatory requirements on those sectors of biotechnology that focus onhuman applications products (e.g., Powell et al., 1996). Consistent with this view,we expect the effect of incumbents’ intellectual property accumulation to be mostinfluential in human sectors (i.e., therapeutics, vaccines, and diagnostics) of thebiotechnology industry, driven by public policies that raise the cost of, and timeuntil, commercialization of human applications relative to other industry sectors.In fact, even after the basic technology is patented, products in the humanapplications sectors are subject to rigorous and lengthy development and clinicaltrials before regulatory approval can be secured for product release. As well,innumerable and complex regulations significantly delay the speed with whichproducts can be brought to market in these sectors (Barley et al., 1992; Powelland Brantley, 1992). Perhaps most important, a firm that is able to “inventaround” an incumbent’s patent in human applications must still incur the costs ofclinical trials and regulatory burdens.

In Canada, conditions for regulatory approval are similarly stringent, costly,and time consuming. The Pharmaceutical Manufacturers Association of Canada(PMAC) reports that it takes 13–15 years and more than C$500 million to obtaina Notice of Compliance from Health Canada and bring a new drug to marketfrom the time it is first produced in a research laboratory (Stewart, 1997). Datafrom Health Canada’s Drug Directorate—the agency responsible for administer-ing the approval process—shows that it took nearly 3 years (1087 days) for theaverage new drug submission to make it through the government approval


process during the 1991–1995 period. Starting in 1996, government efforts beganto improve the timeliness of the approval process. To support this effort,substantial new fees were introduced for the evaluation of drug submissions, anda new policy that established “internationally competitive performance targets”for Health Canada’s Drug Directorate was put in place. Because complete datafor the average approval time for 1996 and 1997 applications are not yetavailable, the impact of this initiative on the duration of the approval processremains unclear. The overall drug development cycle in Canada is summarizedin Fig. 1, which shows that roughly half the patent protection period is spent inextensive development and government approval for human application-basedbiotechnology products and processes.

As a result of regulatory policies, patenting effects are thus likely to beespecially strong in the human sectors of the Canadian biotechnology industrysince the resources required by both incumbents and potential new entrants aremore extensive. Human-sector-specific constraints raise the costs, risks, anddifficulty of bringing human-application products to market. And, as is often thecase in situations where outcome uncertainty is heightened, commonly reliedupon signals for predicting eventual competitive success assume a greater sa-lience (Spence, 1974; DiMaggio and Powell, 1983). Consequently, we anticipatethat incumbents’ patents will disproportionately influence the allocation of en-vironmental resources in the human sectors of the biotechnology industry andthat the effects of incumbents’ patenting activity predicted in H1–H3 will havea disproportionately larger effect in human rather than in nonhuman sectors ofthe biotechnology industry. Therefore, we predict:

H4a: Founding rates of new BFs in human sectors decrease to a greater degree (than innonhuman industry sectors) as the number of recent patents granted to incumbent BFsincreases.H4b: Founding rates of new BFs in human sectors decrease to a greater degree (than innonhuman industry sectors) as the concentration of patents among incumbent BFs in-creases.H4c: Founding rates of new BFs in human sectors decrease to a greater degree (thannonhuman industry sectors) as the ratio of older patents/current patents granted to incum-bent BFs in that sector increases.

FIG. 1. Drug development cycle in Canada. Adapted from Howes, 1997.


INCUMBENTS’ ALLIANCES AND ORGANIZATIONAL FOUNDING

Interfirm alliances have the ability to alter the opportunities and constraintsthat potential entrepreneurs face in organizing attempts (Kogut, Shan, andWalker, 1992). A large body of research on interorganizational linkages hasexamined the benefits and costs that these linkages confer on incumbents (Baumand Oliver, 1991; Miner, Amburgey, and Stearns, 1990; Powell et al., 1996;Uzzi, 1996). A much smaller literature explores the effect of such alliances onnew foundings. Baum and Oliver (1992) show that incumbents’ institutionallinkages with municipal government and community agencies shape foundingpatterns of day-care centers. Kogut, Walker, and Kim (1995) demonstrate thathigh levels of centrality in semiconductor sectors increase founding rates, afinding they attribute to network externalities. Hybels, Ryan, and Barley (1994)show that aggregate alliance activity by incumbent U.S. BFs increases foundingrates. However, researchers have not explored the effects ofdifferent typesofincumbents’ alliances on foundings, instead focusing on a narrow range oflinkages (Baum and Oliver, 1992; Kogut et al., 1995) or implicitly assuminghomogeneity across alliances by aggregating them. We distinguish between thevertical (i.e., symbiotic) andhorizontal(i.e., commensal) alliances of incumbentbiotechnology firms. We do so to suggest that not all alliance types serve thesame function and that dissimilar types of alliances may provide diverse signalsto potential organizational founders about the viability of start-ups.

Vertical Alliances

Vertical alliances link BFs to other complementary industries, both upstreamand downstream. “Downstream” alliances—with pharmaceutical firms, chemicalfirms, or marketing firms—provide BFs with access to complementary assets,such as distribution channels, marketing expertise, and production facilities, thatare necessary for the successful development and commercialization of a productor process (Kogut et al., 1992, 1995). Although such alliances require down-stream partners to give up some control of product and process development,they are likely to go along with such a strategy because it limits their directexposure to product development risks and removes R&D expenses from theirbalance sheets.

“Upstream” alliances—primarily with universities, research institutes, hospi-tals, government labs, and industry associations—provide firms with cutting-edge research know-how that can prove critical to the successful discovery andpatenting of new products or processes. An additional benefit of these upstreamalliances, particularly alliances with universities, is that they may diffuse normsof open science and concomitant cooperation that may benefit all members of asector (Liebeskind, Oliver, Zucker, and Brewer, 1996; Baum and Silverman,1998).

As with the issuance of a patent, the formation of an alliance by an incumbentbiotechnology firm may offer contradictory signals to potential entrepreneurs. In


contrast to patents, however, alliances provide a significantly lower level ofdeterrence to potential founders of biotechnology firms. Whereas patents offerincumbents exclusive control over a particular technology, vertical alliancepartners can and frequently do partner with a number of biotechnology firms (forexample, a pharmaceutical firm is likely to find such a policy desirable as a wayof obtaining “options” on multiple technological trajectories in a sector). Thus,an incumbent firm’s alliance with a vertical partner does not preclude otherbiotechnology firms forming similar alliances. Further, to the extent that adownstream alliance in a particular biotechnology sector also signals to otherdownstream organizations the potential of this sector, one downstream alliancemay lead other downstream firms to follow suit, engaging in “trait-based imita-tion” (Haunschild and Miner, 1997). In such circumstances, the biotechnologyfirm in the original partnership is unlikely to absorb all these resources, althoughit may derive first-mover advantages associated with earlier access to resources.3

Thus, as opposed to patents, spillover effects reduce the potential value ofvertical alliances as mechanisms of control for incumbent biotechnology firms.

In view of the above, we expect that alliances with vertically related partnersoffer at least three inducements to potential entrants, all of which relate toexpanding the resources available for potential new organizations via linkages toother organizational fields (Kogut et al., 1992, 1995).4 First, downstream alli-ances provide a signal of commercial viability that is likely to attract potentialentrants. Such alliances indicate the belief by organizations outside the biotech-nology organizational field—and from an organizational field closer to themarket—that profitable opportunities exist within a given sector. Second, byproviding capital infusions from downstream partners to BFs, such alliancesindicate an expansion of the capital pool available to biotech firms. Such anincrease in the “carrying capacity” of capital should encourage foundings. Third,both upstream and downstream alliances indicate novel ways for potentialentrants to acquire the necessary resources that a start-up BF will require forcompetitive success. Upstream alliances indicate ways to access knowledgewithout having to hire large and costly staffs of scientists. Downstream alliancesindicate ways to commercialize a product without having to invest in costly fixedassets such as distribution networks, marketing departments, or sales forces. Ifliquidity constraints preclude entrepreneurial behavior (Holtzeakin, Joulfaian,

3 Although we emphasize incumbents’ inability to control these spillovers, they can also be seento represent opportunities for entrepreneurs to enter and act to acquire control.

4 We also considered the possibility that such a positive effect of upstream and downstreamalliances within a given sector, might, at very high levels of vertical partnering, become negative. Therationale for such a reversal of the effect of industry sector vertical alliances on the founding rate isthat the capital and managerial capacity limits of upstream and downstream alliance partners to enterinto alliances are being approached (Baum and Oliver, 1992). Preliminary analysis yielded noevidence of the nonmonotonic effects implied by this idea, perhaps given the relative infancy of thebiotechnology industry and the abundance and high capacity for collaboration of potential partners.Therefore, for parsimony, we only point out the possibility.


and Rosen 1994), then reductions in a potential start-up’s direct capital andhuman investments should facilitate foundings. This may be particularly true ofdownstream alliances, which afford access to the most scale- and scope-intensiveassets required.

Overall, then, we expect incumbents’ vertical alliances—both upstream anddownstream—to promote entrepreneurial start-ups by signaling the environmen-tal availability of required technical and financial resources in related industries.Combined, these vertical alliance-derived entrepreneurial inducements, and theirassociated resource spillovers, lead us to expect that increases in incumbents’vertical alliances—both upstream and downstream—will encourage founding:

H5: Founding rates of new BFs in a given industry sector increase as the number of verticalalliances established by incumbent organizations in that sector increase.

Horizontal Alliances

Horizontal alliances link BFs to other BFs operating in overlapping sectors. Incontrast to vertical alliances, such alliances between potential competitors do nottap into the resources and capabilities of nearby industries or related health andresearch sectors. There is some evidence that horizontal alliances are moreproblematic and fraught with difficulty than vertical linkages, notably due to fearof proprietary information leakage and the propensity for learning races. Mow-ery, Oxley, and Silverman (1996) found that alliances involving partners whocompeted in the same primary SIC exhibited lower levels of knowledge transfer,measured by changes in patent cross-citation rates, than did alliances involvingnoncompeting partners. Grindley, Mowery, and Silverman (1994) suggested thatthe semiconductor manufacturers involved in the research consortium SEMAT-ECH were unable to undertake their initial joint research agenda preciselybecause of fears concerning information leakage and learning races and, conse-quently, changed the research agenda to focus on vertical infrastructure issues.Of particular note for this article, Baum and Silverman (1998) found thathorizontal alliances, on average, raised a BFs likelihood of failure—and that thiseffect increased with the relative scope advantage of the BF’s partner.

We expect horizontal alliances to discourage potential entrants in at least twoways. First, Baum and Silverman’s (1998) finding suggests that, while increasesin horizontal alliances will result in an industry sector containing fewer incum-bents, it also implies that these incumbents will be particularly capable rivals whohave successfully appropriated some resources of former alliance partners(Kogut et al., 1992). Of course, noteveryhorizontal alliance between potentialrivals will be overwhelmed by intra-alliance rivalry, and successful horizontalpartnerships will present potential entrants with formidable groups of networkedcompetitors. Hence, increases in horizontal alliances should discourage found-ings by signaling formidable competition from incumbents. Moreover, in con-trast to vertical alliances, horizontal alliances between incumbent biotechnology


firms are unlikely to create resource or technology spillovers that entrepreneurscan appropriate to start new organizations.

Second, despite the risks of horizontal alliances, startups may be driven toforge such alliances because established rivals are repositories of knowledgeneeded by organization builders. Stinchcombe (1965; p. 152), observing that“organizations requiring similar talents and training tend to have interlockingdirectorates,” argued that such linkages to similar (and therefore potentially rival)organizations enable a new organization to tap experience relevant to its forma-tion and survival. Alliances with rivals offer similar benefits. By allying withpotential rivals, new entrants can potentially gain access to uncodified, tacitknowledge about strategy, technology and operations critical to success (Liebe-skind et al., 1996; Powell, 1990; Zucker et al., 1998). However, as the numberof horizontal alliances between incumbents in a given sector increases, potentialentrants face an increasingly limited pool of potential (and desirable) incumbentswith which to partner. Hence, opportunities either to access crucial capabilities orto become a formidable competitor through collaboration in that sector arediminished.

Incumbents’ within-industry alliances thus may signal both the emergence offormidable groups of networked competitors and a more limited opportunity togain access to tacit knowledge about strategy, technology and operations criticalto success, thus discouraging foundings. Taken together, the foregoing argumentsand evidence suggest that horizontal alliances between incumbent biotechnologyfirms may act as particularly effective competitive controls that result in thesuppression of start-up rates. Therefore, we predict:

H6: Founding rates of new BFs in a given industry sector decrease as the number ofhorizontal alliances established by incumbent organizations in that sector increase.

RESEARCH METHODS

Data Description

We tested our hypotheses using data on the 151 foundings of firms in (1)agriculture; (2) aquaculture; (3) horticulture; (4) forestry; (5) engineering; (6)environmental; (7) food, beverage, and fermentation; (8) veterinary; (9) energy;(10) human diagnostics; (11) human therapeutics; (12) human vaccines; and (13)contract research sectors of the biotechnology industry in Canada betweenJanuary 1991 and December 1997.5 We compiled our data usingCanadianBiotechnology,an annual directory of companies active in the biotechnologyfield operating in Canada published since 1991.Canadian Biotechnologyis themost comprehensive historical listing in existence of Canadian BFs, their found-ing, products, growth, performance, and alliances. We cross-checked this infor-

5 Three sectors did not experience a founding during this period, biomaterials, cosmetics andmining.


mation withThe Canadian Biotechnology Handbook(1993, 1995, 1996), whichlists information for a more restrictive set ofcoreCanadian BFs—firms entirelydedicated to biotechnology. We identified patents issued to all Canadian BFs inthe United States between 1975 and 1998 using the Micropatent database (whichbegins in 1975). We used U.S. patent data because Canadian BFs typically filepatent applications in the United States first to obtain a 1-year protection duringwhich they file in Canada, Europe, Japan, and elsewhere (Canadian Biotech ’89andCanadian Biotech ’92).

Two aspects of our industry definition require discussion. One is our use of anational boundary to define the industry boundary, and the second is the inclusionof firms in both human and nonhuman sectors as part of the same industry.

We think several factors point to the value of studying the Canadian biotech-nology industry as a “quasi-independent” organizational population—that is as apopulation that has its own internal dynamic, but is also shaped by, and active in,industry activity beyond the national border. The first is the industry’s signifi-cance, both in size and to the Canadian economy. During the 1991–1997 period,Canada’s national biotechnology industry was second only to the United States’in number of firms. The yearly average number of U.S. BFs during this periodwas approximately 1300 (Amburgey et al., 1996; Ernst and Young, 1997). TheCanadian average was about 460, or roughly one-third the number in the UnitedStates. By way of comparison, Canada’s economy is approximately one-tenth thesize of the U.S. economy. Moreover, between 1989 and 1993, biotechnology ledall Canadian industries in growth in domestic sales (24%), exports (19%), andemployment (14%) (Canadian Biotech ’94).

In addition to its substantial size and economic significance, three furtherfactors suggest the Canadian biotechnology industry should be considered aquasi-independent population. First, Canadian BFs draw almost exclusively onwithin-country sources of financing (i.e., Canadian capital markets, banks, andventure capital firms). U.S. venture capital firms tend to fund start-ups thatoperate within the United States and frequently focus on start-ups that aregeographically proximate (Gompers and Lerner, 1998). Second, there is a veryhigh degree of within-Canada partnering among biotechnology firms. Our dataindicate that such within-country partnerships accounted for more than 90% ofthe horizontalalliances established by Canadian biotechnology firms. Corrobo-rating our data, Barley et al. (1992, p. 325) report few alliances between U.S. andCanadian biotechnology firms (33 alliances at a time when there were over 400Canadian BFs). Our data indicate further that there is a strong tendency towardwithin-country partnerships across many types of alliances and that this isparticularly true of upstream alliances (i.e., hospital, university, government lab,research institute, and industry association). So, to a large degree, the network oforganizations active in the Canadian biotechnology industry is highly localizedwithin Canada’s geographic boundaries. The third supporting factor is the plau-sibility of localized positive externalities resulting from Canadian BFs acquiringhigh-status foreign partners (e.g., large U.S. pharmaceutical companies). That is,


it is reasonable to expect that such events would be strongest proximate to BFswith which such alliances are established, for example, due to localized diffusionof information through media and interpersonal networks. Simply put, if largeU.S. concerns display a willingness to invest in ties to biotechnology start-ups inCanada, that is more likely to trigger start-ups in Canada than in, say, France.

Prior research on the U.S. biotechnology industry has employed two industrydefinition strategies. One approach, employed by Stuart, Hoang, and Hybels(1999); Powell et al. (1996); and Powell, Koput, Smith-Doerr, and Owen-Smith(1998), for example, is to focus on the human sector in isolation. Stuart et al.(1999) study human diagnostics and therapeutics, while Powell and his col-leagues study only human therapeutics. In these studies, the implicit assumptionseems to be that there is something distinctive about the markets and/or theregulatory environment surrounding the human applications sectors of the bio-technology industry that justifies their isolated study. On the other hand, Barleyet al. (1992); Amburgey et al. (1996); and Walker, Kogut, and Shan (1997) haveadopted a similar industry definition strategy as our own that emphasizes “coretechnology/knowledge” as the key defining characteristic of an industry’s bound-ary. As Walker et al. (1997, p. 123) characterize it, for example, “biotechnologyincludes all techniques for manipulating micro-organisms.” Although we adoptthis second approach, we also recognize that human sectors of the biotechnologyindustry do have important features that differentiate them from nonhumansectors (e.g., a much more lengthy and stringent product approval process).Indeed, our recognition of this difference leads us to theorize (H4a–H4c) and testfor stronger suppressing effects of incumbents’ patents on startups in the humansectors. Of course, without data on all biotechnology sectors, we could not testthese hypotheses. As described in detail below, our empirical strategy is sensitiveto sectoral differences, defining all theoretical variables at the sector level, andincorporating time-varying sector-specific controls (e.g., labor supply, financing,and competition), as well as sector-specific fixed effects in our baseline model.

Independent Variables

Sector patents (H1–H4).To test H1 we estimated the effect of the number ofrecent patentsgranted to incumbents in a given industry sector on the rate offounding in that sector. We measured recent patents as patents granted within thepast 5 years. This measure is consistent with cutoffs used in prior research(Podolny et al., 1996).6 We tested H2 using a measure ofinnovation concentra-tion, computed as the ratio ofconcentrationof recent (within 5 years) patents toolder (outside 5 years, i.e., betweent-6 and 1975, the year our patent data begins)patents granted in a given industry sector. We measured concentration using theHirschman–Herfindahl index, which computes the concentration of patenting

6 In unreported results we varied this cutoff from 3 to 7 years, with no substantive change. We alsochecked for effects of incumbents’ patents granted more than 5 years earlier, but preliminary analysisyielded no significant coefficients.


activity in an industry sector as the sum of the squared proportions of all industrysector patents granted to each firm in that sector. Thus, innovation concentrationfor sectori is defined as:

Innovation Concentrationi 5¥ j51

N Sij ~t03t25!2

¥ j51N Sij ~t2631975!

2,

whereSij is the proportion of patents granted to incumbentj in sectori andN isthe total number of firms in sectori . As this ratio increases over time, theconcentration of patenting activity among a few firms increases.

H3 was tested with a measure oftechnological maturity,defined as the ratioof recent to older patents in a given industry sector—a gauge of changes in therate of technological maturity of particular biotechnology sectors. More formally,technological maturity for sectori is:

Technological Maturityi 5Pi ~t03t25!

Pi ~t2631975!,

wherePi is the number of patents granted to incumbents in sectori .To test H4a–H4c, which predict stronger founding effects for incumbents’

patenting activity in the human sectors of biotechnology, we constructed a humansector dummy variable (coded 1 for all human application sectors and 0 other-wise) and interacted it with each of the incumbents’ patenting variables.

Sector alliances (H5–H6).We estimate the effects of a sector’s incumbentBFs’ alliances on the industry-sector founding rate using a set of variables thataggregate the number of alliances of each type (separately) that BFs in eachindustry sector possess at the start of each year. The types of alliances weexamine areHorizontal: with (1) other potential rival BFs operating in the sameor overlapping industry sectors;7 Vertical [downstream]: with (1) pharmaceuticalfirms, (2) chemical firms, and (3) marketing firms; andVertical [upstream]: with(1) universities, (2) research institutes, (3) government labs, (4) hospitals, and (5)industry associations.8 We estimate effects for each type of vertical allianceseparately to allow for the possibility that incumbents’ alliances with differenttypes of vertical partners have effects on the founding rate that differ in magni-tude, significance, or even direction. Of course, our “disaggregated” approachalso permits multiple tests of H5.

Control Variables

Many other factors may influence the founding of BFs. For example, foundingrates may differ across industry sectors as a result of differences in required scale,

7 We found no alliances between BFs innonoverlaping industry sectors. Consequently, we cannotestimate effects for alliances among “nonrival” incumbents.

8 Although all start-ups in our sample areCanadian,our alliance data includes information on allthese firms’ alliances with other organizations worldwide.


sector age, intensity of domestic and foreign competition, or investor interest andexcitement, which if uncontrolled, may lead to spurious findings for our theo-retical variables. Although we cannot account for all imaginable possibilitiesdirectly, controlling for industry- and sector-specific effects that may representplausible alternative explanations for our theoretical findings is vitally important.Accordingly, to avoid such specification bias, we develop a comprehensivebaseline model that includes three sets of control variables. First, we control fora set of time-varying, sector-specific factors that past research indicates mayinfluence organizational founding rates. Second, we control for additional tem-poral effects not directly captured by the time-varying sector controls by includ-ing a calendar time-trend variable (e.g., Barnett, 1990). Third, we account forsector-specific differences not captured by the time-varying industry sectorconditions or time-trend controls by including a set of sector-specific dummyvariables, or “fixed effects” (Greene, 1990). All control variables, defined indetail below, were measured at the start of the year unless otherwise indicated.

First, we obtained yearly information on aggregate financing of BFs from allsources (e.g., venture capital, private placement, IPO, public offering, and other)by biotechnology sector from the National Research Council of Canada. Tocontrol for effects of financing on BF founding we constructed two variables.One, sector biofinancing,measures the total financing (in millions of 1991constant dollars) within a given industry sector. The second variable,other sectorbiofinancingmeasures the total financing (in millions of 1991 constant dollars) inall other industry sectors. Because we expected nonmonotonic effects ofsectorbiofinancing—while initial increases in recent funding stimulate entry by signal-ing a munificent niche to entrepreneurs, high levels of recent funding discourageentry by signaling to entrepreneurs that they are too late—we included bothlinear and squared terms (divided by 100 for rescaling) in the analysis.

Second, we obtained yearly data on M.Sc. and Ph.D. degrees in agriculture andbiology subspecialties (animal science, plant science, soil science, genetics,microbiology, biochemistry, botany, fish and wildlife, food science, veterinarymedicine, veterinary science, zoology, and toxicology) granted by Canadianuniversities. Using this information we created a sector-specific labor supplyvariable computed as the number of M.Sc. and Ph.D.-level graduates in thesubspecialties most relevant to BFs in each sector.9

The resource opportunities available to start-ups depend on the intensity of

9 Sector-specific labor supply variables were defined as follows: (1) human (genetics, microbiol-ogy, biochemistry, and toxicology); (2) agriculture and horticulture (plant science, soil science,genetics, botany, and food science); (3) aquaculture (animal science, genetics, fish and wildlife, foodscience, veterinary medicine, veterinary science, and zoology); (4) forestry (plant science, soilscience, and botany); (5) engineering (microbiology, biochemistry, and toxicology); (6) environmen-tal (microbiology, biochemistry, and toxicology); (7) food, beverage, and fermentation (microbiol-ogy, biochemistry, food science, and toxicology); (8) veterinary (animal science, genetics, microbi-ology, biochemistry, fish and wildlife, veterinary medicine, veterinary science, and zoology); and (9)energy (microbiology, biochemistry, and toxicology).


competition at the time that funding or skilled labor is introduced. If potentialcompetition is not measured explicitly then the effect of adding resources to theenvironment is assumed to be constant over time and estimates for the effects offinancing and labor supply will suffer specification bias. Therefore, we includesector-specific measures of the potential fordirect anddiffusedensity-dependentcompetition (Hannan and Carroll, 1992). We measure the potential fordirectdensity dependent competition based on the number of BFs operating within agiven industry sector.10 We measure the potential fordiffusedensity-dependentcompetition based on the number of BFs operating in all other sectors.

As we characterized it above, the Canadian biotechnology industry is a“quasi-independent” population that, while shaped importantly by its own inter-nal dynamic, does not operate in a vacuum and is likely to be materially affectedby competition and intellectual property accumulation beyond its national bor-ders. Of particular significance is the United States, which is Canada’s closestneighbor and largest trading partner. Not only is the United States home to theworld’s largest national biotechnology industry, Canadian BFs patent their in-novations in and many export directly to the United States. Therefore, wedeveloped two sector-specific control variables. The first controls for sector-specific density-dependent competition from U.S. BFs, and the second forsector-specific intellectual property accumulation by U.S. BFs. We computedthese variables for the start of each observation year based on annual counts ofthe number of U.S. BFs and recent patents granted to U.S. BFs (i.e., granted inthe prior 5 years).11

These counts are disaggregated into six industry subcategories: (1) agriculture,(2) food, (3) diagnostics, (4) therapeutics, (5) veterinary, and (6) other. Since theU.S. subcategories are not as fine-grained as the Canadian sectors, we assignedU.S. BFs and patents to Canadian sectors as follows: (a) U.S.agriculture toCanadian agriculture, aquaculture, and horticulture; (b) U.S.food to Canadianfood, beverage, and fermentation; (c) U.S.human diagnosticsto Canadianhuman diagnostics; (d) U.S.human therapeuticsto Canadian human therapeuticsand human vaccines; (e) U.S.veterinary to Canadian veterinary; and (f) U.S.other to Canadian forestry, engineering, environmental, energy, contract re-search, biomaterials, cosmetics, and mining.

Because BFs and patents in the U.S. agriculture, human therapeutics, and othersubcategories were each assigned to multiple Canadian sectors, we needed amethod for allocating the U.S. counts among the Canadian sectors to formsector-specific U.S. firm densityand recent patentcontrol variables. We em-

10 We did not find any evidence of the nonmonotonic effect of organizational density predicted bythe density dependence model. While it is possible that this is the result of our study design, whichdoes not include information on the early history of the biotechnology industry in Canada (Hannanand Caroll, 1992), the positive coefficients for the linear density terms in our models do not supportsuch a speculation.

11 We are indebted to Terry Amburgey for providing us with these data. For a more detaileddescription of these data, see Amburgey et al. (1996).


ployed two allocation methods. In the first, we simply allocated U.S. firms andpatents equally across the Canadian sectors. In the second we weighted theallocations according to the proportion of Canadian BFs or patents in the sectorsto which the allocations were made. For example, if 20% of Canadian BFs in thehuman therapeutics and vaccine sectors operated primarily in the vaccine sector,then 20% of U.S. human therapeutics BFs would be allocated to the Canadianvaccine sector and 80% to the Canadian therapeutics sector. We based yearlyallocations on proportions computed for that year. In a preliminary analysis, thetwo allocation methods yielded similar estimates, but estimates were moreefficient for the equal allocation method, so we report them below.

In addition to the foregoing controls for resource munificence and competition,the frequencies of foundings and failures in particular industry sectors may serveas signals to entrepreneurs of the attractiveness of those locations (Delacroix andCarroll, 1983). In Delacroix and Carroll’s “rate-dependence model,” recentfoundings and failures are both argued to exert nonmonotonic influences onsubsequent foundings. Foundings initially signal a fertile niche to potentialentrepreneurs, encouraging additional foundings. But as foundings increase fur-ther, competition for resources increases, discouraging foundings. Similarly, atfirst, failures release resources that can be reassembled into new foundings. Butmany failures signal a hostile environment, discouraging founding. To control forsuch effects, we computedsector foundings,defined as the number of BFsfounded in a given sector in the prior year, andsector failures,defined as thenumber of BFs that failed in a given sector during the prior year. To permit theexpected curvilinear effects, we included both linear and squared terms for thesevariables.12

Finally, in addition to the foregoing time-varying controls, to insure thatcoefficients for our theoretical variables are not simply a spurious result of thepassage of time, we include a year time clock in the analysis. And, to account forany sector-specific differences not already captured by the time-varying sectorconditions or time trend controls, we also include a set of sector-specific dummyvariables, or “fixed effects.”13 For efficiency and parsimony, in developing ourbaseline model, we removed fixed effects for sectors that were not significantafter including all time-varying controls and the time-trend variable.

Dependent Variable and Analysis

The dependent variable in our analysis is the yearly number of foundingsoccurring in an industry sector. Because this variable is a count measure, we used

12 We explored the possibility that recent founding and failure inothersectors also influences therate of founding in a given sector but found no evidence to support the idea.

13 The cohort nature of our study design, and the potential mobility of firms across industry sectorsand into biotechnology from related industries, makes it impossible for us to distinguish the “ages”of the various sectors with much accuracy or confidence, so we include only a global time-trendvariable.


the pooled cross-sectional data to estimate the number of foundings expected tooccur within an interval of time (Hausman, Hall, and Griliches, 1984). Poolingacross time (6 years: 7 observation years minus a 1-year lag) and industry sectors(16 sectors) yields 96 sector-year spells. A Poisson process provides a naturalbaseline model for such processes and is appropriate for relatively rare events(Coleman, 1981). The basic Poisson model for event count data is:

Pr~Yt 5 y! 5 expl~ xt!@l~ xt! y/y! #,

where both the probability of a given number of events in a unit interval,Pr(Yt 5 y) and the variance of the number of events in each interval equal therate, l( xt). Thus, the basic Poisson model makes the strong assumption thatthere is no heterogeneity in the sample. However, for count data, the variancemay often exceed the mean. Such overdispersion is especially likely in the caseof unobserved heterogeneity. The presence of overdispersion causes the standarderrors of parameters to be underestimated, resulting in overstatement of levels ofstatistical significance. In order to correct for overdispersion, the negative bino-mial regression model can be used. A common formulation, which allows thePoisson process to include heterogeneity by relaxing the assumption that themean and variance are equal is:

lt 5 exp~p9xt!et,

where the error term,et, follows a gamma distribution. The presence ofetproduces overdispersion. The specification of overdispersion we use takes theform:

Var~Yt! 5 E~Yt!@1 1 aE~Yt!#.

In a preliminary analysis comparing fits of negative binomial and Poissonregression models, the overdispersion parameters were not significantly differentfrom zero (p , .05), indicating that negative binomial models did not improvesignificantly over Poisson models. Therefore, we report estimates from Poissonregression models below.14

Our analysis may be affected by moderate multicollinearity among some ofour explanatory variables [e.g., main effect and interaction terms, which canresult in less precise parameter estimates (i.e., larger standard errors)] for thecorrelated explanatory variables but will not bias parameter estimates (Kennedy,1992). Although moderate multicollinearity does not pose a serious estimationproblem, it may result in conservative tests of significance for correlated vari-ables, making it difficult to draw inferences about the effects of adding particularvariables to our models. Therefore, we estimate and test the significance ofgroups of variables in comparisons of a series of hierarchically nested regression

14 Although, as Barron (1992, p. 216) notes, a quasilikelihood (QL) approach may be preferredwhen lagged counts for autocorrelation are not justified, our inclusion of lagged founding is groundedtheoretically in well-known rate dependence models (Delacroix and Carroll, 1983).


models and examine coefficients’ standard errors for inflation to check thatmulticollinearity is not causing less precise parameter estimates (Kmenta, 1971).Although we do observe some degree of standard error inflation in successivemodels, most notably in relation to the interaction effects, our ability to judge thesignificance of individual coefficients is not materially diminished.

RESULTS

Table 1 presents the Poisson regression estimates of the yearly founding rates ofCanadian biotechnology firms during the 1991 to 1997 period. Model 1 provides abaseline model of BF founding that includes all the control variables.15 For simplic-ity, coefficients for sector fixed-effects are reported in the Appendix. Model 2, whichis a significant improvement over the baseline model (likelihood ratio tests are givenin the table), introduces the three incumbents’ patent variables to test H1, H2, and H3.Model 3, which adds the interactions between the incumbents’ patent variables andthe dummy variable for the human industry sectors to test H4a–H4c, is again asignificant improvement. Finally, to test H5 and H6, we include the horizontal andvertical (upstream and downstream) alliance variables in model 4. Model 4 yields asignificant improvement in fit compared to model 3 and thus represents our best-fitting model. Therefore, we interpret the coefficients from model 4.16

Patenting Effects (H1–H4)

H1, which predicted that incumbents’ recent rate of patenting in a givenindustry sector would lower the founding rate in that sector, is supported by thesignificant, negative coefficient for recent patents granted to sector incumbents.H4a, which predicted a stronger founding suppression effect of incumbents’recent patents in human sectors (i.e., diagnostics, therapeutics, and vaccines) isalso supported by the significant, negative coefficient for the interaction betweenrecent sector patents and the human sector dummy variable. Combined, thenegative main effect and interaction term indicate that while incumbents’ recentrate of patenting in a given industry sector lowered the founding rate in thatsector significantly, the founding-suppression effect was stronger in humansectors (b 5 2.1381 2.5845 2.722) than in nonhuman sectors (b 5 2.138).

In H2 we predicted that as the concentration of incumbents’ patenting activity

15 Preliminary analysis did not support a nonmonotonic effect forsector foundingsand removingthe squared term from the baseline model did not change the likelihood ratio significantly (likelihoodratio test5 .67, 1 df, ns).

16 To maintain a focus on variables of theoretical interest, we comment only briefly here on controlvariable coefficients. Generally, control variable estimates are in expected directions and exhibitrobust effects across the four models. Three exceptions are notable. One is that after adding thealliance variables in model 4 the mutualistic effect ofCanadian sector densityfalls from significance.A second is that the competitive effect ofU.S. sector densityfalls from significance in model 2, afterthe Canadian sector patenting variables are included; the significant negative effect ofU.S. sectorpatentsremains, however. The third is that the mutualistic effect ofother sector densitydisappearsonce the human industry sector interaction terms are introduced in model 3.


in a given sector increased, the founding rate in that sector would decline. Thishypothesis is also supported as indicated by the negative coefficient for the sectorinnovation concentration variable, although this coefficient is significant only atthe p , .10 level. The interaction between sector innovation concentration andthe human sector dummy variable is also significant and negative, supportingH4b. So, while an increasing concentration of incumbents’ patenting activity ina given industry sector lowered the rate of founding in that sector, the suppres-

TABLE 1Poisson Regression Models of BF Founding, 1991–1997: Patent and Alliance Effects

Variable Model 1 Model 2 Model 3 Model 4

Constant 187.6 (254.4) 147.88 (254.3) 139.47 (284.6) 211.83 (323.5)M.Sc. graduates .007* (.003) .007* (.003) .007* (.003) .007* (.004)Ph.D. graduates .013* (.007) .0111 (.007) .0111 (.007) .022* (.013)Sector foundings .022* (.010) .023* (.011) .0171 (.011) .0191 (.012)Sector failures .420* (.118) .432* (.129) .418* (.144) .366* (.172)Sector failures squared/100 2.011* (.003) 2.009* (.003) 2.009* (.003) 2.007* (.004)Year 2.263* (.121) 2.214* (.127) 2.2321 (.161) 2.359 (.247)Sector density .025* (.008) .028* (.008) .028* (.009) .014 (.012)Other sector density .013* (.006) .016* (.007) .008 (.008) .007 (.011)Sector biofinancing .377* (.221) .489* (.273) .533* (.303) 1.226* (.474)Sector biofinancing squared/100 2.0631 (.047) 2.110* (.060) 2.100 (.069) 2.204* (.129)Other sector biofinancing 2.572* (.187) 2.486* (.195) 2.500* (.203) 2.600* (.234)U.S. sector density 2.0061 (.004) 2.004 (.004) 2.002 (.004) .001 (.004)U.S. sector patents 2.011* (.004) 2.009* (.004) 2.008* (.004) 2.011* (.005)Incumbents’ patents

Recent sector patents 2.039* (.015) 2.009 (.042) 2.138* (.057)3 human sector 2.196* (.061) 2.584* (.261)

Sector innovation concentration 2.1661 (.111) 2.463 (1.159) 22.1841 (1.461)3 human sector 23.129* (1.377) 23.175* (1.887)

Sector technical maturity 2.269* (.121) 2.279* (.140) 2.3711 (.235)3 human sector 2.574* (1.014) 2.952* (1.496)

Incumbents’ alliancesHorizontal

Sector biotechnology alliances 2.216* (.067)

Vertical (downstream)Sector pharmaceutical alliances .191* (.079)Sector chemical alliances .250* (.104)Sector marketing alliances .0161 (.010)

Vertical (upstream)Sector university alliances .0581 (.038)Sector government lab alliances 2.116 (.131)Sector research institute alliances .0941 (.061)Sector association alliances 2.021 (.148)Sector hospital alliances .435* (.164)Likelihood ratio 61.69 53.23 42.91 24.32Likelihood ratio test vs model 1 8.46* (3df)Likelihood ratio test vs model 2 10.32* (3df)Likelihood ratio test vs model 3 18.59* (9 df)

Note. Standard errors are in parentheses; coefficients for sector fixed effects are given in theAppendix;N 5 96 sector-years.

1 p , .10.* p , .05.


sion effect was again stronger in human sectors (b 5 22.184 1 23.175 525.359) than in nonhuman sectors (b 5 22.184).

Finally, H3 predicted that as a sector’s technological maturity increased(measured by the predominance of old over recent patents among sector incum-bents) the founding rate in that sector would decline. The negative coefficient fora sector’s technological maturity supports this hypothesis at thep , .10 level.17

In contrast to the prediction in H4c, however, the coefficient for the interactionof technological maturity and human applications sectors is significant andpositive. One possible explanation for this unexpected finding is that in thehuman sectors, which are among the youngest sectors in the industry, anincreasing prevalence of older patents signalsviability—but not yetmaturity—topotential organization founders.

Alliance Effects (H5–H6)

Our theoretical arguments distinguished between incumbents’vertical andhorizontal alliances to suggest that, while vertical alliances in a given sectorwould stimulate founding in that sector (H5), horizontal alliances among bio-technology firms in the same or overlapping industry sectors would depress thefounding rate (H6). The significant, negative coefficient for sector biotechnologyalliances supports the predicted founding-suppression effect of incumbents’alliances with other potential rival BFs.

Our results show that sector vertical alliances do not uniformly increase BFfounding rates. Among the five types of upstream alliances, only one—incum-bents’ hospital alliances—raises the sector-founding rate significantly at thep ,.05 level. University and research institute alliances also raise the founding rate,but their effects are significant only atp , .10, and thecoefficient for hospitalalliances indicates a much larger per-alliance effect. A supplementary analysis(not reported) indicates that the hospital alliance effect is localized to humanapplication sectors of the industry.

In contrast, all three types of downstream alliances significantly increase therate of founding: As the number of collaborations between incumbent BFs andpharmaceutical and chemical firms in a given sector increases, so does thefounding rate. The rate at which BFs are founded increases as a function ofincumbents’ partnerships with marketing enterprises as well, although this effectis significant only at thep , .10 level. Notably, although we have conceivedhospital alliances asupstreambased on their role in information access, they may

17 Multicollinearity likely accounts, in part, for the decline in significance of the coefficient forsector technical maturitysince much of the decline in significance is attributable to the inflation ofthe coefficient’s standard error. It also likely accounts for the jump in the standard error forsectorinnovation concentrationfrom model 2 to model 3. However, the jump in the magnitude of thecoefficient forsector innovation concentrationin model 4 is likely the result of omitted variable biasin model 3 (i.e., the noninclusion of the alliance variables).


also be conceived to have an important downstream role in the clinical trialprocess—and it may be this role that leads them to stimulate foundings.

Overall, then, the vertical alliance coefficients in model 4 indicate thatincumbents’downstreamalliances were more consistently influential in de-termining the rate of BF start-ups than wereupstreamalliances. This patternof results may be an indication that thesignaling content of incumbents’vertical alliances is more important to potential entrants than theirsubstantivecontent. And, prevalence of upstream and downstream alliances may leadpotential entrants to very different conclusions about industry sector attrac-tiveness. In substantive terms, incumbents’ downstream alliances signal theavailability of capital and access to complementary assets, such as distribu-tion channels, marketing, and clinical trials expertise and production facili-ties, all of which are necessary for successful product development andcommercialization. In signaling terms, the prevalence of such alliancesamong incumbents in a given industry sector may be interpreted by potentialentrants as a clear sign of the technical feasibility and potential commercialviability of operations in the sector. Upstream alliances, which, in substantiveterms, can provide early access to cutting-edge research and know-howessential to the discovery and patenting of new products and processes, mayprovide more ambiguous signals, however. In particular, incumbents’ up-stream alliances may be interpreted by potential entrants that a sector remainsin an exploratory mode, far from the realization of commercial success, andperhaps even raising questions about its ultimate commercial viability.

Thus, in contrast to our original arguments, which emphasized the sub-stantive content of vertical alliances, equally—and perhaps more—importantto comprehending the effects of vertical alliances on organization foundingmay be their signal content. This possibility is consistent with the classicsignaling literature in economics (Spence, 1974), Delacroix and Carroll’s(1983) “rate-dependence model,” and more recent research on organizationalstatus in economic sociology (Podolny, 1994; Stuart et al., 1999). Alterna-tively, downstream activities such as marketing or manufacturing may be farmore scale-intensive than upstream activities. Since scale-intensive activitiesare likely to present more severe obstacles to potential entrants, downstreamalliances (which obviate the need for a new entrant to conduct such activitiesitself) may therefore overcome more significant obstacles than upstreamalliances. Whatever the exact mechanism at work, our findings are consistentwith the view that one of the critical problems solved by the biotechnologyindustry’s downstream linkages is of access to production and marketingcapabilities for patented innovations. In hindsight, it isn’t entirely surprisingthat, in a high-technology environment where the problem of competition ismitigated by firms’ internalization of monopoly rents for innovative efforts,potential newcomers to the biotechnology industry are most influenced byincumbents’ downstream rather than upstream alliance activities.


DISCUSSION AND CONCLUSION

In this article, we linked institutional and ecological research emphasizing theenvironmental structuring of entrepreneurial opportunity to Fligstein’s (1996)insight that the structure of market behavior reflects organizations’ efforts tostabilize the competition they face, given the institutional context in which theyare embedded. In doing so, we conceptualized incumbent biotechnology firms’patenting and alliance-building activities as attempts to control potential com-petition and analyzed how incumbents’ technology appropriation and interfirmcollaboration influence start-up rates in the Canadian biotechnology industry. Byconceptualizing incumbents’ actions as attempts to control their competitiveenvironment, we are better able to understand how, and to what effect, high-technology, knowledge-intensive organizations make use of significant elementsin their institutional environments to control competition. In our particular case,we show, and provide a general framework for understanding, the central rolesplayed by intellectual property rights and strategic alliances in the evolution ofthe Canadian biotechnology industry.

Our empirical analysis demonstrates how the evolution of the Canadianbiotechnology industry is conditioned by its underlying patterns of intellectualproperty accumulation and alliance activity. We interpret this industry dynamicas both a reflection of incumbent firms’ attempts to use patent laws to controlcompetition by appropriating monopoly rents to innovation and their use ofstrategic alliances as attempts to manage joint flows of knowledge and commer-cialization competencies. Specifically, we found that prevalent incumbent pat-enting dissuaded entrepreneurs from founding new BFs, and, in addition, thatincreasing concentration of patenting among incumbents further reduced thefounding rate. These effects were stronger in the human applications sectors,where ancillary government policies exacerbate the challenge of overcomingincumbents’ patent protection. We also found that incumbents’ downstreamalliances stimulated BF foundings more consistently than incumbents’ upstreamalliances. We interpreted this to imply that entrepreneurs’ founding decisionswere more strongly influenced by signals of the commercial viability of industrysectors in Canadian biotechnology than by access to cutting-edge research andknow-how. An alternate interpretation is that, compared to upstream alliances,downstream alliances solve a more significant and critical problem for newentrants given the scale-intensiveness of downstream activities.

How generalizable are our results? By its very nature, an approach that ispredicated upon local, market-specific responses to institutions will generateempirical results that are, at first glance, idiosyncratic. Yet we suspect that ourresults are generalizable at least to other high-technology, knowledge-intensiveindustries. Consider semiconductor producers. Semiconductor producers operatein a science- and technology-based milieu broadly similar to that of biotechnol-ogy firms. However, the characteristics of semiconductor technologies lead to atleast three differences in how property rights and interfirm alliances affect firms’


attempts to mediate competition in this industry. First, the appropriability regimesurrounding semiconductor technologies is weaker than in biotechnology (Levinet al., 1987). Second, unlike the biotechnology industry, product life cycles aresufficiently fast in the semiconductor industry that patented innovation is oftenobsolete before the expiration of the patent (Podolny et al., 1996). Third, whereasa biotechnology product typically draws on a single patented innovation, semi-conductor products draw on hundreds or even thousands of prior patents (Grind-ley and Teece, 1997). What mechanisms of control do we think structure marketaction in the semiconductor industry? Typically, incumbents rely on extensivecross-licensing agreements with each other. Given the above-noted characteris-tics of semiconductor technologies, cross-licensing agreements solve problemsassociated with the commercialization of innovations by reducing the need forcostly, time-consuming, and uncertain legal battles. Note that they may alsoeffectively control competition by specifying who is, and who is not, capable ofparticipating in the industry. Without access to other firms’ patents, a semicon-ductor firm is at a decided disadvantage. But, a firm cannot get access to otherfirms’ patents unless it has something of its own that is worth licensing. On thesurface, at least, this cross-licensing pattern may appear to be very different fromthe mechanisms at work in biotechnology. More broadly, however, we see ageneral process in which organizations act to control competition as best they cangiven the set of locally enacted institutional and other constraints they face.

In addition, our results also yield broader implications for the development ofan institutional ecology of organizational founding that are consistent with recentresearch demonstrating that government policy can shape the nature of compe-tition in an organizational population and consequently influence founding rates(e.g., Barnett and Carroll, 1993; Dobbin and Dowd, 1997; for a review, seeBaum, 1996). Our finding that the relationship between incumbent patentingactivity and organizational founding is stronger for those biotechnology sectorsthat are subject to significant regulation of commercialization activities providessupplementary evidence of the manner in which public policies (whether bydesign or not) can dramatically affect competition. More generally, our studyinfuses organizational ecology with institutions—property rights, links to gov-ernment, and relations to upstream and downstream suppliers and customers—that shape interfirm competition and entrepreneurship in both direct and indirectways. As a result, our conception of market processes is more constructionist andnuanced than is typical of ecological research focused on niche overlap andresource dependence.

Further, our results extend the strategic alliance literature. Although priorresearch in this area has demonstrated effects of strategic alliances on a widerange of start-ups’ activities and performance outcomes, including innovativeactivity (Shan, Walker, and Kogut 1994; Stuart, 1998), valuations of initialpublic offerings (Stuart et al., 1999), and rates of growth and survival (Baum,Calabrese, and Silverman, 2000; Baum and Silverman, 1998; Stuart, 1998), fewstudies have analyzed how incumbent firms’ interorganizational alliances shape


patterns of organizational entry (Baum and Oliver, 1992, 1996; Hybels et al.,1994; Kogut et al., 1995), and none have analyzed how dissimilar types ofalliances differentially affect start-up activity. Our findings indicate that aggre-gating all alliances regardless of partner type masks variable founding effectsassociated with different alliance partners (which in turn may stem from distinctconceptions of control associated with each type). Further, although prior re-search has generally assumed (and found) that cooperation among competitorsstimulatesfoundings, on the assumption that cooperation will reduce competitiveintensity in an organizational population (e.g., Dobbin and Dowd, 1997; Flig-stein, 1990), we find that such cooperationdiscouragesfoundings. In this regard,it is interesting to note that prior research finding competition-reducing effects ofcooperation among potential rivals has examined cooperation inproduct oroutputmarkets. By contrast, partnerships between biotechnology firms, which wefind to be competition-enhancing, are typically directed toward cooperation ininput markets—primarily access to skilled scientists and development of newknowledge. Interestingly, a large economic literature proposes that cooperationin input markets—especially R&D—may increase competition, while coopera-tion in output markets decreases it (Katz and Ordover, 1990). Overall, ouralliance-related findings suggest, at a minimum, that different alliance partnertypes have variable founding effects. However, future work may find thatalliance partner-type effects are a function of viable conceptions of control thatexist, and govern incumbents’ behavior, in a particular organizational populationat a particular point in time. Nonetheless, our alliance results point the waytoward more nuanced analyses of how, and under what conditions, specific typesof partner-based interfirm cooperation shape industry competition and evolution.

Finally, our results also inform a growing literature in strategic management andorganization theory that views the competitive dynamics and evolution of organiza-tional populations as shaped by their underlying technologies and technologicalinnovation. Research in this stream has shown how technological processes, includ-ing the emergence of dominant designs (Baum, Korn, and Kotha, 1995; Suarez andUtterback, 1995; Tushman and Anderson, 1986; Utterback and Suarez, 1993),competence-enhancing or -destroying innovations (Henderson and Clark, 1990;Tushman and Anderson, 1986), and technological standards (Wade, 1995) system-atically influence organizational founding and failure rates. Recent studies are alsobeginning to reveal the importance of organizations’ intellectual property develop-ment and appropriation for firm survival and growth (Baum et al., 2000; Baum andSilverman, 1998; Podolny et al., 1996; Powell et al., 1998). In this study we extendedthis line of research by examining how incumbents’ patenting activity and howtechnological maturity affectde novoentry. Our findings complement prior researchby demonstrating not only the importance of technological innovation, but by alsohighlighting the added significance of firms’ appropriation of the returns to knowl-edge development via property rights acquisitions for the evolution of the Canadianbiotechnology industry.

Taken together, our findings underscore the critical role of intellectual


property rights and interfirm alliances in shaping the environment for high-technology, knowledge-intensive entrepreneurial start-ups. Consistent withFligstein (1996), we interpret our results as demonstrating how incumbents’quest to stabilize and control competition configures environmental opportu-nities for organizational founding in the Canadian biotechnology industry.Although we suspect our findings are germane to high-technology, knowl-edge-intensive industries more generally, future work replicating our findingsbeyond a single national industry would contribute to our understanding ofthe coevolution of knowledge-intensive organizations, their underlying tech-nologies and appropriation regimes, and interfirm alliance configurations. Webelieve our study and such follow-up research would help to answer Baumand Powell’s (1995) call for the development of an “institutionally informedecology of organizational evolution,” particularly as it pertains to high-technology, knowledge-intensive industries such as telecommunications,semiconductors, personal computing, and aerospace. In addition, of course,the study of the interrelated dynamics of locally arrived-at conceptions ofcontrol which structure markets, technological evolution and its associatedappropriation patterns, interfirm alliances, and industry evolution should beextended beyond organizational founding to include firm growth, profitabil-ity, and survival, all of which contribute, over time, to the emergence,structuration, transformation, and demise of industries.

APPENDIX

Industry Sector Fixed Effects for Table 1

Sector Model 1 Model 2 Model 3 Model 4

Agriculture 2.941* (.417) 2.930* (.428) 2.998* (.432) 2.442 (.696)Aquaculture — — — —Biomaterials/cosmetics/mining — — — —Contract research organization 9.642* (2.683) 8.118* (2.907) 9.659* (3.134) .463 (6.492)Energy — — — —Engineering — — — —Environment 1.957* (.888) 1.982* (.913) 1.904* (1.088)21.075 (2.289)Food/beverage/fermentation 4.117* (1.376) 3.299* (1.422) 3.138* (1.616) 7.634* (3.413)Forestry — — — —Human 2.433 (.939) 2.240 (1.194) 2.689 (2.033) 2.113 (6.304)Horticulture — — — —Veterinary 2.507* (.971) 2.308* (.961) 2.202* (1.196) 1.830 (2.302)

Note. Standard errors are in parentheses;N 5 96 sector-years. Preliminary analysis yieldednonsignificant coefficients for all excluded sector dummies. Although the human sector dummywas not significant in the preliminary analysis, we retained it to obtain unbiased estimates forinteractions between the patent variables and the human sector dummy (Cohen and Cohen,1983).

* p , .05.


REFERENCES

Aldrich, H. E. (1990). “Using an ecological perspective to study organizational founding rates,”Entrepreneurship Theory and Practice14, 7–24.

Amburgey, T. L., Dacin, T., and Singh, J. (1996). “Learning races, patent races, and capital races:Strategic interaction and embeddedness within organizational fields,”in Advances in StrategicManagement (J. A. C. Baum and J. E. Dutton, Eds.), Vol. 13, 303–322. JAI Press, Greenwich,CT.

Argyres, N. S., and Liebeskind, J. P. (1998). “Privatizing the intellectual commons: Universities andthe commercialization of biotechnology,” Journal of Economic Behavior & Organization35,427–454.

Austin, D. H. (1993). “An event study approach to measuring innovative output: The case ofbiotechnology,” American Economic Review83, 253–258.

Bain, J. S. (1956). Barriers to New Competition, Harvard Univ. Press, Cambridge, MA.Barley, S., Freeman, J., and Hybels, R. (1992). “Strategic alliances in commercial biotechnology,”in

Networks and Organizations: Structure, Form and Action (N. Nohria and R. Eccles, Eds.), pp.311–347. HBS Press, Boston.

Barnett, W. P. (1990). “The organizational ecology of a technological system,” AdministrativeScience Quarterly35, 31–60.

Barnett, W. P., and Carroll, G. R. (1993). “How institutional constraints affected the organization ofearly American telephony,” Journal of Law, Economics and Organization9, 98–126.

Barron, D. N. (1992). “Analysis of count data: Overdispersion and autocorrelation,” SociologicalMethodology22, 179–220.

Baum, J. A. C. (1996). “Organizational ecology,”in Handbook of Organization Studies (S. Clegg, C.Hardy, and W. Nord, Eds.), pp. 77–114. Sage, London.

Baum, J. A. C., Calabrese, T., and Silverman, B. S. (2000). “Don’t go it alone: Alliance networkcomposition and startups’ performance in Canadian biotechnology,” Strategic ManagementJournal21 (Special Issue), 267–294.

Baum, J. A. C., Korn, H. J., and Kotha, S. (1995). “Dominant designs and population dynamics intelecommunications services: Founding and failure of facsimile transmission service organiza-tions, 1965–1992,” Social Science Research24, 97–135.

Baum, J. A. C., and Oliver, C. (1991). “Institutional linkages and organizational mortality,” Admin-istrative Science Quarterly36, 187–218.

Baum, J. A. C., and Oliver, C. (1992). “Institutional embeddedness and the dynamics of organiza-tional populations,” American Sociological Review57, 540–559.

Baum, J. A. C., and Oliver, C. (1996). “Toward an institutional ecology of organizational founding,”Academy of Management Journal39, 1378–1427.

Baum, J. A. C., and Powell, W. W. (1995). “Cultivating an institutional ecology of organizations:Comment on Hannan, Carroll, Dundon, and Torres,” American Sociological Review60,529–538.

Baum, J. A. C., and Silverman, B. S. (1998). “Alliance and patent-based competition in the Canadianbiotechnology industry,” paper presented at the Academy of Management, San Diego CA,August.

Baum, J. A. C., and Singh, J. V. (1994). “Organizational niches and the dynamics of organizationalfounding,” Organization Science5, 483–501.

Biotech ’97, Ernst & Young, San Francisco, CA.Canadian Biotech, various years, Ernst & Young, Toronto.Canadian Biotechnology, various years (F. Haynes, Ed.), Contact Canada, Toronto.Canadian Biotechnology Handbook, various years (P. Winter, Ed.), Canadian Biotech News, Nepean,

Ontario.Carroll, G. R., and Huo, Y. P. (1986). “Organizational task and institutional environments in

ecological perspective: Findings from the local newspaper industry,” American Journal ofSociology91, 838–873.


Caves, R. M., and Porter, M. E. (1977). “From entry barriers to mobility barriers: Conjecturaldecisions and contrived deterrence to new competition,” Quarterly Journal of Economics91,241–261.

Cohen, J., and Cohen, P. (1983). Applied Multiple Regression/Correlation Analysis for the Behav-ioral Sciences, Second Edition. Erlbaum, Hillsdale, NJ.

Coleman, J. S. (1981). Longitudinal Data Analysis, Basic Books, New York.Davis, L. E., and North, D. C. (1971). Institutional Change and American Economic Growth,

Cambridge Univ. Press, Cambridge, UK.Delacroix, J., and Carroll, G. R. (1983). “Organizational foundings: An ecological study of the

newspaper industries of Argentina and Ireland,” Administrative Science Quarterly28,274–291.DiMaggio, P. J., and Powell, W. W. (1983). “The iron cage revisited: Institutional isomorphism and

collective rationality in organizational fields,” American Sociological Review48, 148–160.Dobbin, F., and Dowd, T. J. (1997). “How policy shapes competition: Early railroad foundings in

Massachusetts,” Administrative Science Quarterly42, 501–529.Eisenhardt, K. M., and Schoonhoven, K. (1990). “Organizational growth: Linking founding team,

strategy, environment, and growth among U.S. semiconductor ventures, 1978–1988,” Admin-istrative Science Quarterly35, 504–529.

Fligstein, N. (1996). “Markets as politics: A political-cultural approach to market institutions,”American Sociological Review61, 656–673.

Fligstein, N. (1990). The Transformation of Corporate Control, Harvard Univ. Press, Cambridge,MA.

Gartner, W. B. (1989). “Some suggestions for research on entrepreneurial traits and characteristics,”Entrepreneurship: Theory and Practice14, 27–37.

Gompers, P. A., and Lerner, J. (1998). “What drives venture capital fundraising?” Brookings Paperson Economic Activity; Microeconomics, 149–192.

Greene, W. H. (1990). Econometric Analysis, Macmillan, New York.Grindley, P. C., Mowery, D. C., and Silverman, B. (1994). “SEMATECH and collaborative research:

Lessons in the design of high-technology consortia,” Journal of Policy Analysis and Manage-ment13, 723–758.

Grindley, P. C., and Teece, D. J. (1997). “Managing intellectual capital: Licensing and cross-licensing in semiconductors and electronics,” California Management Review39(2), 8–41.

Hannan, M. T., and Carroll, G. R. (1992). Dynamics of Organizational Populations: Density,Competition, and Legitimation. Oxford Univ. Press, New York.

Haunschild, P. R., and Miner, A. S. (1997). “Modes of interorganizational imitation: The effects ofoutcome salience and uncertainty,” Administrative Science Quarterly42, 472–500.

Hausman, J., Hall, B. H., and Griliches, Z. (1984). “Economic models for count data with anapplication to the patents and R&D relationship,” Econometrica52, 909–937.

Henderson, R. M., and Clark, K. B. (1990). “Architectural innovation: The reconfiguration of existingproduct technologies and the failure of established firms,” Administrative Science Quarterly35,9–30.

Holtzeakin, D., Joulfaian, D., and Rosen, H. S. (1994). “Entrepreneurial decisions and liquidityconstraints,” Rand Journal of Economics25, 334–347.

Howes, P. (1997). “Patent Protection,”in PMAC Annual Review, 1996–97, Chapter 2, Pharmaceu-ticals Association of Canada, Toronto.

Hybels, R. C., Ryan, A. R., and Barley, S. R. (1994). “Alliances, legitimation, and founding rates inthe U.S. biotechnology field, 1971–89,” paper presented at the Academy of ManagementNational Meetings, Dallas, TX.

Katz, M. L., and Ordover, J. A. (1990). “R&D cooperation and competition,” Brookings Papers onEconomic Activity, SI, 137–203.

Kennedy, P. (1992). A Guide to Econometric Methods (3rd edition). MIT Press, Cambridge, MA.Kenney, M. (1986). Biotechnology: The University–Industry Complex, Yale Univ. Press. New

Haven, CT.


Kmenta, J. (1971). Elements of Econometrics, Macmillan, New York.Kogut, B., Shan, W., and Walker, G. (1992). “The make-or-cooperate decision in the context of an

industry network,”in Networks and Organizations: Structure, Form and Action (N. Nohria andR. G. Eccles, Eds.), pp. 348–365, HBS Press, Boston.

Kogut, B., Walker, G., and Kim, D. J. (1995). “Cooperation and entry induction as an extension oftechnological rivalry,” Research Policy24, 77–95.

Kondo, E. K. (1995). “The effect of patent protection on foreign direct investment,” Journal of WorldTrade,29, 97–122.

Lerner, J. (1995). “Patenting in the shadow of competitors,” Journal of Law and Economics38,463–495.

Levin, R. C., Klevorick, A. K., Nelson, R. R., and Winter, S. G. (1987). “Appropriating the returnsfrom industrial research and development,” Brookings Papers on Economic Activity3, 783–833.

Liebeskind, J. P., Oliver, A. L., Zucker, L., and Brewer, M. (1996). “Social networks, learning andflexibility: Sourcing scientific knowledge in new biotechnology firms,” Organization Science7,428–443.

Miner, A. S., Amburgey, T. L., and Stearns, T. (1990). “Interorganizational linkages and populationdynamics: Buffering and transformational shields,” Administrative Science Quarterly35, 689–713.

Mowery, D. C., Oxley, J. E., and Silverman, B. S. (1996). “Strategic alliances and interfirmknowledge transfer,” Strategic Management Journal17(Special Issue), 77–92.

Nelson, R. R., and Winter, S. G. (1982). An Evolutionary Theory of Economic Change, HarvardUniv. Press, Cambridge MA.

Nonaka, I., and Takeuchi, H. (1995). The Knowledge-Creating Company, Oxford Univ. Press, NewYork.

North, D. C. (1991). “Institutions,” Journal of Economic Perspectives5, 97–112.Pisano, G. P. (1990). “The R&D boundaries of the firm: An empirical analysis,” Administrative

Science Quarterly35, 153–176.Podolny, J. M. (1994). “Market uncertainty and the character of economic exchange,” Administrative

Science Quarterly39, 458–483.Podolny, J., Stuart, T. E., and Hannan, M. T. (1996). “Networks, knowledge, and niches,” American

Journal of Sociology102,659–689.Powell, W. W. (1990). “Neither market nor hierarchy: Network forms of organization,” Research in

Organizational Behavior,12, 295–336.Powell, W. W., and Brantley, P. (1992). “Competitive cooperation in biotechnology: Learning

through networks?”in Networks and Organizations: Structure, Form and Action (N. Nohria andR. G. Eccles, Eds.), pp. 366–394, HBS Press, Boston.

Powell, W. W., Koput, K. W., and Smith-Doerr, L. (1996). “Interorganizational collaboration and thelocus of innovation: Networks of learning in biotechnology,” Administrative Science Quarterly41, 116–145.

Powell, W. W., Koput, K. W., Smith-Doerr, L., and Owen-Smith, J. (1998). “Newtork position andfirm performance: Organizational returns to collaboration in the biotechnology industry,”working paper, University of Arizona.

Rosenberg, N. (1982). Inside the Black Box: Technology and Economics, Cambridge Univ. Press.Cambridge, UK.

Shan, W., Walker, G., and Kogut, B. (1994). “Interfirm cooperation and startup innovation in thebiotechnology industry,” Strategic Management Journal15, 387–394.

Singh, J. V., Tucker, D. J., and Meinhard, A. G. (1991). “Institutional change and ecologicaldynamics,” in The New Institutionalism in Organizational Analysis (W. W. Powell and P. J.DiMaggio, Eds.), pp. 390–422, Univ. of Chicago Press, Chicago.

Spence, A. M. (1974). Market Signaling, Harvard Univ. Press, Cambridge, MA.


Stewart, J. (1997). “Drug Development and Approval,”in PMAC Annual Review, 1996–97, Chapter7. Pharmaceuticals Association of Canada, Toronto.

Stigler, G. J. (1971). “The theory of economic regulation,” Bell Journal of Economics2, 3–21.Stinchcombe, A. L. (1965). “Social structure and organizations,”in Handbook of Organizations (J. G.

March, Ed.), pp. 142–193, Rand McNally, Chicago.Stuart, T. E. (1998). “Interorganizational alliances and the performance of firms; A study of growth

and innovation rates in a high-technology industry,” working paper, Graduate School ofBusiness, University of Chicago.

Stuart, T. E., Hoang, H., and Hybels, R. C. (1999). “Interorganizational endorsements and theperformance of entrepreneurial ventures,” Administrative Science Quarterly44, 315–349.

Suarez, F. F., and Utterback, J. M. (1995). “Dominant designs and the survival of firms,” StrategicManagement Journal16, 415–430.

Teece, D. J. (1986). “Profiting from technological innovation,” Research Policy15, 285–305.Tushman, M. L., and Anderson, P. (1986). “Technological discontinuities and organizational envi-

ronments,” Administrative Science Quarterly31, 439–465.Utterback, J. M., and Suarez, F. F. (1993). “Innovation, competition, and industry structure,” Policy

Studies22, 1–21.Uzzi, B. (1996). “The sources and consequences of embeddedness for the economic performance of

organizations: The network effect,” American Sociological Review61, 674–698.Wade, J. (1995). “Dynamics of organizational communities and technological bandwagons: An

empirical investigation of community evolution in the microprocessor market,” Strategic Man-agement Journal16, 111–133.

Walker, G., Kogut, B., and Shan, W. J. (1997). “Social capital, structural holes and the formation ofan industry network,” Organization Science8, 109–125.

Williamson, O. E. (1994). “Transaction cost economics and organization theory,”in The Handbookof Economic Sociology (N. J. Smelser and R. Swedberg, Eds.), pp. 77–107. Princeton Univ.Press, Princeton, NJ.

Zucker, L. G., Darby, M. R., and Brewer, M. B. (1998). “Intellectual human capital and the birth ofU.S. biotechnology enterprises,” American Economic Review88, 290–306.


Canadian Biotechnology Start-Ups, 1991–1997: The Role of Incumbents' Patents and Strategic Alliances in Controlling Competition

Documents