0

Structural Recombination and Innovation:

Unlocking Internal Knowledge Synergy through Structural Change

Samina Karim*

Strategy & Innovation

School of Management

Boston University

595 Commonwealth Ave.

Boston, MA 02215

Tel: 617-353-4289

Email: samina@bu.edu

Aseem Kaul

Strategic Management & Entrepreneurship

Carlson School of Management

University of Minnesota

321 Nineteenth Ave., S.

Minneapolis, MN 55455

Tel: 612-625-8458

Email: akaul@umn.edu

March 29, 2013

*corresponding author

Acknowledgments

We are grateful to seminar participants at the University of Minnesota and the University of Michigan, as

well as attendees of the 2012 Wharton Technology Conference and 2012 Atlanta Competitive Advantage

Conference, for their comments and suggestions on earlier versions of this paper.

1

Structural Recombination and Innovation

ABSTRACT

This paper examines how structural recombination of business units within a firm impacts subsequent

firm innovation. We argue that structural recombination is a means for firms to unlock the potential for

internal knowledge recombination, redrawing the firm’s internal boundaries to bring together

complementary knowledge resources. Structural recombination may thus be beneficial to subsequent firm

innovation, especially where there are substantial untapped synergies between the firm’s knowledge

resources. Specifically, we argue that structural recombination will have a positive effect on firm

innovation where there is unexploited relatedness between the firm’s knowledge resources, and where the

firm has high quality, general purpose knowledge. Results from a 20-year panel of 71 firms operating in

the U.S. medical sector confirm these arguments, with structural recombination having a positive effect

on innovation for firms with unexploited knowledge relatedness and those with high quality knowledge

resources. The study thus highlights the role of structural recombination in realizing untapped knowledge

synergies within the firm, and provides a contingent view of the effects of structural recombination on

firm innovation.

Key words: recombination, innovation, reconfiguration, structure, knowledge based view, contingency

theory

2

INTRODUCTION

Innovation, or the creation of new knowledge, is central to the role of organizations. As proponents of

the Knowledge Based View (KBV) of the firm have argued, not only are knowledge resources a key

source of the firm’s competitive advantage (Grant, 1996a; Liebeskind, 1996), but the creation, transfer

and integration of knowledge is an important reason firms exist (Kogut and Zander, 1992; Grant, 1996b).

By creating social communities of action, firms reduce the cost of communicating and coordinating

knowledge, thus enabling not only the transfer and integration of specialized knowledge, but also the

creation of new knowledge through the combination of existing knowledge to generate new capabilities

(Kogut and Zander, 1992; Grant, 1996a, 1996b). This ‘combinative capability’ of the firm (Kogut and

Zander, 1992) is critical to the success of innovation, which is inherently a recombinative process (Ahuja,

2000; Fleming, 2001; Katila and Ahuja, 2002; Fleming and Sorenson, 2004). Firms may gain competitive

advantage by spanning intra-organizational boundaries (Rosenkopf and Nerkar, 2001) and recombining

knowledge within the firm (Argyres, 1996; Hargadon and Sutton, 1997; Galunic and Rodan, 1998). Such

recombination may be especially valuable for multi-business firms, who may be able to realize

knowledge synergies across businesses (Markides and Williamson, 1994; Cardinal and Opler, 1995),

recombining knowledge across units to generate valuable innovations (Miller, Fern and Cardinal, 2007).

This kind of intra-organizational knowledge sharing has been the subject of considerable study, with a

substantial body of work studying the transfer of knowledge between units within an existing structure

(Tsai, 2002; Hansen and Lovas, 2004, Tortoriello et al., 2011). While treating the structure of the

organization as given and studying inter-unit knowledge transfer is certainly valuable and important, it

may not provide a complete picture of intra-firm knowledge recombination. A substantial body of prior

literature suggests that organizational knowledge may be sticky (Von Hippel, 1994; Szulanski, 1996), so

that the transfer of knowledge across epistemic intra-organizational boundaries may prove difficult

(Brown and Duguid, 2001), and the recombination and transformation of internal knowledge may

encounter semantic and pragmatic difficulties (Carlile and Rebentisch, 2003; Carlile, 2004). Inter-unit

transfer of knowledge may therefore be insufficient in realizing the full value of internal knowledge

3

synergies. Moreover, treating organizational structure as given ignores the reality that organizations can

and do restructure their businesses (Porter, 1987; Bowman and Singh, 1989; Galunic and Eisenhardt,

1996; Brown and Eisenhardt, 1997; Karim and Mitchell, 2004), and that such restructuring may be a

means for firms to redeploy and reconfigure their resources (Garud and Nayyar, 1994; Capron, Dussuage

and Mitchell, 1998; Capron and Mitchell, 1998; Eisenhardt and Brown, 1999; Karim, 2006).

Our study seeks to address these issues, complementing work on inter-unit knowledge transfers by

examining the effect of structural change on firm innovation. While prior studies have focused on intra-

firm knowledge recombination through the transfer of knowledge within an existing structure, we

examine the potential for such knowledge recombination through changes in the organizational structure,

specifically by the merging together of business units, which we refer to as structural recombination

(Karim 2006). Examining the effect of structural recombination on innovation is important not only

because it complements work on knowledge recombination within an existing structure, but because the

received literature offers contradictory perspectives on the effect of such changes in structure on

innovation. On the one hand, as discussed above, work in the dynamic capabilities tradition suggests that

structural recombination may be an important means by which firms recombine their knowledge

resources to develop new capabilities (Eisenhardt and Brown, 1999; Karim and Mitchell, 2004), as well

as a means of integrating knowledge from external sources in a post-acquisition context (Puranam and

Srikanth, 2007; Puranam, Singh and Chaudhuri, 2009). On the other hand, work on structural change

more generally has found evidence for a negative and disruptive effect (Bowman et al., 1999; McKinley

and Scherer, 2000), including a negative effect of structural change on new product innovation (Puranam,

Singh and Zollo, 2006; Karim, 2009).

In the current study, we argue that by dissolving boundaries between units, structural recombination

enables the recombination of internal knowledge resources, helping to realize untapped knowledge

synergies within the firm. More specifically, we predict that structural recombination will have a positive

effect on innovation for firms with untapped internal knowledge synergies, i.e. for firms that have high

quality, general purpose knowledge resources that are related in unexploited ways.

4

These arguments are tested on a sample of 71 firms from the U.S. medical sector, using panel data

from 1978-1997 that combine a detailed record of the evolution of business units with information on

firm patenting. Our results show that the impact of structural recombination is strongly contingent on the

firm’s existing knowledge resources. Consistent with the idea that structural recombination will help

realize untapped knowledge synergies, the effect of structural recombination on innovation is positive

where firms have knowledge resources that are related in unexploited ways and are of high quality.

These findings contribute to the literature on firm innovation and the knowledge based view of the

firm, highlighting the role of structural change in enabling the recombination of knowledge resources

within the firm (Kogut and Zander, 1992, 1996; Grant, 1996a). Our study thus complements the

substantial body of research that has looked at the transfer of knowledge between units (Hansen and

Lovas, 2004; Miller et al., 2007) by recognizing that organizational structure is not static, and suggesting

that the redrawing of intra-organizational boundaries to enable knowledge recombination is an important

alternative to knowledge transfers across such boundaries (Rosenkopf and Nerkar, 2001), and a means to

achieve the transformation of knowledge within the firm (Carlile and Rebentisch, 2003; Carlile, 2004).

Our study also contributes to the literature on dynamic capabilities in general, and organizational

design in particular. While work in the dynamic capabilities tradition has argued for the role of internal

reconfiguration in helping organizations adapt to changing conditions (Capron et al., 1997; Brown and

Eisenhardt, 1999; Galunic and Eisenhardt, 2001; Karim and Mitchell, 2004), studies of organizational

restructuring have generally found a negative effect of such changes (Bowman et al., 1999; McKinley and

Scherer, 2000), including a largely negative effect of organizational reconfiguration on innovation

(Karim, 2009). Our study resolves this apparent contradiction by highlighting the contingent nature of the

impact of structural recombination. Our study suggests that structural recombination both disrupts the

firm’s existing routines and enables the creation of new knowledge combinations, so that the effect of

structural recombination is positive only where unrealized potential for knowledge recombination exists.

By providing a contingent view of structural recombination, our study goes beyond the average effect of

structural change on subsequent innovation (Karim, 2009) to consider the heterogeneity of firms’

5

outcomes from structural recombination. In addition, our study complements work on knowledge

integration in the post-acquisition context (Puranam and Srikanth, 2007; Puranam et al., 2009), showing

that structural recombination may be valuable in recombining internal knowledge resources in addition to

enabling the integration of external knowledge.

THEORY & HYPOTHESES

Innovation, the limits of intra-organizational knowledge transfer, and structural recombination

The creation of new knowledge is a fundamental purpose of organizations (Kogut and Zander, 1992,

1996; Grant, 1996b). By establishing social communities with shared identity, norms, sequences and

patterns, organizations lower the costs of communication and coordination of knowledge, creating a

shared knowledge context that enables the development of new knowledge and capabilities (Kogut and

Zander, 1992, 1996; Grant, 1996b, Hoopes and Postrel, 1999). This knowledge then becomes the source

of the firm’s competitive advantage, since the creation, integration and transfer of knowledge is more

easily accomplished within firm boundaries than through the market (Kogut and Zander, 1992; Grant,

1996a, 1996b; Leibeskind, 1996).

A key element of organizational knowledge creation is the recombination of existing knowledge

resources to create new knowledge and capabilities (Kogut and Zander, 1992; Grant 1996b). Innovation

has been widely recognized as a recombinant process (Ahuja, 2000; Fleming, 2001; Fleming and

Sorenson, 2004), and a substantial body of literature has studied the recombination of knowledge

resources within the firm, with firms developing new capabilities by spanning intra-organizational

boundaries (Hargadon and Sutton, 1997; Galunic and Rodan, 1998; Argote and Ingram, 2000; Rosenkopf

and Nerkar, 2001). Such intra-organizational knowledge sharing may be especially important for

diversified firms, which stand to benefit from realizing knowledge synergies across businesses (Markides

and Williamson, 1994; Cardinal and Opler, 1995; Miller et al., 2007).

While the importance of intra-organizational knowledge sharing has been widely recognized, the

focus of prior research has been on studying the transfer of knowledge between units in a given structure,

6

and the factors that influence these transfers (Hansen and Lovas, 2004). Thus researchers have examined

the role of gatekeepers (Tushman and Katz, 1980), integration routines (Iansiti, 1995), formal and

informal linkages (Hansen, 1999; Tsai, 2000, 2002; Hansen and Lovas, 2004), internal absorptive

capacity (Tsai, 2001; Jansen et al., 2005), the nature of knowledge (Zander and Kogut, 1995; Szulanski,

1996), knowledge relatedness (Hansen, 2002; Hansen and Lovas, 2004) and the structure of internal

networks (Reagans and McEvily, 2003; Fleming, Mingo and Chen, 2007; Tortoriello et al., 2011) as

determinants of the extent of knowledge transfer between units. These studies all take the boundaries

between units as given, and focus on the ways in which knowledge is transferred across these boundaries.

While examining intra-firm knowledge transfer in this way is both important and valuable, it may not

provide a complete picture. There are many reasons why transfers of knowledge between units may be

inadequate to fully realizing the knowledge synergies within the firm. First, to the extent that divisional

boundaries are epistemic, i.e. they define unique communities of practice, each with their own distinct

interpretive schemes and cognitive frames, the transfer of knowledge across these boundaries may be

difficult (Dougherty, 1992; Fiol, 1994; Brown and Duguid, 2001). Intra-organizational boundaries will

result in individuals within a division developing shared identity, perspective and experiences making it

difficult for knowledge to cross these boundaries (Kogut and Zander, 1996).

Second, intra-organizational knowledge transfer may also be compromised by pragmatic and political

barriers within the firm (Argyres, 1996; Carlile, 2004; Kellogg, Orlikowsi and Yates, 2006). Divisions

within an organization will be in competition with each other (Galunic and Eisenhardt, 1996; Tsai, 2002)

and may be reluctant to share knowledge if doing so would compromise their own performance (Porter,

1987) or rob them of credit (Hargadon and Bechky, 2006), or if relations between units are arduous

(Szulanski, 1996). Different divisions may also have different goals and incentives, making it difficult for

them to work towards a common solution (Carlile and Rebentisch, 2003; Carlile, 2004).

Third, transfers of knowledge between units may only be helpful in realizing certain kinds of

knowledge synergies. While knowledge transfers may be helpful in sharing knowledge that is codifiable

and teachable (Zander and Kogut, 1995), they may be less valuable in transmitting knowledge that is tacit

7

and causally ambiguous, and therefore sticky (von Hippel, 1994; Szulanski, 1996). Recombinations of

tacit knowledge may require extensive socialization (Nonaka, 1994) with constant face to face interaction

to enable moments of creativity (Hargadon and Bechky, 2006) that may be hard to achieve across intra-

organizational boundaries. The nature of the innovative or recombinant process may also play a role in

determining the effectiveness of inter-unit transfers. Where the dependence between the knowledge

resources is sequential (Thompson, 1967; Van de Ven et al., 1976), a unilateral, one-time transfer of

knowledge from one unit to the other may be sufficient. But where reciprocal or team interaction

(Thompson, 1967; Van de Ven et al., 1976) is called for, successful innovation may require collective

problem solving (Carlile, 2004), which in turn may require rich communication and concurrent

development (Cardinal et al., 2011), and may be hard to achieve through inter-unit transfers.

Finally, even where knowledge can be successfully transferred across units, it is not clear that such a

transfer is always optimal – as Grant (1996b) puts it, “transferring knowledge is not an efficient approach

to integrating knowledge” (pp. 114). Maintaining knowledge links between units is costly (Hansen,

2002), both in the sense that it uses up managerial time and resources, and because these links may

constrain the ability of individual divisions to adapt to changing conditions. Faced with these costs, firms

may prefer to adopt loosely coupled or modular structures (Orton and Weick, 1990; Sanchez and

Mahoney, 1995), in order to stay flexible (Grant, 1996; Galunic and Eisenhardt, 2001).

None of this is to suggest, of course, that transfers of knowledge across intra-organizational

boundaries are not valuable. There is ample evidence to show that such intra-organizational transfers do

positively impact firm knowledge creation (Tsai, 2001; Miller et al., 2007). Our point is only that while

intra-organizational transfers may be an important means of realizing some of the potential synergies

between organizational knowledge resources, they may be unable to realize all such synergies. Beyond a

certain point, organizations will face a paradox of transfer: the very factors that enable the firm to create,

integrate and transfer knowledge within its firm boundary in a way that is superior to the market (Kogut

and Zander, 1992, 1996; Grant, 1996b) will also limit the firm’s ability to transfer and recombine this

knowledge across its internal boundaries amongst business units.

8

If transfers of knowledge across intra-organizational boundaries prove inadequate, what other means

can firms use to fully realize their potential knowledge synergies? One solution may be to alter their

structure, dissolving boundaries between units, and reshuffling activities from one unit to the other, thus

removing the constraints that limit the transfer of knowledge (Brown and Eisenhardt, 1999; Galunic and

Eisenhardt, 2001). Knowledge is more easily shared within units than between them (Tushman and

Katz,1980; Brown and Duguid, 2001), even when activities within a unit are distant from each other

(Hansen and Lovas, 2004), and prior research has shown that having fewer divisional boundaries helps

enable greater and wider exploration (Argyres, 1995, 1996; Argyres and Silverman, 2004). Structural

recombination, or the merging together of business units within a firm, may therefore enable the

recombination of knowledge resources within the firm, bringing knowledge resources that were

previously separated closer together by removing the internal boundaries between them.

The effect of structural recombination on innovation

Based on the discussion above, we are now in a position to define hypotheses. To begin with, we

consider the main effect of structural change on innovation. More specifically, and in line with our

theoretical argument for the recombination of knowledge resources within the firm, we focus on the

innovation consequences of a particular type of structural change, which we refer to as structural

recombination to distinguish it from other studies1. Structural recombinations are changes in business

units as their resources and market activities are reorganized by merging units together, generally through

the absorption of one unit into another unit or the formation of a new business unit by combining existing

units. Units may have different origins, being either internally developed, acquired or the result of a

recombination.

As discussed above, we believe that recombining units in this way will create the potential for

1 Karim (2006, 2009, 2012) used the general terms ‘reorganization’ and ‘reconfiguration’ when referring to the

addition, deletion, and recombination of business units, or their underlying resources and activities. To remain

consistent with these papers, and to more specifically highlight that our interests are in business units that are being

recombined, we use the term ‘structural recombination’.

9

innovation through successful recombination of existing knowledge (Ahuja, 2000; Fleming, 2001; Katila

and Ahuja, 2002; Nerkar, 2003; Fleming and Sorenson, 2004), specifically the recombination of

knowledge resources within the firm (Galunic and Rodan, 1998; Rosenkopf and Nerkar, 2001; Katila,

2002). This is consistent with both the dynamic capabilities literature that sees organizational

reconfiguration more generally as a means of revitalizing firm capabilities (Brown and Eisenhardt, 1997;

Galunic and Eisenhardt, 2001; Karim and Mitchell, 2004), as well as work on organizational structure and

innovation which argues that fewer divisional boundaries will result in greater exploration (Argyres,

1995, 1996; Argyres and Silverman, 2004). It is also consistent with work on post-merger innovation

which has argued for the importance of structural integration in enabling the recombination of knowledge

(Puranam and Srikanth, 2007; Puranam, Singh and Chaudhuri, 2009). Thus

Hypothesis 1a: Structural recombination will have a positive effect on firm innovation.

While our arguments so far focus on the positive effects of structural recombination in creating new

knowledge and capabilities, it is important to consider that such changes in structure may also prove

costly or disruptive (Miller, 1982; Nadler and Tushman, 1997; Bowman et al., 1999; McKinley and

Scherer, 2000). Changes in organizational structure may disrupt the organizational routines (Cyert and

March, 1963; Nelson and Winter, 1982; Puranam, Singh and Zollo, 2006), communities of practice

(Brown and Duguid, 1991, 2001; Orr, 1996), social relationships (Dougherty and Bowman, 1995; Dess

and Shaw, 2001) and organizational climate (Burton and Obel, 1998) within which organizational

knowledge is situated (Glynn, Lant and Milliken, 1994; Fisher and White, 2000), causing a depletion of

the firm’s existing knowledge. Changes in structure may be especially harmful to the firm’s tacit

knowledge (Nelson and Winter, 1982; Kogut and Zander, 1992; Nonaka, 1994), since such knowledge

may be lost when individuals are moved out of their existing positions (Nixon et al., 2004; Guthrie and

Datta, 2008). Such disruptions of the firm’s existing routines and relationships may in turn compromise

future firm learning (Fisher and White, 2000).

In addition to damaging the firm’s existing knowledge, structural recombination may also lead to

10

increased conflict and uncertainty, further suppressing firm innovation. Changes in organization structure

are associated with increased conflict over incompatible organizational goals and allocation of resources,

as well as greater complexity, uncertainty and ambiguity (Child, 1972; Galunic and Eisenhardt, 1996,

2001; Luscher and Lewis, 2008), so that organizational creativity may be inhibited by a negative and

stressful environment (Amabile and Conti, 1999), and firms risk falling prey to ‘post-restructuring drift’

(Johnson, Hoskisson and Margulies, 1990).

Thus, while structural recombination may benefit the firm by enabling the creation of new knowledge

through recombination, this benefit may be offset by the disruption of the firm’s existing knowledge and

capabilities, causing innovation to decline (Karim, 2009). We therefore offer the competing hypothesis:

Hypothesis 1b: Structural recombination will have a negative effect on firm innovation.

Structural recombination and the potential for internal knowledge recombination

Our discussion so far suggests that structural recombination has two competing effects: on the one

hand, it enables the realization of untapped knowledge synergies within the firm, providing a boost to

firm innovation; on the other hand, it disrupts the firm’s existing knowledge and capabilities, causing

innovation to suffer. While the previous section sets these effects up as competing hypotheses, the logic

of the argument suggests that whether structural recombination is beneficial or harmful to subsequent

innovation will depend upon the extent of untapped knowledge synergies within the firm. Where

significant untapped synergies exist, structural recombination is likely to have a positive effect; where it

does not, the effect of structural recombination will be primarily disruptive. This is similar to arguments

in the literature on post-acquisition integration: targets with strong potential synergies with the parent are

usually integrated, whereas those that are acquired for their stand-alone resources are best left

autonomous (Datta & Grant, 1990; Pablo, 1994; Zollo & Singh, 2004; Puranam and Srikanth, 2007;

Puranam, Singh and Chaudhuri 2009; Zaheer, Castañer and Souder, 2012).

We therefore propose a contingent view of the effect of structural recombination, with its effect on

innovation depending upon the extent to which there are untapped knowledge synergies within the firm.

11

The extent of untapped knowledge synergies within the firm will, in turn, depend upon two factors: the

potential for internal knowledge recombination, i.e. the extent to which knowledge within the firm is

amenable to being recombined; and the value of internal knowledge recombination, i.e. the extent to

which such recombinations are likely to be valuable. In what follows, we discuss these two aspects and

their moderating effect on the impact of structural recombination in more detail.

Potential for internal knowledge recombination

To begin with, the extent of internal knowledge synergies, and therefore the effect of structural

recombination, will depend upon the extent of relatedness between the knowledge resources of the firm. It

is only where knowledge resources in different parts of the firm are related that the recombination of

these resources will yield valuable synergies. If the distinct knowledge resources of the firm are unrelated

to each other, then bringing them together through structural recombination will yield little benefit.

Rather, bringing together disparate knowledge may only amplify the disruptive effect of structural

recombination (Argote and Ingram, 2000).

The importance of relatedness has its roots in the notion of firms as ‘coherent diversifiers’ (Teece et

al., 1994), maximizing the value from internal complementarities by entering into lines of business that

are related, in that they share common technological or market characteristics. Studies have shown that

firms tend to diversify their technologies in coherent ways, “filling all the possible ‘technological gaps’

between [patent] classes” (Breschi et al., 2003, p.86), and that such coherence has a positive effect on

firm innovation performance (Nesta and Saviotti, 2005). In addition, technological relatedness or

coherence has been shown to be critical to achieving performance improvement through process

management (Benner and Veloso, 2008) and realizing technology synergies from mergers (Cassiman et

al., 2005; Makri et al., 2010).

There are several reasons why technological relatedness will enhance the potential for internal

knowledge synergy. First, the greater the relatedness between the knowledge resources of the firm, the

greater the value of innovation from the recombination of this internal knowledge (Teece et al., 1994;

12

Argyres, 1996; Cassiman et al., 2005; Benner and Veloso, 2008; Makri et al., 2010). Second, greater

relatedness between internal knowledge resources will lower the coordination and communication costs

between different units within the firm, enabling them to work together more easily (Tushman, 1978;

Teece et al., 1994; Argyres, 1996; Breschi et al., 2003; Nesta and Saviotti, 2005). As several studies have

pointed out, intra-firm knowledge sharing is critically dependent on the existence of shared language

(Dougherty, 1992), frames (Fiol, 1994), coding schemes (Zander and Kogut, 1995) and mental models

(Galunic and Rodan, 1998) within the firm, and greater technological relatedness implies the existence of

these common elements linking the knowledge of different units. Third, greater knowledge relatedness

may also ameliorate the disruptive effects of structural change. Individuals in related knowledge spaces

may already have informal ties from being members of a shared community of practice (Brown and

Duguid, 1991), and the formation of new ties may be easier given shared language and understanding.

Relatedness between the firm’s knowledge resources, and the synergy associated with such

relatedness, is a necessary but not a sufficient condition for the positive effect of structural recombination

on innovation. To the extent that these potential synergies are already being exploited within the existing

organizational structure through the sharing of knowledge across internal boundaries, structural

recombination may not be required. Instead, as prior literature has suggested, the potential synergies

between the firm’s knowledge resources may be realized through knowledge transfers between units

(Tsai, 2002; Hansen and Lovas, 2004; Miller et al., 2007). Structural recombination is only likely to have

a positive effect on innovation when internal synergies are not being exploited within the existing

structure, i.e. the firm has a set of related knowledge resources, but the connections between these

resources are not being exploited, possibly because knowledge is unable to transcend internal boundaries.

It is in the presence of such unexploited knowledge relatedness that structural recombination can help

unlock the potential for internal knowledge recombination by redrawing the boundaries of the firm. Thus,

Hypothesis 2: The relationship between structural recombination and innovation will be more

positive, the greater the unexploited relatedness between the knowledge resources of the firm.

13

The firm’s internal knowledge may be more amenable to recombination not only if its knowledge is

in related areas, as discussed above, but also if the knowledge itself is capable of being applied more

generally. Other things being equal, the more general the firm’s knowledge resources, the wider the range

of domains in which they may be useful, and therefore the greater the potential for internal knowledge

recombination. If the firm’s knowledge resources are highly specialized to their individual domains, then

it may be difficult to recombine them with knowledge resources from other domains, even if the other

domains are related. Conversely, if the firm’s knowledge resources are more general, i.e. they are capable

of being used more generally than the resources of other firms, then it may be possible to leverage them

across even seemingly distant knowledge domains.

The role of knowledge generality2 has been discussed in the prior literature (Jaffe, Trajtenberg and

Henderson, 1993; Bresnahan and Trajtenberg, 1995) with studies highlighting the role of general purpose

technologies in enabling growth (Moser and Nicholas, 2004; Maine and Garnsey, 2006), driving markets

for technology (Arora et al., 2001; Gambardella and Giarratana, 2012), and increasing both the impact of

a technology and the valuation of the firm that owns it (Lerner, 1994). In particular, greater generality has

been linked to the realization of innovation complementarities (Bresnahan and Trajtenberg, 1995) with

more general technologies being amenable to recombination across a wider range of domains, thus

increasing both the likelihood of successful recombination, and the potential value of such recombinant

innovation.

In line with these arguments, we expect that greater generality of firm knowledge will increase the

value of internal knowledge synergies. The more general the firm’s knowledge resources, on average, the

greater the likelihood that they will be amenable to recombination with other knowledge resources within

the firm. Greater potential for recombination will translate into greater internal knowledge synergies,

which, in turn, will make structural recombination more valuable. Thus,

Hypothesis 3: The relationship between structural recombination and innovation will be more

2 Note that knowledge generality is distinct from the breadth or diversity of firm knowledge (Katila and Ahuja,

2002; Garcia-Vega, 2006) – the former measures how widely a single (average) technology of the firm can be

applied, the latter the range of different technologies within the firm.

14

positive, the greater the generality of the firm’s existing knowledge resources.

Value of internal knowledge recombination

While the arguments above focus on the potential for internal knowledge recombination, the extent of

internal knowledge synergies may also depend upon the value of the combinations made possible. In

particular, the potential for internal knowledge synergies may depend upon the pre-existing quality of the

firm’s knowledge resources. The inherent quality of a firm’s knowledge is a critical part of its overall

combinative capability (Kogut and Zander, 1992), with technologies that recombine high quality

knowledge being more likely to be valuable themselves (Nelson, 1982; Weitzman, 1998). Firms with

higher quality knowledge may have a superior understanding of their technology space, which in turn

may aid in recombinant search, giving clear direction to their search (Gavetti and Levinthal, 2000;

Fleming, 2001), and allowing them to achieve better outcomes in a more efficient manner (Fleming and

Sorenson, 2004). Firms with higher quality knowledge may also have superior absorptive capacity

(Cohen and Levinthal, 1990), and therefore combinative capability (Van den Bosch et al., 1999), being

better able to absorb or engage with technological knowledge from other business units.

Overall, then, we expect the value of internal knowledge complementarity to be greater, the greater

the quality of the firm’s existing knowledge resources. To the extent that structural recombination is a

means to realizing this complementarity, it follows that the effect of structural recombination on

innovation will be strongest for firms with high quality existing knowledge. Thus,

Hypothesis 4: The relationship between structural recombination and innovation will be more

positive, the higher the quality of the firm’s existing knowledge resources.

METHODS & MEASURES

Data and Sample

The empirical setting for this study is the medical sector, including its three main industries of

medical devices, pharmaceuticals, and healthcare service. Data was gathered from several editions of the

15

"Medical & Healthcare Marketplace Guide", namely those published in 1978, 1983, 1986, 1989, 1993

and 1997.3 The guides include information on firms operating in the U.S., and include both domestic and

international firms. Information is available on each firm and its business units, along with the

reconfiguration history of how units were recombined over time. There is also information at the firm

level about performance and degree of diversification (outside of the medical sector). Given the

complexity of this information, and the intricacies of mapping business unit evolution, we limit ourselves

to a sample of firms whose names start with the letters A, B, or C. This gives us an initial sample of 250

firms that we then track from 1978 to 1997, including 181 medical device firms, 38 healthcare service

firms, and 31 pharmaceutical firms. These 250 firms include a total of 866 unique business units over our

study period.

The detailed information on firm organization and reconfiguration for these 250 firms is augmented

with data on firm technology, by matching firms to patents filed with the USPTO using the NBER patent

database (Hall et al., 2001) . Patent data has been widely used in the technology literature to measure firm

knowledge and innovation (Ahuja and Lampert, 2001; Fleming, 2001; Rosenkopf and Nerkar, 2001;

Nerkar, 2003; Miller et al., 2007), and the richness of information available with this data allows us to

construct meaningful measures of the firm’s knowledge characteristics as well as a measure of firm

innovation overall. Note that the use of patent data also allows us to study innovation at a more granular

level than prior work that has examined the effect of structural reconfiguration on firms’ entry into new

product markets (Karim, 2009). Patent data are available for all patents granted up to 2006, allowing us to

study firm innovation for a significant period of time after the structural recombination in our data.

For this study, since our main construct of interest is structural recombination, we sample on the firms

that have multiple business units, are recombination-active, survive for at least two time periods (to

include lagged terms), and for which patent data is available. This sampling reduces our sample from 250

firms to 71 firms. Of these remaining 71 firms, within one time period the average amount of

3 "The Medical & Healthcare Marketplace Guide" was published by International Bio-Medical Information Services, Inc. (Acton,

MA, and Miami, FL; ed. 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).

16

recombination is 1.26 events. The firms had an average (non-logged actual value) of 14,000 employees

and almost $611 million in medical sales.

Measures

Dependent variable

Our dependent variable for all hypotheses is firm innovation in a period. The variable ‘Innovation’ is

measured as the citation-weighted patent stock of a firm at time ‘t+1’, i.e. one time period (approximately

four years) after the structural recombination has been completed4. Patents are counted in the year that

they are applied for, though only patents that are eventually granted are included in our measure.

Following Hall, Jaffe and Trajtenberg (2005) we use an annual discount rate of 15% to create a stock

measure of patents. Thus, the value of patent stock in any year is the sum of the number of patents applied

for in that year and the stock of patents in the previous year discounted by 15%. Moreover, because

patents may vary in their value (Griliches 1990; Trajtenberg 1990; Ahuja and Lampert 2001), we weight

each patent by the number of citations it receives from other patents. This is done by first adjusting the

citations received by the patent by the average citations received by any patent in that same year and

within that same patent class5 (to account for differences in citation rates across patent classes and

cohorts), and then multiplying the patent by this (adjusted) number of citations. The use of citation

weights is a common feature in innovation literature, as is the adjustment of citation counts to account for

cohort and patent class (Hall et al., 2001; Hall et al., 2005; Kaul, 2012). In addition, weighting patents by

citations means that we are closer to measuring the value of the knowledge generated by the firm. Since

we observe the change in patent stock, the measure captures how innovative a firm has been during the

period, or the extent of new knowledge it has created.

4 Our analysis uses a lagged panel structure that includes the firm’s patent stock at time ‘t’ as a predictor, so that our

analysis essentially predicts the change in patent stock between ‘t’ and ‘t+1’. As discussed below we also use the

citation-weighted number of patents applied for between t and t+1 as an alternate dependent variable and find

similar results. 5 Throughout this paper, we use four-digit IPC classes as our measure of patent class. To ensure consistency in

classification, all patents are assigned based on their 2006 patent classes.

17

Independent variables

Our main independent variable is the extent of firm structural recombination during a period. The

variable ‘Recombination’ is a count, for a firm, of the number of events in which a business unit was

merged into another business unit (thus losing its original identity) for the lagged time period ‘t-1’ to ‘t’6.

We measure the ‘Unexploited relatedness’ between a firm’s technologies in two steps. First, we

construct a measure of the overall relatedness of the firm’s knowledge resources, as the average

relatedness between each dyad of firm patents, based on the patent classes to which they belong.

Relatedness between two patent classes is measured using data from the entire population of USPTO

patents, and is calculated as the number of citations between the classes as a proportion of total citations

by patents in either class7. Mathematically,

Knowledge Relatedness8 = ∑pi.pj.rij / ∑pi . ∑pj …(1)

pi is the stock of firm patents in patent class i

rij is the relatedness of patent class i and j (from USPTO data)

This measure theoretically varies from 0 to 1, with higher values indicating greater overall

relatedness, though in practice its range is more limited, since the relatedness between two patent classes

is never 1 (this would require that patents in each class exclusively cite patents in the other class). Our

measure is closely related to measures of technological coherence in the literature (Breschi et al., 2003;

Nesta and Saviotti, 2005), except that we measure relatedness using the pattern of citations between

classes, rather than co-classifications.

The relatedness of the firm’s knowledge resources is then adjusted for the actual sharing of

6 For robustness, we also include the count of recombinations from ‘t-2’ to ‘t-1’. Results from this analysis

(available upon request) shows no evidence of a significant effect of longer lags on innovation. We also find no

evidence for a curvilinear effect of structural recombination as suggested by prior work (Karim, 2009), with the

inclusion of a square term for recombination having no significant effect. 7 In line with prior literature (Hall et al., 2001; 2005) we use only the first listed patent class of the patent. While this

introduces some error in measurement, we believe the extent of the imprecision is small, given the relatively large

size of the patent stock of our sample firms (Benner and Waldfogel, 2008). 8 This operationalization of the knowledge relatedness measure makes use of the fact that, by our definition, the

average relatedness of a patent to all other firm patents is the same for all firm patents in a given patent class. Rather

than considering each patent individually, therefore, we can simply calculate the average relatedness of each patent

class in the firm to every other patent class in the firm.

18

knowledge between patents within the firm, which is calculated as the ratio of self-citations to total

citations by the firm’s patents. This percentage of self-citations is then subtracted from overall relatedness

to give our final measure of unexploited relatedness. Thus,

Unexploited Relatedness = Overall Relatedness – % self-citations …(2)

The intuition is that this measure captures the extent to which the sharing of knowledge between the

technologies within the firm falls short of or exceeds the average sharing of knowledge between these

technologies in general, based on their patent classes. Thus a positive value of unexploited relatedness

means that patents within the firm cite each other less than an average set of patents in the same patent

classes (since knowledge relatedness is greater than percentage self-citations), while a negative value

suggests that they cite each other more. In theory, this measure could vary from -1 to 1.

The variable ‘Knowledge Generality’ is calculated as the average generality of the firm’s patents,

using a measure of generality developed by Jaffe, Trajtenberg and Henderson (1993). The generality of an

individual patent is calculated as 1 minus the Herfindahl index of its forward citations by patent class, i.e.

Generality = 1 - ∑ si2 where si is the ratio of citations received by the patent from patent class i to the total

citations received by the patent. This measure takes values from 0 to 1 with higher values indicating that

the citations received by the patent are more widely spread across patent classes, i.e. the patent is more

general in its use. These patent level generality measures are aggregated to the firm level as the average

generality across the stock of the firm’s patents, using the same depreciation rate as the measure of patent

stock used as the dependent variable.

The variable ‘Knowledge Quality’ is, for a firm in a given period, the average number of citations it

has received per patent, where the number of citations is adjusted for the average citations in the focal

patent’s year and patent class. This measure is similar to what others have used in past papers as

indication of innovative impact or quality (Fleming 2001; Ahuja and Lampert 20019; Rosenkopf and

Nerkar 2001; Nerkar 2003; Miller et al., 2007; Makri et al., 2010), and measures the average quality of

9 Ahuja and Lampert (2001) specifically examined firms’ breakthrough inventions and thus observed the top 1% of

an industry’s citation-weighted patents.

19

the firm’s patents relative to other patents in the same year and patent class. Thus, a knowledge quality

measure of ‘1’ means that the firm’s patents are, on average, neither more nor less impactful than the

patents of other firms in the same technology areas. Values greater than 1 imply that the firm’s patents

are, on average, of higher quality than those of other firms, while a value less than 1 implies that they are

of lower quality.

Control variables

We include a number of controls. First, because our dependent variable is the change in patent stock,

we control for the level of a firm’s patent stock at time ‘t’. This calculation of this measure is described in

the dependent variable section above.

Second, we control for firm performance. One concern with our analysis is that the effect of structural

recombination may vary with the performance context of the recombination. Specifically, if firms are

undertaking structural change as a means of consolidating and streamlining operations in response to poor

performance, then we would expect to see lower subsequent innovation than if structural recombination is

undertaken in order to pursue new strategic opportunities. We therefore include a control for firm growth,

measured as the (winsorized) increase in the firm’s total sales between ‘t-1’ and ‘t’, to account for the

confounding effect of firm performance. Results from using a measure of firm growth based on

employment rather than sales (available upon request) are similar.

In addition to controlling for growth, we try to account for the performance context of the

recombination in two other ways. First, as discussed in the methods section below, we use a two-stage

model to account for the endogeneity of the firm’s recombination decision. Second, as a supplementary

analysis, we examine the effect of firm performance on the moderating effect of the knowledge

characteristics discussed above, by examining the effects of structural recombination in high and low

performance firms separately. While we do not define formal hypotheses for this analysis, our expectation

is that our hypothesized effects will be stronger in high performance firms than in low performance firms.

Third, we include several firm level controls, drawing upon prior work on the impact of structural

recombination (Karim, 2009). ‘Medical Sales’ is the natural log of the monetary sales of the firm (in

20

thousands), and controls for firm size. ‘% Medical” is the ratio of the firm’s medical sales to its total

sales, and indicates extent of non-diversification (i.e. increasing amounts indicate activity predominantly

in the medical sector). ‘Firm age’ is the number of years since the firm was incorporated. ‘Acquired

Units’ is a count of the number of acquired business units at the firm; this controls for the level of a firm’s

acquisition activity.

Finally, we also include a dummy variable to account for periods in which the firm does not patent at

all. One concern with our analysis is that in cases where the firm does not patent, the stock of patents at

t+1 is simply a mathematical function of the stock at t, i.e. the change in patent stock is the result of

depreciation rather than new innovation by the firm. In a sense, this would tend to bias against our results,

since we would not see an improvement in innovation in cases where the firm does not innovate at all.

Nevertheless, to ensure that our results are not driven by periods where the firm does not patent, we

include a dummy variable that takes the value of 1 if the firm has no patents between t and t+1 and 0

otherwise. For further robustness, we also re-run our models using only the subsample of firm-periods

where firms do patent. As reported in the robustness section below, our results continue to hold in this

subsample.

*** Table 1 about here ***

Table 1 provides summary statistics and correlations for our sample. While there are several

significant correlations, the average VIF across our variables is 1.62 with no VIF being higher than 2.1,

suggesting that multicollinearity problems are not severe (Belsley et al., 1980, Belsey 1991).

Methodology

Given the panel nature of our data, as well as the continuous nature of our dependent variable, we use

fixed effects panel regressions (xtreg in STATA) with fixed effects for both firm and period. Since we are

interested in the effect of structural recombination on subsequent innovation, and on the moderating role

of pre-recombination knowledge characteristics, we use a lagged structure, predicting change in patent

stock between t+1 and t as a function of structural recombination between t-1 and t, and knowledge

characteristics at t-1. The basic equation represented by the model is:

21

Patent Stock t+1

= α + β1. Recombination t-1→t + β2. Recombination t-1→t * Knowledge Characteristicst-1

+ β3. Knowledge Characteristicst-1 + β4. Patent Stockt + β. Controls t + γi + δt + ε

….(3)

where γi is a firm fixed effect, δt is a period fixed effect, and ε is an error term.

Endogeneity correction

One issue with this analysis is that our main explanatory variable, structural recombination, may be

endogenous. In particular, firms may vary in their propensity to undertake structural recombination

depending on the nature of their technological knowledge, as well as the size of their patent stock. To

account for this possibility we run a first stage probit regression predicting the likelihood of

recombination as a function of firm technology variables and the lagged values of the other controls in

our main regression (i.e. their values prior to the structural recombination). Following Heckman (1979)

we then calculate the inverse mills ratio from the predicted probability of recombination from this first

stage regression, and include it as a control in our main models.

In addition to including all the independent variables from the second stage analysis, we also include

a measure for the total number of units in the first stage to act as an instrumental variable. The logic for

this instrument is that a firm with greater number of independent units is at greater risk for structural

recombination, but that given similar size, similar number of acquired units, and similar knowledge

characteristics, firms with more units will not be more or less innovative.

The results for this first stage regression are reported in Table 2. Model 1 shows the regression with

all predictors except the instrumental variable, and Model 2 adds the total number of units as a predictor.

A comparison of the two models confirms the strength of our instrument with the addition of total units

significantly improving the fit of the prediction model (Prob (Chi-sqr) <0.05). Following Larcker and

Rusticus (2010) we also test the validity of our instrument by regressing the residuals from our main

(second-stage) equation on the instrumental variable. The results of this analysis (available upon request)

confirm the validity of our instrument, with total units having an insignificant relation with the second

22

stage residuals (F = 0.00, Prob >F = 0.998).

RESULTS

Table 3 reports our main analysis. Model 1 includes only control variables, and model 2 adds our

main independent variable of structural recombination. The coefficient of structural recombination is

insignificant, so neither H1a nor H1b is supported.

*** Tables 2 and 3 and Figure 1a and 1b about here ***

Model 3 then examines the moderating effect of unexploited relatedness, by including the interaction

between structural recombination and unexploited relatedness10. This interaction term enters the model

with a positive and marginally significant (p = 0.053 in a two-sided test) coefficient. Model 3 thus

indicates support for H2; in the presence of unexploited relatedness, structural recombination has a

positive effect on innovation. This positive moderating effect is consistent with our arguments for

structural recombination being a means for firms to realize internal knowledge synergies that are not

being captured through intra-firm transfer.

Model 4 then examines the moderating effect of knowledge generality. While H3 predicted a positive

interaction between knowledge generality and structural recombination, the coefficient of the interaction

term in Model 4 is insignificant. H3 is therefore not supported.

Next, we turn our attention to knowledge quality. Model 5 shows a positive and significant coefficient

for the interaction between knowledge quality and structural recombination, implying that firms with high

quality knowledge are likely to benefit from structural recombination, consistent with H4. Model 5 thus

provides strong support for the idea that the knowledge recombination enabled by structural

recombination will be valuable if the firm has high quality knowledge resources to recombine.

Finally, Model 6 and Model 7 show models with all three interaction terms, Model 6 showing the full

set of interactions without the controls, while Model 7 shows the full model. H2 and H4 continue to be

supported in these models, with the coefficient of the interaction between recombination and unexploited

10 To alleviate multicollinearity issues, all interactions are mean-centered.

23

relatedness, which was only marginally significant in Model 3, achieving significance at conventional

levels once the moderating effect of knowledge quality is accounted for. The effect of structural

recombination is thus seen to be positive for firms with high unexploited relatedness and high quality

knowledge resources, but insignificant otherwise.

The moderating effects of unexploited relatedness and knowledge quality are shown graphically in

Figures 1a and 1b, which show the predicted effects of structural recombination at the high (upper

quartile) and low (lower quartile) levels of the moderating variables, holding all other variables at their

sample mean. The predicted values of innovation are based on Model 7. Thus, figure 1a shows that

structural recombination has a negligible effect on subsequent innovation for firms with unexploited

relatedness at the lower quartile of the distribution, but a strongly positive effect for firms with

unexploited relatedness at upper quartile levels. This is consistent with H2. Similarly, consistent with H4,

figure 1b shows that structural recombination leaves subsequent innovation unaffected for firms with low

quality knowledge, but has a strong positive effect where knowledge quality is high.

Overall, these results highlight the contingent nature of structural recombination on innovation. They

show that structural recombination has a positive effect on subsequent innovation where the firm has

untapped internal knowledge synergies, i.e. where it has not fully exploited the relatedness between its

knowledge resources, and where these resources are of high quality. Where such untapped knowledge

synergies are absent, however, structural recombination has little effect on subsequent innovation.

Supplementary Analysis and Robustness

*** Table 4 about here ***

To further confirm and explore these results, we undertake a number of supplementary analyses,

which are shown in Table 4. Panel A of Table 4 shows a number of alternate analyses, using different

models and samples to ensure that our findings are robust. First, model 8 re-runs our analysis using a flow

rather than a stock measure of our dependent variable, measured as the (citation-weighted) number of

patents applied for between t and t+1, to confirm that our results are not skewed by our choice of a stock

24

variable as a dependent variable11. Second, model 9 re-runs model 7 using a dynamic panel model

(Arrelano and Bover, 1995; Blundell and Bond, 1998) rather than a fixed effects panel model, to ensure

that our results are robust to alternate methods. Third, model 10 re-runs model 7 on the subsample of

firms that have at least one patent between t and t+1, to ensure that our results are not biased by the

inclusion of non-patenting firms. As Table 4 shows, we continue to find support for H2 and H4 across

these alternate analyses. To further confirm that our results are not biased by autocorrelation, we run an

Arellano Bond test for zero autocorrelation, following the dynamic panel model. Results from this test

(available upon request) are unable to reject the hypothesis of no autocorrelation, with z statistics of 0.37

(prob>z = 0.71) for first-order autocorrelation, and -0.88 (prob > z=0.38) for second-order

autocorrelation.

Panel B in Table 4 then undertakes two sets of supplementary analysis to further explore our

findings. First, we examine the relationship between unexploited relatedness and knowledge quality.

While our results so far show that both unexploited relatedness and knowledge quality positively

moderate the impact of structural recombination on subsequent innovation, it is unclear whether they act

as complements or as substitutes. In other words, is it sufficient for a firm to have either unexploited

relatedness or high knowledge quality, or does it need to have both in order to benefit from structural

recombination?

We explore this question in Models 11 and 12 in Table 4, by examining the moderating effect of

knowledge quality in subsamples of firms with high (above median) and low (below median) levels of

unexploited relatedness respectively. As Table 4 shows, the interaction between knowledge quality and

structural recombination is positive and significant in the low unexploited relatedness subsample, but

negative and significant in the high unexploited relatedness subsample, with the difference between them

being statistically significant. We interpret this to mean that unexploited relatedness and knowledge

11 We use a fixed-effects model rather than a count model for three reasons. First, weighting by (class-adjusted)

citations means that the flow measure is not an integer count but a continuous variable. Second, interaction terms in

non-linear models are difficult to test and interpret (Hoetker, 2007). Third, the median number of patents for firms

that did patent in our sample was 19, with a median citation count of 134, so treating our dependent variable as

continuous seems like a good approximation.

25

quality are substitute sources of knowledge synergy, with firms benefiting from structural recombination

in the presence of either high knowledge quality or high unexploited relatedness, but not both. Results

from splitting the sample on the basis of knowledge quality (available upon request) are similar, showing

a positive and significant effect of unexploited relatedness in the low knowledge quality subsample.

In addition to considering the relation between unexploited relatedness and knowledge quality,

we also examine the difference in their moderating effect in high and low performance contexts. As

discussed briefly above, we expect that the firm’s ability to profit from the realization of internal

knowledge synergies following structural recombination will be higher if the structural recombination is

undertaken in the context of strong firm performance. High performing firms are likely to have higher

levels of organizational slack, which may buffer them from the disruptive consequences of change (Cyert

and March, 1963; Bourgeois 1981) and have a positive effect on innovation (Nohria and Gulati, 1996;

Chen and Miller, 2007). Structural recombinations undertaken in the context of strong performance are

also more likely to be focused on the realization of new opportunities between businesses, and thus more

likely to result in successful knowledge recombination. In contrast, structural recombination undertaken

by firms that are struggling with performance issues are more likely to be focused on cost-cutting through

restructuring and downsizing, and are likely to be associated with limited slack and dysfunctional

behavior, which will tend to compromise innovation (Cameron, Whetten and Kim, 1987; Love and

Nohria, 2005). We would therefore expect the role of structural recombination in unlocking internal

knowledge synergies to be stronger in high performance contexts.

This prediction is tested in models 13 and 14 of Table 4, with performance being measured using

a measure of productivity calculated as sales per employee12. Consistent with our expectations, the

coefficients of the interaction terms corresponding to H2 and H4 are positive and significant in the strong

performance case, but insignificant in the weak performance case, with the difference between the two

subsamples being significant in both cases. Therefore, H2 and H4 are supported where structural

recombination is undertaken in the context of strong performance, but not when it is undertaken in

12 Results from measuring performance as the sales growth of the firm are similar.

26

response to weak performance. This supplementary result is only preliminary, given that our measures of

firm performance are fairly crude, but it suggests that the use of structural recombination in unlocking

internal knowledge synergies may be further contingent upon the firm’s prior performance. It also

provides further support for our main argument that structural recombination is a means to unlocking

untapped knowledge synergies, with the effect of unrealized coherence and technological impact being

strongest precisely where our theory would predict.

CONCLUSION

The findings of this study provide strong support for the argument that structural recombination is an

important means by which firms realize untapped knowledge synergies within the firm. In cases where

such untapped internal synergies exist, i.e. where the firm has high quality knowledge resources and

where its knowledge resources are related in unexploited ways, structural recombination has a positive

effect on innovation. Where no such untapped synergies exist, structural recombination has an

insignificant impact on subsequent innovation. Further, the positive effect of structural recombination on

innovation is stronger for firms with strong performance, supporting the idea that structural recombination

enables the realization of internal knowledge synergies. Structural recombination is thus seen to play an

important role in driving firm innovation, albeit one that is contingent upon the existence of untapped

knowledge synergies within the firm.

In highlighting the role of structural recombination in unlocking internal knowledge synergies, our

paper contributes to the literature on firm innovation and the KBV (Kogut and Zander, 1992; Grant,

1996a). The KBV sees the integration and recombination of knowledge as a central reason for the

existence of firms (Kogut and Zander, 1992; Grant, 1996b), and internal knowledge recombination has

been identified as a key source of organizational innovation and competitive advantage (Hargadon and

Sutton, 1997; Galunic and Rodan, 1998; Rosenkopf and Nerkar, 2001). While existing work on internal

knowledge recombination has focused on the transfer of knowledge across divisional boundaries (Tsai,

2002; Hansen and Lovas, 2004; Miller et al., 2007), the epistemic nature of internal boundaries (Kogut

and Zander, 1996; Brown and Duguid, 2001) and the stickiness of firm knowledge (Von Hippel, 1994;

27

Szulanski, 1996) suggest that inter-unit transfer alone may be insufficient to realize the full potential of

knowledge synergies within the firm. In particular, where innovation requires reciprocal interaction across

pragmatic and semantic boundaries then knowledge transfer may prove inadequate, and knowledge

transformation may be required (Van de Ven et al., 1976; Carlile and Rebentisch, 2003; Carlile, 2004).

Our paper suggests that in such cases structural recombination is an important alternate means of enabling

knowledge recombination within the firm, redrawing internal boundaries to exploit unrealized relatedness

between high quality knowledge resources. The recognition that structural recombination may help

unlock internal knowledge synergies has important managerial implications moreover, suggesting that

managers may use structural recombination as a means to driving firm innovation, but only where they

have high quality knowledge resources that are related in unexploited ways.

In highlighting this alternate means of enabling internal knowledge recombination, the paper also

connects to work on structural integration in the post-merger context (Puranam and Srikanth, 2007;

Puranam, Singh and Chaudhuri, 2009). In line with this work, our study shows that structural integration

plays an important role in enabling the coordination and recombination of knowledge (Puranam, Singh

and Chaudhuri, 2009), with such integration being beneficial despite the disruption associated with

structural change where the units being recombined have strong existing knowledge (Puranam, Singh and

Zollo, 2005) and there is substantial potential for knowledge recombination (Puranam and Srikanth,

2007). Our study complements this work in two ways: first, it develops a detailed theoretical account of

the difficulties of inter-unit knowledge transfer that is grounded in the knowledge based view, thus

highlighting the need for structural recombination to capture previously unrealized synergies, such as

those associated with acquisitions. Second, it empirically demonstrates that structural integration enables

knowledge recombination not only in post-merger contexts, but in organizations more generally.

This is not to suggest, of course, that dissolving any and all internal boundaries will be beneficial to

firm innovation. Knowledge resources with little or no connection should be housed in separate units,

since separating these resources will create sub-communities of embedded knowledge and practice

(Brown and Duguid, 1991, 2001; Kogut and Zander, 1992; Karim, 2012) and maximize the depth of

28

innovation (Argyres and Silverman, 2004). Our arguments suggest, and our results show, that structural

recombination is only beneficial where potential but unexploited knowledge synergies exist within the

firm. Nor is it our contention that structural recombination should replace inter-unit transfer as a means of

realizing internal knowledge synergies. Rather, in line with models of punctuated equilibrium (Romanelli

and Tushman, 1994) and organizational balance (Cardinal, Sitkin and Long, 2004), we see structural

recombination as a complement to inter-unit knowledge sharing, with regular transfers across intra-

organizational boundaries being supplemented by occasional redrawing of boundaries to enable more

radical internal knowledge recombinations in the face of changing environmental conditions.

In addition to contributing to the literature on innovation and the knowledge based view, the paper

also contributes to the literature on dynamic capabilities in general, and structural change in particular. In

line with work in the dynamic capabilities tradition, our paper argues and finds evidence for the role of

internal reconfiguration in enabling organizations to develop new capabilities and adapt to changing

conditions (Garud and Nayyar, 1994; Capron et al., 1997; Brown and Eisenhardt, 1999; Galunic and

Eisenhardt, 2001; Karim and Mitchell, 2004). More specifically, the paper sheds new light on the

consequences of changes in organizational structure. While prior work on organizational restructuring has

generally emphasized the negative and disruptive effects of structural change (Bowman et al., 1999;

McKinley and Scherer, 2000), our study offers a more positive view, arguing that under the right

circumstances disruption may have a beneficial effect (Amburgey, Kelley and Barnett, 1993; Zellmer-

Bruhn, 2003). In particular, our study joins Karim (2009) in being among the first to examine the effect of

structural recombination on innovation13. Our work differs from Karim’s (2009) study, however, both

because she focuses on the disruptive effects of recombination while we emphasize its potential to unlock

knowledge synergies, and because, in contrast to her paper, we take a contingent perspective on structural

recombination, moving beyond the average effect of structural recombination to consider the conditions

under which such recombination is more or less beneficial. As such, our study provides a more nuanced

13 Karim (2009) measures innovation as entry into new product lines, a measure that is both less granular and

conceptually distinct from the patent based measures used in our study.

29

perspective on the effects of structural recombination on innovation.

Our study could be improved in several ways. Though our two-stage empirical approach partially

controls for the endogeneity of the structural recombination decision, we cannot entirely account for the

intent behind the recombination. Controlling for the intent behind different recombinations would

certainly be useful, though it’s difficult to know how intent could be measured in this context. In addition,

though we are interested in the recombination of specific business units, our analysis is at the firm level,

due to the difficulty of assigning both patent and financial data to individual units. Our study is also

limited in that we have no information on the structure of R&D within the firm, and therefore cannot

control for the role of a centralized R&D unit in realizing internal knowledge synergies (Cardinal and

Opler, 1995). While this is certainly an important alternative, it is worth noting that such centralization is

not without its difficulties – separating R&D from downstream activities may compromise the firm’s

ability to translate new technologies into marketable products (Cardinal and Hatfield, 2000), and may also

reduce knowledge sharing between units (Tsai, 2002). Even if it were the case that firms had fully

centralized their R&D, moreover, this would only bias against our results, since firms where all

knowledge was centralized would see little or no gains from the recombination of business units. A final

limitation of the study is that our sample is restricted to 71 firms due to the intricate mapping required to

trace the evolution of firms’ business units; a larger sample would serve as a good robustness check of our

results.

In summary, our study highlights the role of structural recombination in unlocking the potential for

knowledge recombination within the firm. We show that firms with high quality knowledge resources that

are related in unexploited ways see an improvement in firm innovation following structural

recombination. Structural recombination is thus seen as a means of realizing knowledge synergies within

the firm, one that complements the more traditional sharing of knowledge across internal boundaries. The

study thus contributes to the literature on firm innovation and the knowledge based view, highlighting an

alternate means of recombining knowledge within the firm, as well as to work on organizational and

structural change, taking a contingent view of the effects of structural recombination on innovation.

30

REFERENCES

Ahuja, G. 2000. Collaboration networks, structural holes, and innovation: A longitudinal study. Administrative Sci.

Quart. 45 (3) 425-455

Ahuja, G., C. M. Lampert. 2001. Entrepreneurship in the large corporation: A longitudinal study of how established

firms create breakthrough inventions. Strategic Management J. 22 521-543.

Amabile, T.M., R. Conti. 1999. Changes in the work environment for creativity during downsizing Acad.

Management J. 42(6) 630-640

Amburgey, T.L, D. Kelly, W.P. Barnett. 1993. Resetting the clock: The dynamics of organizational change and

failure. Administrative Sci. Quart. 38(1) 51-73

Arellano, M., O. Bover. 1995. Another look at the instrumental variable estimation of error-components models. J.

Econometrics 68 29-51.

Argote, L., P. Ingram. 2000. Knowledge transfer: A basis for competitive advantage in firms Organ. Behavior

Human Decision Processes 82(1) 150-169

Argyres, N.S. 1995. Technology strategy, governance structure and interdivisional coordination. J. Econom.

Behavior Organ. 28 337-358

Argyres, N.S. 1996. Capabilities, technological diversification and divisionalization. Strategic Management J. 17

395-410.

Argyres, N.S., B.S. Silverman. 2004. R&D, organization structure, and the development of corporate technological

knowledge. Strategic Management J. 25 929-958

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

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

Benner, M.J., F.M. Veloso. 2008. ISO 9000 practices and financial performance: A technology coherence

perspective. J. Oper. Management 26 611-629

Benner, M.J., J. Waldfogel. 2008. Close to you? Bias and precision in patent-based measures of technological

proximity. Res. Policy 37 1556-1567

Blundell, R., S. Bond. 1998. Initial conditions and moment restrictions in dynamic panel data models. J.

Econometrics 87 115-143.

Bourgeois, L.J. 1981. On the measurement of organizational slack. Acad. of Management Rev. 6(1) 29-39

Bowman, E. H., H. Singh. 1989. Corporate restructuring: Trends and consequences. R. Rock ed. Corporate

Restructuring. New York, McGraw-Hill.

Bowman, E.H., H.Singh, M. Useem, R. Bhadury.1999. When does restructuring improve economic performance?

California Management Rev. 41(2) 33-54

Breschi, S., F. Lissoni, F. Malerba. 2003. Knowledge-relatedness in firm technological diversification. Res. Policy

32 69-87.

Bresnahan, T.F., M. Trajtenberg. 1995. General purpose technologies: 'Engines of growth'? J. Econometrics 65 83-

108

Brown, J.S., P. Duguid. 1991. Organization learning and communities-of-practice: Toward a unified view of

working, learning, and innovation. Organ. Sci. 2(1) 40-57

Brown, J.S., P. Duguid. 2001. Knowledge and organization: A social-practice perspective. Organ. Sci. 12 (2) 198-

213

Brown, S. L., K. M. Eisenhardt. 1997. The art of continuous change: Linking complexity theory and time-paced

evolution in relentlessly shifting organizations. Administrative Sci. Quart. 42 1-34.

Burton, R.M., B. Obel. 1998. Strategic Organizational Diagnosis and Design: Developing Theory for Application.

Kluwer Publishers, Boston, MA

Cameron, K.S., Whetten, D.A., M.U. Kim. 1987. Organizational dysfunctions of decline. Acad. of Management J.

30 (1) 126-138

Capron, L., P. Dussauge, W. Mitchell. 1998. Resource redeployment following horizontal acquisitions in Europe

and North America, 1988-1992. Strategic Management J. 19 631-661.

Capron, L., W. Mitchell. 1998. Bilateral resource redeployment and capabilities improvement following horizontal

acquisitions. Industrial Corporate Change 7(3) 453-483.

Cardinal, L.B., D.E. Hatfield. 2000. Internal knowledge generation: The research laboratory and innovative

productivity in the pharmaceutical industry. J. Eng. Technol. Management 17 247-271

31

Cardinal, L. B., T. C. Opler. 1995. Corporate diversification and innovative efficiency: An empirical study. J.

Accounting Econom. 19 365-381.

Cardinal, L.B., Sitkin, S.B., C.P.Long. 2004. Balancing and rebalancing in the creation and evolution of

organizational control. Organ. Sci. 15 (4) 411-431

Cardinal, L.B., S.F. Turner, M.J. Fern, R.M. Burton. 2011. Organizing for product development across technological

environments: Performance trade-offs and priorities. Organ. Sci. 22(4) 1000-1025

Carlile, P.R. 2004. Transferring, translating and transforming: An integrative framework for managing knowledge

across boundaries. Organ. Sci. 15(5) 555-568

Carlile, P.R., E.S. Rebentisch. 2003. Into the black box: The knowledge transformation cycle. Management Sci.

49(9) 1180-1195

Cassiman, B., M.G. Colombo, P. Garrone, R. Veugelers. 2005. The impact of M&A on the R&D process: An

empirical analysis of the role of technological- and market- relatedness. Res. Policy 34(2) 195-220

Child, J. 1972. Organization structure, environment and performance: The role of strategic choice. Sociology 6 1-22.

Cohen, W., D. A. Levinthal. 1990. Absorptive capacity: A new perspective on learning and innovation.

Administrative Sci. Quart. 35 128-152.

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

Chen, W., K.D. Miller. 2007. Situational and institutional determinants of firm's R&D search intensity. Strategic

Management J. 28 369-381

Datta, D.K., J.H. Grant. 1990. Relationships between type of acquisition, the autonomy given to the acquired firm,

and acquisition success: An empirical analysis. J. Management 16(1) 29-44

Dess, G.G., J.D. Shaw. 2001. Voluntary turnover, social capital and organizational performance. Acad. Management

Rev. 26(3) 446-456

Dougherty, D. 1992. Interpretive barriers to successful product innovation in large firms. Organ. Sci. 3(2) 179-202

Dougherty, D., E.H. Bowman. 1995. The effects of organizational downsizing on product innovation California

Management Rev. 37(4) 28-44

Eisenhardt, K. M., S. L. Brown. 1999. Patching: Restitching business portfolios in dynamic markets. Harvard Bus.

Rev. 77(3) 71-82.

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

Fiol, C.M. 1994. Consensus, diversity and learning in organizations. Organ. Sci. 5(3) 403-420

Fisher, S.R., M.A. White. 2000. Downsizing in a learning organization: Are there hidden costs? Acad. Management

Rev. 25(1) 244-251

Fleming, L. 2001. Recombinant uncertainty in technological search. Management Sci. 47(1) 117-132.

Fleming, L., S. Mingo, D. Chen. 2007. Collaborative brokerage, generative creativity, and creative success.

Administrative Sci. Quart. 52 443-475

Fleming, L., O. Sorenson. 2004. Science as a map in technological search. Strategic Management J. 25 909-928

Galunic, D. C., K. M. Eisenhardt. 1996. The evolution of intracorporate 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 corporate 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

Schumpeterian innovation. Strategic Management J. 19 1193-1201.

Gambardella, A., M.S. Giarratana. 2012. General technological capabilities, product market fragmentation, and

markets for technology. Res. Policy http: dx.doi.org/10.1016/j.respol.2012.08.2002

Garcia-Vega, M. 2006. Does technological diversification promote innovation? An empirical analysis for European

firms. Res. Policy 35 230-246.

Garud, R., Nayyar, P.R. 1994. Transformative capacity: Continual structuring by intertemporal technology transfer.

Strategic Management J. 15 365-385

Gavetti, G., D.A. Levinthal. 2000. Looking forward and looking backward: Cognitive and experiential search.

Administrative Sci. Quart. 45 (1) 113-137

Glynn, M.A., T.K. Lant, F.J. Milliken. 1994. Mapping learning processes in organizations. C. Stubbart, J.P. Meindl,

J.F. Porac eds. Advances in managerial cognition and organizational information processing Greenwich CT,

JAI Press 5 43-83

Grant, R.M. 1996a. Toward a knowledge-based theory of the firm. Strategic Management J. 17 109-122

Grant, R.M. 1996b. Prospering in dynamically-competitive environments: Organizational capability as knowledge

integration. Organ. Sci. 7(4) 375-387

Griliches, Z. 1990. Patent statistics as economic indicators: A survey. J.Econom. Literature 28(4) 1661-1707.

32

Gulati, R., P. Puranam. 2009. Renewal through reorganization: The value of inconsistencies between formal and

informal organization. Organ. Sci. 20 (2) 422-440

Guthrie, J.P., D.K. Datta.2008. Dumb and dumber: The impact of downsizing on firm performance as moderated by

industry conditions. Organ. Sci. 19 (1) 108-123

Hall, B. H., A. B. Jaffe, M. Trajtenberg. 2001. The NBER patent citation data file: Lessons, insights and

methodological tools . NBER Working Paper No.8498. National Bureau of Economic Research: Cambridge,

MA.

Hall, B. H., A. B. Jaffe, M. Trajtenberg. 2005. Market value and patent citations. Rand J. Econom. 36 (1) 16-38

Hansen, M.T. 1999. The search-transfer problem: The role of weak ties in sharing knowledge across organizational

sub-units. Administrative Sci. Quart. 44(1) 82-111

Hansen, M.T. 2002. Knowledge networks: Explaining effective knowledge sharing in multinational companies.

Organ. Sci. 13 (3) 232-248

Hansen, M.T., B. Lovas. 2004. How do multinational companies leverage technological competencies? Moving

from single to interdependent explanations. Strategic Management J. 25 801-822

Hargadon, A.B., B.A. Bechky. 2006. When collections of creatives become creative collectives: A field study of

problem solving at work. Organ. Sci. 17 (4) 484-500

Hargadon, A., R.I.Sutton. 1997. Technology brokering and innovation in a product development firm.

Administrative Sci. Quart. 42(4) 716-749

Heckman, J.J. 1979. Sample selection bias as a specification error. Econometrica 47 153-161

Henderson, R.M., K.B. Clark. 1990. Architectural innovation: The reconfiguration of existing product technologies

and the failure of established firms. Adminstrative Sci. Quart. 35 (1) 9-30

Hoetker, G. 2007. The use of logit and probit models in strategic management research: Critical issues. Strategic

Management J. 28 331-343

Hoopes, D.G., S. Postrel. 1999. Shared knowledge, "glitches", and product development performance. Strategic

Management J. 20 837-865

Iansiti, M. 1995. Technology integration: Managing technological evolution in a complex environment. Res. Policy

24 521-542

Jaffe, A.B., Trajtenberg, M., R. Henderson. 1993. Geographic localization of knowledge spillovers as evidenced by

patent data. Quart J. of Econom. 108 (3) 577-598

Jansen, J.J.P., F.A.J. Van den Bosch, H.W. Volberda. 2005. Managing potential and realized absorptive capacity:

How do organizational antecedents matter? Acad. Management J. 48(6) 999-1015

Johnson, R.A., R.E. Hoskisson, N. Margulies. 1990. Corporate restructuring: Implications for organization change

and development. R.W. Woodman,W.A. Pasmore eds. Research in Organizational Change and Development

Greenwich CT, JAI Press 4 141-165

Karim, S. 2006. Modularity in organizational structure: The reconfiguration of internally developed and acquired

business units. Strategic Management J. 27: 799-823.

Karim, S. 2009. Business unit reorganization and innovation in new product markets. Management Sci. 55 1237-

1254.

Karim, S. 2012. Exploring structural embeddedness of product market activities and resources within business units.

Strategic Organ. 10 333-365

Karim, S., W. Mitchell. 2004. Innovating through acquisition and internal development. Long Range Planning 37

525-547.

Katila, R. 2002. New product search over time: Past ideas in their prime? Acad. Management J. 45(5) 995-1010

Katila, R., G. Ahuja. 2002. Something old, something new: A longitudinal study of search behavior and new product

introduction. Acad. Management J. 45(6) 1183-1194.

Kaul, A. 2012. Technology and corporate scope: Firm and rival innovations as antecedents of corporate transactions.

Strategic Management J. 33 347-367

Kellogg, K.C., W.J. Orlikowski, J. Yates. 2006. Life in the trading zone: Structuring coordination across boundaries

in postbureaucratic organizations. Organ. Sci. 17(1) 22-44

Kogut, B., U. Zander. 1992. Knowledge of the firm, combinative capabilities, and the replication of technology.

Organ. Sci. 3(3) 383-397.

Kogut, B., U. Zander. 1996. What firms do? Coordination, identity, and learning. Organ. Sci. 7 (5) 502-518

Larcker, D.F., T.O. Rusticus. 2010. On the use of instrumental variables in accounting research. J. Accounting

Econom. 49 186-205

Lerner, J. 1994. The importance of patent scope: An empirical analysis. RAND J. Econom. 25(2) 319-333

Liebeskind, J.P. 1996. Knowledge, strategy, and the theory of the firm. Strategic Management J. 17 93-107

33

Love, E.G., N. Nohria. 2005. Reducing slack: The performance consequences of downsizing by large industrial

firms, 1977-93. Strategic Management J. 26 1087-1108

Luscher, L.L., M.W. Lewis. 2008. Organizational change and managerial sensemaking: Working through paradox.

Acad, Management J. 51(2) 221-240

Maine, E., E. Garnsey. 2006. Commercializing generic technology: The case of advanced materials ventures. Res.

Policy 35(3) 375-393

Makri, M., M. A. Hitt, P. J. Lane. 2010. Complementary technologies, knowledge relatedness, and invention

outcomes in high technology mergers and acquisitions. Strategic Management Journal 31 602-628.

Markides, C.C., P.J. Williamson. 1994. Related diversification, core competencies and corporate performance.

Strategic Management J. 15 149-165

McKinley, W., A.G. Scherer. 2000. Some unanticipated consequences of organizational restructuring. Acad.

Management Rev. 25(4) 735-752

Miller, D. 1982. Evolution and revolution: A quantum view of structural change in organizations. J. Management

Stud. 19(2) 131-151.

Miller, D.1986. Configurations of strategy and structure: Towards a synthesis. Strategic Management J.7 233-249.

Miller, D. J., M. J. Fern, L. B. Cardinal. 2007. The use of knowledge for technological innovation within diversified

firms. Acad.Management J. 50(2) 308-326.

Moser, P., T. Nicholas. 2004. Was electricity a general purpose technology? Evidence from historical patent

citations. American Econom. Rev. 94 (2) 388-394

Nadler, D., M. L. Tushman. 1997. Competing by Design: The Power of Organizational Architecture. New York,

Oxford University Press.

Nelson, R.R. 1982. The role of knowledge in R&D efficiency. Quart. J. Econom. 97 (3) 453-470

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

Nerkar, A. 2003. Old is gold? The value of temporal exploration in the creation of new knowledge. Management

Sci. 49(2) 211-229.

Nesta, L., P. P. Saviotti. 2005. Coherence of the knowledge base and the firm's innovative performance: Evidence

from the U.S. pharmaceutical industry. J.Industrial Econom. 53(1) 123-142.

Nixon, R.D., M.A. Hitt, H. Lee, E. Jeong 2004. Market reactions to announcements of corporate downsizing actions

and implementation strategies. Strategic Management J. 25 1121-1129

Nohria, N., R. Gulati. 1996. Is slack good or bad for innovation? Acad. Management J. 39(5) 1245-1264

Nonaka, I. 1994. A dynamic theory of organizational knowledge creation. Organ. Sci. 5 (1) 14-37

Orr, J. 1996. Talking About Machines: An Ethnography of a Modern Job. Ithaca NY, IRL Press.

Orton, J.D., K.E. Weick. 1990. Loosely coupled systems: A reconceptualization. Acad. of Management Rev. 15 (2)

203-223

Paternoster, R., R. Brame, P. Mazerolle, A. Piquero. 1998. Using the correct statistical test for the equality of

regression coefficients. Criminology 36(4) 859-866

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

Puranam, P., Singh, H., S. Chaudhuri. 2009. Integrating acquired capabilities: When structural integration is (un)

necessary. Organ. Sci. 20 (2) 313-328

Puranam, P., Singh, H., M. Zollo. 2006. Organizing for innovation: Managing the coordination-autonomy dilemma

in technology acquisitions. Acad. Management J. 49(2) 263-280

Puranam, P., K. Srikanth. 2007. What they know vs. what they do: How acquirers leverage technology acquisitions.

Strategic Management J. 28 805-825

Reagans, R., B. McEvily. 2003. Network structure and knowledge transfer: The effects of cohesion and range.

Administrative Sci. Quart. 48 240-267

Romanelli, E., M. L. Tushman. 1994. Organizational transformation as punctuated equilibrium: An empirical test.

Acad. Management J. 37(5) 1141-1166.

Rosenkopf, L., A. Nerkar. 2001. Beyond local search: Boundary-spanning, exploration, and impact in the optical

disk industry. Strategic Management J. 22 287-306.

Sampson, R.C. 2007. R&D alliances and firm performance: The impact of technological diversity and alliance

organization on innovation. Acad.Management J. 50(2) 364-386

Sanchez, R., J.T. Mahoney. 1996. Modularity, flexibility and knowledge management in product and organizational

design. Strategic Management J. 17 63-76

Szulanski, G. 1996. Exploring internal stickiness: Impediments to the transfer of best practice within the firm.

Strategic Management J. 17 27-43

Teece, D. J., R. Rumelt, G. Dosi, S.G. Winter. 1994. Understanding corporate coherence. J. Econom. Behavior

34

Organ. 23: 1-30.

Thompson, J.D. 1967. Organization in Action. Chicago: McGraw-Hill

Tortoriello, M., R. Reagans, B. McEvily. 2011. Bridging the knowledge gap: The influence of strong ties, network

cohesion and network range on the transfer of knowledge between organizational units. Organ.Sci. 23 (4) 1024-

1039

Trajtenberg, M. 1990. A penny for your quotes: Patent citations and the value of information. RAND J. Econom. 21

325-342.

Tsai, W. 2000. Social capital, strategic relatedness and the formation of intraorganizational linkages. Strategic

Management J. 21 925-939

Tsai, W. 2001. Knowledge transfer in intraorganizational networks: Effects of network position and absorptive

capacity on business unit innovation and performance. Acad. Management J. 44 (5) 996-1004

Tsai, W. 2002. Social structure of "coopetition" within a multiunit organization: Coordination, competition, and

intraorganizational knowledge sharing. Organ. Sci. 13 (2) 179-190

Tushman, M.L. 1978. Technical communication in R&D Laboratories: The impact of project work characteristics.

Acad.Management J. 21(4) 624-645

Tushman, M.L., P. Anderson. 1986. Technological discontinuities and organizational environments. Administrative

Sci. Quart. 31 (3) 439-465

Tushman, M.L., R. Katz. 1980. External communication and project performance: An investigation into the role of

gatekeepers. Management Sci. 26 (11) 1071-1085

Van de Ven, A.H., Delbecq A.L., R. Koenig. 1976. Determinants of coordination modes within organizations. Amer.

Sociological Rev. 41(2) 322-338

Ven den Bosch, F.A.J., H.W. Volberda, M. de Boer 1999. Coevolution of firm absorptive capacity and knowledge

environment: Organizational forms and combinative capabilities. Organ. Sci. 10 (5) 551-568

Verona, G. 1999. A resource-based view of product development. Acad.Management Rev. 24(1) 132-142

Von Hippel, E. 1994. "Sticky information" and the locus of problem solving: Implications for innovation.

Management Sci. 40 (4) 429-439

Weitzman, M.L. 1998. Recombinant growth. Quart. J. Econom. 113(2) 331-360

Zaheer, A., Castaner, X., D. Souder. 2011. Synergy sources, target autonomy, and integration in acquisitions. J.

Management. published online ahead of print May 2, 2011

Zander, U., B. Kogut. 1995. Knowledge and the speed of transfer and imitation of organizational capabilities: An

empirical test. Organ. Sci. 6(1) 76-92

Zellmer-Bruhn, M.E. 2003. Interruptive events and team knowledge acquisition" Management Sci. 49(4) 514-528

Zollo, M., H. Singh. 2004. Deliberate learning in corporate acquisitions: Post-acquisition strategies and integration

capability in U.S. bank mergers. Strategic Management J. 25 1233-1256

35

Table 1

Summary Statistics & Correlation Matrix

Variable Mean Std. Dev. Min Max 1 2 3 4 5 6 7 8 9 10 11

1 Patent Stock (citation-weighted) 186.44 451.48 0.05 2099.68 1.00

2 Recombination (t-1) 1.26 3.43 0.00 26.00 0.37 1.00

3 Unexploited relatedness 0.12 0.24 -0.43 0.70 -0.55 -0.29 1.00

4 Knowledge generality 0.47 0.16 0.00 0.81 0.04 -0.01 -0.21 1.00

5 Knowledge quality 1.09 0.55 0.11 3.17 -0.11 -0.07 -0.06 0.36 1.00

6 Medical sales 10.48 2.60 6.21 16.02 0.48 0.44 -0.49 0.00 -0.11 1.00

7 % Medical 0.70 0.37 0.00 1.00 -0.45 -0.25 0.46 0.03 0.18 -0.23 1.00

8 Firm Age 51.36 34.55 9.00 159.00 0.51 0.38 -0.30 -0.15 -0.19 0.52 -0.40 1.00

9 Acquired Units 1.42 3.03 0.00 17.00 0.36 0.56 -0.32 -0.01 -0.13 0.54 -0.33 0.40 1.00

10 Sales growth 0.43 1.01 -0.38 4.70 -0.03 -0.03 0.08 0.07 0.13 0.03 0.16 -0.06 -0.04 1.00

11 No Patent (dummy) 0.32 0.47 0.00 1.00 -0.28 -0.21 0.50 -0.15 -0.07 -0.43 0.31 -0.22 -0.25 0.05 1.00

36

Table 2

First Stage Prediction Models

Predictor Model A Model B

Unexploited relatedness

0.05

(0.80)

-0.04

(0.79)

Knowledge quality

0.45

(0.30)

0.45

(0.29)

Knowledge generality

-0.97

(0.99)

-0.97

(0.97)

Patent stock

0.00

(0.00)

0.00†

(0.00)

Employees

0.14

(0.11)

0.11

(0.10)

Medical

0.06

(0.11)

0.03

(0.11)

% Medical

-0.66

(0.55)

-0.81

(0.53)

Firm Age

0.01*

(0.00)

0.01

(0.00)

% Acquired Units

1.37*

(0.54)

0.45

(0.64)

Employee growth

0.08

(0.08)

0.04

(0.08)

Sales growth

-0.39

(0.56)

-0.48

(0.58)

Recombination (lag)

0.00

(0.18)

-0.02

(0.17)

Total Units (instrument)

0.15*

(0.07)

N

(firms)

212

(71)

212

(71)

Log-likelihood -79.466 -76.310

Prob > chi2 0.000 0.000

Model improvement

Chi2 (1) 6.31*

Random-effects probit model. Dependent variable is dummy variable that takes the value 1 if the firm undertook any

structural recombination between t and t+1, 0 otherwise. Independent variables are measured at t, except

recombination lag, which is measured between t-1 and t. Model includes constant term. Figures in parentheses are

robust standard errors. Significance level (two-sided): † <10%, *<5%, **<1%, ***<0.1%

37

Table 3

Main Results

1 2 3 4 5 6 7

Recombination (t-1)

-4.63

(3.55)

3.56

(2.97)

-4.67

(3.45)

-4.25†

(2.28)

6.12

(4.53)

5.56

(4.21)

Recombination *

Unexpltd

relatedness

30.58†

(15.55)

41.36*

(18.34)

37.87*

(16.05)

Recombination x

Knowledge

generality

3.51

(32.80)

30.58

(40.46)

26.61

(39.75)

Recombination *

Knowledge quality

18.70*

(9.01)

18.11*

(6.85)

20.59*

(8.10)

Patent stock (lagged

DV)

0.59***

(0.15)

0.60***

(0.15)

0.54***

(0.13)

0.60***

(0.15)

0.58***

(0.15)

0.45**

(0.14)

0.50***

(0.13)

Unexploited

relatedness

-81.18

(60.63)

-85.60

(59.72)

-42.53

(55.50)

-84.72

(60.52)

-70.12

(53.40)

-6.58

(57.34)

-8.56

(48.98)

Knowledge

generality

43.06

(74.62)

44.78

(75.60)

44.54

(73.94)

47.50

(80.18)

73.13

(65.69)

82.66

(79.44)

96.36

(69.78)

Knowledge quality

-27.05

(32.78)

-22.90

(32.93)

-21.75

(32.59)

-22.95

(33.05)

-20.91

(32.79)

-9.00

(31.62)

-19.73

(32.62)

Medical sales

-38.70

(29.48)

-32.90

(24.18)

-27.43

(22.28)

-32.97

(24.13)

-26.81

(21.25)

-19.88

(19.47)

% Medical

40.75

(84.04)

43.39

(84.57)

30.07

(79.05)

42.46

(82.11)

14.48

(95.07)

-11.98

(80.77)

Firm Age

-0.34

(1.80)

-0.12

(1.98)

0.07

(2.05)

-0.12

(1.99)

-0.49

(1.96)

-0.28

(2.02)

Acquired Units

4.27

(4.21)

1.71

(3.93)

2.00

(3.94)

1.84

(3.40)

-5.11

(4.12)

-4.45

(4.26)

Sales Growth

20.95†

(11.70)

19.24†

(10.27)

17.67†

(9.40)

19.24†

(10.30)

17.80†

(9.50)

15.76†

(8.48)

No Patenting

(dummy)

0.48

(10.25)

0.23

(9.58)

-1.52

(8.98)

0.22

(9.56)

-2.71

(8.93)

-5.24

(9.50)

Inverse Mills

47.46

(65.27)

44.32

(65.53)

47.20

(65.20)

44.58

(65.77)

28.85

(60.65)

46.81

(69.31)

32.87

(60.25)

N

(firms)

212

(71)

212

(71)

212

(71)

212

(71)

212

(71)

212

(71)

212

(71)

R-square (within) 0.322 0.341 0.356 0.341 0.368 0.365 0.391

Dependent variable is patent stock at t+1. All models include firm and period fixed effects, as well as a constant

term. Figures in parentheses are robust standard errors. All models have Prob>F less than 0.1%.

Significance level (two-sided): † <10%, *<5%, **<1%, ***<0.1%

38

Table 4

Robustness and Supplementary analyses

Panel A: Robustness Panel B: Supplementary Analysis

Patent

Flow

Dynamic

Panel

Excluding

no patents

High

Unexpltd

Reltdness

Low

Unexpltd

Reltdness

High

Product.y

Low

Product.y

8 9 10 11 12 13 14

Recombination (t-1)

2.18

(5.29)

5.14

(4.97)

9.96†

(5.51)

0.11

(0.64)

-1.54

(2.38)

11.59

(7.58)

3.40

(2.19)

Recombination *

Unexpltd relatedness

39.98*

(19.24)

48.65**

(15.58)

49.90*

(21.48)

55.88†

(28.36)

-9.14

(11.91)

Recombination x

Knowledge generality

25.30

(45.63)

-17.46

(44.89)

51.97

(63.40)

9.67*

(3.89)

20.24

(77.68)

89.62

(69.81)

9.22

(8.62)

Recombination *

Knowledge quality

18.56*

(9.04)

17.88*

(8.63)

25.05**

(9.10)

-5.46*

(2.27)

22.05†

(11.38)

31.07**

(11.00)

-5.01

(4.03)

Patent stock (lagged

DV)

0.16*

(0.08)

0.76***

(0.09)

0.49***

(0.13)

0.73***

(0.01)

0.68***

(0.13)

0.34*

(0.15)

-0.13

(0.37)

Unexploited

relatedness

-49.15

(82.75)

-84.47

(85.56)

-22.27

(74.69)

1.16

(2.72)

6.85

(241.85)

-44.38

(339.78)

-12.42

(21.60)

Knowledge generality

93.25

(103.89)

81.02

(93.32)

178.81*

(76.18)

-1.59

(9.86)

63.83

(215.46)

59.70

(261.43)

30.51†

(15.89)

Knowledge quality

-19.66

(44.85)

-45.09†

(25.63)

-44.12

(46.19)

-3.00

(4.10)

-75.48

(85.96)

-77.72

(95.33)

-11.26*

(4.71)

Medical sales

-9.11

(13.47)

-4.59

(13.25)

-32.72

(33.85)

2.06*

(1.01)

-83.88*

(31.73)

-26.72

(81.21)

4.49

(4.44)

% Medical

-45.89

(101.45)

-6.53

(89.51)

-32.18

(152.52)

-7.33

(10.17)

253.57

(188.17)

211.24

(359.60)

28.41*

(12.92)

Firm Age

-2.62

(3.09)

7.75***

(1.92)

-0.90

(4.60)

-0.17

(0.17)

-2.88

(5.03)

-2.70

(9.42)

0.13

(0.41)

Acquired Units

3.24

(7.18)

-8.69

(5.36)

-5.82

(6.15)

2.74

(2.04)

-8.72

(7.46)

-4.72

(5.68)

-4.33†

(2.54)

Sales Growth

14.69

(10.09)

9.69

(7.63)

36.53†

(20.52)

-0.74

(0.47)

59.25*

(25.12)

26.62

(25.99)

5.66

(4.17)

No Patenting

(dummy)

-8.58

(13.55)

-25.60

(24.03)

-1.85*

(0.89)

77.51

(64.80)

91.84

(97.86)

0.12

(3.26)

Inverse Mills

61.94

(91.57)

15.59

(58.55)

10.62

(10.09)

9.32

(71.82)

61.37

(151.53)

-20.16***

(4.16)

N

(firms)

212

(71)

212

(71)

143

(53)

106

(44)

106

(39)

104

(52)

108

(49)

R-square (within) 0.257 0.453 0.971 0.458 0.439 0.414

Dependent variable is patent stock at t+1, except in model 8, where it is (citation-weighted) patents between t and

t+1. All models show fixed effect panel regressions, except for model 9, which is a system dynamic panel model

(xtdpdsys in STATA). ‘High’ and ‘Low’ refer to values above and below the sample median respectively. All

models include a constant term. Figures in parentheses are robust standard errors.

Significance level (two-sided): † <10%, *<5%, **<1%, ***<0.1%

39

Figure 1a

Moderating Effect of Unexploited Relatedness

Figure 1b

Moderating Effect of Knowledge Quality

Figures are based on model 7 in Table 2. Figures show change in patent stock based on predicted values when all

other variables are held at their sample means. High and low values refer to upper quartile and lower quartile values

respectively.