Assessing the technological knowledge base of the
firm: a review of breadth and depth measurements
July 1, 2015
Rijksuniversiteit Groningen 1st supervisor: Wilfred Schoenmakers
2nd supervisor: Rene van der Eijk
John de Jong S2583151 MSc BA SIM Word count: 17796
Abstract
According to the knowledge-based view, knowledge is the most important resource to the firm in order to achieve a competitive advantage. Consequently, being able to study a firm’s knowledge base is of the utmost importance. The identified dimensions to measure a knowledge base are breadth and depth. However, various terminologies and measurements exists. Therefore, a literature review is performed in order to, on the one hand, come to an un unequivocal definition of breadth and depth, on the other hand, structure available breadth and depth measurements so that the most appropriate measurements can be identified. Results show that breadth and depth essentially relates to heterogeneity and familiarity of knowledge, respectively. Based on this notion it is found that diversity/concentration indexes, compared to count measurement, can most precisely capture the characteristics of breadth and depth.
1
1
Introduction
In the early 90’s the resource-based view (RBV) gained recognition as an opposing force to the prevailing explanations that take an external perspective of the firm regarding the question why firms achieve a competitive advantage (e.g. Porter, 1985). The RBV takes an internal perspective and posits that superior performance can be traced back to the resources and capabilities of the firm. Barney (1991) states that in order for a firm to achieve a sustainable competitive advantage, a firm must possess valuable, rare, imitable, and non-substitutable resources. Amit and Shoemaker (1993) complement this by stating that uniqueness regarding the development and deployment of those resources also come from managerial decisions taken under uncertainty, complexity, and conflict. Which makes duplication of a firm’s resources by competitors near impossible. Behind all of this lies the underlying belief of the RBV of having an advantage over competitors by holding resources that competitors desire but cannot obtain.
2 into tacit and explicit knowledge, the former uncodified knowledge whereas the latter is codified knowledge. Knowledge is also held on different levels, namely individual-, group-, organization-, and network-level (Kogut and Zander, 1992). These two attributes indicate that the knowledge base that resides within an organization is complex and idiosyncratic. A firm’s knowledge base is hard to replicate because it is embedded in the organization due to knowledge residing on all levels and social relationships being formed on these levels (Kogut and Zander, 1992; Spender, 1996). What is more, nowadays firms also recognizes the importance of knowledge, because competition is increasingly knowledge-based as rivals try to out-learn each other (Lane and Lubatkin, 1998). Specifically, according to the KBV, a firm can be seen as a ‘knowledge creating entity’ (Nonaka, Toyama, and Nagata, 2000), and the creation and application of knowledge is what makes firms able to innovate new products, services, and processes (Bierly and Chakrabarti, 1996; Nonaka et al., 2000; Grand, 1996). Thus, it is ultimately a firm’s knowledge base that can provide them a competitive advantage.
3 depth actually mean and how they best can be measured. This leads to the following research question:
What is breadth and depth of a firm’s knowledge base and how can they best be measured? This research question is answered in two parts. First, literature with regard to breadth and depth is reviewed in order to come to an unequivocal definition. Second, with the definitions in place, it becomes possible to review breadth and depth measurements. Literature is reviewed and structured and subsequently the measurements are chosen that best reflects a firm’s technological base. Consequently, the contribution of this study is also twofold. On the one hand, the literature field is made comparable by defining the dimensions and by structuring and classifying methods to measure a firm’s knowledge base. This is of academic significance, because currently difference in terminology and definitions exists, which might cause confusion what breadth and depth essentially encompass. Especially depth deserves more attention. On the other hand, advantages and disadvantages of measurements are given, followed by suggestions which measurements best to use. In general, this study can help future researcher to get a better understanding what measurement to use with respect to technological knowledge bases. Moreover, given the measurements proposed in this paper, managers will be able to determine their own knowledge base, and subsequently take strategic actions (Katila and Ahuja, 2002).
5
2
Background
The background is composed of two parts. A firm’s knowledge base in general is too broad to review, therefore, it is explained where this review is focused on and why the review is
important. This is followed by a short overview how breadth and depth relate to performance, to build upon the introduced notion of superior performance owing to the knowledge of the firm.
2.1
Setting of the literature review
Considering that knowledge is such a critical resource to the firm, it is of necessity to better understand a firm’s knowledge base and prior research shows breadth and depth are the means to do so. However, the fact still remains that all knowledge that resides within the firm is immense given that this is held on various types and levels, as stated in the introduction. This calls for a focus on the most important types of knowledge.
6 2005). In addition, according to Prencipe (2000) the evolution of the technological knowledge base is “tightly intertwined” (p. 897) with the organizational knowledge base. For example, new technological knowledge can be created due to changes in the NPD process (Prencipe, 2000). Thus, a firm’s technological knowledge base to some extent also reflects other knowledge bases.
8 reason that organizations are interested in codifying knowledge in order to be able to quickly diffuse it within the organization, integrate it with existing knowledge and apply it. Furthermore, by actively searching for explicit knowledge a firm indirectly increases its absorptive capacity, which is critical for its innovative capabilities. Cohen and Levithal (1990) showed that in order to understand and apply external knowledge a firm needs to actively invest in searching for it. In a similar vein, when knowledge is explicit it does not necessarily mean that every firm can understand it, this depends on their absorptive capacity (Cohen and Levithal, 1990). Given that competitors differ in their absorptive capacity, they also differ in their ability to understand certain explicit knowledge (Park et al., 2015). Consequently, having implications with respect to its imitability.
10 unequivocal definition is needed that captures all prior streams and clears confusion that might previously have been caused. Subsequently, an unequivocal definition makes it possible to assess what measurements can best capture the actual knowledge base, given that a definition serves as the base for a researcher’s measurement choice.
2.2
Breadth, depth and performance
11 words, having a broad base, comes at a price as firms might lack the necessary scale to benefit from this wide-ranging knowledge, and they too argue that it entails increased coordination and integration costs. Put simply, a broad knowledge base requires a lot from a firm’s absorptive capacity.
When looking at how depth relates to performance, it appears that less studies have delved into this dimension. Katila and Ahuja (2002) found a positive relationship between depth and inventive performance as well. Boh, et al. (2014) found that inventors with deep expertise, that is, the inventor being able to invent a lot in its core area of expertise, generate highly influential inventions. Which is in contrast with inventors that have a broad expertise, who can generate many inventions but are less influential. According to Moorthy and Polley (2010) this deep expertise implies that firms have a good understanding of risks involved which brings about a more efficient search process and in turn provides higher performance gains.
12
3
Methodology
A firm’s knowledge base is a broad concept as knowledge occurs in many forms and resides on every level in the organization (Kogut and Zander, 1992). The field of expertise which the firm is knowledgeable about clearly depends very much on the type of organization. In this literature review I focus on the technological knowledge base, giving the importance of today’s firm to be inventive (Jansen et al., 2006), and on explicit knowledge given that it can be quickly acquired and diffused within the firm in light of today’s fast-paced competition (Park et al., 2015).
Articles are collected by using the following search terms: Firm Knowledge, Knowledge Base, Knowledge Breadth, Knowledge Depth, Technological Knowledge, Technological Diversity, Technological Diversification and Technological Scope. Further, studies are also found via citations in relevant articles. In consideration of the research question provided on page 3, articles are selected when they define a knowledge base dimension and/or when they measure a knowledge base dimension. Following these criteria, the literature search led to a total sample size of 38 articles. Given this workable amount, no further selection needed to be made.
13 measurements. More specifically, individual-level studies are relevant as an organization is the aggregation of individuals and are included because they bear resemblance to certain firm-level measurements, or because they can be adopted on a firm-level, as performed on 3 measurements. Furthermore, 1 invention-level measurement and 1 technological-field-level measurement have also been transformed to a firm-level measurement, while the remaining invention-level studies are only included to illustrate that certain methods are widely used to measure a knowledge dimension.
To return to the total literature sample, there is a division into two groups: definition articles and measurements articles. Regarding definition articles, a total of 18 studies are found to contain relevant definitions, where 13 of them are also present in the measurement article group as they both define and measure. The group only includes articles that explicitly provide a definition and do not literally draw on another study’s definition. Regarding measurement articles, a total amount of 33 articles are found to contain relevant measurements, 22 of them are on firm-level. The remaining articles are exclusively on another level of analysis. 2 of them are on technological-field-level, 5 are on invention-level, and 4 are on individual-level.
14 cannot be categorized in this dichotomous manner. Besides counting, methods are identified in two other groups, namely, concentration and coherence.
15
4
Literature review
In order to measure a firm’s patent portfolio, according to the literature breadth and depth are the two relevant dimensions to assess. Breadth can be seen as the traditional knowledge base dimension (Ahuja and Katila, 2004). However, a firm’s knowledge base is not unidimensional where firms with a broad knowledge base can be compared with having an exploratory search character and a narrow knowledge base with being exploitative (Katila and Ahuja, 2002). This unidimensional concept relates to whether firms draw from various knowledge sources or merely a few. However, having a narrow knowledge base does not necessarily mean the base is deep, because depth relates to the actual usage of the firm’s existing knowledge (Katila and Ahuja, 2002). In other words, breadth reflects the horizontal dimension of knowledge, whereas depth reflects the vertical dimension (Zhou, 2012). One of the first studies to explicitly define both the breadth and depth dimension was von Tunzelmann (1998) and Wang and von Tunzelmann (2000) who were interested in technological complexity within the firm, along with Prencipe (2000) who focused on technological capabilities. Many knowledge base studies base their definition of breadth and depth on these studies, because Wang and von Tunzelmann (2000) view complexity as bodies of knowledge, and so does Prencipe (2000) with regard to capabilities. All in all, “breadth is more concerned with the degree of heterogeneity and depth with the level of sophistication” (Wang and von Tunzelmann, 2000, p. 806).
17
4.1
Breadth of a technological knowledge base
Breadth, the horizontal dimension of a knowledge base, is discussed in this section. The contribution of this section is twofold. First, it is reviewed how other studies view and define breadth, second, it is reviewed how they measure breadth. The definitions are given in a chronological manner as much as possible in order to give an impression how the concept of breadth evolved throughout the years. Furthermore, different terminology is discussed. It is explicitly mentioned whether breadth refers to the firm’s knowledge base or something else.
4.1.1
Defining breadth
Increased product complexity required firms to possess knowledge of a wide variety of technological fields (von Tunzelmann, 1998; Moorthy and Polley, 2010). When a firm is only active in a few specialized areas there is the risk that its core competencies become core rigidities (Leonard-Barton, 1992). In order to spread those risks a firm needs to be able to place different bets and therefore have a broad technology base (Leten et al., 2007). Furthermore, drawing from diverse set of knowledge increases a firm’s absorptive capacity (Quantana Garcia, 2008) as well as the recombinatory opportunities of knowledge elements which stimulates inventions (Katila and Ahuja, 2002).
18 and gave the following definition “diversity of technologies required to produce or further develop the product range of the firm” (p. 232). Moreover, Prencipe (2000) measured technological capabilities of the firm and understood breadth as “the number of technological fields in which the firm is active” (p. 896). Even though their main objective was not to define knowledge base breadth, they all relate to knowledge within the firm and their definitions are quite similar. However, while Prencipe (200) merely considers the number of fields, von Tunzelmann’s (1998) and Wang and von Tunzelmann (2000) explicitly state it is not about sheer number but about heterogeneity.
19
Table 1. Breadth Definitions
Study Definition
Subba Narasimha and Ahmad (2001)
“ aggregate . . . of dispersion in technological knowledge stock” (p. 24)
Prabhu et al. (2005) “the range of fields over which the firm has knowledge” (p. 115)
Lin et al. (2006) “the extent to which the technology firm diversifies its technological capability on a broadly defined . . . technology area” (p. 19)
Zhang et al. (2007) “the range of knowledge areas that a firm possesses” (p. 517) De Luca (2007) “the number of different knowledge domains with which the
firm is familiar” (p. 97) Quantana-Garcia and
Benavides-Velasco (2008)
“the diversity in the knowledge system and principles underlying the nature of products and their methods of production” (p. 492)
Wu and Shanley (2009) “the scope of scientific and technological domains in which a firm has expertise” (p. 476)
Moorthy and Polley (2010) “learning and search across technological disciplines” (p. 361) Zhou and Li (2012) “the extent to which the firm’s knowledge repository contains
distinct and multiple domains” (p. 1091)
20 for a firm to draw from a diverse technological knowledge base for their innovation process, I define breadth as the diversity in technological fields the firm is knowledgeable about.
4.1.2
Measuring breadth
In this section the different ways of measuring breadth are described. Measurements that hold similarities are grouped together. Two distinct groups can be observed: counting measurements and diversity indexes. Within the group counting measurements, another distinction can be made, namely, a group that counts a patent’s backward citations and a group that counts a patent’s classification codes. The group diversity indexes can also be divided into two subgroups, namely, methods based on the Herfindahl index or the entropy type index.
Counting measurements
Counting measurements are the most straightforward measurements. In the past studies such as the well-known study of Decarolis and Deeds (1999) counted the number of patents of the firm to get an indication of its knowledge stock. This is the traditionally way of measuring knowledge stock and usually, in addition, a yearly obsolescence rate is included as knowledge depreciates quickly (Colombelli and Quatraro, 2014). However, this is a crude way of examining the knowledge the firm has accumulated over the years and it does not explain anything about the characteristics of the knowledge base. Soon it became apparent that in order understand a firm’s knowledge base, one has to examine the relevant dimensions, breadth and depth. This is not to say that counting measures disappeared, however, different variables are taken. With respect to studies that purposely measure breadth, a distinction can be made by studies that count a patent’s backward citations or technological classes.
Counting citations
21 inventor, but subsequently turning it into a firm measurement by totaling it to get an indication of the quality of the team. According to them, one way to measure a firm’s knowledge is to measure the knowledge of key participants. In their view this is the scientific team of the firm.
22 Thus, it is questionable whether this is an appropriate breadth measurement. On the other hand, classification codes are more promising as they indicate what kind of knowledge the patent relates to. Methods based on classes are considered in the following section.
Counting classification codes
23 base, makes a firm more inclined to acquire external knowledge. Even though not explicitly mentioned, this variable also represents depth of the knowledge base. Which illustrates that breadth and depth are interwoven. Because of this, the variable is incorporated in the depth section. Furthermore, Katila (2002) used, besides the citation measurement, another way to measure breadth, by looking at the proportion of subclasses that was new to the firm. This is measured for each year and compared with classes from five previous years under the notion that knowledge depreciates quickly. Her method is related to the study of Katila and Ahuja (2002) from the previous section where newness is claimed to indicate breadth. All in all, unlike citations, the methods above do measure different types of knowledge as each class represents a distinct knowledge field. However, the size of the firm’s patent portfolio is not taken into account. Therefore, Leahey (2006) proposes to include size. Thus, a method that measures the ratio of the cumulative number of unique classification codes and the number of patents. Another study that does take a proportional approach is Özman (2007). His method is difficult to categorize, for example, to some extent the approach also resembles the diversity indexes in the next section. His method is distinct because it is the only study that draws upon the following characteristic of a patent document: within the list of classes assigned to a patent, the list can be further classified by one main code and usually numerous secondary codes (Özman, 2007; Lupu et al., 2011). According to Lupu et al. (2011) the code “which most adequately represents the invention should be listed first” (p. 297). On the notion of the more different fields listed in a patent, the wider its knowledge base, Özman (2007) created the following equation:
24 where is breadth of a patent by counting the number of secondary codes of a patent, other than its main code. This is divided by which is the total number of patents the firm holds.
Diversity indexes
Diversity indexes are by far the most commonly used in measuring knowledge base breadth. These indexes can be performed on multiple levels of analysis (e.g. Trajtenberg et al., 1997), but are typically used on firm-level. Moreover, diversity indexes are not only used to measure breadth, they also measure depth. Because a low score relates to a deep knowledge base (Moorthy and Polley, 2010).
Herfindahl index
The original diversity index most commonly used among scholars measuring knowledge bases is called the Herfindahl index or alternatively the Blau index. The former name is used more often and stems from the field of economics, whereas the later stems from sociology. This index is expressed as:
= 1 − ²
25 diversification chances than a firm with 50 patents. Therefore, they multiplied the index by
, where stands for the number of the firm’s patents. The purpose of this is to give an higher value to firms with fewer patents. They also grouped firms according to the amount of patents to further control for this bias. Polley and Moorthly (2010) also adjusted the index and gave a similar explanation by arguing that the index fails to provide any indication of the spread of patent’s across patent classes. It should be noted that they intended to capture both breadth and depth with their measurement. To overcome the bias they added and subtracted 1/N, which yields: = 1 −
− ∑ [ ² − (
)²]. Furthermore, Leten et al. (2007) transformed the index
26
Entropy type index
Besides the commonly used Herfindahl index, Bierly and Chakrabarti (1996) and Subba Narasimha and Ahmad (2001), as one of the earlier articles to measure breadth, used an index named dispersion index or entropy type index. The index is as follows:
= − ( 1 ∙
)
where is the fraction of patents in ith field. A firm with a high dispersion index indicates a broad knowledge base, whereas a firm with a low value is highly focused on specific domains (in other words, a deep base). The outcome of this index also varies from zero to one. According to Narasimha and Ahmad (2001) this measure has the advantage of considering the number of patents across technology classes. Further, it is not reported that this index experiences the same downward biases as the Herfindahl index does. What is more, as stated on p.19, Prabhu et al.’s (2005) measurement of choice was a count measurement, however, they supplemented this with two alternative measurements, the original Herfindahl Index and an Entropy Index. These measurements were consistent with their main method. Their entropy index is as follows:
" = ∑ & ln (%)
4.1.3
Breadth measurements overview
28 Table 2. Breadth Measurement Framework
Measurement group
Subgroup Studies Aim as described by author(s)
Measurement Possible additions / adjustments Counting Counting
citations
(a) Katila (2002) (b) Katila and Ahuja
(2002)
(a) External search quantity (b) Search scope
(a) Totaling the number of citations of each patent the firm has in its portfolio. (b) For all of the firm’s patents:
'( )*+*,-
.,*+ )*+*,-where new refers to citations in a focal year that were not listed in the previous five years.
Counting classification codes
(a) Lerner (1994); Boh et al. (2014)1
(b) Nesta and Saviotti (2005)
(c) Prabhu et al. (2005) (d) Zhang et al. (2007)
(a) Patent scope; Breadth of expertise (b) Scope of the knowledge base (c) Breadth of knowledge (d) Breadth of knowledge base
Counting the number of
classification codes assigned to each of the firm’s patents.
(a) Count the
cumulative unique codes.
(d) Control for the concentration of the knowledge base (i.e. depth). Counting classes proportionally (a) Katila (2002) (b) Leahey (2006)2 (c) Özman (2007) (a) Technological diversification (b) N/A
(c) Depth of the firm
(a) The proportion of new subclasses entered in a year. (b) The proportion of the
cumulative number of codes and the total of the firm’s
1
These measurement are transformed into a firm-measurement. Lerner (1994) originally measured patent breadth and Boh et al. (2014) originally measured inventor breadth. 2
29 patents.
Table 2. Breadth Measurement Framework (Continued)
Measurement group
Subgroup Studies Aim as described by author(s)
Measurement Possible additions / adjustments (c) The proportion of the ratio
of the sum of the number of secondary codes different from the main code of each firm’s patents and the total patents of the firm:
=∑
Diversity Herfindahl type index
(a) Prabhu et al. (2005) (b) Garcia-Vega (2006) (c) Lin et al. (2006) (d) Avenel et al. (2007) (e) Leten et al. (2007) (f) Quintana- García and
Benavides-Valasco (2008)
(g) Wu and Shanley (2009)
(h) Lahiri (2010) (i) Polley and Moorthly
(2010) (j) Gruber et al. (2013)3 (a) Breadth of knowledge (b) Technological diversification (c) Broad technology diversity (d) Diversity of technological knowledge base (e) Technological diversification (f) Technological diversification (g) Knowledge breadth
1 minus the sum of squares of the proportion of a firm’s patents across
i technology domains for a given
year:
= 1 − ²
(a) Only focus on the proportion and exclude the ‘1-’ part from the formula. (b) Multiply by
to
counterbalance the downward bias regarding firm’s with few patents. (c) Take the square
root, yielding: = 1 − ∑ ²
(e) Take the inverse, yielding: =
3
30 (h) Technological
diversity 1 / ∑ ²
;
Table 2. Breadth Measurement Framework (Continued)
Measurement group
Subgroup Studies Aim as described by author(s)
Measurement Possible additions / adjustments (i) Breadth of technological knowledge (j) Technological recombination breadth
(i) Add and subtract 1/N to control for the size bias, yielding:
1 − − ∑ [ ²
− ()²]
(j) Instead of taking the firm’s total patents, take the firm’s total unique classification codes.
Entropy type index
(a) Bierly and Chakrabarti (1996); Subba
Narasimha and Ahmad (2001)
(b) Prabhu et al. (2005)
(a) Breadth of the knowledge base (b) Breadth of
knowledge
(a) With = the fraction of patents in ith field: = − ( 1 ∙
)
31
4.1.4
Comparison of breadth measurements
32
Table 4. Comparison of Breadth Measurement
Firm 1 Firm 2 Firm 3 Firm 4 Firm 5
Patents Patents Patents Patents Patents
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 Field 1 x x x x x x x x x x Field 2 x x x x x x x Field 3 x x x Field 4 x Field 5 x (1) Counting unique
classification codes 0 (1, min. value) 0.25 (2) 0.25 (2) 0.5 (3) 1 (5, max. value) (2) Unique codes / Total patents 0.33 0.40 0.50 0.6 1 (3) Traditional HI 0 0.48 0.50 0.64 0.80 (4) HI by Garcia-Vega (2006) 0 0.60 0.67 0.80 1 (5) HI by Lin et al. (2006) 0 0.69 0.71 0.80 0.89 (6) HI by Moorthy and Polley (2010) 0 0.52 0.56 0.72 0.96 (7) HI by Leten et al.
(2007) 0 (1, min. value) 0.23 (1.92) 0.25 (2) 0.45 (2.78) 1 (5, max. value) (8) HI by Gruber et
al. (2013) 0 (-8) 0.65 (-2,25) 0.80 (-1) 0.90 (0) 1 (0.8)
(9) EI by Bierly and Chakrabarti (1996) & Subba Narasimha
and Ahmad (2001) 0 0.42 0.5 0.66 1
(10) EI by Prabhu et
33 Counting the number of classes is an easy method to get an understanding of breadth, however, as size is not considered it is also not very precise. Consider the case of firm 2 and 3, although they are active in the same fields, the former is slightly less broad as its base is not evenly dispersed. But this method neglects to capture this. Method 2 does take size into account and correctly gives firm 3 an higher value, but incorrectly scores firm 1 too high. Furthermore, when looking at the traditional Herfindahl index the values are more precise. However, the first thing that stands out is that firm 5 did not get the maximum value while its knowledge base evidently cannot be more diverse. This shows the downward bias at play. The indexes of Lin et al. (2006) and Moorthy and Polley (2010) try to control for this, and indeed, firm 5 gets scored higher but still did not obtain the highest value. While the adjusted index of Garcia-Vega (2006) and Leten et al. (2007) did provide the correct value. This also holds true for the entropy index in method 9. On the other hand, the entropy index of Prabhu et al. (2005) does suffer from the downward bias, for example, when the knowledge base of firm 5 would be twice as large the value would be 2.30.
4.2
Depth of a technological knowledge base
Depth, the vertical dimension of a knowledge base, is discussed in this section. First, the concept of depth and how studies define depth is further examined. Definitions are given so that it
becomes clear how the concept of depth evolved throughout the years. Given that the depth dimension is less widely discussed among previous studies, differences in terminology are less obvious. Thus, possible alternative terms are discussed more elaborately. Second, measurements are reviewed and structured.
4.2.1
Defining depth
34 knowledge is so difficult that firms are being compelled to have an in-depth understanding in their fields (von Tunzelmann, 1998). Furthermore, according to Moorthy and Polley (2010) deep expertise implies that firms have a good understanding of risks involved regarding the projects the firm conducts because they gained experience and are able to learn from their mistakes. Deep understanding also enables the firm to have a more efficient new product development process and to generate more knowledge elements (Subba Narasimha and Ahmad, 2001; Wu and Shanley, 2009).
Von Tunzelmann (1998) defined depth of technological complexity as the “extent of basic scientific and technological knowledge required in each particular area” (p. 232). Wang and von Tunzelmann (2002) defined the depth dimension of firm complexity as “the sophistication of a subject” (p. 806). Prencipe’s (2000) depth of technological capabilities relates to the amount of process stages involved with respect to developing technologies, together with the amount of knowledge related to technology components as well as the combinations of these components. Similar to breadth these three studies have been influential to other depth definitions. What is more, prior to these studies Bierly et al. (1996) discussed and measured breadth or narrowness of the firm’s knowledge base. They used narrowness interchangeably with depth and gave a description similar to others by referring to a firm that has core competences in a few key fields. However, the term narrow should be avoided when describing depth as this wrongly suggest that a knowledge base is unidimensional. Since then, no succeeding studies made use of the term.
35 refers to the exploitation of a familiar set of technology components, as well as Boh et al. (2014) who defined individual depth as “the level of knowledge and skills that an individual holds in a technical domain area” (p. 350). Studies that focused on knowledge base depth of the firm gave similar definitions, as shown in Table 3 below.
Table 3. Depth Definitions
When examining the definitions, familiarity of fields is a recurring view, but this seemingly neglect the level of sophistication as emphasized by von Tunzelmann (1998), Wang and von Tunzelmann (2002) and Prencipe (2000). The definition of Zhou and Li (2012), in turn, does refer to the level of sophistication. Yet, familiarity and sophistication are not unrelated. Being more familiar with a technological field implies more knowledge about components and thus being more capable of handling the field’s complexity. Von Tunzelmann (1998), therefore, also compares depth with the “localization of learning” (p. 232). In short, familiarity is needed to be able to perform on a high level of sophistication, which in turn is needed to enjoy the positive effects of depth. Moreover, two other views stood out. First, Subba Narasimha and Ahmad
Study Definition
Subba Narasimha and Ahmad (2001)
“a firm’s relative capability in each of the technological fields in which it has developed knowledge” (p. 24)
Prabhu et al. (2005)
Lin et al. (2006)
“the amount of within-field knowledge possessed by the . . . firm” (p. 115)
“the extent to which the technology firm diversifies its
technological capability on a narrowly defined . . . technology area.” (p. 19)
Wu and Shanley (2009) “the extent to which a firm is familiar with a particular technological or application domain” (p. 476)
36 (2001) who take an inter-firm view by arguing that “depth connotes a sense of strength” (p. 24) and that this depends on both the extent of the firm’s knowledge in a particular field as well as how knowledgeable other firms are regarding that field. Because according to them, only with this exceptional relative capability are firms able to truly exploit commercial ends in a field. Second, Lin et al. (2006) and Zhou and Li (2012) address that deep knowledge is perhaps not possible in all the fields in which the firm possesses knowledge. They suggest this to be concentrated in a few key fields. When strictly following the other definitions it is possible to be deep in every field, however, this is unlikely the case in practice due to restricted absorptive capacity, bounded rationality, and limited resources (Cohen and Levithal, 1990; Bierly et al., 1996). Being in-depth in a field, i.e. possessing knowledge so that one can operate at a high level of sophistication, requires a lot of effort. Thus, a firm will usually be specialized in a few fields.
Drawing on the previous definitions, I define depth as the extent to which a firm is relatively familiar and is consequently able to comprehend the level of sophistication regarding technological fields in which they hold knowledge. This definition includes the familiarity and sophistication component, as well as how these components hold compared to other firms active in those fields. As subscribed by authors referring to depth, albeit separately.
37 very familiar with this field. In other words, the firm is comfortable and knowledgeable to combine knowledge within this field.
Moreover, Avenel et al. (2007) use a construct which they call hybridization. It refers to diversity at the level of individual items. High hybridization means that different competences can easily be integrated within projects. When a firm’s knowledge base has high hybridization it allows a firm to have a deeper knowledge base. Lastly, another construct used in combination with breath is coherence, which is similar to hybridization and refers to relatedness among technological fields (Nesta and Saviotti, 2005; Leten el al., 2007). For example, “biotechnology and pharmaceuticals are relatively more related to each other than pharmaceuticals and electrical engineering” (Özman, 2007, p. 6). Coherence has the positive effect of being able to fully benefit from complementarities among fields and enjoy spillovers across projects. Therefore, it is assumed that when there is higher coherence there is deeper expertise, as the expertise needed for different fields is related. Or put differently, high coherence implies a concentrated knowledge base. For this reason, studies measuring coherence are included, as they also capture to some extent the portfolio depth. On the other hand, it also says something about breadth, because when a firm operates in highly unrelated fields it can be considered broader than when it operates in closely related fields. However, breadth is more concerned with the diversity of fields. Plus, given depth’s meaning of specialization and deep expertise with respect to a limited amount of technological fields, together with the complementarity benefits of coherence, it can be concluded that a coherent base is a base that facilitates deep learning.
4.2.2
Measuring depth
38 measurement. Same with breadth, a group based on counting and a group using diversity indexes can be observed. The line between these two groups is strictly drawn as the index group uses summation while the counting group does not. Moreover, it is mentioned earlier that Moorthy and Polley (2010) and Bierly and Chakrabarti (1996) claim to measure both breadth and depth with their diversity and dispersion index, respectively. However, as shown during the definition discussion above, depth relates to the concentration of the knowledge base. Therefore, the group is not addressed as diversity but as concentration indexes. Indeed, the antonym of dispersion is concentration, however, concentration indexes singly aimed to capture depth fundamentally differ from breadth indexes: they either take into consideration other firms who are active in technology fields, or they identify the focal firm’s core of expertise. To conclude, the groups are as follows. There is a counting measurements group, which is further categorized as either counting citations or counting classifications. The other group is the concentration indexes. Lastly, coherence measurements are a stand-alone group as they cannot be compared to how methods in the counting group nor the concentration group are performed.
Counting measurements
All but two depth measurements draw on classification codes. The others draw on citations.
Counting citations
39 fields involved, this is less of an issue compared to breadth, as it makes sense that firms are more frequently using fields they are familiar with.
Counting classification codes
Just as with breadth, technology classes assigned to patents are the most common data used for depth measurements. The most straightforward measurements are performed by the following studies. First, Avenel et al. (2007) count the average number of different fields assigned to each patent. As stated, they refer to hybridization. According to them, a high average of fields assigned to the patent base implies they are highly familiar with those fields and can easier enjoy complementarities among them which is a benefit of deep expertise. However, it is arguable whether this construct is appropriate for depth as making use of a high average amount of fields bears a resemblance to breadth as well. Second, Prabhu et al. (2005) and Wu and Shanley (2009) did the opposite by taking the average number of patents in each of the classes found in the firm’s knowledge base. Lastly, Boh et al. (2014) took inventors as level of analysis and focused on their core area of expertise, as determined by the most patented field. They measured depth by counting the total number of patents the focal inventor has published in his core area of expertise, and dividing it by the inventor’s total number of patents. Given that an organization is the aggregation of individuals, this method can easily be converted into a firm measurement by looking at the patents of the firm instead of the patents of an inventor. Basically, the previous two measurements are a rough way to measure the concentration of the knowledge base. More thorough concentration methods are discussed in the next section, thus it has to be determined whether they are equally appropriate.
Concentration measurements
40 summation that is carried on all the firm’s fields in order to determine the concentration of the knowledge base. In the breadth section it was mentioned that diversity measurements can also reflect depth, that is, a low value would indicate a deep knowledge base. This is claimed by Bierly et al. (1996) and Moorthy and Polley (2010) and they use their diversity index (the formula is provided above on p. 24 and p. 23 respectively) to measure both breadth and depth. Lin et al. (2006) also used the same diversity measurement (see p. 23) for breadth as they did for depth except that they only took the most patented field. Similar to Boh et al. (2014), they focused on the firm’s core area of expertise. They subsequently performed the Herfindahl Index by looking at the sub-classification codes of this field. As explained on page 20 Zhang et al. (2007) aimed to control for the concentration of a firm’s knowledge base, even though they did not explicitly address it, this is a representation of depth, so as described in the definition chapter. It is measured in two steps: first, the Relative Technological Advantage of a firm in a particular class is calculated by taking the fraction of the number of patents a firm holds in a technology class, which in turn is divided by the outcome of the total number of firms that have patents in that technology class divided by the total patents of all firms. This yields the following: /.01∑ %%2/ ∑ %2 2
2/ ∑ 322
, with 1 the number of patents by firm i in class t. In the next step the concentration is calculated by dividing the standard deviation with the mean: 4,)'*5+*, =
6789
:789. A high score means the firm is concentrated in one or a few technological classes. Another
41 Moreover, Lin and Wu (2010) also used a more sophisticated depth measurement, which is formulated as follows: < '=*ℎ = ? @∑ 1 1 AB ∙ 1 ∑ 1C AB
it measures the weighted share of patent counts among all classes listed in the firm. 1 expresses the number of applied patents of the firm during a three year window. Consequently, ∑ AB 1 is the total amount of the firm’s patents. Further, ∑ 1 represents the total number of patents among all firms. The calculation is made for every class, which in their study is 36 classes. This equation is based on the notion that depth relates to many of the firm’s patents in a few concentrated classes. To recap, the last three measurements shared a similar characteristic by including other firms. This corresponds to the fact that firms have a higher level of sophisticated knowledge when they are more specialized than their counterparts.
Lastly, on a similar notion of the more a firm utilizes the same field, the more specialized, Özman (2007) created a measurement which acts the opposite as his breadth method, as expressed in the following equation:
= ∑ D
42 assigned to a patent. Therefore, the equation above was extended in order to create a more precise measurement. For the purpose of measuring the extent to which a patent intensively relies on a small number of fields, a Herfindahl Index was included to measure the proportion of a certain field in a particular patent. Thus, depth is finally given by:
= ∑ D(1 − ( )
here, D remains the same, ( is the outcome of Herfindahl Index that measures diversity of a patent, and given that this measurement is used for field depth, expresses the total amount of patents in a field. Transforming this measurement to firm-level simply means taking the group of patents of a firm instead of the group of patents in a particular field. It is left unexplained why the rough measurement is used to measure firm depth, other than that measuring technology field breadth and depth was the article’s main focus and measuring this for firms got allocated a smaller section. To conclude, Özman’s (2007) unique method makes sense as it directly relates to familiarity of components.
Coherence measurements
43 interest: relatedness in the patterns of technological diversification regarding their total firm sample. Ultimately, their final equation explains the probability whether a firm’s core technology field is also active in other fields. This encompasses measuring relatedness between a firm’s core technology field and other fields, together with other explanatory variables that account for remaining diversity opportunity factors. The related study of Nesta and Saviotti (2005) and Colombelli and Quatraro (2014) largely base their methodology on Breschi et al. (2003) but specifically focus on how coherent a firm’s knowledge base is. They measured this as follows. Same with Breschi et al. (2003) they focused on a firm’s patents and their technology field classifications. In order to measure knowledge base coherence, one first needs to understand relatedness among all technology classes. This is defined as:
/ = EG− μ
This results in a matrix that shows relatedness between technology field i and j. E is the number patents with joint occurrence of technology field i and j. The mean μ is defined as:
μ =H<H
where H is the number of patents in class i, H is the number of patents in class j, and K the total amount of patents of all fields combined. This formula indicates the expected value of random co-occurrence. This implies that when E greatly exceeds μ the fields are highly related, and inversely, poorly related when E is much smaller. Or put differently, high or low complementarity between technologies. Variance G is defined as follows:
44 Up until now they used similar measurements with Breschi et al. (2003). Now they depart from them by measuring a firm’s patent portfolio coherence in the following two steps. First, they measure the average relatedness of class i with respect to all other classes within the portfolio weighted by the patent count, as defined below:
L0/ =∑ /∑ M M
where Pj stands for patent count. L0/ measures the expected relatedness of class i with any other given class present within the portfolio. Similar to the overall relatedness of the classes, as explained above, a positive (negative) outcome indicates close (weak) relatedness between classes. Finally, coherence of the firm’s knowledge base is measured as follows:
4HN = I∑
∙ L0/ K
where is the number of patents in class i of a firm’s portfolio. COH is the average relatedness of any class with respect to any other class. Again, a positive value indicates high relatedness of a firm’s technological knowledge base. Inversely, a negative value means poor relatedness.
45 equations to measure a firm’s coherence. In line with their predecessors, first the relatedness between pairs of classes need to be assessed before turning to a firm’s portfolio. This is performed in two steps:
"O = H ∙ PQ/.R
" is the expected number of cited patents of class j in citing patents of class i, Q the amount of
patents classified in class j, and T the total population of patents. H is explained in the following equation: H = ∑ HQ . This is the sum (taking all cited tech classes) of H, which is the number of times patents class i cite patents from class j. This makes H the total amount of times class i is citing. Returning back to the first equation, Q/T implies the proportion of the size of class j. This is included to cover the fact that there is a higher probability of a class being cited if many patents are classified in that class. Still following the similar steps of previous studies, the actual relatedness is calculated as follows:
/ = PH+ HR P" + "R
This lead to a matrix and when the value is greater than 1, classes i and j are more related than the expected random citation patterns.
Same with Nesta and Saviotti (2005) the following step is to calculate the weighted average relatedness, in order to do so the same equation is used:
L0/ =∑ /∑ M M
46 4HN =P∑ 4HN ∙ R
P∑ R
4.2.3
Depth measurements overview
47 Table 3. Depth Measurement Framework
Measurement group
Subgroup Studies Aim as described by authors
Measurement Possible additions / adjustments Counting Counting
classes and taking the average
(a) Avenel et al. (2007) (b) Prabhu et al. (2005); Wu and Shanley (2009) (a) Hybridization in technological knowledge base (b) Knowledge depth
(a) Average number of fields assigned to a firm’s patents. (b) Average number of patents in
the classes listed in the knowledge base. Identify core
field
Boh et al. (2014)4 Depth of expertise Identify most patented class, and divide
the number of patents of this class by the total number of the firm’s patents. Counting
citations
(a) Katila and Ahuja (2002)
(b) Katila (2002)
(a) Search depth (b) Average internal
search age
(a) Divided the sum of the number of times a patent was cited by the total amount of cited patents in a period of five years,
yielding:
∑ /'=**, ),T*
.,*+ )*+*,-(b) The average age of cited patents that are held by the firm in one year.
Concentration Counting classes proportionally
(a) Subba Narasimha and Ahmad (2001) (b) Zhang et al. (2007) (c) Lin and Wu (2010) (d) Özman (2007)5 (a) Depth of technological knowledge (b) Concentration of
the knowledge base
(a) Summation of the proportion of a firm’s patents in a particular class against the total amount of patents, multiplied by the fraction of the firm’s total patents in that class.
(a)
4
This measurement is transformed into a firm-measurement. Boh et al. (2014) originally measured individual depth. 5
48 Table 3. Depth Measurement Framework (Continued)
Measurement group
Subgroup Studies Aim as described by authors
Measurement Possible additions / adjustments (c) Knowledge depth
(d) Depth of the firm
(b) Dividing the number of patents a firm holds in a technology class by the total patents the firm has, which in turn is divided by the outcome of the total number of firms that have patents in that technology class divided by the total patents of all firms. Subsequently, calculate concentration by dividing the standard deviation with the mean.
(c) Square root of the sum of the ratio of the number of patents in a particular class against the total number of patents in the portfolio, multiplied by the ratio of the number of patents in a particular class against the total number of all patents in that class.
(d) =∑ U V
V , where D the depth
of a single patent by counting the number of secondary codes similar to the patent’s main code. is the total firm’s patents.
49 Table 3. Depth Measurement Framework (Continued)
Measurement group
Subgroup Studies Aim as described by authors
Measurement Possible additions / adjustments Diversity6 (a) Bierly et al. (1996)
(b) Lin et al. (2006) (c) Moorthy and Polley
(2010)
(a) Depth of the knowledge base (b) Core field diversity (c) Depth of
technological knowledge
Diversity index. See table 1. (a) Take a firm’s most patented field and specifically apply the index to this field. Coherence (a) Nesta and Saviotti
(2005) (b) Leten et al. (2007) (c) Colombelli and Quatraro (2014) Coherence of a firm’s knowledge base
Identify relatedness among fields i and j, resulting in a matrix. Subsequently, measure relatedness among a firm’s fields i and j. Lastly, averaging the relatedness of every two fields while controlling for the size of the portfolio (see above for the formulas).
(b) Instead of making use of the classes listed in each of the firm’s patents, draw upon the classes listed in each patent’s cited patents.
6
50
4.3
Comparison of depth measurements
Methods using classification codes assigned to patents are also the most common for depth methods. Same with the breadth comparison section, knowledge bases are depicted from deep to broad, and one patent is assigned to a one-level technology field, meaning that the methods of Özman (2007) and Lin et al. (2006) are not illustrated in the table. The depth measurements are depicted in Table 5 below. For the methods 1 and 2 a high score indicates high depth, while for the concentration indexes a low score indicates high depth. All indexes are on a scale of 0 to 1, or are rescaled to values between 0 and 1.
Table 5. Comparison of Depth Measurement
Firm 1 Firm 2 Firm 3 Firm 4 Firm 5
Patents Patents Patents Patents Patents
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 Field 1 x x x x x x x x x x Field 2 x x x x x x x Field 3 x x x Field 4 x Field 5 x (1) Average number of patents in the classes listed in the
knowledge base 1 (3) 0.75 (2,5) 0.5 (2) 0.335 (1,67) 0 (1)
(2) No. of patents in core field / Total
52
5
Discussion
This section further discusses which measurements covered in the literature review are most appropriate based on the present study’s definition of breadth and depth, identified shortcomings, precision, and practicality.
5.1
Breadth
53 provided the most accurate values, together with the entropy index by Bierly and Chakrabarti (1996). Therefore, these three methods are chosen to most appropriately indicate breadth.
5.2
Depth
Measurements that take the average of the classifications of the firm or the average of patents within classes (see Table 3) are too coarse-grained to fully grasp the concept of depth. Regarding the former, taking the average of all fields does not indicate the familiarity per fields. Furthermore, it could also accidently capture breadth of the knowledge base, giving that a high average of the different classes used by the firm would imply that the firm draws from a diverse set of knowledge. Regarding the latter measurement, a high average of patents could also mean the firm is active in a broad range of fields. The studies Boh et al. (2014) and Lin et al. (2006) limit their depth measurement to the core field and take the most patented field as starting point. However, this is also too coarse-grained, for instance, when firms are almost evenly active in three fields, this method would not consider those two other fields.
54 main limitations. Lastly, this measurement is the most complex out all the measurements identified, adopting this measurement might be challenging.
55 other firms is a component exclusively relevant to depth. Therefore, those three methods are chosen to most appropriately indicate depth.
5.3
Measurements for researchers and in practice
Now that the most appropriate methods to measure a firm’s knowledge base have been identified, in order to put them into practice a couple of aspects need to be considered. First of all, studies found an interaction effect between breadth and depth. They are interwoven and can be complementary to each other. Meaning that they can both positively contribute to innovative performance if firms have the right amount of breadth and depth. They can be complementarily in that they both positively affect a firm’s absorptive capacity and both ease managerial decision making by reducing uncertainty. Breadth reduces uncertainty because a firm becomes less dependent on a few fields (Leten et al., 2007). Depth reduces uncertainty because managers get a better understanding of what projects to invest in (Moorthy and Polley, 2010). Furthermore, in order to come up with inventions an understanding of a diverse range of technology fields as well as deep understanding within fields is required (von Tunzelmann, 1998). For these reasons, breadth and depth need to be measured in combination and should not be used separately. This seems intuitive, however, breadth is found to be more often discussed and measured than depth. Second, curvilinear effects have been observed for both dimensions. In order to cover this effect it is recommended to measure for consecutive years. Lastly, when measuring breadth and depth with the recommended methods, the relatedness between technology fields need to be taken into account by considering one of the coherence measurements.
5.4
Limitations and suggestions for further research
57
6
Conclusion
In light of the importance of knowledge to the firm’s competitive position, the aim of this thesis was to review the dimensions breadth and depth of a firm’s knowledge base. This consisted out of two parts as expressed in the research question: What is breadth and depth of a firm’s knowledge base and how can they best be measured? What the dimensions mean was reviewed by examining the given definitions. Subsequently, it became possible to examine the most appropriate measurements. The need for this review can be shown in the different measurements used to measure the same dimensions. Oftentimes authors did not justify or poorly explained why they chose a certain measurement. The importance for a review is also reflected in the different terminology used. Thus, an analysis that determines how their definitions and measurements differ was all the more pressing. The main contribution of this study is that it synthesizes the literature field of a firm’s knowledge base breadth and depth. This is performed by coming to an unequivocal definition so that different concepts and measurements become comparable.
Garcia-58 Vega (2006), Leten et al. (2007), and Bierly and Chakrabarti (1996) that measure breadth are preferable, with respect to depth, the concentration indexes by Subba Narasimha and Ahmad (2001), Zhang et al. (2007), and Lin and Wu (2010) are preferable.
59
References
Adams, R., Bessant, J., & Phelps, R. (2006). Innovation management measurement: A review. International Journal of Management Reviews, 8(1), 21-47.
Alavi, M., & Leidner, D.E. (2001). Review: Knowledge management and knowledge management systems: Conceptual foundation and research issues, MIS Quarterly, 25(1), 107-136.
Alcorta, L., Tomlison, M., & Liang, A.T. (2009). Knowledge generation and innovation in manufacturing firms in China, Industry and Innovation, 16(4-5), 435-461.
Avenel, E., Favier, A.V., & Rieu, V.M. (2007). Diversification and hybridization in firm knowledge bases in nanotechnologies. Research Policy, 36, 864-870.
Amit, R., & Schoemaker, P. J. H. (1993). Strategic Assets and Organizational Rent. Strategic Management Journal, 14(1), 33-46.
Barney, J. (1991). Firm resources and sustained competitive advantage. Journal of Management, 17(1), 99-120.
Bierly, P., & Chakrabarti, A. (1996). Generic knowledge strategies in the U.S. pharmaceutical indutry. Strategic Management Journal, 17, 123-135.
Boh, F., Evaristo, R., & Ouderkirk, A. (2014). Balancing breadth and depth of expertise of innovation: A 3M story. Research Policy, 43, 349-366.
Bontis, N. (2001). Assessing knowledge assets: A review of the models used to measure intellectual capital. International Journal of Management Reviews, 3(1), 41-60.
Breschi, S., Lissoni, F., Malerba, F. (2003). Knowledge-relatedness in firm technological diversification. Research Policy, 32, 69-87.
Brusoni, S., Criscuolo, P. & Geuna, A. (2005). The knowledge base of the world’s largest pharmaceutical groups: what do patent citations to non-patent literature reveal? Journal of Economics of Innovation and New technology, 14(5): 395-415.
Cavusgil, S.T., Calontone, R.J., & Zhao, Y. (2003). Tacit knowledge transfer and firm innovation capability, Journal of Business & Industrial Marketing, 18(1), 6-21.
Chen, C. (2004). The effect of knowledge attribute, alliance characteristics, and absorptive capacity on knowledge transfer performance. R&D Management, 34(3), 311-321.
Coff, R.W., Coff D.C., & Eastvold, R. (2006). The knowledge-leveraging paradox: how to achieve scale without making knowledge imitable. Academy of Management Review, 31(2), 452-465.
60 Colombelli, A., & Quatraro, F. (2014). The persistence of firm’s knowledge base: a quantile approach to Italian data. Economics of Innovation and New Technology, 23(7), 585-610.
Decarolis, D.M., & Deeds, D.L. (1999). The impact of stocks and flows of organizational knowledge on firm performance: an empirical investigation of the biotechnology industry. Strategic Management Journal, 20(10), 953-968.
Eisenhardt, K.; Schoonhoven, C. (1996). Resources-based view of strategic alliance formation: strategic and social effects in entrepreneurial firms. Organization Science, 7(2), 136-150.
Engelsman, E.C., & van Raan, A.F.J. (1992). A patent-based cartography of technology. Research Policy, 23, 1-26.
Fleming, L. (2001). Recombinant uncertainty in technological search. Management Science, 47(1), 117-132.
Fleming, L., Mingo, S., & Chen, D. (2007). Collaborative brokerage, generative creativity, and creative success. Administrative Science Quarterly, 52, 443-475.
Garcia-Vega, M. (2006). Does technological diversification promote innovation? An empirical analysis for European firms. Research Policy, 35, 230-246.
Grant, R. M. (1996). Towards a knowledge-based theory of the firm. Strategic Management Journal, 17, 109-122.
Gruber, M., Harhoff, D., & Hoisl, K. (2013). Knowledge recombination across technological boundaries: Scientists vs. engineers. Management Science, 59(4), 837-851.
Hall, B.H., Jaffe, A.B., & Tratjenberg, M. (2001). The NBER patent citation data file: Lessons, insights and methodological tools. NBER Working paper 8498.
Hitt, M.A., Ireland, R.D., & Lee, H. (2000). Technological learning, knowledge management, firm growth and performance: an introductory essay. Engineering and technology management, 17, 231-246.
Hitt, M.A., Keats, B.W., & DeMarie, S.M. (1998). Navigating in the new competitive landscape: Building strategic flexibility and competitive advantage in the 21st century. Academy of Management Executive, 12(4), 22-42.
Jansen, J. J. P., Van Den Bosch, F. A. J. & Volberda, H. W. (2006). Exploratory innovation, exploitative innovation, and performance: Effects of organizational antecedents and environmental moderators. Management Science, 53, 1661-1674.
Jassimuddin, S.M., Klein, J.H., & Connell, C. (2005). The paradox of using tacit and explict knowledge: Strategies to face dilemmas, Management Decision, 43(1), 102-112.
Katila, R. (2002). New product search over time: past ideas in their prime? Academy of Management Journal, 45(5), 995-1010.
