Strategic Management Journal Strat. Mgmt. J., 30: 711–735 (2009) Published online 30 December 2008 in Wiley InterScience (www.interscience.wiley.com) DOI: 10.1002/smj.746 Received 11 February 2005; Final revision received 9 November 2008

UNBUNDLING COMPETITIVE HETEROGENEITY: INCENTIVE STRUCTURES AND CAPABILITY INFLUENCES ON TECHNOLOGICAL INNOVATION MICHAEL J. LEIBLEIN1 * and TAMMY L. MADSEN2 1 2

Fisher College of Business, Ohio State University, Columbus, Ohio, U.S.A. Leavey School of Business, Santa Clara University, Santa Clara, California, U.S.A.

Many studies argue that the continual creation of new ideas by small and young firms steadily destroys the competitive positions of their larger, more established rivals. Despite this attention, empirical results relating firm size to innovation remain exceedingly fragile. This study proposes three reasons for the empirical inconsistencies in the literature: that small and large firms differ in their: (1) stock of technological experiences, (2) use of own- and partner-firm experiences, and (3) abilities to translate own- and partner-firm experiences into innovation activity. Results from a 10-year study of 463 semiconductor firms demonstrate that the mixed findings generated from prior work are partially attributed to these three general propositions. In particular, resource flows, in the form of operating experience developed internally and accessed through codevelopment partners, positively affect innovation activity; but these benefits diminish as a firm increases in size. The findings broadly support the notion that differences in the incentives and abilities of small and large firms give rise to heterogeneity in the firms’ innovation activity. Copyright  2008 John Wiley & Sons, Ltd.

INTRODUCTION How should firms of different size organize to affect technological innovation? The importance of this question is manifest when one considers the varied opportunities and threats that both small and large firms face in their attempts to develop superior performance through innovation. Early work details the different roles that small, young entrants and large, established firms play in pushing developments along the technological frontier (e.g., Abernathy and Utterback, 1978; Cooper and Schendel, 1976). Historically, work in this tradition has examined how factors associated with firm size affect a firm’s incentives Keywords: organization economics; resource-based view; learning; competitive heterogeneity; innovation *Correspondence to: Michael J. Leiblein, Fisher College of Business, Ohio State University Columbus, OH 43210, U.S.A. E-mail: [email protected]

Copyright  2008 John Wiley & Sons, Ltd.

to invest in research and development (R&D) (see Cohen and Levin, 1989 for a review). More recent efforts shift attention to the relationship between firm size and innovative outputs, such as patents or the development and commercialization of particular types of innovations (e.g., Acs and Audretsch, 1988). Related work linking firm age to innovation or growth shows that a firm’s stock of administrative experience may enable or constrain its behavior (e.g., Barron, Elizabeth, and Hannan, 1994; Sorensen and Stuart, 2000). Although numerous, dramatic examples exist wherein small, young ventures have successfully entered markets previously dominated by large, established firms (e.g., Foster, 1986; Utterback, 1994), drawing definitive conclusions about the relationship between firm size and innovation activity from the empirical literature is difficult. Notable studies show that small and young firms

712

Michael J. Leiblein and Tammy L. Madsen

(Mitchell and Singh, 1993) and large and established firms most often introduce radical technologies (Cohen, 1995). Noting these contradictory conclusions, other work suggests that the relationship between firm size and innovative productivity is either bell-shaped (e.g., Ettlie and Rubenstien, 1987) or U-shaped (e.g., Pavitt, 1990). Summarizing the state of the empirical literature, a recent review states that ‘researchers have not reached a consensus about the role of size. Contradictions abound. Managerial useful generalizations are rare’ (Chandy and Tellis, 1998: 475). Despite substantial theoretical and empirical attention, the state of this literature remains inconclusive. This study contributes to our understanding of firm-level differences in innovation activity by proposing and testing three possible reasons for the inconsistent findings in the extant empirical literature as it relates to firm size and innovation. First, we jointly consider the influence of firm size and experience on innovation. Unbundling the effect of experience from the general effects of firm size and age allows us to account directly for the effects of specific stocks of experience on innovation. Second, controlling for firm size and age, we consider how different stocks of experience, such as those developed internally and those accessed through codevelopment agreements or production sourcing contracts, affect a firm’s innovation activity. In so doing, we account for differences in coordinative capabilities associated with the use of different organizational forms that may affect firms’ innovation activity. Third, we interactively assess how the relationship between experience, the organizational form used to develop or access experience, and the propensity to innovate varies by firm size. In sum, the study argues that small firms have incentive and informational advantages over their larger counterparts. However, the benefits associated with these advantages must be balanced by small firms’ relative lack of experience. This experience shortfall may lead small firms, on average, to access experience via organizational forms that are less well suited for innovation activities. In combination, we argue that the effects of different stocks of experience and incentives may explain the conflicting results observed in extant empirical work. We investigate these ideas using data on 463 international semiconductor firms operating during the 1990 to 1999 time period. The data include Copyright  2008 John Wiley & Sons, Ltd.

information on each firm’s own operating experience and on the operating experience of each firm’s codevelopment partners and production-sourcing partners. Our analysis demonstrates how experience accessed through each of these discrete governance forms affects firms’ propensities to innovate. The findings show that innovation activity in the semiconductor industry is jointly driven by a firm’s stock of experience and the organizational mechanisms it uses to access this experience. In addition, the findings show that the influence of different types of operating experience on innovation activity varies across small and large firms. Together, the results suggest the relative importance of firm-specific capabilities emphasized in the organizational learning and resourcebased view (RBV) literatures and the coordination and incentive alignment issues emphasized in the theory of the firm literature in explaining innovative performance. Subsequent sections provide background theoretical material and develop hypotheses relating different types of experience and organizational form decisions to technological innovation. We then describe the data, measures, and estimation methods. The article concludes by discussing the study’s implications for theory and future technology strategy research.

LITERATURE REVIEW A substantial body of theoretical and empirical research examines the relationship between firm size and innovation activity. Given the variety of definitions and measures of innovation, this study examines process technology innovations at the firm level.1 Process technology innovations are defined as significant changes in the tools, devices, 1 Although this study examines process technology innovations, the existing literature describes innovation in a variety of ways. Studies often discuss innovation using the degree of change, such as incremental or radical (e.g., Henderson and Clark, 1990) or competence-enhancing versus competence-destroying (e.g., Tushman and Anderson, 1986). Other research focuses on the type of change, such as product or process (e.g., Rosenberg, 1982). Innovation also has been defined using different levels of analysis. For instance, Schumpeter (1934) discusses innovation as the introduction of a new method of production (process), which need not be founded on a scientifically new discovery. Other work relates innovation to the timing of an event or change. For example, one might consider innovation only to occur at the first instance that a change is introduced to the world and all remaining events as subsequent ‘adoptions’. These definitions suggest that innovation may involve drastic or evolutionary change, processes or products, and may be defined to

Strat. Mgmt. J., 30: 711–735 (2009) DOI: 10.1002/smj

Incentive Structures, Capability Influences, and Innovation and knowledge that mediate between inputs and outputs (Rosenberg, 1972). Managers perceive new process technology as a means of increasing productivity and flexibility, reducing cycle times, and improving quality (Adler, 1988; Stalk, 1988). Technology scholars state that process technology innovation is crucial to competitiveness (e.g., Ettlie, 1988). Although process innovations do not necessarily require fundamental changes in product design, they often necessitate the development of new manufacturing skill sets and the integration of new manufacturing equipment. When the changes in these skill sets are significant, they drastically affect the cost and/or value of the output. At least three research streams examine the relationship between firm size and the amount of a firm’s innovation activity. The first stream relates firm size and market power to the incentive to invest in innovation activities. This work suggests that large firms typically are better able to control the resources necessary to direct technical change, or to develop barriers that allow them to appropriate the gains from innovation relative to small firms (e.g., Schumpeter, 1942). As such, large firms, on average, will have greater incentives to invest in R&D and, in turn, will be more likely to innovate as compared to small firms. Focusing on the risky nature of R&D investment, Galbraith (1952) extended this basic reasoning by arguing that firms may reduce their exposure to risky investments in innovation activity by spreading investment across a large number of projects. If large firms can spread their investment over a larger number of R&D projects than small firms can, large firms may be able to invest more resources in innovation for a given risk level and, in turn, generate larger amounts of innovation activity than their smaller rivals. Subsequent work refined these arguments by examining how various product- and factor-market imperfections affect the relationship between firm size and innovation activity. For instance, if capital market imperfections exist, large firms may secure financing for risky R&D investments more efficiently than small firms by leveraging internal capital markets (e.g., Armour and Teece, 1981). Similarly, if imperfections exist in the market for technological knowledge, large, and presumably more diversified, firms may be better positioned occur at the instant a change is introduced to the world or to a more local environment. Copyright  2008 John Wiley & Sons, Ltd.

713

to exploit innovations through internal application rather than external licensing (e.g., Teece, 1982), relative to their smaller, and more focused, rivals. Other work argues that scale economies in R&D activities directly favor large firms and/or that complementarities between R&D, manufacturing, and other downstream activities favor large firms (Teece, 1986). The results from this work however, are extremely fragile. Given this fragility, Teece states that, ‘A large but unsatisfactory literature exists in industrial organization on the relationship between market structure and innovation, and between firm size and innovation, but both the theoretical and empirical literature are almost completely silent on interfirm and intrafirm organizational issues. When these issues have been addressed, it is without much of a theoretical foundation (Teece, 1992 : 2). A second body of work, couched within the framework of patent races, explores the trade-offs established firms (incumbents) and entrant firms face when considering investments in innovation. The first of these arguments states that when mechanisms such as the patent system, brand identification, spatial location, or capacity expansion, effectively protect the economic benefits of an innovation, established firms alone stand to reap the benefits of monopoly power once the innovation is introduced. As such, established firms will have greater incentives to invest in innovation than entrants (Gilbert and Newberry, 1982). A second argument conditions the first by pointing out that even when mechanisms exist to protect the economic rents generated from innovation, uncertainty in the invention process, and in subsequent demand, may diminish established firms’ incentives to innovate (Reinganum, 1983). For example, established firms have lower incentives to invest in innovation than entrants when there is uncertainty about whether an innovation will cannibalize a portion of the established firms’ rents (e.g., Reinganum, 1983). This work assumes that established and entrant firms have relatively homogeneous, research capabilities. Yet, few empirical studies test these competing arguments. Indeed, the only empirical article we were able to identify provides a case study that describes how Xerox’s failure to invest in research and development to counteract Canon’s challenge in the early 1980s was due, in part, to Xerox’s unwillingness to cannibalize revenues from its plain paper copier business (Bresnahan, 1985). Strat. Mgmt. J., 30: 711–735 (2009) DOI: 10.1002/smj

714

Michael J. Leiblein and Tammy L. Madsen

Stimulated by the 1961 work of Burns and Stalker that relates to environmental uncertainty and organizational form, a third area of inquiry describes how the complexity and bureaucracy that emerges within firms as they increase in size inhibits subsequent innovation activity. This work examines how the structural and cultural characteristics associated with large firm size influence a firm’s administrative ability to manage innovation (e.g., Dewar and Dutton, 1986; Damanpour, 1996; Haveman, 1992, 1993; Tushman and O’Reilly, 1997). Existing work demonstrates that as firms grow, they develop formal administrative systems and structures that slow their abilities to adjust to shifting environmental conditions (e.g., Haveman, 1992, 1993). For instance, large firms may be more likely to utilize financial control systems over strategic control systems. The use of formal financial control systems may dampen a firm’s commitment to innovation (e.g., Hitt, Hoskisson, and Ireland, 1990). Other work emphasizes that, as firms grow, they also develop norms, values, and intraorganizational social networks for operating (Tushman and O’Reilly, 1997); these factors often become embedded in an organization and, over time, attenuate an organization’s ability to adopt changes in processes. Coupling these conditions with managers’ willingness to maintain the status quo further limits a firm’s ability to adjust to shifting environmental conditions. In combination, these bureaucratic and cultural sources of inertia make it exceedingly difficult and costly for large firms to recognize and estimate the value of new ways of operating (Cohen and Levinthal, 1990; Madsen and Walker, 2007). As a result, large firms may be more susceptible to competence traps (Levitt and March, 1988) and slower to change (e.g., Barnett and Hansen, 1996; Barnett, Greve, and Park, 1994) than small firms. This work suggests that to facilitate innovation, firms need to manage new opportunities outside the formal organizational structure.2 In sum, the preceding arguments emphasize how variance in market imperfections, market power,

2 For instance, work suggests that diversity and experimentation within a large organization facilitates innovation activity (e.g., Ahuja and Lampert, 2001; Burgelman, 1983). Work in organizational learning suggests that the structure of large organizations may limit the achievement of requisite diversity and thereby limit activities related to innovation, such as experimentation (Levinthal and March, 1993).

Copyright  2008 John Wiley & Sons, Ltd.

and broad administrative characteristics may contribute to differences in firms’ innovation activity. However, these arguments downplay how the relationship between firm size and other idiosyncratic firm-level differences, such as heterogeneity in firms’ experience stocks, might affect firms’ innovation activity. The lack of attention paid to the latter firm-level differentials is troublesome for at least two reasons. First, typical measures of firm characteristics only explain a small percentage of the variance in overall innovation rates captured by firm-fixed effects.3 While approximately one-half of the overall variance in innovation activity explained by industry-fixed effects is captured by existing measures such as an industry’s demand and an industry’s appropriability regime, ‘the most widely used measures of firm characteristics . . . jointly explain less than 10 percent of the variance explained by firm effects’ (Cohen and Levin, 1989: 1097). Second, these arguments highlight two effects brought about by large firm size—one associated with firm incentives and the other with organizational bureaucracy. The net effect observed in the empirical literature depends upon which of these two effects dominates. Additional insight might be gained by introducing a theoretical mechanism and empirical variables that distinguish these arguments.

EXPERIENCE AND INNOVATION ACTIVITY We draw from the resource-based view (RBV), experiential learning, and theory of the firm literatures to identify sources of firm-level heterogeneity that affect innovation activity. The RBV suggests that heterogeneity among close competitors may exist because some firms are lucky and/or are provided unique opportunities to purchase or develop valuable resources at a cost below their economic value (e.g., Rumelt, 1984; Barney, 1986). A primary insight offered by the RBV is that the resources that contribute to persistent performance differentials are much broader in nature and more difficult to accumulate than the tangible factors of production typically emphasized in neoclassical 3 Consistent with these ideas, a related work, focusing on research productivity rather than innovation and size, finds that a large portion of the variance in competitors’ research productivity can be attributed to firm-specific effects (Henderson and Cockburn, 1994).

Strat. Mgmt. J., 30: 711–735 (2009) DOI: 10.1002/smj

Incentive Structures, Capability Influences, and Innovation economic theories. Such advantages often evolve through path-dependent learning. Indeed, Winter (2003) defines experience as a crucial input to a firm’s capability.4 The experiential learning literature investigates how different forms of experience affect firm behavior, performance, and survival (for a review see Argote and Kane, 2003). Recent work on the effects of experience provides insights into the manner in which heterogeneous innovation capabilities may be developed. For one, empirical work indicates that different types or sources of aggregate experience such as, organizational experience (direct or indirect), congenital experience and population experience (e.g., Baum and Ingram, 1998), competitive experience (e.g., Barnett and Hansen, 1996), or collaborative experience (e.g., Simonin and Helleloid, 1993) affect a firm’s performance or survival chances. This work often suggests that a positive relationship between experience and performance or experience and survival chances means that learning has occurred. The explanation is that these aggregate forms of experience affect organizational operations outside the cost reducing efficiency improvements within a single technological domain emphasized in the traditional learning curve literature (Ingram and Baum, 1997). While few scholars would dispute that experience is a primary source of learning, past activity may or may not generate learning (Pisano, Bohmer, and Edmondson, 2001). As a result, the linkage between experience and subsequent organizational benefits is uncertain. For instance, established firms often experience difficulty adopting competence-destroying innovations, presumably because such innovations demand very different capabilities than those required by their existing products or processes (e.g., Tushman and Anderson, 1986; Tripsas, 1997). Clark (1987) and Dosi (1988) suggest that the impact of a given innovation on an individual firm (and on competition) likely depends on how the innovation affects the value and applicability of the firm’s established capabilities, how the innovation creates value and 4 Following Winter (2003) we define capabilities as ‘a high-level routine (or collection of routines) that, together with its implementing input flows, confers upon an organization’s management a set of decision options for producing significant outputs of a particular type’ (Winter, 2003: 991). This definition casts learning, experience, resources, and routines as inputs to capabilities. Except for routines, which can also be capabilities, these inputs are not themselves capabilities.

Copyright  2008 John Wiley & Sons, Ltd.

715

risk for the firm, and how the innovation must be managed within the firm. These findings and observations suggest that different types of experience may differentially affect a firm’s ability to innovate. Due to firm-level variation in strategy, resource stock, and operating scale, firms likely vary in the amount and type of experience that they possess. For instance, if firms following a technology leadership strategy are more likely to engage in projects involving specific assets, then transaction cost theory would suggest that these firms are most likely to organize their transactions through hierarchy (e.g., Williamson, 1975; Klein, Crawford, and Alchian, 1978). Similarly, if firm strategy affects the complexity of projects undertaken by a firm, then the problem-solving perspective of Nickerson and Zenger (2004) would suggest that these firms are likely to organize the majority of their exchanges internally. In practice, the operating scale of a firm may also affect the nature of its experience. The CEO of one semiconductor firm impressed upon us the costs of internalizing production by noting an oft stated industry rule of thumb that requires $1.50 of annual revenue for every dollar devoted to capital expenditures. As the cost to build and tool a semiconductor fabrication facility exceeds $1 billion U.S., the application of such a heuristic implies a need for $4 to $5 billion in annual revenue to justify one scale production plant. Aside from the ×86 processor, dynamic access random memory (DRAM), and flash memory product markets, this threshold dwarfs the revenues available in most existing semiconductor product markets. Building on this work and that of King and Tucci (2002), we suggest that heterogeneity in firms’ experience stocks may contribute to variance in their experientially created capabilities and, in turn, their innovation activity. Two forms of experience generally found to benefit organizations are their own operating experience and other organizations’ operating experience (e.g., Ingram and Baum, 1997; Baum and Ingram, 1998). An organization’s own operating experience helps it develop a deep, introspective understanding of the conditions associated with the use of technology and equipment; create enhanced—and often more robust—routines, and develop more efficacious incentive and coordination systems within a given investment trajectory. This deep understanding facilitates technological innovation in at Strat. Mgmt. J., 30: 711–735 (2009) DOI: 10.1002/smj

716

Michael J. Leiblein and Tammy L. Madsen

least three ways. First, experienced employees are likely to develop generalizable troubleshooting and problem-solving skills that enhance their ability to identify, implement, and integrate new tools and methodologies. Second, experienced employees are more likely able to identify promising research proposals and to direct research funding within a trajectory. Third, the existence of own firm operating experience suggests that a firm controls complementary assets that may facilitate innovation outright. Own operating experience also helps firms develop system-wide knowledge. Experienced employees with many cross-functional and crossfacility contacts are able to leverage their contacts to broaden their understanding of a firm’s production system as a whole. Access to system-wide information not only exposes experienced individuals to a broad set of factors when considering new innovations, but also allows them to better track and evaluate the progress of work done by others. Similarly, developing a detailed understanding of the conditions underlying the use of a particular technology implies an understanding of technical specifications as well as the production equipment’s tolerances, operating characteristics, and operating procedures. This knowledge might directly enhance a firm’s competence at transferring new knowledge between engineering and manufacturing functions and its speed at transferring research into practice. In short, own operating experience has been linked to the development of a broad range of valuable routines that support localized learning and innovation. Our first baseline hypothesis is: Hypothesis 1a: Operating experience developed within a focal firm is positively related to the firm’s amount of innovation activity. While own operating experience provides an important means through which firms may develop valuable and proprietary skills, the associated underlying routines are subject to inertial forces. In particular, own firm operating experience may hinder innovations that require large changes in a specific activity or in the ordering of a set of activities. Process innovations frequently require replacement of a particular piece of equipment or task in a manufacturing or production system. Copyright  2008 John Wiley & Sons, Ltd.

Similarly, innovations that require a different production facility layout imply a need to alter facilities and production systems. Such changes often generate an initial decline in system performance as a new activity, piece of equipment, or process sequence is worked out. The cost of negotiating and implementing such changes may hinder the adoption and acceptance of these types of innovations, particularly when firms have built experience with, and in turn, are committed to, an established system.5 One potential means to overcome inertial biases is to develop mechanisms that provide access to other firms’ operating experience. The ability to access external experience provides an opportunity to limit one’s own investment and to draw on a broad, and potentially larger, set of ideas, expertise, and experience than one’s own. By accessing external expertise held by firms with unique histories, organizations may uncover global optima hidden by local peaks (Rivkin, 2001), access complementary knowledge, and develop different knowledge combinations. While the relationship between external experience and innovation is by no means certain, technology transfer and innovation are commonly stated objectives associated with the development of interorganizational arrangements (e.g., Hergert and Morris, 1988; Tsai, 2001). Two of the most frequently utilized mechanisms for accessing external resources are strategic alliances and outsourcing arrangements. Alliances provide a mechanism through which competing firms, with otherwise opposing interests, can enter into mutually beneficial exchanges of ideas, experiences, and/or expertise. The increased cooperation brought on by these relationships alters a firm’s ability to access, coordinate, and control the resources necessary to develop, manufacture, and market goods and services (Hamel, Doz, and Prahalad, 1989). Indeed, the technology management and alliance literatures consistently argue that alliances represent an important channel through 5 In the semiconductor industry, examples of both of these situations exist. For instance, process innovations involving e-beam lithography require extremely precise masks and, at least initially, yield lower throughput values. As a consequence, firms adhered to traditional (photolithography) process methods; this slowed the diffusion of innovations involving e-beam lithography. Similarly, innovations involving X-ray lithography require a different production facility layout than that used with e-beam processes. The need to alter facilities and production systems may have hindered acceptance of these innovations by experienced firms with significant commitments to the old design.

Strat. Mgmt. J., 30: 711–735 (2009) DOI: 10.1002/smj

Incentive Structures, Capability Influences, and Innovation which firms access complementary technological resources (e.g., Powell, Koput, and Smith-Doerr, 1996) and learning opportunities (e.g., Khanna, Gulati, and Nohria, 1998; Rothaermel and Deeds, 2004). This study considers horizontal codevelopment alliances where a focal firm and its partner(s) work together to solve manufacturing process and production problems, typically through substantial personnel interactions. Such codevelopment alliances provide access to other firms’ experiences and, in turn, may alter the focal firm’s incentive and ability to innovate.6 For instance, by spreading the investment required to develop experience across multiple parties, alliances minimize the cost and risk associated with the focal firm’s innovation activity.7 In addition, by exposing a focal firm to a larger and more diverse stock of experiences, alliances may assist the firm in identifying new resource combinations that may be used to solve technical problems. Such exposure also may enhance a focal firm’s ability to detect valuable, innovative solutions. As a result, other firms’ experience accessed through codevelopment alliances is likely to facilitate a focal firm’s innovation activity.

717

Hypothesis 1b: Operating experience developed within a focal firm’s codevelopment partners is positively related to the focal firm’s amount of innovation activity.

into a contractual exchange for production services with its partner(s). Such arrangements are typically referred to as ‘outsourcing’ and are designed to allow a focal firm to contract for the production of specific components or services and thus minimize investment in specific equipment and capabilities. In this study, sourcing partnerships represent arrangements where a focal firm provides cash and/or design license rights in exchange for technological development and/or production services from its partner(s). These pay for service agreements that approximate arm’s-length contracts in which the transacting parties specify the terms of trade and the repercussions for noncompliance. Even though outsourcing merely shifts costs from a focal firm to its suppliers, it may create economic value within a supply chain. By pooling demand across multiple firms, suppliers may achieve scale advantages that support a larger investment in specialized capabilities. Outsourcing agreements also may allow a focal firm to tap into suppliers’ specialized experiences and capabilities (e.g., Dyer, 1996) or to shorten product development cycles through concurrent development (e.g., Clark and Fujimoto, 1991). This process may enhance capability development by allowing the focal firm to avoid the decision-making complexities typically associated with integrated organizational forms (Teece, 1992) or to accelerate its access to complementary knowledge. We hypothesize:

The related literature on outsourcing arrangements directs attention to the benefits of vertical alliances. In the context of manufacturing, a vertical agreement is one where a focal firm enters

Hypothesis 1c: Operating experience developed within a focal firm’s outsourcing partners is positively related to the focal firm’s amount of innovation activity.

6 For example, IBM leverages its technology development alliances with firms such as Advanced Micro Devices (AMD) to support collaborate innovation. AMD works directly with IBM’s research division on R&D, electronic materials, and feasibility studies during the early stages of product development. A goal of the collaboration is for both firms to accelerate the exploration and identification of future technology challenges and, in turn, bring advanced technology to market faster and at a lower cost. AMD views this as a way to better align its process technology innovation activity with future product market needs. 7 For example, in an effort to pool research resources and costs, Infineon of Germany, ASM Lithography of the Netherlands, and three American chip makers—Intel, Advanced Micro Devices, and Motorola—collaborated to develop new technology for semiconductor manufacturing and extreme ultraviolet lithography (EUV). Such collaborative research is essential to sustaining cutting-edge advances in semiconductor manufacturing technology and, in turn, the firms’ strategic positions under conditions of increasing rivalry.

Copyright  2008 John Wiley & Sons, Ltd.

EXPERIENCE, FIRM SIZE, AND INNOVATION ACTIVITY The first set of hypotheses examines whether the fragile correlations between firm size and innovation activity observed in the existing literature stem from the failure of prior work to account for differences in the magnitude of experience and the mode of organization through which this experience is accessed. The extant empirical work, however, identifies a number of additional differences between small and large firms (e.g., Barnett, 1997; Blau and Schoenherr, 1971; Crozier, 1964; Haveman, 1993; Kelly and Amburgey, 1991; Madsen Strat. Mgmt. J., 30: 711–735 (2009) DOI: 10.1002/smj

718

Michael J. Leiblein and Tammy L. Madsen

and Walker, 2007; Zenger and Lazzarini, 2004). These differences also may create systematic differences in the firms’ abilities to benefit from own and other firms’ operating experience. Building on this work, we offer four explanations for why small firms may be privileged in their abilities to utilize own firm operating experience relative to large firms. Small firms may more effectively leverage their own experience than large firms due to the greater uniformity within small firms’ information sets. As the amount and breadth of information within a firm increases, heterogeneous pieces of data are often aggregated into undifferentiated categories. This aggregation obscures qualitative differences of importance to some users within the firm (e.g., Hayek, 1945; Putterman, 1995). The concealment of knowledge, regardless of whether or not it is intentional, increases the costs of negotiating and enforcing rent sharing (Coase, 1937) and of mitigating influence (Milgrom and Roberts, 1990). In addition, large organization size is typically attended by greater structural complexity, such as a large number of hierarchical levels (Blau, 1970; Blau and Schoenherr, 1971). Since coordination costs increase with the number of interfaces that a particular piece of data has to cross, too many levels of hierarchy tend to increase coordination costs associated with information flow (e.g., Merton, 1957; Crozier, 1964). These problems are likely particularly severe for information that requires decision makers to rationalize subjective assessments that are confidently held, but by reason of bounded rationality, difficult to articulate (Polanyi, 1962). To the extent that small firms are able to achieve a finer partition in their measurement of these resources than large firms, small firms will be privileged in their ability to coordinate insights developed through own firm experience in a creative manner that generates valuable innovations. Compared to large firms, small firms also may more effectively leverage own firm experience due to advantages in attracting and hiring talented employees. It has long been accepted that small firms have distinct advantages in the hiring process. For instance, Stigler (1962: 102) argues that top-level managers are more likely to participate in employee recruitment and selection in small firms and, given their understanding of the business, are better able to assess the subjective quality of potential workers. The implication is that these managers are likely to hire exceptionally skilled Copyright  2008 John Wiley & Sons, Ltd.

individuals and are less subject to adverse selection problems than human resource specialists. A related argument suggests that small firms are better able to lure top talent than large firms due to their comparative advantage in utilizing performance contingent contracts. These aggressive reward systems typically offer a ‘higher expected return relative to contracts that pay a fixed amount reflecting some average level of performance’ (Zenger and Lazzarini, 2004: 331). As such, highly talented individuals may selectively apply for work under performance contingent contracts due to the greater expected overall payment. Zenger (1994) demonstrates that small firms are more likely to offer performance contingent contracts than large firms, perhaps due to small firms’ better ability to monitor the efforts of, and allocate contributions to, individual employees. Moreover, a recent study of R&D engineers in Silicon Valley and Route 128 (Zenger and Lazzarini, 2004) found that firms employing more aggressive reward (incentive) systems appear more successful in attracting high-quality, talented engineers. To the extent that high-quality talent is better able to generate unique insights from a given level of experience, small firms may better utilize their own experiences than large firms. Small firms also may be able to more effectively leverage own firm experience due to their advantages in motivating worker effort. As noted in the preceding paragraph, small firms are more likely to utilize high-paying performance contingent contracts than large firms (Zenger, 1994; Zenger and Lazzarini, 2004). The use of performancecontingent contracts is thought to motivate superior worker effort relative to contracts that fail to directly reward performance (Holmstrom, 1979). Research suggests that the use of aggressive rewards directed at individual contributions elicits superior effort from workers as compared to more formal, merit-based incentive systems (Zenger and Lazzarini, 2004). As large firms expand, these formal systems increase organizational complexity and may dampen large firms’ rates of adaptation (Haveman, 1993) and, in turn, their innovation activity.8 8 Of course, performance-contingent contracts impose other costs such as the need to attribute contribution to team production, or psychological costs associated with equity, distributional justice, and influence costs. However, in small firms, both peers and managers have firsthand information about the relative

Strat. Mgmt. J., 30: 711–735 (2009) DOI: 10.1002/smj

Incentive Structures, Capability Influences, and Innovation Related to employee incentive systems and resource measurement, the ability to delineate or define property rights represents a third important distinction between small and large firms. The larger the number and the greater the heterogeneity of competing interest groups within a firm, the more difficult it is to allocate precise property rights within that firm (e.g., Grossman and Hart, 1986; Hart and Moore, 1990). Since informational diseconomies increase with the number of data categories that exist within a firm and with the number of organizational boundaries a particular datum has to cross, moral hazard problems tend to amplify in large firms. As a result, as a firm grows, it becomes increasingly difficult to appropriately link innovators’ contributions to their standing within the firm. Indeed, small firms often are characterized as more effective in protecting an innovator’s property rights than large competitors (National Academy of Engineering, 1995). Last, work also demonstrates that large firm size buffers the effects of competitive pressures but may not necessarily strengthen a firm’s innovation or technical competence (e.g., Barnett, 1997). One explanation is that compared to small firms, large firms, on average, have invested more in building and maintaining ties with legitimizing institutions (Barnett, 1997; Meyer and Rowan, 1977). However, efforts to build legitimacy shift a large firm’s focus away from building technical competence (Selznik, 1949). As a result, such efforts potentially reduce the amount and diversity of internal activities related to innovation, such as experimentation (Levinthal and March, 1993). Barnett calls this phenomenon ‘compensatory fitness’ (Barnett 1997: 138 italics in original) and asserts that, as firms increase in size, investments in building legitimacy substitute for investments in building technical competence (see Meyer and Rowan, 1977). Taken together, the informational diseconomies, weaker employee selection and incentive systems, and compensatory fitness that emerge as firms increase in size suggest that, holding a firm’s experience constant, the relationship between own operating experience and innovation activity will be attenuated as firms increase in size.

performance of all employees. As a result, reward schemes that base pay on performance, and thereby avoid the need to differentially measure performance, are likely more effective the smaller the firm (Zenger, 1994). Copyright  2008 John Wiley & Sons, Ltd.

719

Hypothesis 2a: The positive influence of own operating experience on a focal firm’s amount of innovation activity is negatively associated with the size of the focal firm. While experiences accessed through codevelopment and outsourcing agreements represent a viable method for accessing distant knowledge, learning and capability acquisition is not equally available to all firms. The following underscores how two differences in organizational attributes associated with firm size may give rise to variance in firms’ abilities to leverage the experience of their partners. These differences suggest an important modification to the association between access to external operating experience and innovation proposed by Hypotheses 1b and 1c.9 The first difference is due to the possibility that small and large firms vary in their abilities to bear the information costs required to evaluate potential codevelopment or sourcing partners. Due to their limited scale, small firms often are unable to justify investment in formal corporate development functions. As a result, functional managers in small firms are more likely to share the activities associated with evaluating potential alliance and sourcing partners than managers with corresponding positions in large firms. In his analysis of international joint ventures, Buckley notes: ‘The horizons of small firms are limited by managerial constraints . . . Therefore, when an opportunity presents itself, it is often seized without proper evaluation . . . Consequently, small firms with inexperienced managers can behave in a na¨ive fashion’ (Buckley, 1997: 72). In sum, the problematic consequences of adverse selection likely are particularly acute for small firms. A second and related difference that may affect small firms’ abilities to leverage operating experience accessed through external linkages is small firms’ limited experience and/or ability with managing the alliance process. Several authors suggest that firms differ in their capability to manage alliances. For instance, Anand and Khanna 9 Anand and Khanna point out the importance of this interactive effect but are unable to address the issue with their data. They state ‘[I]f firm effects are larger for older firms, then we cannot disentangle the effects of past experience from the effects of ability in explaining the unobserved heterogeneity. If the reverse were the case. . . then differences in underlying ability must be large enough to offset the advantage of experience’ (Anand and Khanna, 2000: 300).

Strat. Mgmt. J., 30: 711–735 (2009) DOI: 10.1002/smj

720

Michael J. Leiblein and Tammy L. Madsen

(2000) suggest that repeated use of alliances may result in learning capabilities. Delios and Henisz (2000) suggest that over time, firms develop hazard mitigating capabilities that allow them to utilize collaborative agreements in the presence of exchange hazards. Kale, Dyer, and Singh (2002) show that the ability to codify knowledge is dependent on the formation of a formal alliance function. However, small firms typically lack resources that would allow their general managers to develop a refined means of filtering through these complex arrangements and decisions. For instance, small firms often are not able to hire specialist executives to manage their corporate development activities or to engage in multiple partnerships to build managerial expertise in partnership management. Given the complexity of managing opportunism and other challenges in alliances, small firms’ lack of administrative skills and supporting resources suggests that they may benefit less from an alliance at the margin. These firm-level differences in the ability to evaluate partners and manage ongoing external relationships suggest that small firms are comparatively disadvantaged in leveraging experience from their codevelopment and outsourcing efforts, relative to large firms. We hypothesize: Hypothesis 2b: The positive influence that operating experience developed within a focal firm’s codevelopment partners has on the focal firm’s amount of innovation activity is positively associated with the size of the focal firm.

Hypothesis 2c: The positive influence that operating experience developed within a focal firm’s outsourcing partners has on the focal firm’s amount of innovation activity is positively associated with the size of the focal firm.

INDUSTRY CONTEXT AND DATA The preceding discussion emphasizes how operating experience as well as the organization of operating experience may differentially affect small and large firms’ innovation activity. Since the success of specific types of innovation models differ substantially across types of innovations and industries (e.g., Pavitt, 1984), our empirical design is based Copyright  2008 John Wiley & Sons, Ltd.

on observable behavior involving related innovations in a single industry. We define ‘innovation’ or more precisely, ‘process innovation,’ as the application of new materials, equipment, procedures, or knowledge that alters the cost and/or value of a product or service. This definition is consistent with several sources. For instance, Schumpeter (1934) describes innovation as the introduction of a new method of production, which need by no means be founded upon a new scientific discovery. Thompson (1965: 2) refers to innovation as ‘the generation, acceptance, and implementation of new ideas, processes, products or services.’ More recently, building on work by Ettlie and Reza (1992) and Utterback and Abernathy (1975), Damanpour defines innovation as ‘the adoption of an idea or behavior new to the adopting organization’ (Damanpour, 1996: 694). The empirical context for this study is the global semiconductor industry. The data include 463 private and public firms active in the global semiconductor industry during the period 1990 through 1999. The primary data stem from the Integrated Circuit Engineering Corporation’s (ICE) annual reports entitled ‘Profiles: A Worldwide Survey of IC Manufacturers and Suppliers’ (Integrated Circuit Engineering Corporation (ICE), 1991–2000). A firm’s ICE report provides summary financial data, manufacturing plant locations, manufacturing plant capacity, and the process technology employed by the firm in its manufacturing facilities. The reports also detail a firm’s annual codevelopment and sourcing agreements. Codevelopment partners are defined in this study as those firms involved in horizontal arrangements in which firms collaboratively work together to solve technical problems associated with the development of a new process technology. Outsourcing partners are those firms involved in vertical contracts that provide access to process technology via production of a device or family of devices, typically in exchange for monetary compensation. An example of a codevelopment agreement taken from our sample is an agreement between TEMIC Semiconductors and the Fraunhofer-Institut (ISIT). In this close relationship, the two companies agreed to jointly develop 0.5 micron and smaller complementary metal–oxide–semiconductor (CMOS) process technologies by collaborating on work at ISIT’s fabrication facility in Itzehoe, Germany. Outsourcing agreements, as defined in our sample, Strat. Mgmt. J., 30: 711–735 (2009) DOI: 10.1002/smj

Incentive Structures, Capability Influences, and Innovation typically do not involve this level of teamwork. For instance, in February 1997, SST (Silicon Storage Technology) signed an agreement with foundry partner TSMC (Taiwan Semiconductor Manufacturing Company) for the production of its 2M (flash memory) products using 0.35 and 0.5 micron processes. In this agreement, SST designed a flash memory device using its patented ‘SuperFlash’ technology and contracted for production of a specific family of devices with TSMC, an independent ‘foundry’ or manufacturing partner. We focus on the semiconductor industry for several reasons. For one, both small, entrepreneurial firms and large, established firms play a critical role in the industry (Eisenhardt and Schoonhoven, 1996). In addition, industry data allow us to capture a clear and reliable measure of process innovation activity based on the development and use of new process technologies (Martin and Solomon, 2003). Throughout the industry’s history, and particularly since the early 1980s, successive innovations in process technology have improved product performance by reducing the physical dimensions of semiconductor devices (i.e., scaling). The near universal benefits of process innovation in the industry mitigate concerns regarding unobserved variance in the motivation, intentionality, and strategy of the sampled firms. Numerous studies demonstrate that the introduction of process innovations in this industry improve key product performance attributes, such as speed and heat dissipation (e.g., Gruber, 1994), facilitate entry into premium product markets (e.g., Nanda and Bartlett, 1994), and generate operational efficiencies that yield strong and persistent price reductions (e.g., Gruber, 1994). As a result, the industry provides an environment where little variance is likely to exist across firms in qualitative matters such as the intentionality underlying a specific investment trajectory. In sum, the technical challenges associated with the development of new processes in semiconductor manufacturing (e.g., the incorporation of a new energy source, lithography lens, mask, or resist material into the manufacturing process) has created a dynamic wherein clear differences exist between more and less innovative firms.10 10 For instance, as deep ultraviolet lithography replaced traditional optical lithography equipment techniques, many firms struggled to develop shallow trench isolation techniques that were required to improve the planarity or surface characteristics

Copyright  2008 John Wiley & Sons, Ltd.

721

MEASURES Dependent variable Since we are interested in the number of times a firm has engaged in process innovation activity, we use a cumulative counter design for the dependent variable. A similar counter design has been used in the extant empirical literature on new product innovation (e.g., Ahuja and Katila, 2004; Geroski, Machin, and Van Reenen, 1993; Hitt et al., 1996; also see Hagedoorn and Cloodt, 2003). In applying this measure to semiconductor process technology innovations, we follow work by Martin and Salomon (2003) that measures innovation activity as the production of products incorporating a new lithography process technology. During the time period observed, industry process innovation activity focused on the development and integration of new lithography technology (e.g., i-line stepper, DUV 248nm stepper, DUV 193nm stepper), energy sources (e.g., krypton fluoride laser, argon fluoride laser), lens materials (e.g., refractive fused silica, calcium fluoride), masks (e.g., introduction of quartz material, various resolution enhancement technologies), and resist materials (e.g., various means of chemically amplifying resist properties) into the overall manufacturing process. These activities resulted in nine distinct process technology innovations. These nine innovations correspond to devices produced with 1.0 micron, 0.8 micron, 0.65 micron, 0.5 micron, 0.35 micron, 0.25 micron, 0.18 micron, 0.15 micron, and 0.12 micron feature sizes. We count the number of these new generations that a single firm introduces over time and take the natural log of a firm’s cumulative count to reduce skewness in the distribution. The measure is specified as: t −1 Innovation activityit = [ln (
Tit )] t0 where Tit is the number of new process technologies adopted by firm i in year t, and t0 is the first year of firm i’s existence. The counter is initiated at 1 for each firm.

of the semiconductor device (e.g., Davari, Dennard, and Shahidi, 1995: 602). Strat. Mgmt. J., 30: 711–735 (2009) DOI: 10.1002/smj

722

Michael J. Leiblein and Tammy L. Madsen

Independent variables Firm size is defined as the natural log of a firm’s sales. We use three measures to capture a firm’s capability development efforts or resource stocks, own production experience, experience of codevelopment partners, and experience of sourcing partners. First, firm i’s own production experience is defined as: Own production experiencei,t−1 t−1


= ln (

OEit /discount)

t0

where t0 is the first year that firm i operated a fabrication facility, t − 1 is the year before the current year, OEit is the total amount of fabrication (production) experience accumulated by firm i’s fabrication facilities in year t, and discount is equal to the age of the experience squared. The experience measure begins with the first year that a firm operated a production facility (e.g., for firms founded prior to 1990, our data imputes the production history of each firm’s fabrication facilities). The discount is used to account for the potential antiquation of a firm’s own production experience or for the decay of knowledge gained from the firm’s past production experience (e.g., Argote, 1999; Baum and Ingram, 1998).11 The rate of change in process technology in the industry (i.e., ‘Moore’s Law’ indicates that the data density of integrated circuits doubles approximately every 18 months) suggests that the value of operating knowledge decays rapidly. Given this, we apply a discount equal to the age of the experience squared; this discount assumes that the value of experience depreciates more rapidly than linear initially, and subsequently accelerates. The second two experience measures capture learning from external sources and are defined in a similar fashion. The experience of firm i’s 11 We tested four discount rates: 1) a discount equal to 1 which assumes no depreciation of an organization’s experience; 2) a discount equal to the age of the experience, which assumes that experience depreciates at a linear rate, 3) a discount equal to the square root of the age of the experience, which assumes that the experience initially depreciates at a lower rate than a linear form and slows at an increasing rate over time; and 4) a discount equal to the age of experience squared, which assumes that the value of experience depreciates more rapidly then linear initially, and subsequently accelerates. The results of our hypotheses tests are consistent across discount rates.

Copyright  2008 John Wiley & Sons, Ltd.

codevelopment partners is defined as: Experience of codevelopment partnersi,t−1 t−1
= ln ( CEit /discount) t0

where t0 is the first year that firm i’s codevelopment partners operated fabrication facilities and t − 1 is the year before the current year, CEit is the total amount of fabrication experience accumulated by all of firm i’s codevelopment partners based on their fabrication (production) in year t, and discount is the age of the experience squared. We use a natural log transformation of the sum to reduce skewness in the distribution. The experience of firm i’s sourcing partners is defined as: Experience of sourcing partnersi,t−1 t−1
= ln ( SEit /discount) t0

where t0 is the first year that firm i’s sourcing partners operated fabrication facilities and t − 1 is the year before the current year, SEit is the total amount of fabrication experience accumulated by all of firm i’s sourcing partners based on their fabrication in year t, and discount is equal to the age of the experience squared. We use a natural log transformation of the sum to reduce skewness in the distribution. Similar to the measure for own production experience, the above two experience measures begin with the first year that a partner firm operated a production facility (e.g., for partner firms founded prior to 1990, our data imputes the production history of each firm’s fabrication facilities). Before defining the control variables, we provide some descriptive details on the experience variables for the sample. Figure 1 shows the average magnitude of a firm’s own production experience as well as experience accessed from codevelopment and sourcing partners for the sample from 1990 to 1999. The annual averages for own experience and experience from sourcing partners exceed the annual average for experience from codevelopment partners. These differences, along with slight variations in the trend lines, suggest that disaggregating external sources of experience by type may help uncover additional unobserved heterogeneity among firms. The notable decline in the average Strat. Mgmt. J., 30: 711–735 (2009) DOI: 10.1002/smj

Incentive Structures, Capability Influences, and Innovation

723

Figure 1. Average Ln(Experience/Discount), Semiconductor Industry, 1990–1999 Figure 3.

Figure 2.

Average Experience Mix, Large Firms, Semiconductor Industry, 1990–1999

fabrication experience held by a firm from 1993 to 1999 is consistent with the formation of specialized foundries in the early 1990s and in turn, the

Figure 4.

Average Experience Mix, Small Firms, Semiconductor Industry, 1990–1999

industry’s deintegration. Figures 2 and 3 chronicle changes in the average firm experience mix by year for small and large firms. In these charts, firms were classified as large when their ln(size) was greater than the sample mean and small when their ln(size) was less than or equal to the sample mean. A comparison of the charts indicates that, on average, large firms obtain a greater proportion of their total experience from internal efforts and projects with codevelopment partners compared to the experience mix of small firms. The reverse is true for experience from sourcing partners. Aside from cyclical fluctuations, the experience mix for small and large firms is relatively stable across time.

Average Cumulative Sum of Process Technology Innovations by Firm Size, Semiconductor Industry, 1990–1999

Copyright  2008 John Wiley & Sons, Ltd.

Strat. Mgmt. J., 30: 711–735 (2009) DOI: 10.1002/smj

724

Michael J. Leiblein and Tammy L. Madsen

Control variables We include four firm-specific control variables that might simultaneously influence the underlying motivations of alliance formation and a firm’s process innovation activity: (1) firm age, (2) country location of a firm’s headquarters, (3) a firm’s initial technology stock, and (4) whether a firm experienced a merger or acquisition (M&A) in its history. Firm age is defined as the natural log of the number of years since a firm was founded. To capture geographic region effects, we created four dummy variables to specify whether a firm is headquartered in the United States, Japan, other Asian Countries, or in the rest of the world (e.g., Europe, the states of the former Soviet Republic, etc.). The latter category is excluded from our regression analyses. A firm’s initial technology stock is defined as the process technology in use by the focal firm at the onset of data collection. Last, we also reviewed every firm’s history to determine whether the firm experienced an M&A event involving the transfer of semiconductor fabrication facilities to the focal firm. We used a dummy variable to capture these effects: if a firm engaged in an M&A event at time t, the dummy was set to 1 at time t and thereafter; otherwise, the variable was set to 0. Accounting for these firm-specific factors allows us to reduce the likelihood that any unobserved heterogeneity is correlated with our process innovation activity measure. We further control for variance attributable to specific technical subfields through a measure of competitive intensity. Within industry, variation in firm asset stocks contributes to variance in the competitive strength of firms (Barnett and Amburgey, 1990; Barnett, 1997). Including a variable that captures heterogeneity in competitive strength among firms provides an indicator of whether an increase in the size of rival firms has a competitive effect on the focal firm’s subsequent process innovation activity. Following Barnett (1997), we control for potential heterogeneity in the competitive strength of firms by including a measure for competitive intensity, defined as sum of the ln(size) of all the firms at time t minus the ln(size) of the focal firm. All measures, except the initial technology stock measure and the dummy variables for firm location, are lagged one year to avoid simultaneity. We also include a dummy variable for each calendar year of observation (omitting one year from Copyright  2008 John Wiley & Sons, Ltd.

the regression analysis). Finally, our initial model specifications included measures for product market controls; the effects for these parameters were not statistically significant and, therefore, are not reported.

METHOD We test our hypotheses on cross-sectional time series data, or panel data. Given the panel data structure, the error terms may not be independent across time or within a firm (panel) (Greene, 1993). The latter might occur, for instance, if heterogeneity in process innovation activity among firms occurs over time due to systematic differences in the firms’ abilities to leverage own experience or experience from partners. A firm fixed- or random-effects model may address this issue (Greene, 1993). We use a random-effects specification for two reasons. First, the average number of time period observations per panel, T, in our sample is 8 whereas the number of panels, m, is 463. Generating consistent estimates on panel data using a fixed-effects generalized least squares (GLS) specification requires panel data where T
m. When m
T and a fixed-effects GLS specification is applied, the estimates may be inconsistent (Hsiao, 1986). In contrast, random effects estimators are asymptotic in the number of panels, m. Second, three of our control variables—a firm’s initial technology resource stock, the indicator variables for a firm’s location, and the M&A indicator variable—are time invariant. As a result, the effects of these variables will be perfectly colinear with the fixed effect for a given firm; such colinearity will harm the fixed effects estimates. In contrast, a random-effects specification can accommodate time invariant parameters. Given the above, a random-effects specification is preferred (Hsiao, 1986; Kennedy, 1998; for a similar approach see Barnett and Salomon, 2006; Yin and Zajac, 2004). The model is specified as: I nnovation activityit+1 = δ + aEit + βSit + γ .Eit x Sit / + jXit + ui + εit

(1)

where δ is the intercept; a is the coefficient representing the effects associated with E, the vector capturing a firm’s own production experience, the Strat. Mgmt. J., 30: 711–735 (2009) DOI: 10.1002/smj

Incentive Structures, Capability Influences, and Innovation experience of its codevelopment partners and the experience of its sourcing partners; β is the coefficient associated with the covariate for a firm’s size, S; g is the vector of coefficients associated with the interactions between the experience measures and firm size; j is the vector of coefficients associated with the control variables and X is the corresponding vector of covariates; ui is a random disturbance characterizing the ith observation and is constant through time; and εit is the error term specific to a particular observation. In sum, we use a fixed-year effect (via indicator variables) and a random-firm effect to control for any remaining unobserved heterogeneity. The efficient estimator used is GLS. We employ the Swamy-Arora method for unbalanced panels derived by Baltagi and Chang (1994). The method includes an initial step to cleanse the data of first order autocorrelation (Baltagi and Wu, 1999). While the emphasis on process innovation in this industry mitigates the likelihood that significant unobserved variance exists in the innovative strategies of firms in our sample, it is possible that some unobserved firm-specific heterogeneity exists. In general, the adverse effects of unobserved firmspecific heterogeneity may be treated through the use of instrumental variables or two-stage models (e.g., Heckman, 1979). The presence of panel data allows the use of additional techniques such as fixed- or random-effects regression (Wooldridge, 2002: 248). However, important trade-offs exist across these alternatives. For instance, while the use of instrumental variables allows unobserved firm characteristics to vary across time, the selection of candidate instruments is subjective, and the efficacy of the estimation is dependent on the validity and strength of the underlying instruments. Although panel data techniques can only address selection bias due to (individual- or firm-) fixed effects, they do not require any subjective assessment. Since no single approach dominates its alternatives, the final choice frequently depends on the specifics of the research design. Lacking a clear and specific means to distinguish between experience portfolios and a theory to guide selection of instrumental variables that identify whether and when firms choose one portfolio strategy over another, we address potential selection and unobserved heterogeneity issues using a random firm-effects specification, several firm-level control variables (e.g., firm age, country location, initial Copyright  2008 John Wiley & Sons, Ltd.

725

technology stock, M&A history), and year fixed effects. Table 1 presents the correlation matrix, means, and standard deviations for the variables. The matrix reports a correlation of 0.36 between the ln(cumulative sum of technology innovations) and ln(own production experience), a correlation of 0.45 between ln(own production experience) and ln(experience of codevelopment partners), a correlation of 0.48 between ln(own production experience) and firm size, and a correlation of (−0.44) between the dummy variable for a Japanese firm and the dummy variable for a U.S. firm. All remaining correlations are less than 0.41. High levels of colinearity among the independent variables might give rise to less precise parameter estimates (generally indicated by higher standard errors) for the explanatory variables without necessarily harming them for the purposes of hypothesis testing (Belsley, Kuh, and Welsch, 1980: 115; Greene, 1993). Given the correlations, we examined the variation inflation factors for all the variables. The variance inflation factor values among the variables are less than 1.64, with the average equal to 1.22. These results suggest a lack of multicolinearity. Robustness checks We conducted several tests to verify the robustness of the results. First, we tested our model specifications using two alternative designs for the dependent variable: 1) a normalized measure of a firm’s innovation activity defined as a firm’s innovation activity at time t divided by the industry average innovation activity at time t; and 2) a measure for how far (or near) a firm’s process technology is relative to the industry’s process technology frontier. The results with these dependent variables are consistent with the results reported here. Second, of the 2,599 firm-year observations in the entire sample, 655 observations existed where all three experience measures were reported as zero. One possible explanation for this condition is that these observations represent firms whose products are in development or who partner with less established firms that may be absent from the ICE database. This explanation seems plausible considering 631 of the 655 observations with this characteristic may be characterized as small firms in our sample. To verify the robustness of our results, we Strat. Mgmt. J., 30: 711–735 (2009) DOI: 10.1002/smj

1.0 −0.00 −0.04 .00 −0.08 −0.19 −0.14 −0.03 0.07 1.0 0.24 1.0 0.31 0.38 1.0 0.09 0.04 0.10 −0.14 −0.19 −0.21 0.14 0.39 0.32 0.07 −0.07 0.12 0.33 0.30 0.42 0.23 0.09 0.13 0.36 0.40 0.48 −0.33 −0.07 −0.23 0.17 0.07 0.16 0.93 2.60 4.74 1289.95 0.66 0.08 0.06 2.06 3.44 4.04 1.13 0.07 1 2 3 4 5 6 7 8 9 10 11 12

Ln(cumulative sum of technology innovations) Firm Ln(age) Ln(size) Industry competitive intensity US firm Japanese firm Other Asian firm Ln(experience of codevelopment partners/D) Ln(experience of sourcing partners/D) Ln(own production experience/D) Initial technology resource stock Merger/acquisition

0.28 1.06 1.56 325.92 0.47 0.28 0.24 4.37 5.14 4.46 0.16 0.25

10 9 8 7 6 5 4 3 2 1 Std dev Mean Variable

Table 1. Means, standard deviations, and correlations

1.0 −0.44 1.0 −0.37 −0.08 1.0 −0.17 0.19 0.10 1.0 0.16 −0.01 −0.10 0.23 1.0 −0.29 0.30 0.14 0.45 0.06 1.0 0.07 −0.10 −0.08 −0.27 −0.01 −0.29 1.0 0.03 0.02 −0.03 0.15 0.04 0.23 −0.04

Michael J. Leiblein and Tammy L. Madsen 11

726

Copyright  2008 John Wiley & Sons, Ltd.

reestimated our model specifications omitting the cases where all three experience measures for a firm were equal to zero. If this data characteristic adversely biased our results, the results would have likely lost significance when the cases were omitted. On the contrary, the results of our hypotheses tests remained the same. Third, we reviewed every firm’s history to identify instances where an M&A event may have affected the ownership of semiconductor fabrication facilities. This process yielded 65 events that involved the transfer of fabrication facilities. We conducted various sensitivity analyses to understand whether these events influenced our findings. First, we constructed an indicator variable that is set to one if a firm engaged in an M&A event that involved the transfer of fabrication facilities to the focal firm, and zero otherwise. We tested this variable two ways: 1) time-varying indicator: the variable was set to one for the year (t) in which the M&A event occurred, and set to zero otherwise; 2) indicator retained as one: if a firm engaged in an M&A event at time t, the variable was set to one at time t and thereafter; the variable was set to zero if a firm never engaged in an M&A event. The coefficient for the timevarying indicator variable was not significant in any of our models. However, the coefficient for the ‘indicator retained as one’ variable was positive and significant in all steps of our analysis. Importantly, the results with these variables included are consistent with our previous results. As a third test, we omitted the firms that gained fabrication facilities via M&A from our sample and reran the model specifications. Again, the results were consistent with our prior results. Given the above findings, and as noted in the control variable section of the article, the revised study reports results with the M&A ‘indicator retained as one’ variable.

RESULTS Table 2 lists the results of the regression analyses. Model 1 reports the results of the baseline model, which includes the control variables and firm size. The Model 2 adds the main effects for own production experience, experience of codevelopment partners, and experience of sourcing partners; Models 3 to 5 include the interactions of these experience stocks and firm size, respectively; Model 6 presents the full model. We first Strat. Mgmt. J., 30: 711–735 (2009) DOI: 10.1002/smj

Copyright  2008 John Wiley & Sons, Ltd.

0.05•••• (0.008) −0.21•••• (0.05) −0.046• (0.02) 0.07 (0.04) 0.03 (0.043) 0.12•••• (0.01) −0.002 (0.002) 0.17 836.83 2599

0.04•••• (0.007) −0.17••• (0.05) −0.039 (0.02) 0.05 (0.04) 0.01 (0.04) 0.10•••• (0.01) −0.002 (0.001) 0.232 879.04 2599

0.01•••• (0.001) 0.004•••• (0.001) 0.002•• (0.001)

0.003 (0.003)

3.08 (1.95)

(2)

0.04•••• (0.007) −0.16••• (0.05) −0.038 (0.021) 0.04 (0.04) 0.01 (0.04) 0.10•••• (0.01) −0.002 (0.002) 0.233 885.38 2599

−0.003•• (0.001)

0.01•••• (0.001) 0.004•••• (0.001) 0.002•• (0.001)

0.01• (0.003)

2.71 (1.96)

(3)

• p < 0.05, •• p < 0.01, ••• p < 0.001, •••• p < 0.0001. Standard errors are in parentheses. D is the discount used for the experience variables. Model specifications include dummy variables for year effects and control for first-order autoregressive disturbances.

Rsq Wald X2 N

Industry competitive intensity

Merger/acquisition

Other Asian firm

Japanese firm

US firm

Initial technology resource stock

Firm and industry characteristics Firm Ln(age)

Ln(experience of sourcing partners/D) × Ln(size)

Ln(experience of codevelopment Partners/D) × Ln(size)

Resource flows & firm size Ln(own production Experience/D) × Ln(size)

Ln(experience of sourcing partners/D)

Ln(experience of codevelopment partners/D)

Resource flows Ln(own production experience/D)

0.01• (0.003)

2.99 (1.93)

Intercept

Firm size Ln(Size)

(1)

0.04•••• (0.007) −0.16••• (0.05) −0.039 (0.022) 0.04 (0.04) 0.02 (0.04) 0.10•••• (0.01) −0.002 (0.002) 0.233 887.74 2599

−0.003•• (0.001)

0.01•••• (0.001) 0.004•••• (0.001) 0.002•• (0.001)

0.004 (0.003)

3.02 (1.96)

(4)

0.04•••• (0.007) −0.16••• (0.05) −0.038 (0.022) 0.05 (0.04) 0.01 (0.04) 0.10•••• (0.01) −0.002 (0.002) 0.232 881.41 2599

0.001 (0.001)

0.01•••• (0.001) 0.004•••• (0.001) 0.002•• (0.001)

0.003 (0.003)

3.08 (1.95)

(5)

Dependent variable = Ln(cumulative sum of technology innovations)

Random effects panel GLS regression estimating the effects of experience on innovation, Global Semiconductor Industry, 1990–1999

Variable

Table 2.

0.04•••• (0.007) −0.16••• (0.05) −0.038 (0.022) 0.04 (0.04) 0.02 (0.04) 0.10•••• (0.01) −0.002 (0.002) 0.234 893.95 2599

−.003•• (0.001) −.003•• (0.001) 0.001 (0.001)

0.01•••• (0.001) 0.004•••• (0.001) 0.002•• (0.001)

0.006 (0.003)

2.73 (1.96)

(6)

Incentive Structures, Capability Influences, and Innovation 727

Strat. Mgmt. J., 30: 711–735 (2009) DOI: 10.1002/smj

728

Michael J. Leiblein and Tammy L. Madsen

discuss the results for firm size beginning with the baseline model and then shift our attention to the results of the hypotheses tests. The results in Model 1 indicate that the coefficient for the main effect of firm size is positive and significant. The positive effect of firm size is consistent with the general notion that large scale facilitates innovation activity. Descriptive data provides additional insights into the amount of process innovation activity engaged in by small and large firms during the period of observation. Figure 4 shows the average ln(cumulative sum technology innovations) by firm size; for illustration purposes, we use the sample mean ln(size) to define large and small firms. On average, small firms’ process innovation activity increased from 1990 to 1997 but lagged that of large firms. However, Figure 4 also shows that the difference in the process innovation activity of large firms and small firms declines over time. By 1998, large and small firms, on average, engage in approximately the same amount of process innovation activity. These observations suggest that scale alone does not account for the variance in process innovation activity. The test results for Hypotheses 2a through 2c speak to this issue more explicitly. Hypotheses 1a, 1b, and 1c predicted that own, codevelopment partner-, and sourcing-partner experience are positively related to process innovation after controlling for firm size. Consistent with the hypotheses, Model 2 shows that the regression coefficients for each experience term are positive and significant. Overall, the findings indicate that firms benefit from their own experience and from the experience accessed through codevelopment partners and sourcing partners. Although the coefficient on firm size remains positive in this model, the explanatory power of the experience variables swamps the significant effect of firm size. The results for the interactions of experience and size suggested in Hypotheses 2a through 2c are reported in Models 3 through 6. Models 3 and 6 show that the coefficient for the interaction of own production experience and firm size is negative and significant, consistent with Hypothesis 2a. The benefits of own production experience diminish with increases in firm size. Models 4 and 6 show that the coefficient for the interaction of codevelopment partners’ experience and firm size is negative and significant. This result, counter to Hypothesis 2b, suggests that the benefits of codevelopment partners’ experience also diminish as a Copyright  2008 John Wiley & Sons, Ltd.

firm increases in size. Importantly, Model 6 shows that significant variance is explained by firm size; the coefficient on firm size is positive and significant. Taken together, the findings imply that the benefits of operating experience developed by a firm’s codevelopment partners on the firm’s innovation activity are positive but increasingly muted as the focal firm increases in size. Moving to Model 5, the coefficient on the interaction term for the experience of sourcing partners and firm size is positive, as predicted, but not significant; this finding is consistent in Model 6. Quantifying the impact of the experience effects and their interactions may help interpret the findings. We begin with the main effects for the three experience measures reported in Model 2. The findings suggest that the average firm will generate one more process innovation by increasing annual own production experience by approximately 1.71 million wafers (thin slices of semiconductor material) per year [= 0.33/0.01 × 1000 × 52] or 33K wafers per week.12 Achieving the same effect through collaboration requires that a firm’s partners’ hold a comparatively larger stock of experience. For instance, a firm’s innovation counter will increase by one when the total production experience of its codevelopment partners increases by 4.29 million wafers annually (∼82.5 K wafers per week) or when the total production experience of its sourcing partners increases by 8.58 million wafers annually (∼165 K wafers per week). These differences suggest that firms engaging in codevelopment or sourcing arrangements may benefit more by partnering with one large firm or with a large number of medium-small- sized firms (e.g., 4.29M and 8.58M are approximately equal to (2.63 to 5) × 1.71M). To illustrate the interaction effects, we consider the overall influence of an average amount of each experience type for firms of varying sizes. We begin with a firm with average (ln)own firm production experience equal to 4.04 (∼56.82 K wafers/week). The estimated process innovation activity is then given by [0.01(4.04) + (−0.003) (4.04)(firm size)]. In addition to demonstrating that the influence of own firm experience on process innovation activity diminishes with firm size, 12 Given that the average innovation activity for the sample is 0.93, an increase of one process innovation for the average firm equates to a 0.33 unit increase in the dependent variable.

Strat. Mgmt. J., 30: 711–735 (2009) DOI: 10.1002/smj

Incentive Structures, Capability Influences, and Innovation this equation suggests that operating experience positively influences process innovation activity for firms with less than ∼$28 million in annual sales. Using a similar approach, we can assess the size interaction term for the experience of codevelopment partners. The experience held by the average firm’s codevelopment partners decreases the focal firm’s process innovation activity according to [= 0.004(2.06) + (−0.003)(2.06)(firm size)]. This equation suggests that operating experience positively influences process innovation activity for firms with less than ∼$3.8 million in annual sales. In sum, the effects of firm size vary by the type of partner experience accessed. Large firm size, compared to small firm size, reduces the benefits associated with the experience of codevelopment partners but enhances the benefits associated with the experience of sourcing partners. A next step is comparing the relative strength of the effects using the standard deviation of each variable. Beginning with the results from Model 3, for a firm with an average level of own production experience (= 4.04) and average firm size (= 4.74), a one standard deviation increase in own production experience decreases process innovation activity by 1.9 percent [= 4.46[(0.01) + (−0.003)(4.74)]. For the same firm, a one standard deviation increase in ln(size) yields a 0.3 percent decrease in the firm’s innovation activity [= 1.56 [(0.01) + (−0.003)(4.04)]. Innovation activity is dampened slightly more by increases in own production experience than by increases in firm size (1.9% versus 0.3%). Moving to Model 4, for a similar firm, a one standard deviation increase in the experience of codevelopment partners decreases process innovation activity by 4.4 percent whereas a one standard deviation increase in ln(size) contributes a 0.34 percent decrease. Similar to the effects of own production experience, process innovation activity is dampened more by increases in the experience of codevelopment partners than by increases in firm size. In sum, the coefficients on the interaction terms suggest that both firm size and the different forms of experience contribute to variation in firm process innovation activity. Last, the effects for the control variables are fairly consistent across all models. The findings suggest that a firm’s process innovation activity tends to increase with age; the coefficient for firm (ln)age is positive and significant in all models. A firm’s initial technology resource stock has a Copyright  2008 John Wiley & Sons, Ltd.

729

negative imprint on the firm’s subsequent innovation activity; the coefficient for initial technology resource stock is negative and significant in all models. For the most part, the indicator variables indicating the location of a firm’s headquarters are not statistically significant; exceptions include the negative and significant coefficient associated with the variable for U.S. firms in Model 1. Innovation activity is also positively associated with firms that experienced an M&A event in their history; the coefficient for the M&A indicator variable is positive and significant in all models. The effects for the remaining variables are not significant.

DISCUSSION This study contributes to the management and strategy literatures by investigating how firm size and different forms of operating experience affect innovation activity. Three main findings emerge. First, the results indicate that operating experience benefits firm innovation activity. This suggests that empirical analyses of the association between innovation activity, firm size, and firm age that exclude operating experience may suffer from omitted variable bias. Second, the results indicate that different forms of experience are associated with different amounts of innovation activity. For instance, the coefficient for own firm experience (β = 0.01, from Model 2) is larger in magnitude than the coefficient for experience of codevelopment partners (β = 0.004) and the coefficient for the experience of sourcing partners (β = 0.002). Wald tests indicate that these differences are statistically significant (own experience versus experience of codevelopment partners, χ 2 = 9.09, p < 0.002; own experience versus experience of sourcing partners, χ 2 = 16.62, p < 0.0001). Since large and small firms tend to access operating experience through different organizational governance forms, the type of governance form used to access operating experience represents a second important omitted variable that may affect innovation activity. Third, the results show that the benefits of different forms of experience on innovation activity vary with firm size. In sum, our analysis suggests that firm size affects innovative activity in a more complicated fashion than demonstrated in much of the extant empirical work. The findings also have a variety of theoretical implications for the study of innovation activity. Strat. Mgmt. J., 30: 711–735 (2009) DOI: 10.1002/smj

730

Michael J. Leiblein and Tammy L. Madsen

Existing models often suggest that investment in innovation activity is driven by market power or broad administrative characteristics. This emphasis is justified by the often unstated assumption that firms have equivalent levels of experience and innovative ability. For instance, the patent race literature often assumes, for the sake of tractability, that organizations do not retain significant memory or knowledge between innovations (i.e., Reinganum, 1989: 864). By demonstrating the importance of prior experience, our findings indicate that the ‘memorylessness’ assumption is untenable. More broadly, the relationships between various sources of operating experience and innovation activity suggest one way to address the critique of Cohen and Levin that prior empirical work has ‘failed to take systematic account of more fundamental sources of variation in the innovative behavior and performance of firms and industries’ (Cohen and Levin, 1989 : 1061). In sum, the finding that operating experience explains a significant fraction of the variance in innovation rates suggests the importance of unpacking the sources of idiosyncratic research capabilities. These findings also indicate the broad importance of integrating theories of organization with theories of experiential learning and capability development. At the exchange level of analysis it is widely understood that organizational forms vary in their costs and benefits. While the use of more hierarchical forms of organization entail greater administrative cost, they offer potentially offsetting advantages in their ability to direct and coordinate behavior (e.g., Williamson, 1985; Williamson, 1991). Indeed, recent transaction-level studies in the semiconductor industry demonstrate the performance consequences of maintaining a discriminating alignment between exchange level challenges due to specificity, uncertainty, and complexity and the use of particular types of organization (e.g., Leiblein, Reuer, Dalsace, 2002; Macher, 2006). At the firm or portfolio level, additional costs and benefits become salient. In addition to maintaining cooperation and coordination among actors in a single exchange, it is necessary to efficiently direct and coordinate activity across exchanges within the portfolio. This brings added challenges such as how to create incentives for multiple partners to share experiences, whether and how to reduce risk through investment in a diverse set of experiential forms, and how to coordinate information throughout the portfolio. The outcome of these Copyright  2008 John Wiley & Sons, Ltd.

decisions is reflected in the firm’s experience portfolio mix. The theory and findings presented in this study suggest the link between firm experience portfolios and heterogeneity in organizational form choice has important implications for firm innovation activity. Our analysis also has implications for the study of the relationship between different categories of experience and innovation activity. While own firm experience and experience accessed through codevelopment or sourcing agreements all enhance innovation activity, our results indicate that the effects differ in magnitude across the different categories of experience. The different organizational forms associated with these experience stocks provide managers with different mechanisms for allocating resources, designing incentives, and managing synergies. These differences affect the manner in which experience may be used to identify and implement innovations. This suggests that beyond unpacking internal sources of heterogeneity, work also should consider how variance in external collaborative linkages affects innovation activity. Finally, the study shows that innovation-oriented benefits associated with particular experience accumulation mechanisms vary with firm size. These findings complement recent empirical work that examines how firm size moderates the effectiveness of formal and informal interfirm learning mechanisms (e.g., Rosenkopf and Almeida, 2003). Our data do not allow us to directly measure finegrained sources of firm heterogeneity that exist across small and large firms. However, our findings are consistent with logic suggesting that small and large firms employ different employee selection, incentive, and monitoring systems that affect their abilities to translate operating experience into innovation activity. This implies that unbundling heterogeneity in innovation activity requires that work examine how the effects of different types of experience on innovation are influenced by the different selection, incentive, and monitoring devices associated with specific types of organizational form. Some managerial implications also emerge from the results. At a broad level, the results highlight the trade-offs that managers of small and large firms may face when attempting to access experience via alternative mechanisms. They also demonstrate that the optimal method for accessing experientially developed knowledge may differ for small and large firms. More importantly, the specific strengths and weaknesses associated Strat. Mgmt. J., 30: 711–735 (2009) DOI: 10.1002/smj

Incentive Structures, Capability Influences, and Innovation with small and large firms might influence the efficacy with which they are able to access and leverage internally and externally developed experience. For instance, a typical small firm’s dependence on external experience, coupled with its lack of partner selection and hazard mitigating capabilities, suggests that it is crucial that managers of small firms focus their attention on corporate development and alliance management activities. In contrast, the likely existence of muted incentives and coarse measurement instruments at the operational level of large firms, suggests that managers of large firms should direct their attention to identifying opportunities for divisions focused on specific process activities. Several limitations of the study are noteworthy. For one, we use a traditional, albeit narrow, measure of innovation activity—a count of a firm’s innovation activity. Hagedoorn and Cloodt (2003: 1367) suggest that, ‘Innovative performance in the narrow sense refers to results for companies in terms of the degree to which they actually introduce inventions into the market, i.e. their rate of introduction of new products, new process systems or new devices’ (Freeman and Soete, 1997). While extant research shows that our measure effectively underlies a firm’s innovation activity and performance (e.g., Hagedoorn and Cloodt, 2003), combining it with other measures, such as patent citation counts or a count of new product announcements, might strengthen the generalizability of our findings. Second, our data sources lack detailed intrafirm data that would allow a definitive test of the influence of informational diseconomies and specific microlevel employee selection, incentive, and monitoring systems on innovation activity. Although we are not aware of any resources that provide access to this type of data for large scale, systematic, longitudinal empirical analysis, existing theoretical and survey-based work links these fine-grained mechanisms and firm size (e.g. Barnett, 1997; Blau and Schoenherr, 1971; Crozier, 1964; Haveman, 1993; Kelly and Amburgey, 1991; Madsen and Walker, 2007; Zenger and Hesterly, 1997; Zenger and Lazzarini, 2004). As an intermediate solution, future work might explore how specific institutional features, such as the existence of a formal alliance function or a human resource management function, affect a firm’s ability to leverage various forms of operating experience into innovation activity. Third, one might argue that our findings for own production experience Copyright  2008 John Wiley & Sons, Ltd.

731

are due to selection effects, because surviving firms may hold a greater stock of technological innovations. Given the importance of innovation activity, it is certainly possible that firms that are ‘poor’ at leveraging own operating experience or operating experience accessed through codevelopment or outsourcing arrangements might exit the sample. However, some surviving firms also might be weak—as such, slow to accumulate technological innovations. For instance, large firm size, for a variety of reasons, can enhance a firm’s viability (survival) but lower its competitive intensity (Barnett, 1997; Madsen and Walker, 2007). As a result, large firms might survive as weak competitors; such weak competitors might fall behind industry leaders in technological innovation. Indeed, data trends show that, on average, small firms catch up to large firms in their innovation activity over the time period observed. The average innovation activity values for large firms in our sample might pick up weak large firms, thereby lowering the average for the large firm population as a whole. In addition, as a sensitivity analysis, we omitted firms that exited the industry prior to 1999 from our data and reran the model specifications. The findings are consistent with those reported here. Finally, our analysis covers a single, and somewhat unique, empirical context. Although the semiconductor industry provides an excellent context for this study in many respects, the single industry context limits generalizability. Several potential research avenues emerge from the study. Future work might complement our emphasis on the relationship between one type of experience, operating experience, and innovation along a well-established technological trajectory by incorporating fundamentally different measures of firm-level experience (e.g., experience managing innovation transitions) and innovation (e.g., comparing disruptive and nondisruptive innovations). Other extensions include examining how the efficacy of internal and external learning mechanisms vary across different types of organizations such as young and old, private and public, or founder managed and professionally managed firms. Our general arguments also propose a more nuanced view of the need for alignment between organizational structures and the mechanisms used to develop innovation expertise. For instance, future work might consider two-stage modeling techniques that explicate how variables such as size, experience, and age affect the breadth Strat. Mgmt. J., 30: 711–735 (2009) DOI: 10.1002/smj

732

Michael J. Leiblein and Tammy L. Madsen

and depth of external search, and how the breadth and depth of external search affects innovation in a system of equations. In summary, this study emphasizes how managerial investments and actions influence a firm’s innovation behavior. In so doing, we provide potentially valuable insights regarding the relationship between operating experience and organizational form. For one, heterogeneity in firm size influences how different types of experience affect subsequent innovation activity. Compared to larger firms, small firms benefit more from their own operating experience and the operating experience of their codevelopment partners. The benefits of sourcing partners’ operating experience however, appear to be amplified as firms increase in size. These findings have important practical implications for small, resource constrained, firms that often are driven to outsource key activities. As these firms grow, they may benefit by devoting more attention to the development of administrative capabilities and the crafting of agreements that enhance their ability to continually learn from codevelopment partners.

ACKNOWLEDGEMENTS We thank Don Antunes, Jay Barney, Jay Dial, Nicolai Foss, Mike Hitt, Bob Hoskisson, Stephen Phelan, and Harbir Singh for comments on earlier versions of the manuscript. We also wish to acknowledge the helpful comments provided by seminar participants at the Society for Entrepreneurship Scholarship, and the Center for Science, Technology and Society, Santa Clara University.

REFERENCES Abernathy WJ, Utterback JM. 1978. Patterns of industrial innovation. Technology Review 80: 41–47. Acs ZJ, Audretsch D. 1988. Innovation in large and small firms: an empirical analysis. American Economic Review 78(4): 678–690. Adler PS. 1988. Managing flexible automation. California Management Review 31: 34–56. Ahuja G, Katila R. 2004. Where do resources come from? The role of idiosyncratic situations. Strategic Management Journal , August–September Special Issue 25: 887–907. Ahuja G, Lampert CM. 2001. Entrepreneurship in the large corporation: a longitudinal study of how Copyright  2008 John Wiley & Sons, Ltd.

established firms create breakthrough inventions. Strategic Management Journal , June–July Special Issue 22: 521–543. Anand BN, Khanna T. 2000. Do firms learn to create value? The case of alliances. Strategic Management Journal , March Special Issue 21: 295–315. Argote L. 1999. Organizational learning: creating, retaining and transferring knowledge. Kluwer: Norwell, MA. Argote L, Kane AA. 2003. Learning from direct and indirect experience in organizations: the effects of experience content, timing, and distribution. In Group Creativity: Learning through Innovation, Paulus PB, Nijstad BA (eds). Oxford University Press: New York; 277–303. Armour HO, Teece DJ. 1981. Vertical integration and technological innovation. Review of Economics and Statistics 62: 470–474. Baltagi BH, Chang Y. 1994. Incomplete panels: a comparative study of alternative estimators for the unbalanced one-way error component regression model. Journal of Econometrics 62: 67–89. Baltagi BH, Wu PX. 1999. Unequally spaced panel data regressions with ar(1) disturbances. Econometric Theory 15: 814–823. Barnett ML, Salomon RM. 2006. Beyond dichotomy: the curvilinear relationship between social responsibility and financial performance. Strategic Management Journal 27(11): 1101–1122. Barnett WP. 1997. The dynamics of competitive intensity. Administrative Science Quarterly 42: 128–160. Barnett WP, Amburgey TL. 1990. Do larger organizations generate stronger competition? In Organizational Evolution: New Directions, Singh JV (ed). Sage: Newbury Park, CA; 78–102. Barnett WP, Greve HR, Park DY. 1994. An evolutionary model of organizational performance. Strategic Management Journal , Winter Special Issue 15: 11–28. Barnett WP, Hansen MT. 1996. The Red Queen in organizational evolution. Strategic Management Journal , Summer Special Issue 17: 139–157. Barney JB. 1986. Strategic factor markets: expectations, luck and business strategy. Management Science 32: 1231–1241. Barron DN, Elizabeth W, Hannan MT. 1994. A time to grow and a time to die: growth and mortality in credit unions in New York City, 1914–1990. American Journal of Sociology 100: 381–421. Baum JAC, Ingram P. 1998. Survival-enhancing learning in the Manhattan hotel industry, 1898–1980. Management Science 44: 996–1016. Belsley DA, Kuh E, Welsch RA. 1980. Regression Diagnostics: Identifying Influential Data and Sources of Collinearity (Wiley Series in Probability and Statistics). John Wiley & Sons: New York. Blau PM. 1970. A formal theory of differentiation in organizations. American Sociological Review 35: 201–208. Blau PM, Schoenherr RA. 1971. The Structure of Organizations. Basic Books: New York. Strat. Mgmt. J., 30: 711–735 (2009) DOI: 10.1002/smj

Incentive Structures, Capability Influences, and Innovation Bresnahan TF. 1985. Post-entry competition in the plain paper copier market. American Economic Review 75(2): 15–19. Buckley P. 1997. International technology transfer by small- and medium-sized business enterprises. Small Business Economics 9(1): 67–79. Burgelman RA. 1983. A process model of internal corporate venturing in the diversified major firm. Administrative Science Quarterly 28: 223–244. Burns T, Stalker GM. 1961. The Management of Innovation. Free Press: New York. Chandy RK, Tellis GJ. 1998. Organizing for radical product innovation: the overlooked role of willingness to cannibalize. Journal of Marketing Research 35(4): 474–487. Clark KB. 1987. Investing in new technological and competitive advantage. In The Competitive Challenge: Strategies for Industrial Innovation and Renewal , Teece D (ed). Ballinger: Cambridge, MA; 59–82. Clark KB, Fujimoto T. 1991. Product Development Performance: Strategy, Organization, and Management in the World Auto Industry. Harvard Business School Press: Boston, MA. Coase RH. 1937. The nature of the firm. Economica 4: 386–405. Cohen WM. 1995. Empirical studies of innovative activity. In Handbook of the Economics of Innovation and Technological Change, Stoneman P (ed). Blackwell: Cambridge, MA; 182–264. Cohen WM, Levin RC. 1989. Empirical studies of innovative and market structure. In Handbook of Industrial Organization (Vol. 2), Schmalensee R, Willig RD (eds). Elsevier Science: Amsterdam, The Netherlands; 1059–1107. Cohen WM, Levinthal DAL. 1990. Absorptive capacity: a new perspective on learning and innovation. Administrative Science Quarterly 35: 128–152. Cooper AC, Schendel D. 1976. Strategic responses to technological threats. Business Horizons 19(1): 61–69. Crozier M. 1964. The Bureaucratic Phenomenon. University of Chicago Press: Chicago, IL. Damanpour F. 1996. Organizational complexity and innovation: developing and testing multiple contingency models. Management Science 42(5): 693–716. Davari B, Dennard RH, Shahidi GG. 1995. CMOS scaling for high performance and low power: the next ten years. Proceedings of the IEEE 83(4): 595–606. Delios A, Henisz WJ. 2000. Japanese firms’ investment strategies in emerging economies. Academy of Management Journal 43: 305–323. Dewar R, Dutton J. 1986. The adoption of radical and incremental innovations: an empirical analysis. Management Science 32(11): 1422–1433. Dosi G. 1988. Sources, procedures and microeconomic effects of innovation. Journal of Economic Literature 26: 1120–1171. Dyer JH. 1996. Specialized supplier networks as a source of competitive advantage: evidence from the auto industry. Strategic Management Journal 17(4): 271–291. Copyright  2008 John Wiley & Sons, Ltd.

733

Eisenhardt KM, Schoonhoven CB. 1996. Resource-based view of strategic alliance formation: strategic and social effects in entrepreneurial firms. Organization Science 7: 136–150. Ettlie JE. 1988. Taking Charge of Manufacturing. JosseyBass: San Francisco, CA. Ettlie JE, Reze E. 1992. Organizational integration and process innovation. Academy of Management Journal 35(4): 795–827. Ettlie JE, Rubenstein A. 1987. Firm size and product innovation. Journal of Product Innovation Management 4: 89–108. Foster R. 1986. The timing of technological transitions. In Technology in the Modern Corporation: A Strategic Perspective (2nd edn). M. Horwitch (ed). Pergamon Press: New York; 35–50. Freeman C, Soete L. 1997. The Economics of Industrial Innovation. Pinter: London, UK. Galbraith JK. 1952. American Capitalism: The Concept of Countervailing Power . Houghton Mifflin: Boston, MA. Geroski P, Machin S, Van Reenen J. 1993. The profitability of innovating firms. Rand Journal of Economics 24: 198–211. Gilbert RJ, Newbery DMG. 1982. Preemptive patenting and the persistence of monopoly. American Economic Review 72: 514–526. Greene WH. 1993. Econometric Analysis. Macmillan: New York. Grossman SJ, Hart OD. 1986. The costs and benefits of ownership: a theory of vertical and lateral integration. Journal of Political Economy 94: 691–720. Gruber H. 1994. The yield factor and the learning curve in semiconductor production. Applied Economics 26(8): 837–843. Hagedoorn J, Cloodt M. 2003. Measuring innovative performance: is there an advantage in using multiple indicators? Research Policy 32: 1365–1379. Hamel G, Doz YL, Prahalad CK. 1989. Collaborate with your competitors and win. Harvard Business Review 67: 133–139. Hart O, Moore J. 1990. Property rights and the nature of the firm. Journal of Political Economy 98(6): 1119–1159. Haveman HA. 1992. Between a rock and a hard place: organizational change and performance under conditions of fundamental environmental transformation. Administrative Science Quarterly 37: 48–75. Haveman HA. 1993. Organizational size and change: diversification in the savings and loan industry after deregulation. Administrative Science Quarterly 38: 20–50. Hayek FA. 1945. The use of knowledge in society. American Economic Review 35(4): 519–531. Heckman JJ. 1979. Sample selection bias as a specification error. Econometrica 47(1): 153–161. Henderson RM, Clark KB. 1990. Architectural innovation: the reconfiguration of existing product technologies and the failure of established firms. Administrative Science Quarterly 35: 9–30. Henderson R, Cockburn I. 1994. Measuring competence? Exploring firm effects in pharmaceutical research. Strat. Mgmt. J., 30: 711–735 (2009) DOI: 10.1002/smj

734

Michael J. Leiblein and Tammy L. Madsen

Strategic Management Journal , Winter Special Issue 15: 63–84. Hergert N, Morris D. 1988. Trends in international collaborative agreements. In Cooperative Strategies in International Business: Joint Ventures and Technology Partnerships between Firms, Contractor FJ, Lorange P (eds). Lexington Books: New York; 99–110. Hitt MA, Hoskisson RE, Ireland RD. 1990. Mergers and acquisitions and managerial commitment to innovation in M-form firms. Strategic Management Journal , Summer Special Issue 11: 29–47. Hitt MA, Hoskisson RE, Johnson DD, Moesel RA. 1996. The market for corporate control and firm innovation. Academy of Management Journal 39: 1084–1119. Holmstrom B. 1979. Moral hazard and observability. Bell Journal of Economics 10: 74–91. Hsiao C. 1986. Analysis of Panel Data. Cambridge University Press: New York. Ingram P, Baum JAC. 1997. Opportunity and constraint: organizations’ learning from the operating and competitive experience of industries. Strategic Management Journal , Summer Special Issue 18: 75–98. Integrated Circuit Engineering (ICE). 1991–2000. Profiles: A Worldwide Survey of IC Manufacturers and Suppliers. Integrated Circuit Engineering: Scottsdale, AZ. Kale P, Dyer J, Singh H. 2002. Alliance capability, stock market response, and long-term alliance success: the role of the alliance function. Strategic Management Journal 23(8): 747–767. Kelly D, Amburgey TL. 1991. Organizational inertia and momentum: a dynamic model of strategic change. Academy of Management Journal 34: 591–612. Kennedy P. 1998. A Guide to Econometrics (3rd edn). MIT Press: Cambridge, MA. Khanna T, Gulati R, Nohria N. 1998. The dynamics of learning alliances: competition, cooperation, and relative scope. Strategic Management Journal 19(3): 193–210. King A, Tucci C. 2002. Incumbent entry into new markets: the role of experience and managerial choice in the creation of dynamic capabilities. Management Science 48: 171–186. Klein B, Crawford RG, Alchian AA. 1978. Vertical integration, appropriable rents, and the competitive contracting process. The Journal of Law & Economics 21(2): 297–326. Leiblein MJ, Reuer JJ, Dalsace F. 2002. Do make or buy decisions matter? The influence of organizational governance of technological performance. Strategic Management Journal 23(9): 817–833. Levinthal DA, March JG. 1993. The myopia of learning. Strategic Management Journal , Winter Special Issue 14: 95–112. Levitt B, March JG. 1988. Organizational learning. Annual Review of Sociology 14: 319–400. Macher JT. 2006. Technological development and the boundaries of the firm: a knowledge-based examination in semiconductor manufacturing. Management Science 52(6): 826–843. Copyright  2008 John Wiley & Sons, Ltd.

Madsen TL, Walker G. 2007. Incumbent and entrant rivalry in a deregulated industry. Organization Science 18: 667–687. Martin X, Salomon R. 2003. Tacitness, learning, and international expansion: a study of foreign direct investment in a knowledge-intensive industry. Organization Science 14(3): 297–311. Merton RK. 1957. Social Theory and Social Structure. Free Press: New York. Meyer JW, Rowan B. 1977. Institutionalized organizations: formal structure as myth and ceremony. American Journal of Sociology 83: 340–363. Milgrom P, Roberts J. 1990. The economics of modern manufacturing: technology, strategy, and organization. American Economic Review 80(3): 511–528. Mitchell W, Singh K. 1993. Death of the lethargic: effects of expansion into new technical subfields on performance in a firm’s base business. Organization Science 4(2): 152–180. Nanda A, Bartlett C. 1994. Intel Corporation: leveraging capabilities for strategic renewal . Harvard Business School, case study number 9-394-141. Harvard Business School: Boston, MA. National Academy of Engineering. 1995. Risk & Innovation: The Role and Importance of Small HighTech Companies in the U.S. Economy. National Academy Press: Washington, D.C. Nickerson JA, Zenger TR. 2004. A knowledge-based theory of the firm: the problem-solving perspective. Organization Science 15(6): 617–632. Pavitt K. 1984. Sectoral patterns of technical change: towards a taxonomy and a theory. Research Policy 13: 343–373. Pavitt K. 1990. What we know about the strategic management of technology. California Management Review 32: 17–26. Pisano GP, Bohmer RMJ, Edmondson AC. 2001. Organizational differences in rates of learning: evidence from the adoption of minimally invasive cardiac surgery. Management Science 47: 752–768. Polanyi M. 1962. Personal Knowledge: Towards a PostCritical Philosophy. Harper & Row: New York. Powell WW, Koput KW, Smith-Doerr L. 1996. Interorganizational collaboration and the locus of innovation: networks of learning in biotechnology. Administrative Science Quarterly 41(1): 116–145. Putterman L. 1995. Markets, hierarchies, and information: on the paradox in the economics of organization. Journal of Economic Behavior & Organization 26: 373–393. Reinganum JF. 1983. A note on the strategic adoption of new technology. Journal of Optimization Theory and Applications 39: 133–141. Reinganum JF. 1989. The timing of innovation: research, development, and diffusion. In Handbook of Industrial Organization (Vol. 1), Schmalensee R, Willis R (eds). Elsevier: Amsterdam, The Netherlands; 849–908. Rivkin JW. 2001. Reproducing knowledge: replication without imitation at moderate complexity. Organization Science 12: 274–293. Strat. Mgmt. J., 30: 711–735 (2009) DOI: 10.1002/smj

Incentive Structures, Capability Influences, and Innovation Rosenberg N. 1972. Technology and American Economic Growth. Harper Torch Books: New York. Rosenberg N. 1982. Inside the Black Box: Technology and Economics. Cambridge University Press: Cambridge, UK. Rosenkopf L, Almeida P. 2003. Overcoming local search through alliances and mobility. Management Science 49(6): 751–766. Rothaermel FT, Deeds DL. 2004. Exploration and exploitation alliances in biotechnology: a system of new product development. Strategic Management Journal 25(3): 201–221. Rumelt RP. 1984. Towards a strategic theory of the firm. In Competitive Strategic Management, Lamb RB (ed). Prentice Hall: Englewood Cliffs NJ; 556–570. Schumpeter JA. 1934. The Theory of Economic Development. Harvard University Press: Cambridge, MA. Schumpeter JA. 1942. Capitalism, Socialism and Democracy. Harper Brothers: New York. Selznik P. 1949. TVA and the Grass Roots: A Study in the Sociology of Formal Organization. University of California Press: Berkeley, CA. Simonin BL, Helleloid D. 1993. Do organizations learn? An empirical test of organizational learning in international strategic alliances. In Academy of Management Best Paper Proceedings, 1993 , Moore D (ed). Omnibus: Madison, WI; 222–227. Sorensen JB, Stuart TE. 2000. Aging, obsolescence and organizational innovation. Administrative Science Quarterly 45: 81–112. Stalk G. 1988. Time: the next source of competitive advantage. Harvard Business Review 66(4): 41–51. Stigler GJ. 1962. Administered Prices and Oligopolistic Inflation. University of Chicago Press: Chicago, IL. Teece DJ. 1982. Towards an economic theory of the multiproduct firm. Journal of Economic Behavior & Organization 3: 39–64. Teece DJ. 1986. Profiting from technological innovation: implications for integration, collaboration, licensing, and public policy. Research Policy 15: 285–305. Teece DJ. 1992. Competition, cooperation, and innovation: organizational arrangements for regimes of rapid technological progress. Journal of Economic Behavior & Organization 18: 1–25. Thompson VA. 1965. Bureaucracy and innovation. Administrative Science Quarterly 10(1): 1–20. Tripsas M. 1997. Unraveling the process of creative destruction: complementary assets and incumbent

Copyright  2008 John Wiley & Sons, Ltd.

735

survival in the typesetter industry. Strategic Management Journal , Summer Special Issue 18: 119–142. Tsai W. 2001. Knowledge transfer in intraorganizational networks: effects of network position and absorptive capacity on business unit innovation and performance. Academy of Management Journal 44: 996–1004. Tushman ML, Anderson P. 1986. Technological discontinuities and organizational environments. Administrative Science Quarterly 31: 439–465. Tushman ML, O’Reilly C. 1997. Winning through Innovation: A Practical Guide to Leading Organizational Change and Renewal . Harvard Business School Press: Boston, MA. Utterback JM. 1994. Mastering the Dynamics of Innovation: How Companies Can Seize Opportunities in the Face of Technological Change. Harvard Business School Press: Boston, MA. Utterback JM, Abernathy WJ. 1975. A dynamic model of product and process innovation. Omega 3(6): 639–656. Williamson OE. 1975. Markets and Hierarchies: Analysis and Antitrust Implications. Free Press: New York. Williamson OE. 1985. The Economic Institutions of Capitalism. Free Press: New York. Williamson OE. 1991. Comparative economic organization: the analysis of discrete structural alternatives. Administrative Science Quarterly 36: 269–296. Winter SG. 2003. Understanding dynamic capabilities. Strategic Management Journal , October Special Issue 24: 991–995. Wooldridge JM. 2002. Econometic Analysis of Cross Section and Panel Data. MIT Press: Cambridge, MA. Yin X, Zajac EJ. 2004. The strategy/governance structure fit relationship: theory and evidence in franchising arrangements. Strategic Management Journal 25(4): 365–383. Zenger TR. 1994. Explaining organizational diseconomies of scale in R&D: agency problems and the allocation of engineering talent, ideas, and effort by firm size. Management Science 40: 708–729. Zenger TR, Hesterly WS. 1997. The disaggregation of corporations: selective intervention, high-powered incentives, and molecular units. Organization Science 8(3): 209–234. Zenger TR, Lazzarini SG. 2004. Compensating for innovation: do small firms offer high-powered incentives that lure talent and motivate effort? Managerial & Decision Economics 25(6/7): 329–345.

Strat. Mgmt. J., 30: 711–735 (2009) DOI: 10.1002/smj