Decision Making

jroero

2.-Business_Unit_Reorganization_a.pdf

Home >Business & Finance homework help >Operations Management homework help >Decision Making

MANAGEMENT SCIENCE Vol. 55, No. 7, July 2009, pp. 1237–1254 issn 0025-1909 � eissn 1526-5501 � 09 � 5507 � 1237

informs ® doi 10.1287/mnsc.1090.1017

Business Unit Reorganization and Innovation in New Product Markets

Samina Karim School of Management, Boston University, Boston, Massachusetts 02215,

[email protected]

This paper empirically examines how business unit reorganization affects innovation, and explores how thelearning process may mediate this relationship. Unit reorganization is the creation, deletion, or recombina- tion of business units within a firm. Innovation is radical and involves product market entry by a firm into markets in which it was not previously active. I test competing hypotheses that predict either a U-shape or inverted U-shape relationship between reorganization and innovation to determine whether and how learning occurs in the presence of unit-level structural change. Theoretical support is drawn from literature on dynamic capabilities and organizational learning. The sample studied is 250 medical firms belonging to the pharmaceu- tical, healthcare-service, and medical-device industries, studied over a 20-year period. The findings are twofold. First, reorganization is found to exhibit a U-shape relationship with innovation, supporting learning arguments that stress the importance of experiencing a cohort of multiple events. Second, only reorganization experiences within a current period affect future innovation; past experiences do not impact future innovation, implying that firms may face constraints in organizational memory. The study concludes by exploring the structural origin (i.e., from internal, acquired, or recombined units) of innovative activity within firms.

Key words: reorganization; innovation; organizational learning; reconfiguration; market entry; mergers and acquisitions; innovation origin; internal development; organizational memory; organic growth

History: Received March 8, 2007; accepted February 9, 2009, by Pankaj Ghemawat, business strategy.

1. Introduction This paper examines how business unit reorganiza- tion affects innovation, and explores how the learn- ing process may mediate this relationship. We often hear in the media about firms reorganizing. Firms face the constant challenge of judging whether or not they are organized optimally. Structural organization determines resource allocations within firms, use of internal routines and communication networks, and the flow of information and tasks (Chandler 1962, Levitt and March 1988, Galunic and Eisenhardt 1996, Helfat and Eisenhardt 2004). Reorganization changes these circumstances, creating a new environment for learning and the potential for resources to be used in new combinations leading to innovation (Schumpeter 1934, Penrose 1959, Kogut and Zander 1992, Galunic and Rodan 1998). The goal of this paper is to deter- mine if there are any innovative benefits to reorga- nization and, in general, to determine how learning occurs in the presence of unit-level structural change. Do firms learn immediately from a few reorganization attempts? If not, how many experiences are needed before there is learning? I observe how innovative firms are as an indication of learning having occurred. My motivation for this study comes from prior

work on dynamic capabilities that have found firms to be highly reorganization-active. Reorganization has

been explored at the resource level, divisional level, and activity level. At the resource level, scholars have observed how resource recombination within the firm may result in innovation (Schumpeter 1934, Henderson and Clark 1990, Kogut and Zander 1992, Galunic and Rodan 1998) and how redeployment of resources between targets and acquirers can create value from acquisitions (Capron et al. 1998, Capron and Mitchell 1998). Strategy research on diversifica- tion has also studied restructuring, which refers to the addition or deletion of divisions as firms choose what businesses (i.e., market activities) to be involved in (Porter 1987, Bowman and Singh 1990, Hoskisson and Johnson 1992). Recently, the dynamic capabili- ties literature has focused on the reorganization of market activities, a process that involves reorganiz- ing both the underlying resources and the divisions that are responsible for these activities. This body of work has addressed the reassignment of business charters between divisions of a firm, and the patching or reconfiguration of business units and their asso- ciated activities (Galunic and Eisenhardt 1996, 2001; Brown and Eisenhardt 1997; Eisenhardt and Brown 1999; Karim 2006). Based on this significant body of literature, I won-

der how much value is actually created or destroyed from reorganization. In a prior study, I found that,

1237

more often than not, firms pursued reorganization of units before divestiture (Karim 2006). These find- ings suggest that firms are attempting to create value by realigning units’ resources and activities within the firm, and that there is an underlying belief that reorganization is potentially beneficial. Does this potential benefit come to fruition? To date, there are two empirical studies that have examined unit-level structural change and firms’ financial performance. Brickley and Van Drunen (1990, pp. 251–252) focus on firms that reorganize by “merging, splitting, or liquidation of existing units or creating new units.” In their study of 222 reorganization announcements involving 179 firms, they find that most firms’ rea- sons for pursuing unit reorganization are either to increase efficiency or to change their “scale of invest- ments” (i.e., growth opportunities into new prod- uct markets or exiting current product markets).1

Similarly, Barkema and Schijven (2008) study how Dutch firms’ organizational restructuring announce- ments from annual reports affect future firm perfor- mance, finding a nonsignificant relationship. There is still a gap in the literature exploring how unit reor- ganization affects innovation. Brown and Eisenhardt (1997) expressed the importance of studying this rela- tionship; they noted, “innovation is intimately related to broader organization change. Yet research to date has revealed very little about the underlying struc- tures and processes by which firms actually achieve continuous innovation and ultimately, change” (p. 2).2

Furthermore, earlier empirical papers only observed whether or not reorganization occurred; there is an opportunity to study more granular degrees of unit reorganization, reflecting different amounts of experi- ence with structural change. This paper contributes to the strategy field by

informing both the dynamic capabilities perspective and organizational learning theories. First, it provides rich quantitative, multi-industry empirical evidence of how business unit reorganization affects perfor- mance, complementing the qualitative, case-based

1 Brickley and Van Drunen (1990) also find that reorganization announcements increase shareholder wealth; however, in the three- year period following the restructurings there is a decline in earn- ing performance. 2 Related to dynamic capabilities, modularity literature has referred to how unit reorganization may affect innovation. However, the main question addressed by this body of work is, “How do firms partition tasks inside and outside the organization?” Studies have focused on organizational forms and how changes in product architecture may lead to innovation or efficiency (Langlois and Robertson 1992, Garud and Kumaraswamy 1995, Sanchez 1995, Ciborra 1996, Sanchez and Mahoney 1996, Lei et al. 1996, Baldwin and Clark 2000, Schilling 2000, Schilling and Steensma 2001). My interest in this paper is to focus on the reorganization of tasks (via the business units that are responsible for them) that the firm itself manages.

papers in the dynamic capabilities literature. Second, it studies innovative performance into new prod- uct markets, an outcome variable that has yet to be explored. Furthermore, by doing so, it allows the observation of whether the growth opportuni- ties mentioned by Brickley and Van Drunen (1990) come to fruition. Lastly, it contributes to organiza- tional learning theories by exploring how the learn- ing process mediates the reorganization-innovation relationship. Currently, there is no clear theory addressing the

reorganization phenomenon. My interest, motivated by Brown and Eisenhardt’s (1997) suggestion, is to develop theory that informs us about the mecha- nisms at work during reorganization. This is impor- tant for us to better understand how firms create value and pursue successful change. Because the- ories regarding structural change are still ambigu- ous, I present competing hypotheses, one predicting a U-shape and the other an inverted U-shape relation- ship between reorganization and future innovation. The main distinction that is explored is whether unit reorganization results in immediate benefits followed by eventual decline or whether learning and innova- tion occur only after experiencing multiple reorgani- zation events. Some terminology should be clarified, because “re-

organization” and “innovation” are broad terms. In this study, reorganization refers to change at the business unit level. Business unit reorganization (see Figure 1 for example) is the creation of units, dele- tion of units, or recombination of units within the firm; the resources and activities of these units still remain within the organization, but the structural location where these resources and activities reside have been altered somehow.3 Unit acquisitions and divestitures, which are choices made by firms to obtain new or discard existing resources and activi- ties, are not considered reorganization events in this study;4 reorganization is the change in structural loca- tion within the firm of what the firm already has

3 A target that is acquired and added as a new, autonomous unit to the firm is not considered a reorganization event in this study. By creation of unit I am referring to when a firm creates a unit whose boundaries did not exist earlier. Thus, a created unit is developed by splitting out parts of existing units (either internal or acquired) into a newly named unit, completely merging (i.e., recombining) several existing units into a newly named unit, or a bit of both. Similarly, divesting a unit is not considered deletion. By deletion I mean the case of when an existing unit’s identity is dissolved because its resources and activities have recombined into another unit. Thus, creating and deleting units are two examples of recombination. The third example of recombination is when a unit that existed earlier still retains its identity, but its resources and activities have changed somehow (either parts were split out or parts were merged in). Examples of recombination are highlighted in Karim (2006). 4 Karim (2006) defines reconfiguration as encompassing the process of unit reorganization as well as unit acquisition and divestiture.

Figure 1 Example of Reorganization and Innovation at Johnson & Johnson Inc.

Codman & Shurtleff Inc. (acquired pre- 1975) - implantables

Kees Surgical Specialty (acquired 1987) - arthroscopes

Applied Fiberoptics (acquired 1982) - electrodes

- neurostimulators

Codman & Shurtleff Inc. - implantables - arthroscopes - electrodes - lighting equip. - n. stimulators

By 1989 reorganize KSS into CS.

J&J Orthopedics Inc. (created internal unit 1986) - implantables

J&J Professional Inc. (created recombined unit 1993) - implantables - ortho. devices - electrodes - lighting equip. - n. stimulators

1978

Codman & Shurtleff Inc. - implantables - electrodes - lighting equip. - n. stimulators

199319891986

By 1986 reorganize AF into CS.

By 1993 reorganize CS and J&J Ortho into new J&J Prof.

By 1986 reorganize parts of CS into new J&J Ortho.

- lighting equip.

- ortho. devices

to work with and continues to work with. Acquired units that experience structural change at a later time (i.e., when their activities are no longer new to the firm) are considered to be reorganization events.5

In this manner, the reorganization of units both inter- nally developed and acquired are observed. Inno- vation is generally considered to be the creation of new resources by the firm. In this paper, innova- tion is observed at the product level and is radi- cal (Abernathy and Clark 1985), representing internal entry (i.e., not via acquisition) into new product mar- kets. Incremental innovations, namely changes within existing product lines, are not addressed here. This distinction is important; although incremental learn- ing may be occurring (some of which manifests in incremental innovations and some learning that does not manifest in innovation at all), my interest in this paper is to study substantial learning through radical innovations. The empirical setting for this paper is the medi-

cal sector and its three industries: healthcare services, pharmaceuticals, and medical devices. The sample

Thus, reconfiguration addresses not only reorganization, but also the choice to add/shed resources and activities to/from the firm. 5 In Figure 1, the acquisition of Kees Surgical and its arthroscope products in 1987 is not considered a reorganization event. In 1989, when the Kees Surgical unit is recombined into the Codman and Shurtleff business unit, this is considered a reorganization event.

observed is of 250 firms, their units and product markets, from 1975 to 1997. Each unit’s evolution is tracked as it is reorganized and any radical innovative activity (i.e., internal entry into new product markets) is noted. Units and market entries obtained via acqui- sition are also observed. Each firm is then represented by a history of unit reorganizations and innovations over time. The main findings of the paper are twofold. First,

the relationship between reorganization and future innovation exhibits a U-shape relationship, support- ing arguments that stress the importance of experi- encing multiple events from which to learn. Second, although learning from multiple events within a three to four-year period influences innovation, the experi- ence prior to that period does not. This suggests that there may be limits to organizational memory and the learning that can be retained. Lastly, exploratory work suggests that internal units and those combined with acquired units (as compared to acquisitions alone) are key to innovative activity. This paper is organized as follows. It begins

with background on relevant literature and presents hypotheses. This is followed by methods and mea- sures that outline the data and sample, operational- ization of variables, and the methodology of this study. Finally, the paper concludes with results and discussion.

2. Background and Hypotheses 2.1. Literature on Reorganization Reorganization has been explored in various con- texts within the strategy field, including innovation, acquisition, diversification, and dynamic markets. The chosen unit of analysis has varied between the resource level, divisional level, and activity level. At the resource level, key terms include recom-

bination and redeployment. Innovation literature has addressed how new resource combinations or similar resources combined in new ways (i.e., recombination) may result in innovation. The first question explores the content of what is brought together, whereas the latter examines the process by which value is created from the content. Stemming from Schumpeter’s con- cept of creative destruction (1934), architectural inno- vation refers to new links between components within product architectures (Henderson and Clark 1990). Other scholars have theorized about the character- istics of knowledge affecting resource recombination and innovation (Galunic and Rodan 1998), as well as knowledge itself (comprised of information and know-how) being a recombined resource that leads to market opportunities (Kogut and Zander 1992). In the context of acquisitions, studies have explored how best to manage resources between targets and acquir- ers. Capron and colleagues found that resource rede- ployment between target and acquirer is greater for resources that face greater market failure, and usu- ally moves in the direction from the stronger business to the weaker business (Capron et al. 1998, Capron and Mitchell 1998). Others have observed that firms may reuse resources for different purposes—referred to as “organizational modularity”—as they exit and enter different markets, achieving economies of scope (Helfat and Eisenhardt 2004). The divisional level of reorganization is usually

referred to as restructuring and has received much attention in the diversification literature. Restructur- ing refers to the addition (either internal or acquired) or deletion of divisions as firms choose what busi- nesses (i.e., product markets) to be involved in. Papers addressed what made firms more or less division- alized, how best to restructure and refocus diversi- fied firms, and how divisional restructuring affected financial performance (Hoskisson and Galbraith 1985, Hoskisson 1987, Porter 1987, Keats and Hitt 1988, Bowman and Singh 1990, Hoskisson and Johnson 1992, Johnson 1996). Bringing together both resource and divisional lev-

els of reorganizing, the dynamic capabilities perspec- tive has explored activity-level reorganizing as firms respond to turbulent markets by realigning their busi- nesses frequently (Teece et al. 1997, Eisenhardt and Brown 1999, Eisenhardt and Martin 2000). This body

of work advocates a recombinant view of organiza- tions, and can be partitioned into two main areas dealing with direct reorganization of market activi- ties and indirect reorganization by reorganizing the divisions (also known as business units) within which they reside. First, scholars have explored the direct reassignment

of divisional charters within the firm between divi- sions (Galunic and Eisenhardt 1996, 2001; Birkinshaw and Lingblad 2005). Charters are businesses (prod- uct and market arenas) in which a division partic- ipates and for which it is responsible (Galunic and Eisenhardt 1996), and is synonymous with my use of the term “product market activity” in this paper. Studying one large Fortune 100 firm, Galunic and Eisenhardt (1996) found that there were three patterns of charter loss; the life-cycle phase of a division (start- up, growth, and mature phases) along with its per- formance and executives’ perceptions influenced the division’s loss of a charter. They also examined char- ter gains and found three patterns to exist; charters were usually awarded to weaker divisions to encour- age them and give them an opportunity (labeled “new charter opportunities”), stored in safe-haven divisions as orphaned charters (labeled “charter foster homes”), or vied for by multiple divisions through interdivi- sional competition (labeled “charter wars”) (Galunic and Eisenhardt 2001). More recently, Birkinshaw and Lingblad (2005) have focused on this latter competi- tion where charter overlap may exist between busi- ness units, and developed a theoretical framework to identify the relationship between units based on the amount of overlap and the fungability of charter scope. The second area of study within dynamic capabil-

ities literature involves patching and reconfiguration— the indirect realignment of business activities (and their underlying resources) through the reorganiza- tion of the business units that are responsible for them. Eisenhardt and Brown (1999, p. 73) observe patching, “� � � the adding, deleting, splitting, transfer- ring, or combining chunks of businesses,” as firms frequently remap their divisional structures to react to changing market opportunities. Studying six firms in the computer industry, they find that innovation is the result of continuous change involving three firm char- acteristics: some structural design freedom to impro- vise (labeled “semistructures”), a balance between planning and reacting by having probes into the future (labeled “links in time”), and rhythmic, time- paced transition processes between product activities (labeled “sequenced steps”) (Brown and Eisenhardt 1997). Building on this work, Karim and Mitchell (2004) study reconfiguration that includes patching to changing market opportunities, but also involves pur- poseful experimentation and search for new oppor- tunities. Business unit reconfiguration is defined as

“� � � the addition of units to the firm, deletion of units from the firm, and recombination of units within the firm” (Karim 2006, p. 799). Carroll and Karim (2009) explore theories that predict reconfiguration and those factors that hinder it. In an earlier study I focused on the recombination aspect of reconfig- uration, examining what forms of unit recombina- tion were most common and when recombination occurred; I found that acquisitions were often recom- bined together or merged into internal units, and that firms often pursued recombination before divestiture in an attempt to create value (Karim 2006). Although there has been a good deal of qualitative

work studying direct and indirect activity-level reor- ganization, there is still room for examining reorgani- zation empirically with a large sample and in multiple industries. Furthermore, all the papers mentioned above refer to some value creation or benefits result- ing from reorganization; however, even if we assume that these benefits may arise from the reorganization of content (i.e., the resources held within units that make activities feasible), there is still no clear theory addressing the process by which these benefits may accrue. Brown and Eisenhardt (1997) theorize, based on complexity theory, that an inverted U-shape rela- tionship may exist between structural flexibility and innovation; however, this has not yet been empiri- cally examined or the generalizability explored. The goal of this paper is to draw upon dynamic capa- bilities and organizational learning theories to dis- cern the process by which the learning mechanism is actually at work during reorganization. Literatures addressing the reorganization-innovation relationship support two competing hypotheses, one of immediate positive performance followed by a decline, the other of declining performance followed by improvements. Below, I present arguments for each.

2.2. Business Unit Reorganization and Innovation The first hypothesis argues that although business unit reorganization may facilitate the recombination of resources in new ways to enhance innovation (Schumpeter 1934, Penrose 1959, Kogut and Zander 1992, Galunic and Rodan 1998), excessive amounts of reorganization may be harmful to organizations. Studying firms’ reorganization patterns, scholars have found evidence suggesting that there is a perception or belief that reorganization may be beneficial. Firms proactively experiment with structure and attempt reorganization in the hopes of finding opportunities (Ciborra 1996). Furthermore, units being divested are usually first recombined in some way in an attempt to create value (Karim 2006). The organizational learning literature refers to this recombination as the “grafting” together of components from which to obtain knowl- edge (Huber 1991). Empirical papers that explore

grafting and its effects on learning include studies on joint ventures and alliances (Lyles 1988, Lane and Lubatkin 1998).6 Business unit reorganization that involves recombination can be viewed as units being grafted onto each other to potentially lead to new opportunities and innovation. Another positive conse- quence of reorganizing is its role in bringing different groups of people together. Research has shown that cross-functional teams lead to faster product devel- opment (Eisenhardt and Tabrizi 1995); this is a result of social interaction and trust significantly affecting the extent of resource exchange, which in turn affects product innovation (Tsai and Ghoshal 1998, Figueroa and Conceicao 2000). One motivation for reorganiz- ing may be to increase the interaction of employees by recombining business units and, in doing so, facil- itating innovation. Both dynamic capabilities and organization learn-

ing theories caution against too much change or disor- der. Recall that Brown and Eisenhardt (1997), drawing upon dynamic capabilities and complexity theories, expect that an inverted U-shape relationship exists between structural flexibility (i.e., potential for patch- ing or reorganizing) and innovation. Complexity the- ory describes systems as interconnections between agents following certain rules that operate at the edge of chaos (i.e., instability) (Gleick 1987; Waldrop 1992; Kauffman 1991, 1992; Stacey 1995). “If the inter- connections become too rich, however, the system becomes so hyperactive that it becomes difficult to make any sense of it” (Stacey 1995, p. 487). Thus, business unit reorganization can be viewed as creat- ing new interconnections between units that in excess, beyond the threshold of stability, may be too complex for fruitful innovation to occur. Within structure, these interconnections are the lines of authority and com- munication, and the knowledge (i.e., information and data) that flows through these lines (Chandler 1962). Organization learning literature highlights the impor- tance of coordinating these interconnections to foster learning; one way is via the accessibility (i.e., location and proximity within structure) of resources (Levitt and March 1988, Dodgson 1993). Although reorgani- zation may serve as a mechanism to alter resource accessibility, there still remain other obstacles to coor- dination. Huber (1991, p. 103) warns that firms may experience information overload if “the information to be interpreted exceeds the units’ capacity to pro- cess the information adequately.” Reorganization cre- ates new interconnections of information and data as

6 Lyles (1988) found that in environments of uncertainty, joint ven- ture organizations attempted to operate in a decentralized mode. Lane and Lubatkin (1998) found that alliance partners’ abilities to learn from each other depends upon similarities in knowledge bases, “dominant logics” regarding firms’ objectives (Prahalad and Bettis 1986), and organizational structure.

units are combined or activities are moved in or out of units; if the capacity to process this amount of infor- mation does not exist, then the unit may be unable to learn or innovate. Thus far, I have spoken of reorganization as a pur-

poseful mechanism that brings resources together by changing their location and proximity to one another. Fiol and Lyles (1985, p. 806) remind us, however, that change may be caused “merely by a need to do something” and may not result in learning. Alter- natively, some degree of change (even if it is busy work that is not thoughtful) could still result in unin- tentional learning. In their theory piece, they pre- dict an inverted U-shape relationship between level of change and level of learning, regardless of whether or not the intent for change was thoughtful. In this paper, I draw similar parallels between degree of reorganization and degree of innovation. As with general change and learning, excessive amounts of reorganization may lead to disruption, inefficiency, and misinterpreted information that result in declin- ing innovative performance. This logic supports the idea that there is some optimal level of reorganization that enhances innovation.

Hypothesis 1A. Business unit reorganization (within a given period) will have an inverted U-shape impact on future innovation (in the next period).

2.3. Learning as a Mediating Factor Although firms may want to innovate or have the potential to innovate, learning how to reorganize alone may not be enough to innovate in a new prod- uct market. There could be deterrents such as resource constraints, inertia, risks from cannibalization, dis- agreements over objectives, and managerial power struggles, to name a few. Thus, learning can certainly occur without innovation, but I argue that innovation (namely, internal entry into a new product market) cannot occur without learning. Given the extent of differences between product markets and the distinc- tion that the innovations referred to in this paper are not incremental extensions of existing products, entry reflects the firm’s ability to create something new. This creation of products that are new to the firm is only feasible through significant learning. Workers must learn what content (i.e., resources) to use, con- figurations (i.e., the architecture of these resources) to design, and processes (i.e., the procedures to deal with problemistic and opportunistic search and selec- tion) to routinize to facilitate the creation of a new product (Penrose 1959, Cyert and March 1963, Ansoff 1965, Nelson and Winter 1982, Henderson and Clark 1990). Furthermore, managers must assess the value of a new innovation and decide whether and how to commercialize it; commercializing a product in a mar- ket that is new to the firm involves learning about

the new market as well. Based on these characteris- tics, this study assumes that learning is a mediator between the reorganization-innovation relationship, and that innovative market entry reflects learning. Scholars define learning, like innovation, as an

active (i.e., not passive7) process. Organizational learning involves the processes of knowledge acqui- sition, information distribution, information interpre- tation, and functioning of an organizational memory (Huber 1991). These are processes for obtaining, shar- ing, interpreting, storing, and retrieving knowledge. Several scholars stress the role of interpretation; learn- ing is viewed as a new response or action based on the interpretation of data (i.e., giving data meaning) (Daft and Weick 1984), and also as the “develop- ment of insights, knowledge, and association” (Fiol and Lyles 1985, p. 811). Arrow (1962, p. 155) makes explicit that, “Learning is the product of experience. Learning can only take place through the attempt to solve a problem and therefore only takes place during activity.” These learning processes involving knowl- edge and experience are all necessary to innovate; in essence, innovation (or market entry) can be viewed as a culmination of these processes and be indicative of some learning having taken place.8

Classic learning literature has focused on a model of learning-by-doing, exemplified by the learning curve where the inputs needed per item of output decline at a declining rate due to experiential learning (Arrow 1962). Empirical studies on the learning curve have been well suited for examining the improvements in process development and manufacturing productiv- ity (Pisano 1994, Argote 1996). In this paper, my goal is to examine innovative productivity and to explore how learning occurs in the presence of structural unit change. Based on experiential learning, the literature supports a U-shape relationship between change and performance, acknowledging that experience may not

7 Lane and Lubatkin (1998) describe three methods of learning: pas- sive, active, and interactive. However, their description of passive is different from my use above. They associate passive learning with acquisition of knowledge from journals, seminars, and consultants. Active learning involves benchmarking and observing competitors. Interactive learning occurs when firms have a student-teacher rela- tionship in that one firm is close enough to the other to learn the tacit components of their capabilities. 8 Learning, like innovation, may be categorized into different degrees. Argyris and Schön (1978) categorize three types of learn- ing: single-loop learning (occurs when errors are detected and corrected, without altering the fundamental nature of the organi- zation’s activities), double-loop learning (involves modification of norms, policies, and objectives), and deutero learning (occurs when both single and double-loop learning occur). Fiol and Lyles’ (1985) categories of learning are similar: lower-level learning (focuses on immediate change in behavior or an activity; is a result of repetition and routine and involves association building), and higher-level learning (aims at adjusting overall mission, beliefs, rules, or norms rather than a specific activity or behavior).

result in immediate learning, but instead may take a roundabout process before learning is achieved. It is important to distinguish that this experience may take two forms: one referring to prior time and past expe- riences, and the other referring to the number or type of events that may be experienced in cohort within a current period. Time affects learning because learning is built on

memories, associations, and the ability to recognize useful information (Huber 1991). Association and recognition are influenced by an organization’s past learning, thus, learning is highly path dependent. This is widely acknowledged by scholars. Absorp- tive capacity is largely a function of a firm’s level of prior related knowledge (Cohen and Levinthal 1990). Interpretations, part of Daft and Weick’s (1984) model of learning, are influenced by one’s previous experi- ence. Furthermore, any positive feedback from these experiences can lead to stronger path dependence (March 1991) and the risk of firms falling into com- petency traps or making erroneous inferences (Levitt and March 1988). Competency traps occur when a routine that led to good performance in the past is used repeatedly although it may be far from optimal. Organizations also face “superstitious” learning when incorrect connections are made between actions and outcomes (Levitt and March 1988). Organizations can overcome the pitfalls of path

dependence through “higher-level learning,” which involves discriminating, unlearning, and experimen- tation (Hedberg 1981, Lyles 1988). Discriminating is the ability to discern when a certain behavior or action is appropriate to a particular situation; i.e., when to apply what is learned from one event to another. Unlearning involves discarding obsolete and misleading knowledge that is no longer appropriate. Experimentation is the ability to move away from the learned responses of successful programs and to experiment with new approaches. It is through the experience of time and multi-

ple events that firms learn to discriminate and un- learn from other successful experiences. Learning also occurs from unsuccessful experiences (i.e., mis- takes); Hayward (2002) found that a moderate num- ber of mistakes during acquisition experiences led to improved acquisition performance. Studying this same relationship, Haleblian and Finkelstein (1999) examined the effect of acquisition experience (i.e., the number of acquisition events an acquirer experienced within their study period) on acquisition performance (i.e., abnormal returns averaged around acquisition announcement) and found it to be U-shaped. Using behavioral learning theory, they attribute the initial declining returns of experience on performance to inappropriate generalizations made to other acquisi- tions, and the increasing returns to experiencing a sig- nificant number of multiple events and discrimination

among acquisitions. Other studies have also found that learning and the skill of discriminating develop as firms are faced with variation and experience het- erogeneous events (Haunschild and Sullivan 2002, Schilling et al. 2003). Based on these arguments, firms may not have suc-

cess innovating after reorganization until they have experienced several reorganization events and prac- ticed the appropriate application of what is learned to various circumstances. Drawing parallels to Haleblian and Finkelstein’s (1999) study of acquisition experi- ence, reorganization experience may similarly involve improper generalizations followed by accurate dis- crimination, resulting in a U-shape relationship with innovation.

Hypothesis 1B. Business unit reorganization (within a given period) will have a U-shape impact on its future innovation (in the next period).

3. Methods and Measures 3.1. Data and Sample I gathered data from several editions of The Medical & Healthcare Marketplace Guide, published in 1975, 1978, 1983, 1986, 1989, and each year thereafter.9 The guides include information concerning U.S. and non-U.S. firms operating in the U.S. medical industry, which can be further divided into the categories of health- care services, medical instruments, and pharmaceuti- cal drugs. For each firm there is information regarding its product markets (see Online Appendix 1, provided in the e-companion)10 and the business units pos- sessed each year. A business unit is a structural com- ponent whose identity is recognized by the firm with a unique address and some responsibility for one or sev- eral product markets. Note that several units of a firm may be involved with the same product market. The guide includes descriptions and histories of firms and their units. The unit-level information includes what product or services a unit offers, how the unit came to be in the firm (whether it was internally developed, acquired, or a new unit that resulted from the recom- bination of several other units), if participation in cer- tain product markets ceased (either by divestiture or dissolution), if there was entry into a new product market (either by acquisition or internal innovation), top personnel, and more. Information regarding the

9 The Medical & Healthcare Marketplace Guide was published by Inter- national Bio-Medical Information Services Inc. (Acton, MA, and Miami, FL; eds. Adeline B. Hale and Arthur B. Hale) in 1975, 1978, 1983, 1986, and 1989. Subsequent editions have been published by MLR Publishing Company (Philadelphia, PA) and by Dorland’s Biomedical Publications (Philadelphia, PA). 10 An electronic companion to this paper is available as part of the on- line version that can be found at http://mansci.journal.informs.org/.

Figure 2 Sample Firms’ Survival in Medical Industry, 1978–1997

250

85 98117132

158

100

150

200

250

300

1975 1980 1985 1990 1995 2000 Year

N um

be r

of e

xi st

in g

fi rm

divestiture or liquidation of a unit is available at the firm-level description, which also states general infor- mation for the entire organization such as nationality, date of incorporation, public or private status, level of revenue, key officials, and more. This paper examines the origin and evolution

(of structure and activities) of all units (both internal and acquired) within firms, and further notes when and where (within which units) innovation occurs. For each unit, I observe whether the company created it internally, obtained it through acquisition, or derived it by recombining several existing units. Unit evolu- tion is tracked during the 1975–1997 period to observe business unit reorganization and to determine which units are innovating. For structural evolution, I trace how all units are created, merged/recombined, split, dissolved, or divested. Similarly, the entry, exit, and continuation in product markets is traced for all units over the study period. The study sample11 consists of 250 firms that exist in

the 1978 panel. Of the 250 firms, only a third (85 firms) were still alive in 1997;12 the remaining firms were either acquired (97 firms) or shut down (68) (Figure 2). The 250 firms in the sample have a total of 866 units and 1,274 product lines over the time period of the study. (Table 1 lists descriptive statistics of the sam- ple.) The sample consists mostly of firms that produce medical devices (72%), have domestic parents (90%), and are not diversified (74%). Less than half the sam- ple is firms that are acquisition active (34%) and have multiple units (44%).

3.2. Variables and Operationalization

3.2.1. Dependent and Independent Variables. The dependent variable for the competing hypotheses is degree of innovation �doi� within a time period. This

11 This sample and its descriptive information (Figures 2 and 4 and Table 1) are the same as those used in the study by Karim (2006). 12 Although firms exit the sample at various times between 1978 and 1997, there is no survivor bias because this study is testing each firms’ innovative ability postreorganization, independent of other firms’ activities or survival. Furthermore, the study controls for firms’ nuances through a fixed-effects model.

doi is a count variable that is calculated as the num- ber of entries (via internal innovation, not acquisition) into new product markets by the firm within a time period. Similarly, the independent variable, degree of reorganization �dor�, counts the number of unit bound- ary changes that occur in a firm within a time period. More specifically, if two units (regardless of their ori- gin) are combined, both are reorganized in some way; thus, this would be counted as two reorganization events. If a unit is split into two newly created units and loses its original identity, this also is counted as two reorganization events. If a unit loses part of itself to another existing unit, both units are changed, and thus this also is counted as two reorganization events. A simple way to imagine the operationaliza- tion is as follows: All of the units of a firm were charted and tracked across time with arrows indicat- ing any unit changes; the degree of reorganization for a time period was calculated by counting the num- ber of arrows within that period. There are no arrows (reorganization events) drawn for a unit acquisition or divestiture at time t however, if an acquired unit is structurally changed in a subsequent time t + x (e.g., recombined with an internal unit or another acquired unit), then arrows were drawn and this was measured as reorganization.

3.2.2. Control Variables. There are several in- dustry- and firm-level controls13 included in the study of testing the relationship between reorganization and radical innovation. At the industry level there is the industry category in which the firm earns the most revenues, and thus operates in chiefly. This is a dummy variable indicating category of phar- maceutical or healthcare services (with comparison being made to the omitted category of medical devices). Due to the rarity and specialization of block- buster drugs, I expect that pharmaceutical firms inno- vate less into new product markets than medical device firms. Furthermore, I expect healthcare ser- vice firms to innovate more compared to medical device firms that need to undergo lengthy processes of FDA approval. Firm-level controls include number of employees, medical sales, percent medical sales (to eval- uate diversified firms), age, public status, and foreign status. I expect that innovation is greater for firms that are large, earning higher revenues and are focused on the medical industry. Furthermore, because innova- tion is measured at the product market level within the United States, I expect more innovations (i.e., entry into new markets) from mature, public, and domestic firms.

13 Dummy controls are described here and included in the corre- lation matrix because they were used to test the appropriateness of various models (random versus fixed effects) and to generate descriptive statistics.

Table 1 Sample Firms’ Descriptive Statistics

Total no. of firms in sample = 250 Total no. of units in sample = 866

Total no. of product markets in sample = 1�274 Summary of 250 firms:

Survive to 1997? 85 survive 165 do not survive (68 shut, 97 acquired) Foreign? 24 foreign parents 226 U.S. parent/firm Industry category? 181 medical devices 38 healthcare services 31 pharmaceuticals Multiunit? 109 multiunit firms 141 single-unit firm Acquisition active? 84 acquisition-active 166 no acq. activity Diversified? 64 diversified 186 only medical

Min/Max statistics for 250 firms over study period, 1975–1997

Min Max Mean

No. of employees 3 230,293 6,503 Total sales ($) 159 thousand 11 billion 386 million Medical sales (%) 1 100 81 Profit margin (%) −102 258 6 Age in 1978 1 (founded 1978) 311 (founded 1668) 24 No. of competitorsa 1 2,258 143 No. of product markets 1 49 3 No. of innovations 0 36 3 No. of unit reorganizations 0 55 2 No. of unique units 1 48 3.5 No. of acquired units 0 33 1.7

market as reported by Medical Guides source.

Additional controls are added that may effect how much innovation occurs within a firm. First, con- sider a firm’s level of acquisition activity. As acqui- sitions are a useful mechanism for obtaining new technologies and entry into new product markets (Granstrand and Sjolander 1990, Capron et al. 1998, Karim and Mitchell 2000), firms may substitute inno- vation with acquisition (Pitts 1977, Constable 1986, Hitt et al. 1990). Hitt et al. (1990) provide a theoretical model predicting that acquisitions lead to increased substitution for innovation due to constraints on resources (being already allocated to acquisitions) and managerial risk aversion (internal development has higher failure rates). Earlier work has also pro- posed that acquisition strategies may serve to avoid risky internal R&D investments (Pitts 1977, Constable 1986). Others have found contrary empirical evidence; Karim and Mitchell (2000) found that firms that were acquisition active produced more new innovations to an industry than did nonacquirers. The variable num- ber of acquired units is the count of the number of units that the firm possesses that were originally acquired. I also control for the ratio of acquired to total units to compare firms that have greater or lesser percent- age of acquired units. Next, consider a firm’s level of product market (PM) activity. The variable num- ber of product markets is the count of the number of unique product markets being managed by the firm. Firms already involved in other product areas will

likely face resource constraints, and I expect that they will innovate less into new product markets. Lastly, the level of innovation may also depend on the firm’s level of complexity. The variable number of total units is the number of units being managed by the firm. As the number of units increases, so may the complexity within the organization and coordination necessary to manage the units.14 Instead of viewing complexity as the total number of units to manage, some schol- ars have also measured it as the relative size of each unit. Eisenhardt and Brown (1999) highlight that unit size matters; size determines agility, and the abilities to focus and exploit growth opportunities. To explore how this may affect innovation, I control for average unit size, which is each firm’s number of employees divided by the total number of units.

3.3. Methodology Hypotheses 1A and 1B predict that the degree of inno- vation by a firm is a function of the degree of unit reor- ganization it has experienced in the past. Because the dependent variable, degree of innovation (the number of product markets entered internally), is a nonneg- ative count variable, a linear model’s assumptions of

14 Research from the structuralist perspective on innovation has found mixed results regarding complexity’s (measured as the num- ber of business units) effect on innovation (Edwards 2000, Blau and McKinley 1979, Damanpour 1996); thus, it is not clear if this vari- able will positively or negatively affect degree of innovation.

normally distributed errors and homoskedasticity (where variance of the residuals is constant) are violated, producing inefficient, inconsistent, and biased coefficient estimates. For count models, one should use either a Poisson or negative binomial regression model15 (Long 1997). After running a Poisson model, one can perform a goodness-of-fit test of the model to determine if the Poisson process is inappropriate and if a negative binomial model should be used instead.16

To capture the effect that reorganization has on future innovation, time-varying lagged values of independent variables are used as regressors in the estimated equation (Judge et al. 1988, Ahuja and Katila 2001). This technique is appropriate for panel data in which one can observe the dependent and independent variables changing over time for differ- ent groups. In this study, a group is a firm (j), X rep- resents control variables, and the lagged independent variable is the degree of reorganization �dor�. Thus, a linear lagged model resembles:

degree of innov�j� t�

= exp(B0 + B1dor�j� t − 1� + B2dor�j� t − 2� + B3X3�t − 1� + · · · + BnXn�t − 1�)� (1)

Note that the model expressed in Equation (1) includes two lags of the independent variable; this allows me to explore two types of experience. The hypotheses of the paper explore how experiencing multiple events within some given period (t − 1) affects a future out- come (Haleblian and Finkelstein 1999), in this case, innovation (at time t). To explore whether past expe- rience (t − 2) affects innovation, I include a second lag. Thus, innovation during period t is tested as a

15 A Poisson maximum-likelihood regression model (Long 1997) assumes that the incidence of an event is distributed following a Poisson process, where the mean of this process is determined by the regressors, and the mean and variance of the distribution are equal. Changes in the independent variables affect the frequency at which an event occurs by altering the mean of the Poisson distri- bution. A Poisson model estimates the probability of an observed event, conditional on an expected mean �. In such a model, the expected number of events (in this case being the number of inno- vations) is calculated as an exponential function. A model’s signifi- cance can be tested using the model log-likelihood chi-square, and coefficient estimates can be tested for significance by performing a test of the Wald statistic. 16 Most count variables exhibit overdispersion, and a negative bino- mial model is more appropriate than a Poisson. In a negative binomial count model, the count variable is generated by a Poisson- like process except that there is greater dispersion (the variance is greater than the mean (�) of the distribution). The negative bino- mial model replaces the constant mean (�) with a variable mean that is represented as the mean with some level of error (��). The probability of the observed count then becomes conditional on the distribution of the error, which is usually a gamma distribution.

function of reorganization during periods t − 2 and t − 1, and controls for period t − 1. Because the study is comprised of seven panels (1975, 1978, 1983, 1986, 1989, 1993, and 1997) with six periods of data observed (period 1 = 1975–1978, period 2 = 1978–1983, etc � � � ), there are at most four observations per firm due to the double-lagged approach. Incorporating a two- period lag, the following four equations (at most) are observed for each firm j:

degree of innov�j� 1983–1986�

= f (dor�j� 1975–1978�� dor�j� 1978–1983�� controls�j� 1978–1983�

) � (2)

degree of innov�j� 1986–1989�

= f (dor�j� 1978–1983�� dor�j� 1983–1986�� controls�j� 1983–1986�

) � (3)

degree of innov�j� 1989–1993�

= f (dor�j� 1983–1986�� dor�j� 1986–1989�� controls�j� 1986–1989�

) � (4)

degree of innov�j� 1993–1997�

= f (dor�j� 1986–1989�� dor�j� 1989–1993�� controls�j� 1989–1993�

) � (5)

The four observations above are included in the study if the firm in question survives to the end of the study. For firms that do not survive, as many observations as possible are incorporated, as is typical with panel data (Ahuja and Katila 2001). Because the competing hypotheses predict a

U-shape or inverted U-shape relationship between degree of reorganization and future innovation, the regression model represents a quadratic model where the degree of reorganization is included as a linear term and a squared term. Thus, the complete model tested resembles

doi�j� t� = exp B0 + B1�dor�j� t − 1�� + B2�dor�j� t − 2�� + B3�dor�j� t − 1��2 + B4�dor�j� t − 2��2 + B5X5�t − 1� + · · · + BnXn�t − 1� � (6)

where j denotes a firm and t denotes a time period between two panels.

4. Results and Discussion Of the 1,274 product markets that firms were active in during the study period, 815 (64%) were innova- tions, 443 (35%) were acquisitions, and 16 (1%) were via joint ventures (Figure 3). The sample firms had a total of 866 units, of which 370 (43%) were internally created, 413 (48%) were acquired, 63 (7%) came into being as merged recombinations of existing units, and 20 (2%) came into being as joint ventures (Figure 4).

Figure 3 Product Market Origins

150

300

450

600

750

900

16 443 815

Acquired Internal

64%

35% 1%

Joint venture

Total PMs (1978–1997) innovations

Most of the units faced reorganization at least once (namely 13% of internally created units, 53% of acqui- sitions, and 36% of units that originated as recombi- nations), and many were reorganized several times. To test reorganization’s effect on future radical in-

novation, I first ran a Poisson regression model and then discarded it in favor of a negative binomial regression model based on a goodness-of-fit test.17

Although the sample began with 250 firms, this was reduced to 132 surviving firms in 1986. This reduc- tion was necessary to keep a minimum of two lag periods for the independent variables, and resulted in 434 observations. The sample was further reduced to 83 firms and 306 observations due to 17 firms hav- ing only one observation and 32 firms having all zero outcomes.18

A glance at the correlation matrix (Table 2) reveals that there is a high correlation between the num- ber of total units and acquired units; however, tests for multicollinearity do not indicate this to be prob- lematic.19 The matrix also informs us that firms with

17 A goodness-of-fit test on the Poisson model indicated that the innovation process exhibited greater variance (overdispersion) and was appropriate for study using negative binomial regression. By performing a Hausman specicification test comparing the fixed- effects and random-effects model, I was able to reject the random- effects model (that dispersion may vary across the firms because of unidentified firm-specific reasons). 18 For the fixed-effects models, STATA set a minimum of two obser- vations per group (i.e., firm) and dropped groups with zero inno- vations across all time periods. This model assumes independence of observations between firms, and adjusts for correlations within firm observations over time (similar to robust standard errors). 19 Due to the high Pearson correlation between “number of acquired units” and “number of total units,” I performed various tests of multicollinearity (where correlated predictors may inflate standard errors) by calculating variance inflation factors (VIFs) and condi- tion indices for predictors of each model. The mean VIFs for the models are not especially large (below 6) (Belsley 1991), and none of the individual VIFs or condition indices exceeded 30; Belsley et al. (1980) suggest that 30 or higher indicate collinearity problems. To test for robustness of the main independent variables beyond Model 7, I include Models 8 and 9, which do not have number of acquired and total units in the same model; in both of these models the main independent variable’s estimated coefficients’ sig- nificance and magnitude remain unchanged. Because Model 8 had the lowest mean VIF, I continue with these predictors when testing Models 10–12.

Figure 4 Business Unit Origins and Change

Units by source 370 413 63 20

13 53 36 59

2 16 21 12

24 9 27 14

Acquired

48%

43%

created

Entry as reconf.

Joint venture

Units by source

% continue w/o reconf.% divest w/o reconf.

% reconfigured

100

150

200

250

300

350

400

450

10%

20%

30%

40%

50%

60%

70%

Reconfigured (%)

Divest w/o reconf. (%)

Continue w/o reconf. (%)

Internally

many units (or many acquired units) are involved in more product markets and have a large percentage of acquired units as compared to internal units. Further, more product markets correlates with more reorga- nizations, and larger firms have larger average unit size. Tables 3 and 4 report the models’ results of the

reorganization-innovation analysis. Model 1 includes only the control variables of firm-level attributes. Models 2, 3, and 4 iteratively add controls for firms’ acquisition activity, product market activity, and com- plexity respectively. Model 5 adds the linear term of the independent variables, namely, degree of unit reorganization lagged one time period; and Model 6 adds the square term to test the quadratic relationship. Model 7 adds the second lagged time period variables for reorganization and its quadratic term. Models 8 and 9 explore the effects of controlling for ratio of acquired to total units and average unit size. Models 10 and 11 exclude various outliers from the full sam- ple to test for model sensitivity.20 Finally, Model 12 adds a control for the level of profitability, net income,

20 To test model sensitivity against influential data points, I tested the sample for data points that had (A) high leverage or (B) large residuals. (A) An observation, (x� y), in which the x-value is far from the center of mass of the other xs is a potentially influen- tial data point; to determine such points’ leverage, one needs to estimate the regression model excluding each observation one at a time to record significant changes. I found five potential obser- vations with x-values moderately removed from the others, and reran Model 8 excluding each observation one at a time; I found that only one of the five observations changed the estimated results

to further test the reorganization-innovation relation- ship; however, the sample is reduced because of miss- ing data. Model 1 is not significant, Model 12 is barely significant, and all other models are significant. The models are highly robust because the magnitude, direction, and significance of each estimated coeffi- cients remain fairly consistent. The results in Table 4 support Hypothesis 1B (the

independent variable is negative and its squared term is positive) and do not support Hypothesis 1A. This suggests that reorganization and innovation have a U-shape relationship;21 firms benefit from reorgani- zation only after experiencing a certain number of events.22 For the sample studied, the results estimate a U-shape where the inflection point is at eight23 reor- ganization events (see Figure 524). This is interesting because most firms in the sample experience fewer than eight reorganizations, implying that their unit reorganizations will not result in radical innovative benefits. However, those firms that were highly reor- ganization active (beyond the turning point) had the opportunity to learn from enough events to boost innovativeness and new product market entry. This result implies that the process of learning requires a certain number of trials within a period (in this case within a three- to four-year window); these trial events allow firms to compare and contrast differ- ent reorganization scenarios, to discriminate, unlearn, and experiment as they overcome path dependence.

by reducing the significance of the square (quadratic) term—this one observation is excluded from Model 10. To detect outliers (i.e., large residuals), I plotted a residual-versus-fitted plot and found three potential outlier points (one of which was also the high leverage observation from Model 10); these three observations were excluded in Model 11. The results in Model 11 are similar to Model 10. 21 More-stringent econometric tests were applied to confirm a U-shape relationship following the suggestions of Lind and Mehlum (2007); the results indicate that the relationship is a U-shape (see Online Appendix 2, provided in the e-companion). 22 To avoid bias from those firms that never participate in reor- ganization, I reran Model 8 with only reorganization-active firms (143 observations) and find that there is still a U-shape relationship between reorganization and innovation. 23 Interestingly, in their study of acquisition experience’s effect on acquisition performance, Haleblian and Finkelstein (1999) found a U-shape relationship with an inflection point as eight acquisition events. Models 10 and 11, excluding various outliers, reduce the significance of the square term slightly, but estimates still result in an inflection point of eight events. 24 Figure 5 depicts that an increase in reorganizations from 0 to 8 results in a decrease in innovative activity from 2 to 1, and further reorganizations from 8 to 16 result in a subsequent increase from 1 to 2. Although these magnitudes may seem small, the significance of one innovative market entry is evident when you consider that firms only enter three markets, on average, in the data. Thus, a change in the degree of reorganization can have a strong impact on a firm’s potential market activities. Ta bl e 2

Pe ar so n Pa

ir w is e Co

rr el at io n M at ri x 1

2 3

4 5

6 7

8 9

10 11

12 13

14 15

1. D eg re e of

in no

va tio

n (t )

1� 00

00 2.

D eg re e of

re or ga ni za tio

n (t

− 1)

0� 12

26 1� 00

00 3.

D eg re e of

re or ga ni za tio

n (t

− 2)

0� 14

41 0� 54

25 1� 00

00 4.

Ca te go

ry :P

ha rm

ac eu tic al

0� 04

90 0� 31

11 0� 31

62 1� 00

00 5.

Ca te go

ry :H

ea lth

ca re

se rv ic es

0� 08

80 0� 06

96 0� 05

39 −0

�1 37

0 1� 00

00 6.

N um

be r of

em pl oy ee s (t

− 1)

0� 14

09 0� 34

41 0� 29

45 0� 22

23 0� 11

35 1� 00

00 7.

M ed ic al sa le s (t

− 1)

0� 18

35 0� 45

17 0� 29

40 0� 49

73 0� 04

93 0� 38

38 1� 00

00 8.

Pe rc en tm

ed ic al sa le s (t

− 1)

−0 �0 73

6 −0

�2 39

2 −0

�2 85

0 −0

�1 26

0 −0

�0 35

5 −0

�4 46

1 −0

�0 66

1 1� 00

00 9.

Ag e (t

− 1)

0� 21

35 0� 28

21 0� 30

21 0� 39

76 −0

�1 52

7 0� 29

16 0� 28

03 −0

�3 58

2 1� 00

00 10

. Pu

bl ic st at us

0� 01

75 0� 20

67 0� 22

49 0� 37

48 0� 17

05 0� 21

48 0� 32

37 −0

�3 71

7 0� 15

35 1� 00

00 11

. Fo re ig n st at us

0� 14

90 0� 18

21 0� 17

12 0� 31

10 −0

�1 32

4 0� 37

50 0� 10

30 −0

�2 45

5 0� 48

67 0� 05

90 1� 00

00 12

. N um

be r of

ac qu

ire d un

its (t

− 1)

0� 32

05 0� 55

43 0� 41

90 0� 38

70 0� 02

62 0� 37

40 0� 45

41 −0

�3 33

7 0� 40

48 0� 20

91 0� 27

21 1� 00

00 13

. R at io of

ac qu

ire d to

to ta lu ni ts (t

− 1)

0� 21

75 0� 37

33 0� 30

14 0� 27

57 0� 07

58 0� 27

49 0� 37

06 −0

�3 69

5 0� 33

91 0� 32

61 0� 23

15 0� 75

97 1� 00

00 14

. N um

be r of

pr od

uc tm

ar ke ts (t

− 1)

0� 26

72 0� 73

93 0� 61

02 0� 40

04 −0

�0 01

2 0� 37

85 0� 57

59 −0

�2 45

8 0� 37

40 0� 23

13 0 � 19

46 0� 75

05 0� 50

42 1� 00

00 15

. N um

be r of

to ta lu ni ts (t

− 1)

0� 34

21 0� 60

64 0� 49

48 0� 45

82 0� 08

04 0� 43

76 0� 53

53 −0

�3 46

9 0� 50

13 0� 24

17 0� 37

62 0� 93

96 0� 69

56 0� 79

78 1� 00

00 16

. Av er ag e un

it si ze

(t − 1)

−0 �0 16

5 0� 02

70 0� 02

54 0� 00

74 0� 01

35 0� 77

27 0� 07

26 −0

�2 98

4 0� 13

27 0� 17

74 0� 26

74 0� 00

88 0� 01

68 0� 00

13 0� 01

77 1� 00

N ot e. Se

e Fo ot no

te 19

fo r te st s of

m ul tic ol lin ea rit y.

Table 3 Reorganization’s Effect on Innovation, Models 1–6 (Conditional Fixed-Effects Negative Binomial Regression)

Model 1 Model 2 Model 3 Model 4 Model 5 Model 6

Variable Full sample Full sample Full sample Full sample Full sample Full sample

Degree of unit reorganization (t − 1) −0�033 −0�172∗ Square (dor (t − 1)) 0�008† Degree of unit reorganization (t − 2) Square (dor (t − 2)) Firm-level controls

Number of employees (t − 1) 0�000∗ 0�000∗ 0�000∗ 0�000∗ 0�000† 0�000∗ Medical sales (t − 1) −0�000 −0�000 0�000 0�000 0�000 0�000 Percent medical sales (t − 1) 1�619† 1�798∗ 1�960∗ 1�894∗ 1�862∗ 1�954∗ Age (t − 1) −0�008 −0�006 −0�005 −0�006 −0�006 −0�005

Level of acquisition activity Number of acquired units (t − 1) 0�166∗∗ 0�301∗∗ 0�429∗ 0�451∗ 0�335†

Level of PM activity Number of product markets (t − 1) −0�088∗∗ −0�086∗ −0�075∗ −0�094∗

Level of complexity Number of total units (t − 1) −0�129 −0�154 −0�042

Wald �2 10�35 17�37 21�16 21�63 23�15 27�28 Prob > �2 0�241 0�043 0�020 0�027 0�027 0�011

Notes. Full sample number of observations = 306, number of groups = 83. Observations per group: min. 2, avg. 3.7, max. 4. †p < 0�1; ∗p < 0�05; ∗∗p < 0�01; ∗∗∗p < 0�001.

Perhaps with greater events there is greater hetero- geneity from which to learn; I do not control for dif- ferent characteristics of the reorganizations, however, intuitively we expect that with increased number of trials that there will be some new experiences that each firm faces. The importance of multiple events also highlights that it is not enough to simply bring together resources (an issue of content); instead, firms should learn (an issue of process) from a cohort of reorganization events about what combinations were successful given the circumstances and characteris- tics of each reorganization. For managers, the impli- cation of a U-shape relationship is that firms with limited opportunities (due to time or resources) for change should be cautious when pursuing reorgani- zation because it may not be beneficial for radical innovation.25

I included the double-period lagged variables to explore whether past experience, beyond the cohort of events within a current period, influenced the reorganization-innovation relationship. Do reorgani- zations from three to four years prior also affect future innovation? Models 9–12 indicate that only the single- period lagged independent variables are significant, and the double-period lags are insignificant. In other

25 On the other extreme, it is not realistic to assume that infinite reorganization events within a three- to four-year period will result in infinite innovations. Because the average number of reorgani- zations in the sample was two, it is clear that it is the minority of firms that reorganized past the inflection point. The logic from Hypothesis 1A of too many changes being disruptive (more for some than others) and reorganization requiring resources and time still applies as constraints to successful change.

words, although a cohort of experiences within a cur- rent three- to four-year period may affect future inno- vation, past experiences from three to four years prior do not have a significant effect. There are two possible explanations for this finding: it could be the case that firms’ do not have the capacity within their organi- zational memory to preserve what they learned from past events, or the case that past trials mirror cur- rent trials and provide no new insights. Given that events (whether current or past) usually have dif- fering scenarios (i.e., market conditions, size of the firm, involvement in product markets, etc� � � � change over time), I doubt it is the latter explanation. More likely, this lack of significance is due to some capacity constraint. Several of the control variables are also signifi-

cant. Larger firms (with greater number of employ- ees) and more profitable firms (with greater net income) radically innovate more than smaller, less- profitable firms. This may be indicative of these firms’ resource bounty. Consistent with this logic is the find- ing that less-diversified firms with a greater per- centage of medical sales also innovate more in the medical industry than more diversified firms. This is not surprising because less-diversified firms are able to concentrate their resources mainly in the medi- cal sector. Examining further controls reveals that a greater number of total units enhances innovation, but the average size of each unit does not have an impact. This implies that the number of modular units has a larger effect on innovation than the size of each unit; perhaps this is because greater modularity allows for more reorganization possibilities (Karim

Table 4 Reorganization’s Effect on Innovation, Models 7–12 (Conditional Fixed-Effects Negative Binomial Regression)

Model 7 Model 8 Model 9 Model 10 Model 11 Model 12

Variable Full sample Full sample Full sample Partial samplea Partial sampleb Partial samplec

Degree of unit reorganization (t − 1) −0�240∗ −0�283∗∗ −0�237∗ −0�248∗ −0�255∗ −0�363∗∗ Square (dor (t − 1)) 0�015∗∗ 0�017∗∗ 0�016∗∗ 0�014† 0�014† 0�019∗ Degree of unit reorganization (t − 2) 0�202 0�071 0�164 0�048 0�039 0�154 Square (dor (t − 2)) −0�031 −0�018 −0�029 −0�010 −0�008 −0�021 Firm-level controls

Number of employees (t − 1) 0�000∗ 0�000∗ 0�000 0�000† −0�000 0�000∗ Medical sales (t − 1) 0�000 −0�000 0�000 0�000 0�000 −0�000∗ Percent medical sales (t − 1) 2�063∗ 2�010∗ 1�849† 1�983∗ 1�511∗ 5�817∗∗ Age (t − 1) −0�007 −0�004 −0�007 −0�004 0�009 −0�002

Level of acquisition activity Number of acquired units (t − 1) 0�449∗ 0�493∗∗∗ Ratio of acquired to total units (t − 1) −0�538 −1�771 −0�479 −0�327 −2�289

Level of PM activity Number of product markets (t − 1) −0�115∗ −0�100∗ −0�131∗∗ −0�115∗ −0�169† −0�104∗

Level of complexity Number of total units (t − 1) −0�105 0�288∗∗ 0�279∗∗ 0�319∗∗ 0�240∗ Average unit size (t − 1) 0�000

Level of profitability Net income (t − 1) 0�000∗∗

Wald �2 28�30 26�22 30�18 25�69 25�57 24�33 Prob > �2 0�020 0�036 0�017 0�041 0�043 0�083

Notes. Full sample number of observations = 306, number of groups = 83. Observations per group: min. 2, avg. 3.7, max. 4. aModel 10 used the full sample excluding one outlier data point that had high leverage. bModel 11 used the full sample excluding three outlier data points that had large residuals. cModel 12 used the full sample, but because of missing data on NI was cut to number of observations = 137 and number of groups = 38. †p < 0�1; ∗p < 0�05; ∗∗p < 0�01; ∗∗∗p < 0�001.

2006).26 Furthermore, firms’ levels of product mar- ket activity and acquisition activity were significant. As expected, firms already involved in many product markets innovate less in the following period. Firms that are spreading their resources too thin are either unable to use their limited resources towards further radical innovations or are choosing to not pursue fur- ther markets because they are already overcommitted. Regardless of which scenario is driving this result, they are both consistent with the idea of firms facing resource constraints. Finally, the models estimate that acquisition-active firms radically innovate more than firms with fewer acquired units (and finds no signifi- cance in the percentage of acquired units). This refutes the theory that firms use acquisitions to replace inno- vation (Hitt et al. 1990). On the contrary, it reinforces earlier findings that acquisition-active firms possess more innovations within an industry than do nonac- quirers (Karim and Mitchell 2000).

26 For example, if Firm A has 10,000 employees in one unit, Firm B has 10,000 employees between two units, and Firm C has 20,000 employees between two units, then the models predict that Firms B and C innovate more than Firm A. However, the limitation of the unit-size variable should be noted—it is not a true average of distinct unit sizes, but instead an approximated average size calcu- lated by the ratio of total firm size to number of total units.

This last finding prompted me to explore where innovation is occurring in firms. If acquisition-active firms innovate more than less acquisition-active firms (as found here) and more than nonacquirers (Karim and Mitchell 2000), is it the case that these acquired (target) units are the ones innovating? Of the 815 innovations from all 250 firms, the majority (661 prod- uct markets) are created in internally created units (Figure 6). Also, note that recombinations of internal and acquired units innovate more (62 product mar- kets) than either acquired units alone (31) or recombi- nations of acquired units (47). Forming conclusions on the full sample of 250 firms would be biased because the sample includes firms that are nonacquirers and thus can only innovate in internally created units or combinations of these units. It is more appropriate to look at the subsample of firms that are acquisi- tion active; these firms have the potential to innovate in all kinds of units. Taking this abridged sample of 83 firms (also in Figure 6) alters the number of inno- vations originating in internally created units to 239 and those from a mix of internal units to 9. A proper, unbiased t-test of the origin of innovation should con- trol for the number of each type of unit (e.g., there are few internal units that are recombined together com- pared to other recombinations; Karim 2006). Figure 7 depicts the average number of innovations from the

Figure 5 Reorganization and Innovation (Based on Model 8 Estimates)

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29

Degree of reorganization

D eg

re e

of i

nn ov

at io

Figure 6 Origin of Innovations

239

9 62

31 47

661

14 62

31 47

100

200

300

400

500

600

700

239 9 62 31 47

All 250 firms 661 14 62 31 47

Int Int + Acq Acq

N um

be r

of i

nn ov

at io

Int + Int Acq + Acq

Unit origin

83 firms w/ int. and acq. units

83 firms when controlling for the number of units of each type. There is a greater average number of radical innovations from units containing some internal portion (1.35, 0.43, 0.45 innovations, respec- tively, for I, II, IA units) versus those from units consisting of only acquisitions (0.07 and 0.22 inno- vations, respectively, for A and AA units).27 This is consistent with earlier acquisition studies that found that bilateral resource redeployment between target and acquirer (i.e., between acquired units and inter- nal units) has the highest impact on the merged firm’s performance as compared to unilateral deployment (Capron and Mitchell 1998). These observations indi- cate that acquisition-active firms still rely on their internally created units to innovate; acquisitions may provide key ingredients for internal units and, in this

27 A paired t-test of the number of total innovations from I, II, and IA units compared to the number of total innovations from A and AA units for each firm after controlling for the number of each type of unit possessed finds the difference to be significant (p < 0�001).

Figure 7 Average Number of Innovations by Origin for 83 Firms

0.43 0.45

0.07

0.22

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.35 0.43 0.45 0.07 0.22

Int Int + Int Int + Acq

A ve

ra ge

n um

be r

of i

nn ov

at io

Unit origin

Average number of innov. per unit

Acq Acq + Acq

1.35

way, best serve the firm. From a learning perspective, this finding would support the logic that innovation and learning are enhanced when a unit contains some internally developed portion consisting of the firm’s familiar routines and resources.

5. Conclusion The goal of this paper was to empirically study busi- ness unit reorganization’s affect on innovative perfor- mance, and to explore the learning mechanism that may affect this relationship. There are several conclu- sions. Reorganization enhances innovation only after eight events are experienced within a three- to four- year period, exhibiting a U-shape phenomenon. This study reinforces Haleblian and Finkelstein’s (1999) findings that firms need several events from which to discriminate, unlearn, and experiment before there are positive outcomes. Thus, the findings here sup- port an organizational learning perspective of how the learning process mediates the reorganization- innovation relationship. Furthermore, studying the double-lagged regressors of unit reorganization reveal that firms have capacity constraints on their organi- zational memory because past reorganization experi- ence from a prior period does not have an impact on future innovation; only reorganization events within a current period are significant. As expected, resource availability and constraints may also affect innova- tiveness. Larger, more-focused, and profitable firms radically innovate more in the medical industry than their counterparts. Also, firms active in fewer prod- uct markets innovate more into future new markets. Although it is not distinguishable if these results are driven by firms’ strategic choice, they may also be the result of resource-based decisions. Finally, this paper supports previous literature that

highlight the role of acquisitions as mechanisms for

obtaining new resources (Capron et al. 1998, Karim and Mitchell 2000). As acquisition-active firms inno- vate more than less acquisition-active firms, these “new resources” are not only those that come bundled with the target, but also firms’ potential for future innovation that are developed due to these acqui- sitions. Investigation into the origin of innovations reveals that, on average, it is not acquisitions alone, but instead the units with some internal portion, that innovate the most. Acquisitions that are recombined with internal units contribute a significant amount of innovations, and some innovations also do emerge from acquired units. These observations strikingly reinforce that, even for acquisition-active firms, inter- nal units are still key to the organization. There are several important managerial implica-

tions of this study because business unit reorga- nization has become more prevalent through the years (Karim 2006). First, managers should distin- guish between reorganization events and be wary not to generalize. They should realize that reorganization may only be fruitful if the organization is able to learn from a cohort of several experiences; thus, reorgani- zation may not be the best innovation strategy for organizations with limited time and resources. Fur- thermore, managers should try to increase the capac- ity of organizational memory to make use of past experiences by codifying what was learned. Finally, I encourage managers to be acquisition-active, but not to neglect their internal units as they are the main source of innovation. This study has made an initial exploration into how

learning occurs within the phenomenon of reorgani- zation; there is still vast potential for future study. An obvious limitation of this paper is the ability to observe radical innovation only at the product market level. A fruitful study would be to investi- gate more granular, incremental innovations within product markets and to compare whether the find- ings are consistent with the findings in this paper. It would also be useful to add a control indicating level of relatedness between acquired activities and the activities of firms, and to determine if this affects innovativeness. Furthermore, I was only able to docu- ment reorganization and innovation in three- to four- year periods, having to aggregate a cohort of multiple events into this time frame. A more accurate timing of events at yearly markers would inform us about the value (or not) of having events simultaneously or more frequently. A key issue that is not addressed in this study is that of intent: How much is the lack of market entry attributable to the lack of learning versus rational reasons for not wanting to invest in new markets? Further market controls, interviews, and “real-time” observations of firms undergoing reorganization and innovation will undoubtedly be

informative and shed light on the observations that are made here. My goal was to observe the intricate changes of

organizational structure over a long period, explore its effects on innovation, and inform the dynamic capabilities and organizational learning literatures. It was through the simultaneous tracking of business unit evolution and product market development that allowed me to empirically test competing hypotheses regarding the relationship between reorganization and innovation. I hope the findings in this paper shed new insights on structural evolution and its consequences, and contribute to our understanding of the broader concepts of how firms learn and effectively change.

6. Electronic Companion An electronic companion to this paper is available as part of the online version that can be found at http:// mansci.journal.informs.org/.

Acknowledgments The author thanks Jay Anand, Nick Argyres, Tim Carroll, Jo Thori Lind, Anita McGahan, Halvor Mehlum, Will Mitchell, Anne Parmigiani, Margie Peteraf, M. B. Sarkar, Mario Schijven, and Scott Turner for their detailed suggestions on this paper. The author also thanks the 2008 participants of the Wharton Technology and Innovation conference, DRUID Fundamental on Strategy and Organization, sem- inars at London Business School, University of Maryland, College Park, University of Texas at Austin, and University of Washington, Seattle, all for their thoughtful comments. Finally, the author thanks Pankaj Ghemawat, the associate editor, and the two anonymous reviewers for their invalu- able guidance through the development of this project.

References Abernathy, W. J., K. Clark. 1985. Innovation: Mapping the winds of

creative destruction. Res. Policy 14 3–22. Ahuja, G., R. Katila. 2001. Technological acquisitions and the inno-

vation performance of acquiring firms: A longitudinal study. Strategic Management J. 22 197–220.

Ansoff, H. I. 1965. Corporate Strategy. McGraw Hill, New York. Argote, L. 1996. Organizational learning curves: Persistence, trans-

fer and turnover. Internat. J. Tech. Management 11(7, 8) 759–769. Argyris, C., D. A. Schön. 1978. Organizational Learning: A Theory of

Action Perspective. Addison-Wesley, Reading, MA. Arrow, K. 1962. The implications of learning by doing. Rev. Econom.

Stud. 29 155–173. Baldwin, C. Y., K. B. Clark. 2000. Design Rules: The Power of Modu-

larity. MIT Press, Cambridge, MA. Barkema, H. G., M. Schijven. 2008. Towards unlocking the full

potential of acquisitions: The role of organizational restructur- ing. Acad. Management J. 51(4) 696–722.

Belsley, D. A. 1991. Conditioning Diagnostics, Collinearity and Weak Data in Regression. Wiley, New York.

Belsey, D. A., E. Kuh, R. Welsch. 1980. Regression Diagnostics. Wiley, New York.

Birkinshaw, J., M. Lingblad. 2005. Intrafirm competition and charter evolution in the multibusiness firm. Organ. Sci. 16(6) 674–686.

Blau, J., W. McKinley. 1979. Ideas, complexity, and innovation. Admin. Sci. Quart. 24(2) 200–219.

Bowman, E. H., H. Singh. 1990. Corporate restructuring: Trends and consequences. R. Rock, ed. Corporate Restructuring. McGraw-Hill, New York, 8–22.

Brickley, J. A., L. D. Van Drunen. 1990. Internal corporate restruc- turing: An empirical analysis. J. Accounting Econom. 12(1–3) 251–280.

Brown, S. L., K. M. Eisenhardt. 1997. The art of continuous change: Linking complexity theory and time-paced evolution in relent- lessly shifting organizations. Admin. Sci. Quart. 42(1) 1–34.

Capron, L., W. Mitchell. 1998. Bilateral resource redeployment and capabilities improvement following horizontal acquisitions. Indust. Corporate Change 7(3) 453–483.

Capron, L., P. Dussauge, W. Mitchell. 1998. Resource redeploy- ment following horizontal acquisitions in Europe and North America, 1988–1992. Strategic Management J. 19(7) 631–661.

Carroll, T., S. Karim. 2009. Predicting organizational reconfigura- tion. A. Bøllingtoft, D. Døjbak Håkonsson, J. F. Nielsen, C. C. Snow, J. Ulhøi, eds. New Approaches to Organization Design: The- ory and Practice of Adaptive Enterprises. Springer, New York.

Chandler, A. 1962. Strategy and Structure: Chapters in the History of the American Industrial Enterprise. MIT Press, Cambridge, MA.

Ciborra, C. U. 1996. The platform organization: Recombining strate- gies, structures, and surprises. Organ. Sci. 7(2) 103–118.

Cohen, W., D. A. Levinthal. 1990. Absorptive capacity: A new per- spective on learning and innovation. Admin. Sci. Quart. 35(1) 128–152.

Constable, J. 1986. Diversification as a factor in U.K. industrial strategy. Long Range Planning 19 52–60.

Cyert, R. M., J. G. March. 1963. A Behavioral Theory of the Firm. Prentice-Hall, Englewood Cliffs, NJ.

Daft, R. L., K. E. Weick. 1984. Toward a model of organizations as interpretation systems. Acad. Management Rev. 9(2) 284–295.

Damanpour, F. 1996. Organizational complexity and innovation: Developing and testing multiple contingency models. Manage- ment Sci. 42(5) 693–716.

Dodgson, M. 1993. Organizational learning: A review of some lit- eratures. Organ. Stud. 14(3) 375–394.

Edwards, T. 2000. Innovation and organizational change: Develop- ments towards an interactive process perspective. Tech. Anal. Strategic Management 12(4) 445–464.

Eisenhardt, K. M., S. L. Brown. 1999. Patching: Restitching business portfolios in dynamic markets. Harvard Bus. Rev. 77(3) 71–82.

Eisenhardt, K. M., J. A. Martin. 2000. Dynamic capabilities: What are they? Strategic Management J. 21 1105–1121.

Eisenhardt, K. M., B. N. Tabrizi. 1995. Accelerating adaptive pro- cesses: Product innovation in the global computer industry. Admin. Sci. Quart. 40(1) 84–110.

Figueroa, E., P. Conceicao. 2000. Rethinking the innovation process in large organizations: A case study of 3M. J. Engrg. Tech. Man- agement 17(1) 93–109.

Fiol, C., M. Lyles. 1985. Organizational learning. Acad. Management Rev. 10(4) 803–813.

Galunic, D. C., K. M. Eisenhardt. 1996. The evolution of intracor- porate domains: Divisional charter losses in high-technology, multidivisional corporations. Organ. Sci. 7(3) 255–282.

Galunic, D. C., K. M. Eisenhardt. 2001. Architectural innovation and modular coporate forms. Acad. Management J. 44(6) 1229–1249.

Galunic, D. C., S. Rodan. 1998. Resource recombinations in the firm: Knowledge structures and the potential for Schumpete- rian innovation. Strategic Management J. 19(12) 1193–1201.

Garud, R., A. Kumaraswamy. 1995. Technological and organiza- tional designs for realizing economies of substitution. Strategic Management J. 16(Summer Special Issue) 93–109.

Gleick, J. 1987. Chaos: The Making of a New Science. Heinemann, London.

Granstrand, O., S. Sjolander. 1990. The acquisition of technology and small firms by large firms. J. Econom. Behav. Organ. 13 367–386.

Haleblian, J., S. Finkelstein. 1999. The influence of organizational acquisition experience on acquisition performance: A behav- ioral learning perspective. Admin. Sci. Quart. 44(1) 29–56.

Haunschild, P. R., B. N. Sullivan. 2002. Learning from complexity: Effects of prior accidents and incidents on airlines’ learning. Admin. Sci. Quart. 47(4) 609–643.

Hayward, M. L. A. 2002. When do firms learn from their acquisition experience? Evidence from 1990–1995. Strategic Management J. 23(1) 21–39.

Hedberg, B. 1981. How organizations learn and unlearn. P. Nystrom, W. Starbuck, eds. Handbook of Organizational Design, Vol. 1. Oxford University Press, Oxford, UK, 3–27.

Helfat, C. E., K. M. Eisenhardt. 2004. Inter-temporal economies of scope, organizational modularity, and the dynamics of diversi- fication. Strategic Management J. 25(13) 1217–1232.

Henderson, R. M., K. B. Clark. 1990. Architectural innovation: The reconfiguration of existing product technologies and the failure of established firms. Admin. Sci. Quart. 35(1) 9–30.

Hitt, M. A., R. E. Hoskisson, R. D. Ireland. 1990. Mergers and acqui- sitions and managerial commitment to innovation in M-form firms. Strategic Management J. 11(Summer) 29–47.

Hoskisson, R. E. 1987. Multidivisional structure and performance: The contingency of diversification strategy. Acad. Management J. 30(4) 625–644.

Hoskisson, R. E., C. S. Galbraith. 1985. The effect of quantum versus incremental M-form reorganization on performance: A time- series exploration of intervention dynamics. J. Management 11(3) 55–70.

Hoskisson, R. O., R. A. Johnson. 1992. Corporate restructuring and strategic change: The effect on diversification strategy and R&D intensity. Strategic Management J. 13(8) 625–634.

Huber, G. P. 1991. Organizational learning: The contributing pro- cesses and the literatures. Organ. Sci. 2(1) 88–115.

Johnson, R. 1996. Antecedents and outcomes of corporate refocusing. J. Management 22(3) 439–483.

Judge, G. G., R. C. Hill, W. E. Griffiths, H. Lutkepol. 1988. Introduc- tion to the Theory and Practice of Econometrics. Wiley, New York.

Karim, S. 2006. Modularity in organizational structure: The recon- figuration of internally developed and acquired business units. Strategic Management J. 27(9) 799–823.

Karim, S., W. Mitchell. 2000. Path-dependent and path-breaking change: Reorganizing business resources following acquisi- tions in the U.S. medical sector, 1978–1995. Strategic Manage- ment J. 21(10–11) 1061–1081.

Karim, S., W. Mitchell. 2004. Innovating through acquisition and internal development: A quarter-century of boundary evolu- tion at Johnson & Johnson. Long Range Planning 37(6) 525–547.

Kauffman, S. A. 1991. Antichaos and adaptation. Sci. Amer. 265(2) 78–84.

Kauffman, S. A. 1992. The Origins of Order. Oxford University Press, Oxford, UK.

Keats, B. W., M. A. Hitt. 1988. A causal model of linkages among environmental dimensions, macro organizational characteris- tics, and performance. Acad. Management J. 31(3) 570–598.

Kogut, B., U. Zander. 1992. Knowledge of the firm, combinative capabilities, and the replication of technology. Organ. Sci. 3(3) 383–397.

Lane, P. J., M. Lubatkin. 1998. Relative absorptive learning and interorganizational learning. Strategic Management J. 19(5) 461–477.

Langlois, R. N., P. L. Robertson. 1992. Networks and innovation in a modular system: Lessons from the microcomputer and stereo component industries. Res. Policy 21(4) 297–313.

Lei, D., M. A. Hitt, J. D. Goldhar. 1996. Advanced manufactur- ing technology: Organizational design and strategic flexibility. Organ. Stud. 17(3) 501–523.

Levitt, B., J. G. March. 1988. Organizational learning. Ann. Rev. Sociol. 14 319–340.

Lind, J. T., H. Mehlum. 2007. With or without U? The appropriate test for a U shaped relationship. Working paper, Univerity of Oslo, Norway.

Long, J. 1997. Regression Models for Categorical and Limited Dependent Variables. Sage Publications, London.

Lyles, M. A. 1988. Learning among joint-venture sophisticated firms. Management Internat. Rev. 28 85–98.

March, J. G. 1991. Exploration and exploitation in organizational learning. Organ. Sci. 2(1) 71–87.

Nelson, R. R., S. G. Winter. 1982. An Evolutionary Theory of Economic Change. Belknap Press, Cambridge, MA.

Penrose, E. T. 1959. The Theory of the Growth of the Firm. Wiley, New York.

Pisano, G. P. 1994. Knowledge, integration, and the locus of learn- ing: An empirical analysis of process development. Strategic Management J. 15(Winter) 85–100.

Pitts, R. A. 1977. Strategies and structures for diversification. Acad. Management J. 20(2) 197–208.

Porter, M. E. 1987. From competitive advantage to corporate strategy. Harvard Bus. Rev. 65(3) 43–59.

Prahalad, C. K., R. Bettis. 1986. The dominant logic: A new linkage between diversity and performance. Strategic Management J. 7(6) 485–501.

Sanchez, R. 1995. Strategic flexibility in product competition. Strate- gic Management J. 16(Summer Special Issue) 135–159.

Sanchez, R., J. T. Mahoney. 1996. Modularity, flexibility, and knowl- edge management in product and organization design. Strate- gic Management J. 17(Winter Special Issue) 63–76.

Schilling, M. A. 2000. Toward a general modular systems theory and its application to interfirm product modularity. Acad. Man- agement Rev. 25(2) 312–334.

Schilling, M. A., H. K. Steensma. 2001. The use of modular organi- zational forms: An industry-level analysis. Acad. Management J. 44(6) 1149–1168.

Schilling, M. A., P. Vidal, R. E. Ployhart, A. Marangoni. 2003. Learn- ing by doing something else: Variation, relatedness, and the learning curve. Management Sci. 49(1) 39–56.

Schumpeter, J. A. 1934. The Theory of Economic Development. Trans- action Publishers, New Brunswick, NJ.

Stacey, R. D. 1995. The science of complexity: An alternative per- spective for strategic change processes. Strategic Management J. 16(6) 477–495.

Teece, D. J., G. Pisano, A. Shuen. 1997. Dynamic capabilities and strategic management. Strategic Management J. 18(7) 509–533.

Tsai, W., S. Ghoshal. 1998. Social capital and value creation: The role of intrafirm networks. Acad. Management J. 41(4) 464–476.

Waldrop, M. M. 1992. Complexity: The Emerging Science at the Edge of Order and Chaos. Viking, London.

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.

<< /ASCII85EncodePages false /AllowTransparency false /AutoPositionEPSFiles true /AutoRotatePages /None /Binding /Left /CalGrayProfile (Dot Gain 20%) /CalRGBProfile (sRGB IEC61966-2.1) /CalCMYKProfile (U.S. Web Coated \050SWOP\051 v2) /sRGBProfile (sRGB IEC61966-2.1) /CannotEmbedFontPolicy /Error /CompatibilityLevel 1.3 /CompressObjects /Tags /CompressPages true /ConvertImagesToIndexed true /PassThroughJPEGImages true /CreateJDFFile false /CreateJobTicket false /DefaultRenderingIntent /Default /DetectBlends true /DetectCurves 0.0000 /ColorConversionStrategy /CMYK /DoThumbnails false /EmbedAllFonts true /EmbedOpenType false /ParseICCProfilesInComments true /EmbedJobOptions true /DSCReportingLevel 0 /EmitDSCWarnings false /EndPage -1 /ImageMemory 1048576 /LockDistillerParams false /MaxSubsetPct 100 /Optimize true /OPM 1 /ParseDSCComments true /ParseDSCCommentsForDocInfo true /PreserveCopyPage true /PreserveDICMYKValues true /PreserveEPSInfo true /PreserveFlatness true /PreserveHalftoneInfo false /PreserveOPIComments true /PreserveOverprintSettings true /StartPage 1 /SubsetFonts true /TransferFunctionInfo /Apply /UCRandBGInfo /Preserve /UsePrologue false /ColorSettingsFile () /AlwaysEmbed [ true ] /NeverEmbed [ true ] /AntiAliasColorImages false /CropColorImages true /ColorImageMinResolution 300 /ColorImageMinResolutionPolicy /OK /DownsampleColorImages true /ColorImageDownsampleType /Bicubic /ColorImageResolution 300 /ColorImageDepth -1 /ColorImageMinDownsampleDepth 1 /ColorImageDownsampleThreshold 1.50000 /EncodeColorImages true /ColorImageFilter /DCTEncode /AutoFilterColorImages true /ColorImageAutoFilterStrategy /JPEG /ColorACSImageDict << /QFactor 0.15 /HSamples [1 1 1 1] /VSamples [1 1 1 1] >> /ColorImageDict << /QFactor 0.15 /HSamples [1 1 1 1] /VSamples [1 1 1 1] >> /JPEG2000ColorACSImageDict << /TileWidth 256 /TileHeight 256 /Quality 30 >> /JPEG2000ColorImageDict << /TileWidth 256 /TileHeight 256 /Quality 30 >> /AntiAliasGrayImages false /CropGrayImages true /GrayImageMinResolution 300 /GrayImageMinResolutionPolicy /OK /DownsampleGrayImages true /GrayImageDownsampleType /Bicubic /GrayImageResolution 300 /GrayImageDepth -1 /GrayImageMinDownsampleDepth 2 /GrayImageDownsampleThreshold 1.50000 /EncodeGrayImages true /GrayImageFilter /DCTEncode /AutoFilterGrayImages true /GrayImageAutoFilterStrategy /JPEG /GrayACSImageDict << /QFactor 0.15 /HSamples [1 1 1 1] /VSamples [1 1 1 1] >> /GrayImageDict << /QFactor 0.15 /HSamples [1 1 1 1] /VSamples [1 1 1 1] >> /JPEG2000GrayACSImageDict << /TileWidth 256 /TileHeight 256 /Quality 30 >> /JPEG2000GrayImageDict << /TileWidth 256 /TileHeight 256 /Quality 30 >> /AntiAliasMonoImages false /CropMonoImages true /MonoImageMinResolution 1200 /MonoImageMinResolutionPolicy /OK /DownsampleMonoImages true /MonoImageDownsampleType /Bicubic /MonoImageResolution 1200 /MonoImageDepth -1 /MonoImageDownsampleThreshold 1.50000 /EncodeMonoImages true /MonoImageFilter /CCITTFaxEncode /MonoImageDict << /K -1 >> /AllowPSXObjects false /CheckCompliance [ /None ] /PDFX1aCheck false /PDFX3Check false /PDFXCompliantPDFOnly false /PDFXNoTrimBoxError true /PDFXTrimBoxToMediaBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXSetBleedBoxToMediaBox true /PDFXBleedBoxToTrimBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXOutputIntentProfile () /PDFXOutputConditionIdentifier () /PDFXOutputCondition () /PDFXRegistryName () /PDFXTrapped /False /Description << /CHS <FEFF4f7f75288fd94e9b8bbe5b9a521b5efa7684002000410064006f006200650020005000440046002065876863900275284e8e9ad88d2891cf76845370524d53705237300260a853ef4ee54f7f75280020004100630072006f0062006100740020548c002000410064006f00620065002000520065006100640065007200200035002e003000204ee553ca66f49ad87248672c676562535f00521b5efa768400200050004400460020658768633002> /CHT <FEFF4f7f752890194e9b8a2d7f6e5efa7acb7684002000410064006f006200650020005000440046002065874ef69069752865bc9ad854c18cea76845370524d5370523786557406300260a853ef4ee54f7f75280020004100630072006f0062006100740020548c002000410064006f00620065002000520065006100640065007200200035002e003000204ee553ca66f49ad87248672c4f86958b555f5df25efa7acb76840020005000440046002065874ef63002> /DAN <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> /DEU <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> /ESP <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> /FRA <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> /ITA <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> /JPN <FEFF9ad854c18cea306a30d730ea30d730ec30b951fa529b7528002000410064006f0062006500200050004400460020658766f8306e4f5c6210306b4f7f75283057307e305930023053306e8a2d5b9a30674f5c62103055308c305f0020005000440046002030d530a130a430eb306f3001004100630072006f0062006100740020304a30883073002000410064006f00620065002000520065006100640065007200200035002e003000204ee5964d3067958b304f30533068304c3067304d307e305930023053306e8a2d5b9a306b306f30d530a930f330c8306e57cb30818fbc307f304c5fc59808306730593002> /KOR <FEFFc7740020c124c815c7440020c0acc6a9d558c5ec0020ace0d488c9c80020c2dcd5d80020c778c1c4c5d00020ac00c7a50020c801d569d55c002000410064006f0062006500200050004400460020bb38c11cb97c0020c791c131d569b2c8b2e4002e0020c774b807ac8c0020c791c131b41c00200050004400460020bb38c11cb2940020004100630072006f0062006100740020bc0f002000410064006f00620065002000520065006100640065007200200035002e00300020c774c0c1c5d0c11c0020c5f40020c2180020c788c2b5b2c8b2e4002e> /NLD (Gebruik deze instellingen om Adobe PDF-documenten te maken die zijn geoptimaliseerd voor prepress-afdrukken van hoge kwaliteit. De gemaakte PDF-documenten kunnen worden geopend met Acrobat en Adobe Reader 5.0 en hoger.) /NOR <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> /PTB <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> /SUO <FEFF004b00e40079007400e40020006e00e40069007400e4002000610073006500740075006b007300690061002c0020006b0075006e0020006c0075006f00740020006c00e400680069006e006e00e4002000760061006100740069007600610061006e0020007000610069006e006100740075006b00730065006e002000760061006c006d0069007300740065006c00750074007900f6006800f6006e00200073006f00700069007600690061002000410064006f0062006500200050004400460020002d0064006f006b0075006d0065006e007400740065006a0061002e0020004c0075006f0064007500740020005000440046002d0064006f006b0075006d0065006e00740069007400200076006f0069006400610061006e0020006100760061007400610020004100630072006f0062006100740069006c006c00610020006a0061002000410064006f00620065002000520065006100640065007200200035002e0030003a006c006c00610020006a006100200075007500640065006d006d0069006c006c0061002e> /SVE <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> /ENU (Use these settings to create Adobe PDF documents best suited for high-quality prepress printing. Created PDF documents can be opened with Acrobat and Adobe Reader 5.0 and later.) >> /Namespace [ (Adobe) (Common) (1.0) ] /OtherNamespaces [ << /AsReaderSpreads false /CropImagesToFrames true /ErrorControl /WarnAndContinue /FlattenerIgnoreSpreadOverrides false /IncludeGuidesGrids false /IncludeNonPrinting false /IncludeSlug false /Namespace [ (Adobe) (InDesign) (4.0) ] /OmitPlacedBitmaps false /OmitPlacedEPS false /OmitPlacedPDF false /SimulateOverprint /Legacy >> << /AddBleedMarks false /AddColorBars false /AddCropMarks false /AddPageInfo false /AddRegMarks false /ConvertColors /ConvertToCMYK /DestinationProfileName () /DestinationProfileSelector /DocumentCMYK /Downsample16BitImages true /FlattenerPreset << /PresetSelector /MediumResolution >> /FormElements false /GenerateStructure false /IncludeBookmarks false /IncludeHyperlinks false /IncludeInteractive false /IncludeLayers false /IncludeProfiles false /MultimediaHandling /UseObjectSettings /Namespace [ (Adobe) (CreativeSuite) (2.0) ] /PDFXOutputIntentProfileSelector /DocumentCMYK /PreserveEditing true /UntaggedCMYKHandling /LeaveUntagged /UntaggedRGBHandling /UseDocumentProfile /UseDocumentBleed false >> ] >> setdistillerparams << /HWResolution [2400 2400] /PageSize [612.000 792.000] >> setpagedevice