Taking technologies off the shelf: The technological impact of
reactivating shelved technologies
University of Groningen Faculty of Economics and Business
Department of Innovation Management and Strategy Research Master Thesis
Supervisors: D.L.M. Faems and P.M.M. de Faria
Holmer Jan Kok S1771663 22 June 2014
Abstract
Creating new inventions is a difficult and inherently uncertain process. The firm cannot predict how soon, or whether, the new inventions it creates will be used as inputs for subsequent recombinant efforts. As a consequence, some technologies emerge which the firm will not immediately further develop upon, in which case they will be figuratively put on the shelf. Nevertheless, these technologies remain an integral part of the firm’s technological base; therefore the firm should seek to exploit them to their maximum potential. The fundamental question which this study aims to answer is: Can firms create impactful new inventions by reactivating shelved technologies? In this study, shelved technologies are technologies which the firm created but has not further developed for an extended period of time. Accordingly, reactivating a shelved technology implies that the originating firm recombines the shelved technology for the first time after creation. To examine the recombinant value of these technologies, we constructed a database containing 45512 patent applications which were subsequently granted to 186 R&D-intensive firms from five high-tech sectors between 1995 and 2003. Our findings indicate that reactivating shelved technologies has a negative impact on technological performance but that external use of these technologies influences their recombinant value positively. Our study demonstrates that the frequency of use of old knowledge is an important determinant of its recombinant value.
1
INTRODUCTION
The recombination of existing knowledge is the principal source of inventive novelty (Henderson and Clark, 1990; Galunic and Rodan, 1998; Fleming, 2001; Schoenmakers and Duysters, 2010; Carnabuci and Operti, 2013; Gruber et al., 2013). Inventors disentangle existing systems, withdrawing and recombining useful components to create novel combinations (Hargadon and Sutton, 1997). In this context, components refer to the elements that constitute an invention (Fleming, 2001), e.g., the rotating blades in mechanical fans or the magnetic needle in modern compasses.1
Since inventions are rapidly progressing and becoming more complex, firms are increasingly required to search for new knowledge components in more distant and less obvious places (March, 1991; Katila and Ahuja, 2002; Laursen and Salter, 2006; Leiponen and Helfat, 2010). Firms engage in search activities (Cyert and March, 1963; Katila and Ahuja, 2002) across different dimensions; they search across and within familiar technological domains, organizational boundaries, country borders, but also over time (see Li et al., 2008 for a review). Once accessed, the recombination of these different components can yield important strategic outcomes for the firm (Rosenkopf and Nerkar, 2001; Katila and Ahuja, 2002; Nerkar, 2003; see Laursen, 2012 for a review).
Out of the four search dimensions described in the literature, the technological dimension has received the most attention (e.g., Rosenkopf and Nerkar, 2001; Miller et al., 2007; Nemet and Johnson, 2012; Gruber et al., 2013) while the temporal dimension has received the least (e.g., Katila, 2002; Nerkar, 2003; Heeley and Jacobson, 2008). As noted by Heeley and Jacobson (2008), prior research provides limited insights into the difference in recombinant value of old knowledge compared to new knowledge. Compared to recent knowledge, old knowledge is generally well-understood because it has been around for longer
1
2
(Ahuja and Lampert, 2001; Heeley and Jacobson, 2008) and it can unveil previously-ignored paths of development (Garud and Nayyar, 1994; Nerkar, 2003). At the same time, old knowledge is often perceived as obsolete and unrepresentative of current alternatives (Sørensen and Stuart, 2000; Ahuja and Lampert, 2001; Katila, 2002). Jointly, this would suggest that recombining old knowledge can either benefit or hurt technological performance. Despite this, indications are present that firms can benefit from sourcing and recombining old knowledge. For example, Katila (2002) found that using old outer-industry knowledge contributes to new product introductions in the robotics industry. Moreover, Nerkar’s (2003) findings indicate that recombining old knowledge with new knowledge has a positive impact on technological performance in the pharmaceutical industry. Recently, Wang and Hagedoorn (2014) found that firms can increase the returns to internal R&D by reusing old knowledge. To give more concrete examples, revisiting the old idea of using heat as a technological component proved to be decisive for the creation of Hewlett-Packard’s thermal ink-jet (Fleming, 2002). More recently, the rejuvenation of tactile feedback screens by Apple was an evident success, showing that old ideas need not necessarily be obsolete or unrepresentative of current alternatives (Nerkar, 2003).
3
uncertain process (Fleming, 2001). The firm cannot foresee whether it will be able to immediately further develop upon the outcomes of recombinant processes (Garud and Nayyar, 1994). As a result, some internal technologies will remain undeveloped for a prolonged period of time. However, if the firm is able to reuse such old technologies in new and useful recombinant efforts, then it can effectively increase its returns to internal R&D (Wang and Hagedoorn, 2014). Thus, in this study we will show whether firms can create significant positive value from reevaluating their technological base and revisiting old unused technologies. Our second contribution is that we will examine the technological impact of varying frequencies of use of old technologies on their recombinant value. Earlier studies have covered several contingencies upon which the value of old knowledge depends (e.g., Katila, 2002; Nerkar, 2003), but have not explicitly studied the impact of frequency of use on the value of old knowledge. By specifically focusing on the reactivation of shelved technologies, we can examine this effect in more detail. To be more specific, these are technologies for which the effect of frequency of use on recombinant value will be clearly visible because their internal frequency of use is low by definition. In addition, even though they were shelved by the originating firm, this does not imply that these technologies will not be recombined by other parties. Therefore, whereas their internal frequency of use is low, their external frequency of use can vary. Taken together, the primary aim of this study is to find out whether reactivating shelved technologies has a significant impact on technological performance.
4
of knowledge search by focusing on the reactivation of shelved technologies. In the process, we will provide important insights into the extent to which the internal and external frequency of use of old knowledge influences its recombinant value. Relying on a consolidated database containing 45512 patents applied for, and subsequently granted, by 186 firms in five key high-tech industries at the European Patent Office (EPO) in the period between 1995 and 2003, we will test the technological performance implications of reactivating shelved technologies.
Our findings indicate that reactivating shelved technologies has a negative impact on the technological performance of a new invention. However, our findings also indicate that the external use of shelved technologies contributes positively to their recombinant value. Hence, our findings imply that prior knowledge use is an important determinant of recombinant value. We contribute to the innovation literature by showing to what extent the value of old knowledge is influenced by its internal but also external frequency of use. Moreover, embedded in the broad literature on open innovation (Chesbrough, 2003; Laursen and Salter, 2006), our findings suggest that firms should not be overly protective of their own knowledge (Laursen and Salter, 2014) and should instead seek to leverage their existing technological basis by making shelved technologies more easily accessible to other firms (Henkel, 2006).
5
THEORY AND BACKGROUND
Knowledge recombinationFirms thrive by creating and introducing new technologies to the market allowing them to sustain their competitive advantages and maintain satisfactory performance (Barney, 1991; Nonaka, 1994). Hence, understanding the origins of new inventions is important (Ahuja and Lampert, 2001; Fleming, 2001). In this study, an invention is the creation of a new technology whereas an innovation is the commercialization of a new technology (Arthur, 2007). Scholars (e.g., Henderson and Clark, 1990; Galunic and Rodan, 1998; Hargadon, 2002; Fleming, 2001; Nerkar, 2003) have conceptualized the creation of inventions as a recombinant process whereby firms create new inventions by recombining existing knowledge components with each other. Henceforth, we will rely on the theory of recombinant invention (Fleming, 2001) to study the technological impact of inventions which reactivate shelved technologies.
6
7 Temporal dimension of knowledge search
The identification of a problem triggers the firm to search for solutions to it (Cyert and March, 1963; Katila and Ahuja, 2002; Nickerson and Zenger, 2004). The search for new components for recombination constitutes such a problem-solving search. Searching for new components has been described as occurring along a continuum, where one end corresponds to distant exploratory search and the other end to local exploitative search (March, 1991; Katila and Ahuja, 2002; Laursen and Salter, 2006). Broadly speaking, distant search refers to searching for new and unfamiliar components, whereas local search refers to searching for familiar components (Stuart and Podolny, 1996; Fleming, 2001). While useful, this conceptualization of knowledge search as falling on a continuum between local and distant search is incomplete if it is not complemented with the various dimensions along which knowledge is sought. Scholars have shown that firms need to balance knowledge search simultaneously across technological (Rosenkopf and Nerkar, 2001; Miller et al., 2007; Nemet and Johnson, 2012; Gruber et al., 2013), organizational (Rosenkopf and Nerkar, 2001; Cassiman and Veugelers, 2006; Rothaermel and Alexandre, 2009), geographical (Phene et al. 2006), but also temporal spaces (Katila, 2002; Nerkar, 2003; Heeley and Jacobson, 2008). The search over time for components has received the least attention. Specifically, searching for old components has often been dismissed as a poor strategy because old components tend to generate less impact due to their supposed obsolescence (Sørensen and Stuart, 2000; Ahuja and Lampert, 2001; Rosenkopf and Nerkar, 2001; Fabrizio, 2009). However, indications are present that recombining old knowledge can produce positive results (Katila, 2002; Nerkar, 2003; Wang and Hagedoorn, 2014).
8
Wang and Hagedoorn, 2014). For example, a specific advantage of recombining old knowledge is that it may result in uncovering valuable overlooked paths of development (Garud and Nayyar, 1994; Nerkar, 2003). At the time of creation of the technology, such paths may have been neglected due to inventors’ bounded rationality or because complementary assets were missing (Nerkar, 2003). At the same time, an important disadvantage of recombining old knowledge is that, especially in technological domains evolving at a rapid pace (Fabrizio, 2009; Nadkarni and Chen, 2014), firms need to incorporate the most recent and current technological alternatives into new inventions to keep up with competitors (Sørensen and Stuart, 2000; Katila, 2002; Nerkar, 2003; Fabrizio, 2009) but also to avoid getting excluded from early developments in emerging markets (McGrath, 1997; Katila, 2002). In these rapidly-evolving markets, new technologies whose attributes are not well-aligned with the current or future state of technology will not be used very often and will only have limited value. Finally, it should be noted that the value of old knowledge is generally dependent on the firm’s ability to recollect important details pertaining to the recombination of this knowledge. If the firm fails to retain important details pertaining to its recombination over time, then recombinant outcomes based on the old knowledge will tend to be rather poor (Garud and Nayyar, 1994; Marsh and Stock, 2003, 2006).
Aside from these relatively specific pros and cons, temporal search scholars have primarily based their arguments for and against using old knowledge on the relation between the frequency of use of old knowledge and the benefits and costs associated with its recombinant use. Curiously however, these scholars do not explicitly recognize frequency of use as a principal dimension of old knowledge recombination. As a result, they only provide a limited conceptualization of its role in affecting technological value. In this study, we emphasize to what extent the frequency of use of old knowledge affects its recombinant value.
9
its value. Firstly, the recombination of unfrequently used old knowledge is more unpredictable and uncertain. It has been argued that recombining recent knowledge is less difficult because it “conserves cognitive capabilities” (Katila, 2002, p. 996). Moreover, firms tend to use embedded routines when using recent knowledge and therefore tend to make fewer errors (Nerkar, 2003). In contrast, firms tend to engage in effortful trial-and-error processes when using old knowledge (Nerkar, 2003). Thus, from these points of view, using old knowledge is difficult because it requires considerable cognitive efforts on the part of inventors. This can be explained by the fact that the old knowledge which Katila and Nerkar discuss has not been used frequently. Specifically, it has been argued that frequently using the same components makes their use more predictable and certain and reveals their underlying relationship with other components (Fleming, 2001; Katila and Ahuja, 2002). As implied above, the old knowledge in question requires considerable efforts to recombine and so its recombinant use is likely unpredictable and uncertain. Therefore, we can argue that compared to frequently used old knowledge, the recombination of unfrequently used old knowledge is more unpredictable and uncertain.
10
contributes to product observability and consequently makes the value and recombinant potential of the technology become clearer and certain over time (Zander and Kogut, 1995). As a technology becomes used more often in new products, it becomes easier to access relevant information and knowledge related to it. For example, the more often a technology is used to create a new product, the more products the firm can acquire and reverse engineer to obtain knowledge pertaining to the technology (Zander and Kogut, 1995). In sum, for knowledge to become legitimized, easily evaluated, and well-known throughout an industry, its frequency of use must be high and its use should be widely spread rather than confined to a single firm.
Thirdly, unfrequently used old knowledge is easily forgotten and difficult to access and may therefore be perceived as more unique by most firms (Barney, 1991; Katila, 2002). Generally, the firm and its close competitors search within the same knowledge space and become familiar with the same knowledge components (Katila and Chen, 2008). Thus, in relation to competitors’ search activities, the firm will regard knowledge located outside this knowledge space (i.e., unfrequently used old knowledge) as more unique and valuable. Assuming the knowledge has successfully been accessed, this perceived uniqueness will stimulate the firm to spend more resources on recombining it. As a result, recombinant outcomes based on this type of knowledge will be more useful and novel on average. For similar reasons, the value of recombining frequently used and easily-accessible old knowledge might be lower.
11
reach its limits quite rapidly (Fleming, 2001; Fleming and Sorenson, 2001; Katila and Ahuja, 2002). Specifically, since old technologies have been around for longer and generally been used more frequently, scholars suggest that the recombinant potential of old technologies is more likely to have been exhausted already compared to new technologies (Ahuja and Lampert, 2001; Fleming, 2001; Katila, 2002; Heeley and Jacobson, 2008; Kelley et al., 2013).
In sum, the value of old knowledge is contingent upon the frequency at which it has been used. Yet, none of the studies on the temporal dimension of search have paid sufficient attention to this aspect. As a result, it is as of yet unclear whether the frequency of use of recombined old knowledge has implications for technological performance. We think it is important to examine this topic in more detail.
Shelved technologies
Firms invest considerable resources into the creation of new technologies (Artz et al., 2010). However, the outcomes of this process are uncertain (Fleming, 2001). The firm cannot foresee whether the new technologies will be immediately useful in subsequent recombinant efforts (Garud and Nayyar, 1994). Consequently, some technologies which the originating firm will not subsequently recombine for a long period of time are bound to emerge. We refer to these types of technologies as shelved technologies.
12
early’ and should therefore be retained until they can finally be useful (Garud and Nayyar, 1994). Thus, we study the technological impact of reactivating shelved technologies to show whether and when firms can create significant value from unused technologies in their stock of technologies. Another reason to study the reactivation of shelved technologies is that it allows us to show whether the frequency of use of old knowledge significantly influences its recombinant value. Specifically, since the internal frequency of use of shelved technologies is inherently low, examining these technologies allows us to identify whether the internal frequency of use of old knowledge has an impact on its recombinant value. Moreover, the fact that these technologies are not necessarily inaccessible to other firms lets us examine the impact of external use on the recombinant value of old internal knowledge.
To better understand the reactivation of shelved technologies, we must know why firms would shelve them in the first place. The following part will present five distinct reasons for shelving technologies. Firstly, to further develop technologies, accessing assets complementary to their exploitation is vital (Garud and Nayyar, 1994; Nerkar, 2003). An individual resource by itself cannot yield many benefits and therefore needs to be complemented by additional resources when it is deployed on the market (Kraaijenbrink et al., 2010). For example, marketing knowledge is an essential complement to most technological knowledge (Fang et al., 2013). Thus, in the absence of complementary resources, it is possible that the created technology simply cannot be further developed internally.
13 useful (Fleming, 2001).
Thirdly, it has been argued that bounded rationality limits inventors’ ability to recognize certain paths of development (Nerkar, 2003). This is because rationally bounded individuals cannot make perfect decisions due to the limited amount of information they possess (Cyert and March, 1963). Failure to consider the potential paths of development which new technologies could create may be a direct result of such cognitive limitations (Levinthal and March, 1993; Nerkar, 2003). If these paths remain hidden, the firm will not find immediate use for the new technology and will therefore not use it for some time. It should be noted that this does not necessarily imply that the technology is not valuable but merely that the firm cannot yet recognize its real value.
Finally, it must be noted that the firm may shelve a technology simply because it is, and will probably never be, very useful. As we established before, the recombinant process is rather uncertain (Fleming, 2001). Therefore, it is possible that the firm creates technologies which end up being of low value. Since it considers these technologies to be of low value, rather than to spend efforts on trying to sell them on the market to other firms, the firm will simply shelve them2.
Thus, we contend that certain technologies are considered as shelved when they are not further developed internally after creation for an extended period of time. This creates two possibilities: (i) the technologies remain shelved over the entire length of their existence or (ii) the originating firm reactivates them at a certain point in time. In this study, we examine the latter possibility and discuss the performance implications of reactivating shelved technologies.
HYPOTHESES
Taking technologies off the shelf
In this study, our main contention is that reactivating shelved technologies is a valuable strategic move for the firm because it can create new paths of development. Earlier, we
2
14
discussed that an important reason for why old technologies are perceived as less valuable is that they have been used very frequently such that their technological potential has already been exhausted (Ahuja and Lampert, 2001; Heeley and Jacobson, 2008). We also established that shelved technologies have not been recombined internally after creation. Consequently, this increases the likelihood that their technological potential has not yet been exhausted (Fleming, 2001). In addition, firms sometimes shelve technologies because they lack the resources to further develop them or the information to recognize how they could be further developed. Garud and Nayyar (1994) and Nerkar (2003) argued that old technologies can often open up new paths of development which were neglected at the time due to missing complementary resources or inventors’ bounded rationality. However, once critical complementary resources and information are accessed by the firm, the reactivating of the shelved technology becomes a possibility. Thus, once concerns of missing resources and bounded rationality, which in the past blocked the firm’s ability to further develop the technology, are reduced, the firm will probably be able to identify and create new paths of development. In sum, these insights culminate in the following hypothesis:
H1a: New inventions that reactivate shelved technologies will generate a higher technological impact than new inventions which do not reactivate shelved technologies.
15
place (Nerkar, 2003). We also noted that using a technology repeatedly makes its use more reliable and predictable, and it creates a deeper understanding of its underlying concepts (Fleming, 2001; Katila and Ahuja, 2002). Hence, because shelved technologies have not been further developed internally, the focal firm might find its use more unreliable, unpredictable and difficult to understand. Successfully further developing upon them might therefore be problematic.
Finally, we argue that the firm may not be able to retrieve all important information and knowledge relevant to the shelved technology when it reactivates it. As a result, key aspects of the shelved technology may not be fully understood and properly incorporated in the new technology. Consequently, the new invention that reactivates the shelved technology will be of lower quality than if the shelved technology had been better understood and incorporated. This may occur because successfully integrating a technology of the past into a new recombinant effort strongly depends on the proper maintenance of knowledge pertaining to this technology over time (Garud and Nayyar, 1994; Katila, 2002; Nerkar, 2003; Marsh and Stock, 2003, 2006). This is pertinent because knowledge depreciates rapidly losing much of its value within 5 years of creation (Dierickx and Cool, 1989; Darr et al. 1995; Katila and Ahuja, 2002). Hence, resources have to be spent on maintaining knowledge so that it retains its quality up until its future retrieval (Garud and Nayyar, 1994; Argote, 2012). Maintaining the quality of shelved technologies can however become difficult if knowledge pertaining to it is poorly maintained or lost altogether. For example, in case important knowledge pertaining to the shelved technology inadvertently exits the firm due to organizational downsizing (Schmitt et al., 2012), the possibility that reactivating shelved technologies will generate a low technological impact is increased. In sum, the new invention that reactivates a shelved technology may end up being of relatively low value:
16 External use of shelved technologies
As the name suggests, shelved technologies were figuratively put on the shelf by the firms who created them. Nevertheless, this does not necessarily imply that other firms were not able to access and subsequently recombine these technologies. We will argue that the external frequency of use of shelved technologies can have a positive impact on its recombinant value.
Our arguments are based on the idea that the frequency of use of knowledge is directly related to its availability and salience (Levitt and March, 1988; Fleming, 2001). Shelved technologies which have not been widely diffused throughout the industry will be poorly understood and difficult to access and evaluate for other firms (Levitt and March, 1988; Katila, 2002). Moreover, these technologies will not have reached sufficient levels of acceptance by customers in the industry (Heeley and Jacobson, 2008). Thus, other firms may attribute a low value to these shelved technologies, perceiving them as less valuable and unrepresentative of current alternatives (Katila, 2002; Nerkar, 2003). As a result, there will be little understanding of how the new invention that recombines the shelved technology works but also little incentive to explore its recombinant opportunities. In contrast, when the external frequency of use of shelved technologies increases, the technological potential of these technologies is revealed and the focal firm and its competitors will be given a better understanding of how they work and how they can be further built upon (Zander and Kogut, 1995; Yang et al., 2010). Moreover, the invention that reactivates this shelved technology will be perceived as more legitimate and will more easily achieve industry-wide acceptance (Heeley and Jacobson, 2008). These mechanisms will increase firms’ incentive to further develop the invention that reactivates the shelved technology.
17
will have retained some unique insights about how these technologies can be recombined. After all, the focal firm is where these technologies were originally developed. Consequently, it is possible that the original inventors of these technologies still possess unique insights about their architectural design (Henderson and Clark, 1990). This unique internal knowledge, combined with the knowledge accumulated through observing the product being used externally, will lead to a recombination of the shelved technology by the originating firm which is of superior quality and usefulness. In sum, we hypothesize:
H2a: The more frequently the shelved technology has been used externally, the higher the technological impact of new inventions that reactivate the shelved technology will be.
Although external use can increase the value of shelved technologies, we posit that this causal relationship can also run in the opposite direction. Consider the following example. A firm creates an inherently valuable technology which it cannot further recombine due to a lack of important complementary resources. Consequently, it is forced to shelve this technology. However, other firms with different resource configurations will also recognize the value of the now shelved technology and use it in new recombinant efforts. Since they perceive the technology as valuable, they will devote more resources to recombining it. As a consequence, the recombinant potential of shelved technologies which from the moment of creation are inherently valuable will be exhausted more quickly through external use. Firms will quickly run down the recombinant potential of valuable technologies, cherry-picking the best characteristics of the technology and using these for the creation of new inventions. Accordingly, rather than to increase the value of shelved technologies, external use will decrease the recombinant potential and value of shelved technologies. Thus, we hypothesize:
18
METHODOLOGY
Sample and data
The analysis was carried out using data on 186 firms from five key R&D-intensive sectors which were drawn from the ‘2004 EU Industrial R&D Investment Scoreboard’. The five key sectors were pharmaceuticals and biotechnology, electronics and electrical machinery, chemicals, IT hardware, and engineering and general machinery. The 186 firms are the largest R&D spenders in each of these sectors.
We studied patents applied for, and subsequently granted, by these 186 firms between 1995 and 2003. However, we made an important caveat by only including patent applications and patent citations made by these 186 firms. That is to say, while the data on applied and granted patents were accurate and complete, the backward citations in our sample were solely based on those made by the firms in the sample. As a result, we cannot exclude the possibility that our data underestimated the number of backward patent citations on each patent. Moreover, all variables constructed therefrom, most notably the frequency of use of patents, may have been lower than if patent data of all firms in the population had been available3. Nevertheless, since we focused on the largest R&D spenders in each sector, we are reasonably confident that the majority of patent applications in all five sectors in the sample were covered. More exactly, it has been noted that large firms are responsible for the majority of patenting activities (Arundel and Kabla, 1998; Ahuja and Lampert, 2001; Nerkar, 2003). For example, in his sample of patents belonging to the 33 largest R&D spenders in the pharmaceutical industry, Nerkar (2003) estimated that his data accounted for 70% of all patent grants in that industry.
In this research, we thus relied on patent applications to study inventions. Although patents are a frequently used appropriation mechanism, firms also rely on other mechanisms such as secrecy and lead time to protect their knowledge assets (Cohen et al., 2000). Nevertheless, to the extent that patent applications accurately reflect the creation of inventions
3
19
(Ahuja and Lampert, 2001; Fleming, 2001), they can reliably be used as indicators of inventive activities, especially since our sample only includes firms from high-tech sectors wherein patent propensity rates are generally high (Arundel and Kabla, 1998).
We restricted our analysis to patent applications which were subsequently granted (Belderbos et al., 2014). When studying the technological impact of inventions, it is more useful to study granted patents because these have passed patent examiners’ evaluation of patentability and thus better reflect the novelty, non-obviousness, and usefulness of inventions. To be clear, we only included patents which were applied for and subsequently granted between 1995 and 2003. Therefore, the data is truncated to the right because EPO patent applications on average take four years to be granted (Belderbos et al., 2014). However, patent applications can receive citations well before they are granted; hence, our citations count will not be affected. To control for the truncation effect, we include application year dummies in all models.
20 Measures
Dependent variable
To measure technological impact, we relied on forward citations. Forward citations have often been used to indicate the usefulness of inventions (Trajtenberg, 1990; Sørensen and Stuart, 2000; Ahuja and Lampert, 2001; Fleming, 2001; Rosenkopf and Nerkar, 2001; Reitzig, 2004). That is, a high citation count indicates that a technology is frequently being used as an input for new inventions. To avoid erroneous findings based on this variable, we applied a fixed four-year window to forward citations. Schoenmakers and Duysters (2010) discuss that most EPO forward citations occur within four to five years, so this particular fixed window was appropriate to measure the technological impact of inventions. For example, if a patent was applied in 1995, we counted the number of times it was cited in the years 1995 up until 1999.
Independent variables
21
technology for patents that had at least 1 shelved technology as a backward citation.
Approximately 3.5% of all patents qualified as such and the share of reactivating patents remained relatively stable over the years as indicated below in table 1. However, the total number of patents declined over the years due to the truncation effect of patent grants described earlier.
TABLE 1
Share of reactivating EPO patents (1995-2003)
1995 1996 1997 1998 1999 2000 2001 2002 2003 Reactivating patents 194 199 194 220 201 206 162 126 100 % total 3,25% 3,23% 3,11% 3,65% 3,49% 3,89% 3,61% 3,82% 4,37%
Total grants 5971 6162 6228 6035 5753 5289 4485 3300 2289
To measure the external frequency of use of these shelved technologies, we counted the number of times a shelved technology had been cited externally by the other 185 firms in the sample between its application date and first internal citation. We called this variable external
use shelved technology and assigned each reactivating patent the respective number of times
the shelved technology it recombined had been used externally.
Control variables
22
intellectual property than others” (p. 193). Conversely, Ahuja and Lampert (2001) found that patents which had few technological antecedents generated a high technological impact. They argue that truly novel technologies have no or few technological antecedents. Hence, to control for these confounding effects, we included a control variable backward citations that counts the number of backward citations of a patent application. We also included a variable that counts the number of three-digit IPC classes in which the patent was applied. Patents that cover many technological fields likely receive more forward citations as their scope of application is larger (Nerkar, 2003; Miller et al., 2007; Belderbos et al., 2014). To control for this we included a control variable technological scope. Finally, in each regression we added firm dummies, year dummies, and a full set of three-digit IPC class dummies.
Analytical method
23 Results
Descriptive statistics
Fig. 1 indicates that the distribution of forward citations was highly skewed to the right. That is, 16.7% of patents received 0 forward citations in the first four years after application while the majority of patents received fewer than 10 forward citations (about 71%), and only 1% of patents received more than 60 forward citations in the first four years. These results are fairly consistent with existing research on patent citations (e.g., Ahuja and Lampert, 2001; Schoenmakers and Duysters, 2010; Nemet and Johnson, 2012).
FIGURE 1
Distribution of fixed four-year window forward citations
Tables 2 and 3 indicate that pairwise correlations were sufficiently low to rule out multicollinearity issues. VIF analyses, not reported here, also confirmed that multicollinearity was not an issue in our models. We split the data up into two parts to be able to test the moderating effect of the external use of shelved technologies on technological impact. Thus, table 3 only concerns the patents which recombined at least one shelved technology
n = 45512 0 .01 .02 .0 3 .04 .05 .06 Den sity 0 50 100 150 200 250 300 350
24 TABLE 2
Descriptive statistics full sample
Variables Mean SD 1 2 3 4 5 1 Technological impact 8,8050 13,1412 2 Shelved technology 0,0352 0,1843 -0,0496 3 Age 5,4996 3,7052 -0,0983 0,2691 4 Technological scope 1,4572 0,7344 0,0265 -0,0017 0,0242 5 Backward citations 1,7826 1,3289 0,0669 0,0641 -0,0191 0,0286 6 Self-citations 0,4375 0,4561 -0,0333 0,1679 -0,1499 0,0378 0,0558 TABLE 3
Descriptive statistics reactivating patents sample
Variables Mean SD 1 2 3 4 5
1 Technological impact 5,3933 7,5444
2 External use shelved technology 0,7878 1,5060 0,1018
3 Age 10,7193 3,3863 -0,1167 -0,1253
4 Technological scope 1,4507 0,7450 0,065 -0,0341 -0,0093
5 Backward citations 2,2285 1,9774 0,0748 0,093 -0,3216 0,0458
6 Self-citations 0,8384 0,2407 -0,097 -0,2649 0,2911 -0,012 -0,314
Hypotheses testing
Table 4 provides the results of the negative binomial regressions. Models 1 and 4 are the baseline models which only included the control variables. Consistent with prior research, the mean age of backward citations had a negative impact on the dependent variable (Rosenkopf and Nerkar, 2001; Miller et al., 2007). Moreover, the number of three-digit IPC classes and number of backward citations had a positive impact on the dependent variable (Nerkar, 2003; Belderbos et al., 2014), though they were not significant in the models restricted to shelved technologies. Finally, the number of self-citations was negatively related to the number of forward citations a patent received, again consistent with prior research (Sørensen and Stuart, 2000).
25
support for hypothesis 1b which predicted a negative relationship between reactivating a shelved technology and technological impact. Finally, model 5 tested the relationship between the external frequency of use of shelved technologies and technological impact. The significant and positive (p = 0.017) sign of external use shelved technology provides support for hypothesis 2a.
TABLE 4
Robust negative binomial regressionsa
Model 1 Model 2 Model 3 Model 4 Model 5 All patents All patents All patents Reactivating
patents
Reactivating patents
Shelved technology (H1) -0.297*** -0.149*** (0.0333) (0.0353)
External use shelved technology (H2) 0.0510**
(0.0213) Age -0.0279*** -0.0259*** -0.0197* -0.0201** (0.00192) (0.00203) (0.0103) (0.0102) Technological scope 0.0330** 0.0329** 0.0321** -0.131 -0.132 (0.0156) (0.0156) (0.0156) (0.0835) (0.0834) Backward citations 0.0501*** 0.0537*** 0.0515*** 0.0260 0.0260 (0.00476) (0.00482) (0.00477) (0.0241) (0.0241) Self-citations -0.0635*** 0.00207 -0.0511*** -0.376** -0.316** (0.0152) (0.0150) (0.0156) (0.154) (0.158) Observations 45512 45512 45512 1602 1602 Pseudo R2 0.022 0,022 0.022 0.049 0.050 Log Likelihood -142099.8 -142206.4 -142090.1 -4183.9 -4181.3 a
Significance levels are indicated by *10%, **5%, ***1%. Robust standard errors in brackets. Firm, three-digit IPC, and year dummies included in all models.
26 TABLE 5
Mean forward citations of reactivating patents All patents Reactivating patents Type of Patent Mean forward citations External use shelved technologya Mean forward citations Regular patent 8,93 0 4,87 Reactivating patent 5,39 1 5,59 2 6,62 3 5,86 4 7,28 5 9,33 6 6,47 7 6,4 a
We only included external use of shelved technologies between 0 and 7 because very few shelved technologies had been used more than 7 times externally in our sample (only 10 out of 1602 shelved technologies).
Sensitivity analyses
TABLE 6 Sensitivity analysisa
Model 6 Model 7 Model 8 Model 9 Model 10 Model 11 Model 12
Full sample (1997-2003) Full sample (internal forward citations) Full sample (external forward citations) Full sample (Poisson regression) Full sample (minus unused shelved technologies) Only internal backward citations (internal use = 0)
Only internal backward citations (internal use > 0)
Shelved technology -0.116*** -0.158*** -0.146*** -0.155*** -0.0237 (0.0417) (0.0484) (0.0352) (0.0372) (0.0549)
Age -0.0236*** -0.0185*** -0.0276*** -0.0283*** -0.0259***
(0.00220) (0.00262) (0.00209) (0.00216) (0.00206)
Age internal components -0.0406*** -0.0290***
(0.00358) (0.00374)
External use internal components 0.0812*** 0.0461***
(0.0150) (0.00845) Age internal components*External
use internal components 0.00201 -0.00323*
(0.00330) (0.00188) Technological scope 0.0237 0.0281 0.0362** 0.0410** 0.0364** 0.0601* 0.00677 (0.0173) (0.0234) (0.0152) (0.0173) (0.0157) (0.0307) (0.0309) Backward citations 0.0448*** 0.0666*** 0.0469*** 0.0503*** 0.0522*** 0.0504*** 0.0228*** (0.00560) (0.00652) (0.00474) (0.00473) (0.00483) (0.0171) (0.00624) Self-citations -0.0313* 0.212*** -0.116*** -0.0467*** -0.0498*** -0.122* -0.105* (0.0181) (0.0217) (0.0155) (0.0171) (0.0157) (0.0680) (0.0540) Observations 33379 45512 45512 45512 44507 11667 12060 Pseudo R2 0.023 0.034 0.027 0.124 0.022 0.027 0.025 Log Likelihood -103430.7 -76231.4 -132272.6 -308181.7 -139469.2 -35122.2 -38218.7 a
28
Our next step was to examine in more detail the relationship between the frequency of use of technologies, their age, and technological impact in models 10, 11, and 12. First, we created new variables including external use internal components and internal use internal
components which respectively measured the number of times internal backward citations had
been used externally and internally. Moreover, we created a variable age internal components which measured the age of the internal backward citations of a patent. Finally, since we interacted these variables with each other, we also mean-centered them to facilitate interpretation and reduce multicollinearity concerns (Aiken and West, 1991).
The results in model 10 indicate that external use of shelved technologies effectively strongly reduced the negative influence of reactivating shelved technologies on technological performance. Once we excluded all shelved technologies which had not been used by anyone after their original creation until their first internal recombination, the dummy variable shelved
technology turned insignificant (p = 0.666). Thus, external use would seem to be a positive
29
for hypothesis 2a as it indicates that external use contributes to the recombinant value of old internal technologies as long as these have not yet been subsequently recombined by the focal firm yet.
FIGURE 2
Interaction between age of internal components and external use (internal use > 0)
DISCUSSION AND CONCLUSION
FindingsThe existing literature on the temporal dimension of knowledge search has examined whether the firm can produce impactful inventions by relying on old knowledge (Katila, 2002; Nerkar, 2003; Heeley and Jacobson, 2008). However, it has not studied whether the frequency of use of old knowledge has an impact on its recombinant value. To fill this research gap, we examined the technological performance of new inventions that reactivate shelved technologies. Our main contention was that firms can find value in their existing stock of technologies and reactivate technologies which they had put on the shelf long ago. Our primary findings indicate that reactivating shelved technologies has a negative impact on technological performance. However, we found that this negative impact is reduced when the external use of shelved technologies increases. We find additional support for both these results in our sensitivity analyses. It would therefore seem that the value of old internal knowledge is contingent upon the frequency at which it has been used internally and externally.
0 1 2 3 4 5 6
New knowledge Old knowledge
30
Firstly, our results convincingly show a robust negative relation between reactivating shelved technologies and technological impact. Even after controlling for age, we find that the reactivation of shelved technologies has a strong negative impact on the technological impact of inventions. This finding resonates with the notion that technologies which are shelved are, in fact, shelved for a good reason. That is to say, it would seem that the firm will find further use for truly useful inventions soon after their creation. In contrast, inventions which take a long time to finally be internally recombined (i.e., ranging between 9 and 12 years according to our thresholds; see appendix) may not be very useful.
Recently, Wang and Hagedoorn (2014) suggested that firms can increase their returns to internal R&D by reusing useful but under-utilized old knowledge. However, they merely focused on the quantity of output (i.e., number of patents) rather than the quality of the knowledge that is created. While reusing old knowledge to increase returns to internal R&D is certainly possible, this does not have any direct implications for the quality of knowledge creation. Instead, our results show that reusing old and under-utilized knowledge has a negative impact on technological performance. This weakens the arguments that shelved technologies have much useful recombinant potential left because they have not been internally recombined yet or because they emerged ‘too early’ (Garud and Nayyar, 1994; Nerkar, 2003). Instead, our results would suggest that firms ‘dig into empty holes’ when reactivating these technologies (Heeley and Jacobson, 2008, p.726), or that they end up discovering technological paths of
development which are not relevant anymore.
31
also to be more easily evaluated. As a result, upon internal reactivation, other firms will be better able to access and recombine the technology which embeds the unshelved technology.
On top of this, as real options logic suggests (McGrath, 1997), it is possible that the focal firm will have observed how the technology performs externally hence reducing uncertainty about how it should be further recombined. Then, upon reactivation, the firm can combine what it learned from these external developments with its own unique knowledge about the technology. As a consequence, the recombinant outcome will be accessible to other firms but will also include unique knowledge elements added by the focal firm. This finding is particularly interesting in view of debates on the value of open innovation (Chesbrough, 2003; Laursen and Salter, 2006). It would seem that opening the firm’s technological base up to other firms and seeking synergies between internal and external knowledge is key to increasing the technological impact of reactivating technologies. In other words, the implication is that shelved technologies, for which the focal firm may not have found immediate further use for upon creation, may eventually become more useful through external use.
This finding is in line with arguments recently proposed by Yang et al. (2010) and Belenzon (2012). They argue that it is possible that firms can recapture external developments of internal technologies. By doing so, these firms effectively recapture value from knowledge spillovers. Similarly, Yang and Steensma (2014) found that firms can more easily explore new technological domains when their internal knowledge is linked to those domains through external parties’ knowledge recombination efforts. In that way, the firm is able to increase its scope of knowledge search by observing how its own technology is used in external technological developments.
Implications
32
implication is that firms should strike a careful balance between observing how the technology performs on the market, hence saving resources otherwise spent on costly trial-and-error processes and lowering technological uncertainty, and capturing the remaining recombinant potential of the technology in a timely manner. In the same vein, Katila and Chen (2008) found evidence that firms should conduct their knowledge search activities ‘out of sync’ with their rivals. More precisely, they find that firms who search before or ahead of competitors tend to have better innovation performance. Our study complements their findings and indicates that firms can benefit from recombining their own knowledge for the first time after external parties have used it already.
With regards to the literature on temporal search, earlier studies found that the source of the old knowledge is a determinant of its technological value (Katila, 2002; Petruzelli et al., 2012) and that old knowledge becomes more valuable when it is recombined with new knowledge (Nerkar, 2003). To this emerging literature, we add that the frequency of use of old knowledge can influence its recombinant value. We particularly note that unused internal technologies benefit greatly from external use. This finding adds another layer of depth to the broad local versus distant search discussion (e.g. March, 1991; Nerkar, 2003; Laursen and Salter, 2006; Lavie et al., 2010). Specifically, our results would suggest that the distinction between the two constructs is not clear-cut for temporal search. Nerkar (2003) earlier defined temporal exploration as recombining old knowledge (unfamiliar knowledge) and temporal exploitation as recombining new knowledge (familiar knowledge). However, we argue that this is too simplistic for two reasons.
33
a firm becomes more familiar with internal knowledge as it uses the knowledge more frequently (e.g., Fleming, 2001; Katila and Ahuja, 2002). However, we established that external use of knowledge can also contribute to the firm’s familiarity with internal knowledge. Through observing how the firm’s own technology is being used externally, the focal firm can learn to become more familiar with its own technology. It would thus appear that the cognitive distance between a firm’s current and past activities is contingent upon more than just its own search activities. Our findings therefore implicate that the degree of knowledge familiarity should be determined based upon both internal and external frequencies of use.
Our study also has practical implications. We studied in depth whether and when firms should exploit their shelved technologies. To our knowledge, no studies have explicitly examined the value of reactivating shelved technologies. Garud and Nayyar (1994) noted 20 years ago that some technologies, though perhaps valuable, are not immediately further developed upon and are instead shelved by the firm. However, since the firm invested resources in the development of these technologies, we advise it to continuously reevaluate their recombinant potential and determine whether it is worth taking them off the shelf. Although the outcomes may be relatively less valuable, in doing so the firm utilizes its technological base more effectively resulting in greater returns to internal R&D (Wang and Hagedoorn, 2014). These returns to internal R&D could be amplified when we consider that external use can raise the value of a firm’s unused knowledge. Indeed, our findings suggest that the firm should carefully observe how other firms make use of its shelved technologies before recombining them. This will tend to reduce the firm’s own recombinant uncertainty with these technologies. Additionally, assuming the firm shelved the technology because it was incapable of further recombining it, these monitoring activities also save the firm considerable resources otherwise spent on costly trial-and-error processes involved with attempts to internally develop the internal technology.
34
originating from its own technologies. To support this process, we suggest that managers should increasingly consider opening up their unused technological base to other firms so as to increase its dissemination and external acceptance but also to reveal important recombinant opportunities (Alexy et al., 2013). To make this process quicker, the firm should consider seeking external partnerships to exploit and reactivate shelved technologies (Cohen and Levinthal, 1990; Harrigan and Newman, 1990; Tsang, 2000; Alexy et al., 2013).
However, it is important that managers realize that these selective revealing strategies should be carried out delicately. As argued by Henkel (2006), the firm should carefully consider which elements of the technology it reveals to other parties. On the one hand, if too many important elements are revealed, external parties will quickly exhaust the recombinant potential of the technology. Moreover, if too much knowledge is revealed, the costs of knowledge spillovers (e.g., increased competition) in some cases may offset the benefits accruing from learning from these spillovers (Yang et al., 2010). Consequently, the firm will find it more difficult to recapture value from external developments through its own recombinant efforts because there will be fewer remaining unique modifications it can make to these technologies. Alternatively, it is also possible that too much openness will serve as a signal to other firms that the revealed technologies are not very valuable (Laursen and Salter, 2014). As a result, the dissemination of knowledge which the firm finds difficult to further recombine will be slower. However, on the other hand, if the firm reveals too few technological elements, external parties will not be able to further develop the technology and the firm will not be able to recapture any knowledge from external developments of its own technologies. Thus, we advise managers to strike a careful balance between how much of the shelved technology the firm reveals and how much it keeps for itself.
Limitations
35 replicate our analysis with different samples.
Secondly, we relied solely on patent data to study the recombination of old technologies. Though patent data is publicly available and is generally reliable (Criscuolo and Verspagen, 2008), studies have shown that some firms tend to rely more on alternative appropriation mechanisms (Arundel and Kabla, 1998; Cohen et al., 2000). Nevertheless, since we sampled firms from industries in which patenting propensity rates are generally elevated, these concerns are alleviated (Arundel and Kabla, 1998).
Thirdly, Criscuolo and Verspagen (2008) show that most of EPO patents’ backward citations are added by patent examiners, rather than inventors themselves. Thus, one may argue that we mostly captured examiners’ evaluation of technological antecedents of inventions. However, since EPO patent examiners tend to have lower workloads (Belderbos et al., 2014), their scrutiny of patents’ technological antecedents is likely to be very accurate. In fact, EPO patents tend to contain mostly strictly technologically relevant backward citations compared to those of other patent systems (Michel and Bettels, 2001). Moreover, it should be considered that backward citations from other patent systems often contain a lot of noise. For example, USPTO patents contain more backward citations added by the firms themselves, rather than those objectively added by the patent examiner. It is possible that these backward citations may simply have been added for strategic reasons, rather than to indicate the technological antecedents of the invention (Criscuolo and Verspagen, 2008). In sum, we are relatively certain that the backward citations in our data accurately reflected the technological antecedents of inventions.
36
only looked at the frequency of use of citations made in the past 5 years, while Miller et al. (2007) only looked at the frequency of use of citations in the past 10 years. Instead, we looked at all the citations made to a certain patent between its application year and the year in which it is recombined. However, taking into consideration that the average shelved technology in our sample was 12.32 years old (i.e., the average age of the shelved technologies in our list of backward citations; see appendix), these concerns are negligible. That is to say, compared to the 10-year threshold used by Miller et al. (2007), we estimate that we only marginally overestimate the number of citations made to shelved technologies by not including a specific threshold (i.e., the number of citations equivalent to, on average, 2.32 more years during which citations made to the technology were counted).
Future research
Future research should continue examining how the frequency of use of knowledge, its age, and its recombinant value are intertwined. Our findings illustrate that the relationship between a technology’s frequency of use, age, and value is not clear-cut. For instance, there seem to be important differences between internal and external use of technologies and how either impacts the value of old technologies. Future work should uncover what exactly the origins of these differences are and how they may exert an influence on technological performance. For example, it is possible that industry-wide acceptance of knowledge is more important in some industries than others for knowledge recombination (e.g., some industries may be particularly driven by pressures for legitimacy). Alternatively, these differences can perhaps best be explained by the degree to which firms engage in boundary-spanning learning activities such as the formation of alliances or the acquisition of particular firms (Rosenkopf and Almeida, 2003). For example, some inter-firm cooperative activities, by granting access to complementary resources, may trigger the reactivation of unfrequently used knowledge (Harrigan and Newman, 1990; Tsang, 2000). This is an important avenue for future research considering that we know little about the antecedents of resource recombination (Carnabuci and Operti, 2013).
37
answer to the question whether or not firms consciously shelve and unshelve technologies. With patent data it is difficult to identify which technologies have been shelved for future use and which have simply been forgotten about. We also argued that technologies need to be maintained over time to keep their value (Garud and Nayyar, 1994; Nerkar, 2003; Argote, 2012). However, knowledge that remained unused for a long period of time will likely be quickly forgotten and will consequently rapidly deteriorate in quality. In that case, the maintenance of the knowledge becomes even more important. Scholars should investigate to what extent the maintenance of shelved technologies (or old technologies in general) influences the value of knowledge recombination. Finally, it should be investigated how recombinant efforts based on internal technologies can be improved through opening the technological base up to other parties. For instance, it would be interesting to see what balance firms should strike between revealing too much of their own knowledge, and revealing just enough of it so as to be able to fruitfully recapture external developments (Henkel, 2006).
Conclusion
38
REFERENCES
Ahuja, G., & Morris Lampert, C. (2001). Entrepreneurship in the large corporation: A longitudinal study of how established firms create breakthrough inventions. Strategic
Management Journal, 22(6‐7), 521-543.
Aiken, L.S., S.G. West. (1991). Multiple Regression: Testing and interpreting interactions. Sage Publishing.
Alexy, O., George, G., & Salter, A. J. (2013). Cui bono? The selective revealing of knowledge and its implications for innovative activity. Academy of Management Review, 38(2), 270-291. Almeida, P., Song, J., & Grant, R. M. (2002). Are firms superior to alliances and markets? An empirical test of cross-border knowledge building. Organization Science, 13(2), 147-161. Argote, L. (2012). Organizational learning: Creating, retaining and transferring knowledge (2nd ed.). Springer.
Arthur, W. B. (2007). The structure of invention. Research Policy, 36(2), 274-287.
Artz, K. W., Norman, P. M., Hatfield, D. E., & Cardinal, L. B. (2010). A longitudinal study of the impact of R&D, patents, and product innovation on firm performance. Journal of Product
Innovation Management, 27(5), 725-740.
Arundel, A., & Kabla, I. (1998). What percentage of innovations are patented? Empirical estimates for European firms. Research Policy, 27(2), 127-141.
Barney, J. (1991). Firm resources and sustained competitive advantage. Journal of
Management, 17(1), 99-120.
Belderbos, R., Cassiman, B., Faems, D., Leten, B., & Van Looy, B. (2014). Co-ownership of intellectual property: Exploring the value-appropriation and value-creation implications of co-patenting with different partners. Research Policy 43(5), 841-852.
Belenzon, S. (2012). Cumulative innovation and market value: Evidence from patent citations*. The Economic Journal, 122(559), 265-285.
Cameron, A. C., & Trivedi, P. K. (1990). Regression-based tests for overdispersion in the Poisson model. Journal of Econometrics, 46(3), 347-364.
Carnabuci, G., & Operti, E. (2013). Where do firms' recombinant capabilities come from? Intraorganizational networks, knowledge, and firms' ability to innovate through technological recombination. Strategic Management Journal, 34(13), 1591-1613.
Cassiman, B., & Veugelers, R. (2006). In search of complementarity in innovation strategy: Internal R&D and external knowledge acquisition. Management Science, 52(1), 68-82. Chesbrough, H. W. (2003). Open innovation: The new imperative for creating and profiting
from technology. Harvard Business Press.
39
Cohen, W. M., Nelson, R. R., & Walsh, J. P. (2000). Protecting their intellectual assets:
Appropriability conditions and why US manufacturing firms patent (or not) (Working Paper
No.7552). Retrieved from National Bureau of Economic Research website:
http://www.nber.org/papers/w7552
Criscuolo, P., & Verspagen, B. (2008). Does it matter where patent citations come from? Inventor vs. examiner citations in European patents. Research Policy, 37(10), 1892-1908. Cyert, R. M., & March, J. G. (1963). A behavioral theory of the firm. Prentice Hall. Darr, E. D., Argote, L., & Epple, D. (1995). The acquisition, transfer, and depreciation of knowledge in service organizations: Productivity in franchises. Management Science, 41(11), 1750-1762.
Dierickx, I., & Cool, K. (1989). Asset stock accumulation and sustainability of competitive advantage. Management Science, 35(12), 1504-1511.
Fabrizio, K. R. (2009). Absorptive capacity and the search for innovation. Research
Policy, 38(2), 255-267.
Fang, Y., Wade, M., Delios, A., & Beamish, P. W. (2013). An exploration of multinational enterprise knowledge resources and foreign subsidiary performance. Journal of World
Business, 48(1), 30-38.
Fleming, L. (2001). Recombinant uncertainty in technological search. Management
Science, 47(1), 117-132.
Fleming, L. (2002). Finding the organizational sources of technological breakthroughs: The story of Hewlett‐Packard's thermal ink‐jet. Industrial and Corporate Change, 11(5), 1059-1084.
Fleming, L., & Sorenson, O. (2001). Technology as a complex adaptive system: Evidence from patent data. Research Policy, 30(7), 1019-1039.
Galunic, C., & Rodan, S. (1998). Resource recombinations in the firm: Knowledge structures and the potential for Schumpeterian innovation. Strategic Management Journal, 19(12), 1193-1201.
Garud, R., & Nayyar, P. R. (1994). Transformative capacity: Continual structuring by intertemporal technology transfer. Strategic Management Journal, 15(5), 365-385.
Gruber, M., Harhoff, D., & Hoisl, K. (2013). Knowledge recombination across technological boundaries: Scientists vs. engineers. Management Science, 59(4), 837-851.
Hargadon, A. B. (2002). Brokering knowledge: Linking learning and innovation. Research in
Organizational Behavior, 24, 41-85.
Hargadon, A., & Sutton, R. I. (1997). Technology brokering and innovation in a product development firm. Administrative Science Quarterly, 716-749.
Harrigan, K. R., & Newman, W. H. (1990).Bases of interorganization co‐operation: Propensity, power, persistence*. Journal of Management Studies, 27(4), 417-434.
40
Heeley, M. B., & Jacobson, R. (2008). The recency of technological inputs and financial performance. Strategic Management Journal, 29(7), 723-744.
Henderson, R. M., & Clark, K. B. (1990). Architectural innovation: The reconfiguration of existing product technologies and the failure of established firms. Administrative Science
Quarterly, 35(9), 9-30.
Henderson, R., & Cockburn, I. (1994). Measuring competence? Exploring firm effects in pharmaceutical research. Strategic Management Journal, 15(S1), 63-84.
Henkel, J. (2006). Selective revealing in open innovation processes: The case of embedded Linux. Research Policy, 35(7), 953-969.
Katila, R. (2002). New product search over time: Past ideas in their prime? Academy of
Management Journal, 45(5), 995-1010.
Katila, R., & Ahuja, G. (2002). Something old, something new: A longitudinal study of search behavior and new product introduction. Academy of Management Journal, 45(6), 1183-1194. Katila, R., & Chen, E. L. (2008). Effects of search timing on innovation: The value of not being in sync with rivals. Administrative Science Quarterly, 53(4), 593-625.
Kelley, D. J., Ali, A., & Zahra, S. A. (2013). Where do breakthroughs come from? Characteristics of high‐potential inventions. Journal of Product Innovation
Management, 30(6), 1212-1226.
Kraaijenbrink, J., Spender, J. C., & Groen, A. J. (2010). The resource-based view: A review and assessment of its critiques. Journal of Management, 36(1), 349-372.
Laursen, K. (2012). Keep searching and you’ll find: What do we know about variety creation through firms’ search activities for innovation? Industrial and Corporate Change, 21(5), 1181-1220.
Laursen, K., & Salter, A. (2006). Open for innovation: The role of openness in explaining innovation performance among UK manufacturing firms. Strategic Management
Journal, 27(2), 131-150.
Laursen, K., & Salter, A. J. (2014). The paradox of openness: Appropriability, external search and collaboration. Research Policy, 43(5), 867-878.
Lavie, D., Stettner, U., & Tushman, M. L. (2010). Exploration and exploitation within and across organizations. The Academy of Management Annals, 4(1), 109-155.
Leiponen, A., & Helfat, C. E. (2010). Innovation objectives, knowledge sources, and the benefits of breadth. Strategic Management Journal, 31(2), 224-236.
Levinthal, D. A., & March, J. G. (1993). The myopia of learning. Strategic Management
Journal, 14(S2), 95-112.
Levitt, B., & March, J. G. (1988). Organizational learning. Annual Review of Sociology, 14(1), 319-338.
