Radical versus incremental innovations in the U.S. telecom industry : a study drawing on patent data.

(1)

Radical versus incremental innovations in the U.S. Telecom Industry:

A study drawing on patent data

Master's Thesis

Student: Pauline Hogenkamp (6413730)

University of Amsterdam, Faculty of Economics and Business

Supervisor: Dr. R.M. Singh

(2)

Abstract

In today's rapid changing environments, competitive advantage dissipates over time, as competitors disrupt market equilibrium and gradually undermine existing business models. Firms must be possess innovative capabilities in order to adjust to environmental changes and to sustain superior performance. Innovation can be divided into two types of innovation: radical innovation and incremental innovation. These types of innovation draw on different knowledge sources and different organizational capabilities. Existing literature on innovation has acknowledged the difference in capabilities required for both types of innovation.

However, authors have been rather inconclusive on the underpinnings, or so-called micro foundations, of the firm-level capabilities required to generate radical or incremental innovation. In an attempt to relate the firm's micro foundations with its innovative

capabilities, the current study brings forward four organizational elements as a predictor of the character of the firm's innovative capabilities. The results show that radical innovative capabilities are positively related with the acquisition of technology, and that incremental innovative capabilities are positively related with the internal development of technology. Furthermore, there is an inverted curvilinear relationship between the deployment of patent attorneys and the character of the innovative capabilities. This implies that there is an optimal percentage of external attorneys in order to develop a radical innovation. Though, this

relationship is not significant, so no conclusions can be made. Also the age of the technologies that are used to develop the innovation, and the deployment of external inventors, have both no significant relationship with the type of the firm's innovative capabilities.

(3)

1. Introduction

In today’s rapidly changing business environments, with growing global competition, increasing technological advances, and frequently changing customer preference changes, economies of scale and scope are no longer sufficient to achieve a sustained competitive advantage. The underpinnings of these advantages are eroded by innovations by competitors. Creative innovations will be set in motion when competitors try to level the playing field, and as a result competitive advantages will dissipate (Christensen, 2001). Therefore, for many existing organizations the ability to develop new ideas and innovation is one of the top priorities (Porter and Stern, 1999).

To adjust to the rapidly changing environment, firms in dynamic environments are in a constant pursuit of adding, shedding, renewing and reconfiguring resources and capabilities in order to sustain a superior performance (Teece et al., 1997; Eisenhardt et al., 2000). This firm level perspective of the need to respond to environmental changes is explicated by the

Dynamic Capabilities View. According to Teece (2007) dynamic capabilities consist of three capabilities: the ability 1) to sense opportunities, 2) to seize opportunities and 3) to

reconfigure resources. The micro foundations, or the underpinnings of these firm level capabilities, are defined by Teece as "distinct skills, processes, procedures, organizational structures, decision rules, and disciplines". A good understanding of these micro foundations helps a firm to develop and to enhance its innovative capabilities. Therefore, micro

foundations are the core of understanding the creation and the preservation of competitive advantage (Gavetti, 2005). In the literature there is a significant gap in understanding the micro foundations and their effect on an organization's innovative capability (Abell et al., 2008). This study will move beyond the firm level analysis and will focus on micro level processes involving human capital.

(5)

In order to respond to rapidly changing environments, firms must develop new marketable products and processes, which requires inventive and innovative capabilities. Schumpeter (1947) made a differentiation between invention, innovation and diffusion. He defines invention as the conception of new ideas. As opposed to invention, innovation is defined as the part of the change process that involves the development of the new generated ideas into marketable products and processes. Last, diffusion is the process in which the new products and processes are diffused across the market. Competitors sense the opportunity and adopt the new technology. A distinction can be made between endogenous innovation (emanating from invention), and exogenous innovation (emanating from adoption) (Bourreau and Doğan, 2001).

Research on invention focuses on the creation of opportunities, whereas research on innovation focuses on the exploitation of opportunities. In this study, the emphasis lies on the development and commercialization of opportunities, corresponding to innovation. Invention does not necessarily lead to innovation, and thus it has no technological impact. However, innovation is the implementation of the invention. It is an important driver behind resource heterogeneities between organizations. Resource heterogeneity is associated with differential value creation (Barney, 1991), enabling firms with superior resources to have a competitive advantage over their competitors.

Innovations can also be divided by their degree of radicality. There are overall two types of innovation: incremental, or conservative, and radical innovation (Abernathy and Utterback, 1978). Incremental innovations requires knowledge that build further on the existing

knowledge and the existing products on the market will remain competitive. In contrast, radical innovations will involve large technological involvements which require knowledge that is different from the existing technological knowledge. It transforms the market or it

(6)

creates a new one, and the existing products on the market will become non-competitive and obsolete.

Radical and incremental innovations are different in nature and must be investigated separately to examine the innovation characteristics (Dahlin and Berens, 2005). Research attempts to describe the differences between radical and incremental innovative capabilities (SUBRAbernathy and Clark, 1985) and noted it that both types of innovative capabilities draw on different knowledge sources (Cardinal, 2001). Due to the basic differences between radical and incremental innovations, logically, the capabilities required to create them are also distinct. Firms find it difficult to develop both incremental and radical innovations. Most firms have an incremental innovation approach, while they need both type of innovative capabilities to respond properly to the changing environment.

Several researchers conducted research on the differences in elements underlying the incremental and radical innovative capabilities. Germain et al. (1996) shows that the size of a firm is related to its radical innovative capabilities However, the size appears to be related to a firm's incremental innovative capabilities (Germain et al., 1996). The same authors reveal that decentralization of innovation adoption is positively related to incremental innovative

capabilities, because it promotes the initiation phase of innovation adoption (Dewer et al., 1986) by exposing the decision makers to information about novel inventions (Hage and Aiken, 1970). However, decentralization of innovation adoption is unrelated to radical innovative capabilities (Germain et al., 1996). This can be explained by the resistance that emerges from centralized financial decision-makers (Thompson, 1969). The more radical the innovation, the larger the resistance of the centralized financial domain (Germain et al., 1996) and hence, the less likely it is that the novel product or process is implemented.

(7)

Although there is much research conducted on the firm-level elements underlying the incremental and radical innovative capabilities, there is hardly any research done to the micro-foundations of innovative capabilities. A microeconomic analysis may be able to explain the difference between incremental and radical innovative capabilities in terms of the behaviour of individual agents, e.g. the firm's employees. Subramaniam and Youndt (2005) recognized the need for a better understanding of the micro-foundations and they conducted research on the influence of intellectual capital on the character of innovations. Since organizations invest significant resources in the development of their intellectual capital, often with a strategic need to enhance either radical or incremental innovative capabilities, it is important how intellectual capital influences the firm's innovative capabilities (Tushman and O'Reilly, 1997). Subramaniam and Youndt (2005) make a distinction between human, organizational and social capital. However, Subramaniam and Youndt assess a firm's intellectual capital at firm level. Human capital is measured by the overall skill, expertise and knowledge levels of the employees, organizational capital is measured by the organization's ability to appropriate and store knowledge in organizational-level repositories, and social capital is measured by the organization's overall ability to share and leverage knowledge among and between networks of employees. Therefore, they do not succeed in linking the behaviour of individual agents to firm's innovative capabilities.

In absence of studies testing the effect of micro level elements on the character of the firm's innovative capabilities, this study will try to answer the research question "To what

extent is there a difference between innovative capabilities required for radical versus incremental innovations in terms of micro-organizational elements?”. The organizational

elements that will be tested are the firm's capability of acquiring technologies externally versus developing technologies internally, the age of the technologies underlying the innovation, external versus internal inventors and external versus internal patent attorneys.

(8)

The study will make use of patent data to measure the effect of the above mentioned organizational elements on the innovation's character. Since there is no direct measure of a firm's innovation output, mostly, patents are used as a indirect measure of a firm's innovation output (Lee et al., 2012). Patents contain detailed information about elements as the inventors, the attorneys, the geographical location, the application and granting date and the fields of technology to which the innovation belongs. This information can be used to analyze several features of innovations. Moreover, patents have citations to previous patents and literature, which make it possible to trace linkages between the documents. It provides information about technology spillovers, since a linkage between patents from different organizations demonstrates a knowledge flow between both (Branstetter, 2000; Maurseth and Verspagen, 2002). Furthermore, the citations create indicators for the technological importance of the patent.

The patents used for this study are from the U.S. wireless telecommunications industry, registrated by the United States Patent and Trademark Office (USPTO) which is an U.S.-based agency that issues patents for inventions. The U.S. wireless telecommunications industry is a fast growing, dynamic market. Dynamic markets are characterized by a high speed of innovations (Bourreau and Dogan, 2001). With the market leaders battling for market share, this industry is highly interesting for research on innovations. Telecom corporations are

searching for new technologies to disrupt the current market dynamics (Christensen, 1997), characterized by a high level of competition.

The incumbents of the telecom industry must make the decision whether they devote resources to radical or incremental innovation activities, or to a combination of both types of innovation. Being able to determine what capabilities are more likely to generate one type

(9)

over another, makes it in the future possible for firms to make targeted investments in certain resources and capabilities, to generate the desired type of innovation. It will help managers to make decisions in which type of resources they should invest in order to achieve either one of both types of innovation. This enables firms to create an ideal balance between radical and incremental innovations.

(10)

2. Literature review

Through a specific path of capability development, organizations may gain competitive advantage at a certain point in time. However, in an ever changing, dynamic environment the organization's former competitive advantage may erode (Li and Liu, 2012). As a result, firms must constantly update their resources and capabilities to meet the future needs and requirements of the environment. The process leading to solutions to the new environmental requirements is called innovation. So through innovation firms can preserve their competitive advantage.

Innovative capabilities are defined as the capacity of developing and adopting new products and processes to satisfy future requirements (Adler and Shenbar, 1990). Firms that constantly invest in innovative capabilities are more likely to achieve sustainable innovation outputs. However, it is not yet known which organizational aspects constitute innovative capabilities. As Wolf (1994) says: "the most consistent theme found in the organizational innovation literature is that its research results have been inconsistent".

Dynamic Capabilities

In today's rapidly changing environments, the sustainability of a firm's competitive advantage is dependent on its capacity to innovate and to adapt to the environmental changes (Brown and Eisenhardt, 1997; Tushman and O'Reilly, 2008). An organization's short-term competitive advantage based on its current resource base might erode when new technologies are introduced in the market. The source of long-term competitive advantage is a firm's ability to integrate, build and reconfigure internal and external competences to address rapidly

changing environments (Teece et al., 1997). This ability is reflected by a firm's dynamic capabilities. Teece and Pisano (1994) define dynamic capabilities as the capabilities which allow the firm to create new products and processes and respond to changing market

(11)

circumstances. Dynamic capabilities are the high-order capabilities that continuously create and adapt the resource configuration to create competitive advantage (Eisenhardt et al., 2000). They differ from the organization's operational capabilities which apply to the current

operations of the organization. In contrast, dynamic capabilities are the firm's capabilities enabling it to reconfigure its operational capabilities, in order to react to a changing environment.

The dynamic capabilities can be divided into three components: adaptive, absorptive and innovative capabilities (Wang and Ahmed, 2007). Adaptive capabilities are the ability to identify opportunities in the external environment (Wang and Ahmed, 2007), absorptive capabilities are the ability to absorb external knowledge through learning processes (Lane et al., 2006), whereas innovative capabilities are the ability to transform the knowledge into new products, processes and systems for the benefit of the firm and its stakeholders (Lawson and Samson, 2001).

Also Teece et al. (2007) deconstruct dynamic capabilities into three subcategories: the ability 1) to sense and shape opportunities and threats, 2) to seize opportunities and 3) to reconfigure the firm's assets in response to threats. To sense and shape opportunities, organizations must constantly scan the environment. However, sensing involves not only scanning activities but also learning and interpretive activities (Teece et al., 2007), enabling the firm to properly value the opportunities. These sensing capabilities are thus closely related to the absorptive and adaptive capabilities identified by Wang and Ahmed (2007). Once a new opportunity is sensed, the opportunity must be seized through development and

commercialization activity. In other words, the new acquired knowledge must be assimilated, integrated with the current knowledge and used for commercial ends. Therefore, seizing capabilities suit with Wang and Ahmed's definition of innovative capabilities. Teece et al. add a third dimension to the definition of dynamic capabilities, namely the ability to respond

(12)

adequately to threats. This is in contrast with the subdivision of dynamic capabilities made by Wang and Ahmed (2007) which is limited to the sensing and seizing of opportunities.

Barreto (2010) explains the dynamic capabilities in a similar way as Teece (2007). He proposes that dynamic capabilities consist of four dimensions, namely the ability to sense opportunities and threats, to make timely decisions, to make market-oriented decisions and to change its resource base. However, this explanation is criticized for two reasons. First, to make "market-oriented decisions" may not be valid for economies where the market

mechanism is not so perfect (Li and Liu, 2012).. Second, Barreto's definition only includes the identification of an opportunity and the formulation of a response to it, but it does not include the implementation of a course of action (Li and Liu, 2012).

Most current explanations of dynamic capabilities decompose dynamic capabilities into three dimensions, namely strategic sense-making capacity, timely decision-making capacity and change implementation capacity (Li and Liu, 2012). Strategic sense-making capacity is the capacity to analyse the usefulness of the current resource base and to sense opportunities and threats in the external environment. The timely decision-making capacity is the capacity to quickly formulate, evaluate and choose from alternative strategies to timely adapt to environmental changes (Shafman and Dean, 1997). The change implementation capacity is the ability to execute and to coordinate the chosen strategies in the decision-making process.

According to Teece et al. (1997) differences in the firm's ability to proactively search for and to seize opportunities, are the result of differences in the firm's dynamic capabilities. Most researchers assume that these differences in dynamic capabilities are caused by the context-specific and path-dependent evolution of capabilities (Dierickx and Cool, 1989). This implies that dynamic capabilities have firm-specific or idiosyncratic features. On the other

(13)

hand, Eisenhardt et al. (2000) assume that dynamic capabilities are "best practices" and that there are certain commonalities in dynamic capabilities across firms, which contradicts the theorem that they are firm-specific.

Firms facing similar industry conditions are likely to develop capabilities to perform similar activities, so commonalities in capabilities arise from the requirements of the

operating environment (Jantunen et al., 2012). Since sensing capabilities are aimed at scanning the environment, these capabilities are probably dependent on the type of environment. It is therefore that practices comprising sensing capabilities are likely to be similar across firms within one single industry (Jantunen et al., 2012). The development processes in the practices comprising seizing and reconfiguring types of capabilities are more context-specific and path-dependent and may therefore differ in their manifestation (Jantunen et al., 2012).

Organizational learning

A subset of an organization's innovative capabilities is its ability to recognize the value of new information, to assimilate it and to apply it to commercial ends (Cohen and Levinthal, 1990). This ability collectively constitutes the organization's ' absorptive capacity' (Cohen and Levinthal, 1990). According to Camisón et al. (2010) absorptive capacity has four dimensions: acquisition, assimilation, transformation and application. Zahra and George (2002) divide absorptive capacity into 'potential absorptive capacity' and 'realised absorptive capacity'. The potential absorptive capacity is the capacity to identify and to acquire the external knowledge and the realised absorptive capacity is the capacity to transform and to exploit the external knowledge (Zahra and Geroge, 2002). However, absorptive capacity is not static, but it evolves over time through the firm's learning processes (Todorova and

(14)

the firm's innovative capability is closely tied with the firm's ability to utilize its knowledge resources (Subramaniam and Youndt, 2005). Since the firm's existing knowledge resource base is embedded in its past experiences and its resulting routines, the innovation process within a firm is path-dependent (Phene et al., 2010).

The firm can increase its absorptive capacity by extending its existing knowledge base. The knowledge can be extended by either acquisition of new knowledge or by the modification of the existing knowledge (Kostopoulos, 2011). This process, of acquiring and modifying knowledge resources, is called learning. A firm that is able to learn and to extend its existing knowledge base has the capacity to develop its absorptive capacity and will be able to use new information for innovation. Strategy scholars recognize two types of learning: exploration and exploitation. Exploration enables the creation of new knowledge, whereas exploitation bolster the refinement and use of existing knowledge (Levinthal and March, 1993; March, 1991; Levinthal, 1997; Katila and Ahuja, 2002). Exploratory learning requires potential absorptive capacity (Zahra and George, 2002), because it involves acquisition of new, external knowledge (Gebauer et al., 2012). In contrast, exploitative learning requires 'realised absorptive capacity' (Zahra and George, 2002), because it reflects the application of existing knowledge (Gebauer et al., 2012).

Firms must seek for a balance between the ability to exploit their existing assets and the ability to explore new opportunities and to adapt to them (Gibson and Birkinshaw, 2004). They can choose for temporal separation of the two activities, which means that the firm explore at one point in time and exploit at another (Brown and Eisenhardt, 1997). Another approach is organizational separation of both types of learning, whereby firms divide themselves into separate organizational units exclusively dedicated to exploitation or

exploration (Benner and Tushman, 2003). Despite the potential merits of the balance between exploitation and exploration, firms must manage trade-offs when pursuing the two

(15)

contradicting types of learning simultaneously (Lavie et al., 2010). Exploitation and

exploration rely on distinctive modes of organizational behaviour and routines (March, 1991). However, both types of learning activities are supported by the same scarce resources,

constraining managers to make decisions whether to invest in exploitation or exploration activities.

If organizations search locally for new knowledge, they are likely to innovate in the neighbourhood of their existing competencies (Prahalad and Hamel, 1990). Hence,

exploitation leads to innovations that are build on the firm's established knowledge base (Nelson and Winter, 1982). Exploration is an initial step toward developing new, yet unknown areas, and requires a more distant search for knowledge. Distant knowledge may lack

salience and significance, since it does not reinforce the established organizational routines, resources and capabilities appropriate for the firm's existing technological trajectory.

According to Levinthal and March (1993) the failure to access and utilize distant knowledge is called the "myopia of learning". Firms must overcome the myopia of learning in order to pursue exploratory search.

Incremental and radical innovative capabilities draw differently upon organizational knowledge. Since incremental innovative capabilities are defined as "the capabilities to generate innovations that refine and reinforce existing products and services" (Subramaniam and Youndt, 2005), it is likely that this type of innovation draws upon knowledge that builds on and reinforces existing knowledge (Abernathy and Clark, 1985). In contrast, radical innovative capabilities are "the capabilities to generate innovations that significantly transform existing products and services" (Subramaniam and Youndt, 2005). This type of innovation draws upon knowledge that destroys the value of the existing knowledge base (Abernathy and Clark, 1985) and transforms the knowledge base into something significantly

(16)

new. In short, incremental innovation is positively related with exploitation and radical innovation is positively related with exploration (Gupta et al., 2006).

An organization's myopia of learning tends to limit the organization to knowledge search in the local environment. This knowledge is easier to absorb and to recombine with the existing knowledge. For that reason, it is expected that incremental innovations, which

reinforce the existing knowledge, are developed by the firm itself. In contrast, exploratory innovation output is positively related with technology-related acquisitions (Wagner, 2011). Technology-related acquisitions are related to the exploration of the knowledge bases previously inaccessible to the acquirer (Gupta et al., 2006). Acquisitions that focus on exploitation are rather marketing-related (e.g. increased market-power), distribution-related (e.g. access to distribution channels) or production-related (e.g. efficiency gains) than technology-related (Wagner, 2011). Therefore, firms that acquire technologies from external sources are expected to generate exploratory innovation output. Since the technology was previously inaccessible to the acquiring firm, the firm was not able to assemble the external knowledge internally and to develop it internally to commercial ends. By acquiring the technology, the firm is able to generate innovations that are not build on the firm's existing knowledge base. The greater the depth of knowledge, the better the firm's capabilities in product and process differentiation (Zahra, Ireland and Hitt, 2000). Therefore, it is expected that radical innovative capabilities are positively related with the acquisition of technologies by interactions with external organizations.

Hypothesis 1a. Radical innovative capability is positively related to external acquisition of

new technologies.

Hypothesis 1b. Incremental innovative capability is positively related to internal development

(17)

Scope of search

Incremental innovations are positively related with exploitation, which concerns the employment of current assets and capabilities (March, 1991) in order to increase in the efficiency of current technologies (Henderson, 1993). Exploitative learning requires local knowledge that is closely related to the firm's pre-existing knowledge base (Helfat, 1994; Martin and Mitchell, 1998; Stuart and Podolny, 1996). This type of learning improves the incremental innovative capabilities, but it simultaneously reduces the incentives for high-impact radical innovations (Ahuja and Lampert, 2001). In contrast, radical innovations are positively related with explorative learning, which is concerned with the development of new assets and capabilities (March, 1991). In other words, exploratory learning involves a

conscious effort to move away from the organization's current routines and knowledge bases (March, 1991; Miner, Bassoff and Moorman, 2001).

Firms search across the landscape of possible technologies to develop an innovation (Nelson and Winter, 1982). They search for new products (Ahuja and Katila, 2002), for superior organizational designs (Bruderer and Singh, 1996), for optimal manufacturing methods (Jaikumar and Bohn, 1992) and for the best methods to implement novel

technologies (Von Hippel and Tyre, 1995). Knowledge search can be characterized in terms of 'depth' and in terms of 'scope' (Ahuja and Katila, 2002). The depth of knowledge search is the degree to which existing knowledge is reused or exploited. The scope of knowledge is the degree to which the search entails exploration. An increase in the scope enriches the knowledge pool by adding distinctive new variations of technologies (Ahuja and Katila, 2002). The larger the scope of search, the less related are the technologies to the existing knowledge base.

According to Ahuja and Katila (2004), knowledge search has two dimensions: geographic and scientific search. The technological landscape is sector-specific (Ahuja and

(18)

Katila, 2004) and is differentiated across geographical space (Freeman and Soete, 1997; Katila and Ahuja, 2002; Hansen and Lovas, 2004). Organizations that pursue exploration can break away from local search by crossing the technology-science boundary (Cockburn et al., 2002) or by crossing the geographical boundaries. Besides the 'science' and 'geography'

dimensions mentioned by Ahuja and Katila (2004), 'time' is another dimension that influences the technological landscape. Over time, firms can sustain heterogeneous resource positions (Helfat, 1994; Knott, 2003). Technologies originating back in history are located further away from the organization's current knowledge base than the technologies used in its current routines. Organizations that increase their search scope to aged technologies are able to recombine non-local knowledge sources in order to generate explorative innovations. Therefore, it is expected that radical innovative capabilities are positively related with the reuse of aged technologies. In contrast, incremental innovative capabilities are expected to have a positive relationship with the use of new technologies, which are in the neighbourhood of the firm's current knowledge base.

Hypothesis 2a. Radical innovative capability is positively related with the (re)use of aged

technologies.

Hypothesis 2b. Incremental innovative capability is positively related with the use of new

technologies.

Intellectual capital

Since the process of innovation is commonly equated with the practice of utilizing existing and searching for new knowledge sources, an organization's ability to innovate is closely tied with its intellectual capital (Subramaniam and Youndt, 2005). An organization's intellectual capital can be divided into three distinct aspects - namely, human, organizational,

(19)

and social capital (Davenport and Prusak, 1998). The first aspect, human capital, is defined as the knowledge, skills and abilities residing with and utilized by individuals (Schultz, 1961). Second, organizational capital is the institutionalized knowledge and codified experiences residing within organizations and utilized through databases, patents, manuals, structures, systems and processes (Youndt et al., 2004). The third aspect, social capital, is the knowledge residing within and utilized by interactions among an organization's individuals and their networks of interrelationships (Nahapiet and Ghoshal, 1998). Each of these three aspects of intellectual capital leads to a different approach for reinforcing and transforming knowledge. Therefore, it is likely that different types of innovative capabilities, that vary in the kinds of knowledge they draw upon, are selectively influenced by different aspects of intellectual capital (Subramaniam and Youndt, 2005).

Research shows strong evidence that institutionalized knowledge accumulated in an organization's routines help to reinforce the existing knowledge base, and consequently enhance the organization's incremental innovative capabilities (Subramaniam and Youndt, 2005). In other words, organizational capital has a positive effect on incremental innovative capabilities. Human capital by itself has little effect on incremental innovative capabilities. It has even a negative effect on radical innovative capabilities, suggesting that fiercely

independent experts are not able to transform knowledge on their own (Subramaniam and Youndt, 2005). However, the interaction of human and social capital positively influences radical innovative capability (Subramaniam and Youndt, 2005) indicating that, unless individual knowledge is networked through interrelationships, it provides little benefit in terms of innovative capabilities.

(20)

The human capital involved in the innovation process can be categorized into three types: 1) pure scientists; 2) pure inventors; and 3) bridging scientists (Subramaniam, 2012). Pure scientists are the human capital solely engaged in scientific research in pursuit of a publication of their research findings, while pure inventors are solely engaged in the translation of the scientific knowledge into commercial products and processes

(Subramaniam, 2012). Inventors are not only involved in the translation of the knowledge into products and processes, but they are also concerned with the protection of the innovations, enabling the organization to benefit from its innovation efforts. The third type of human capital, bridging scientists, comprises the individuals that bridges the scientific and technological domain.

Different types of human capital contribute differently to an organization's learning activities (Allen 1977; Subramaniam, 2012). Pure scientists are exploring the scientific knowledge base in hope to find new knowledge that could potentially lead to innovations. Hence, they are hypothesized to contribute directly to pioneering innovation (Subramaniam, 2012; Baba et al., 2009). Pioneering innovation is defined as innovation that emerges from the scientific knowledge base and that is built on no prior technologies. Meanwhile, pure

inventors have experience with applied technologies and are expected to be able to recombine the applied technological knowledge into a new applied (product or process) technology (Subramaniam, 2012). This type of innovation is known as recombinatory innovation. It draws upon the search for knowledge from multiple technologies in an attempt to recombine that knowledge into a novel innovations (Henderson and Clark, 1990; Subramaniam, 2012; Ahuja and Katila, 2004; Fleming, 2001; Yang et al., 2010). The pattern through which a firm's inventors exchange knowledge influences which recombinant innovations the firm generates (Carnabuci and Operti, 2013).

(21)

The recombination of aged technologies can be valuable, because of their reliability (Katila, 2002), or because of its unexploited potential (Nerkar, 2003). As recombinatory innovation derives benefits from the extension of the scientific knowledge base, pure scientists facilitate the recombination activities of the pure inventors (Subramaniam, 2012; Baba et al., 2009). Pure inventors help firms to translate the scientific capabilities facilitated by the pure scientists into valuable innovations (Subramaniam, 2012). 'Bridging inventors' are needed to bridge the scientific and the technological domain within an organization

(Subramaniam, 2012; Gittelman and Kogut, 2003).

Collaboration with external human capital gives a firm access to distant knowledge resources and enables the firm to recombine a broader perspective of technologies into a new concept. By providing the firm with novel knowledge sources, external scientists contribute to pioneering innovations and external inventors contribute to recombinatory innovations. Hence, it is expected that collaboration with external inventors is positively related with radical innovative capabilities. In contrast, organizations that exclusively utilize internal human capital do not have access to other knowledge resources and have a relatively small perspective of technologies to recombine. These organizations are more likely to search for and to recombine local knowledge which is close to the existing knowledge base. Naturally, it is expected that these organizations are more likely to deliver innovations that are built upon their existing technologies. In other words, it is expected that the use of exclusively internal inventors is positively related with incremental innovative capabilities.

Hypothesis 3a. Radical innovative capability is positively related with collaboration with

external human capital.

Hypothesis 3b. Incremental innovative capability is positively related with use of internal

(22)

Patent attorneys

Patents are widely used as indicators of inventive and innovative activities. The difference between invention and innovation is that invention refers to the creation of a novel product or process, whereas innovation refers to the commercializing of that novel concept (Hitt et al., 1993). Before the invention can become an innovation, entrepreneurial efforts are required to develop, manufacture and market the invention. A firm's innovative capacity can be defined as the successful outcomes of its inventions (Suarez-Villa, 1990). Since patents do not encompass the development of the invention, it must be interpreted as an indicator of a firm's inventive activity rather than a key measure of a firm's innovative output (Griliches et al., 1986; Ma, 2008).

Logically, invention would not take place if the inventor is not able to reap the rewards of it. According to the development model, patents enable organizations to appropriate the gains from their inventions, either through developing it themselves or by selling it to others for them to develop, which is itself the incentive to invest in invention (MacDonald and Lefang, 1998). In this model, the society is rewarded with innovation. Another way of looking at the patent system is that of the information model, arguing that patents allow organizations to contribute their private information to a public knowledge source from which information for innovations may be drawn (MacDonald and Lefang, 1998). In this case, the disseminated information is society's reward and innovations emerge from the use of this information.

In the patent application process, most inventors are assisted by patent attorneys. The main function of the patent attorney is to assist its clients in the protection of their patents, to ensure that the client can appropriate the gains from its inventions (MacDonald and Lefang, 1998). This implies that there is no place for dissemination of information to third parties and that the inventor to reap awards from the invention, either by developing it himself or by selling it to other parties. This is in line with the development model described by MacDonald

(23)

and Lefang (1998) rather than the information model which requires dissemination of

information to the society. To put it differently, attorneys can easier justify their function when invention contributes to innovation enabling the inventor to reap awards from the invention (development model), rather than on the grounds that invention contributes to dissemination of private information (information model), which indirectly contributes to the innovation of third parties (MacDonald and Lefang, 1998). If a patent attorney is not be able to protect the firm's invention, the firm would not be able to reap the rewards of it and hence, it would not develop the invention into a commercial innovation. In that perspective, the patent attorney is not only involved in the invention process, but it is also concerned with the development of the invention, or in other words, with innovation. Therefore, the quality of a patent attorney is not only an indicator for inventive activity, but also for innovative activity.

For the application and the protection of patents, organizations can either choose to deploy patent law expertise internally or to employ external patent attorneys. Whereas external patent attorneys tend to specialize by broad technology areas, in-house patent attorneys are likely to be specialized in the firm's existing technology base and have a deep understanding of the technologies (Somaya et al., 2007). Since internal patent attorneys have a direct relationship with the firm's scientists and inventors, the attorneys are able to observe technology developments at an early stage and to interact with them to explore patentable inventions (Somaya et al., 2007). Because of their firm-specific knowledge, internal patent attorneys are better qualified to protect incremental inventions drawing on local knowledge, than to protect radical inventions drawing on distant knowledge. As mentioned earlier, invention will not happen if the firm cannot appropriate the gains from the invention, or in other words, if the attorney is not able to prevent the competitors from imitating or from inventing around. Therefore, internal patent attorneys are expected to contribute to

(24)

patent attorneys have extended knowledge of a broad spectrum of technologies. Hence, they are better qualified to protect radical inventions, which are produced by the recombination of distant knowledge sources. To put it another way, external patent attorneys are expected to contribute more to radical innovations than to incremental innovations.

Hypothesis 4a. Radical innovative capability is positively related with the employment of

external patent attorneys.

Hypothesis 4b. Incremental innovative capability is positively related with the deployment of

(25)

3. Methodology

The research conducted in this thesis attempts to answer the question to what extent there is a difference between innovative capabilities required for radical versus incremental innovations. In the previous chapter, the theoretical framework underlying the research question was presented. This framework provided several hypotheses that predict the

relationship between an organization's operational and innovative capabilities (radical versus incremental). These hypotheses will be tested by use of secondary data comprising patent data from the U.S. wireless telecom industry. Patents are useful in conducting research to an organization's innovative capabilities, since they can be seen as the output of a firm's

innovative process (Hagedoorn, 2003). Although some organizations might choose to protect their inventions in an alternative way (e.g. secrecy) (Cohen et al., 2000), in the cellular industry patents are the predominant form of intellectual property rights and are regarded as meaningful indicators of innovative activity (Trajtenberg, 1990). Patents contain detailed information about the inventive process, enabling researchers to analyze different aspects of innovations. For example, patent citations enable researchers to analyze knowledge spillover, since a citation is an indicator of knowledge flows between organizations (Maurseth and Verhagen, 2002).

3.1 Dataset

The original sample contains in total 20991 patents on the wireless technologies: Advances Mobile Phone System (AMPS), Global System for Mobile Communications (GSM), Time Division Multiple Access (TDMA) and Code Division Multiple Access

(CDMA). AMPS is 1G analog standard, GSM is a 2G analog standard and TDMA and CDMA are 2G digital standards. The "G" is this categorization refers to the generation of the wireless

(26)

network technology. The patents from the database are granted in the United States, ranging in granting years from 1980 till 2006.

Besides firm assigned patents, the original data contains 367 individual owned patents. Non-firm assigned inventors do not possess sufficient capital to develop the invention into a commercialized innovation (Dahlin et al., 2004). Furthermore, compared with firm-assigned inventors, non-firm assignees are less profit-oriented. Hence, they generate inventions that are hard to protect and therefore less useful for profit-oriented organizations. Including non-firm assigned patents would bring an unnecessary increase in the variance of the outcome, and lower the practical relevance for policy makers of R&D departments. That is why non-firm assigned patents are excluded from the sample that is used in this research.

The original sample of patents has a truncation error. A truncation error is an error made by truncating an infinite sample and approximating it by a finite sample. Since patents are subjected to a "grant lag", which is the examination time between the application date and the grant date (Trajtenberg, 1990), patent applications during the last years of the data's time frame might not be granted yet and are not included in the data set. This results in missing observations in the data. To preclude the truncation error, the time window must be shortened with the average grant lag between the application date and the granting date. This average grant lag appears to be three years. However, since there are possible outliers in the duration between the application date and the granting date, the average grant lag must be corrected with a safety factor. With a safety factor of 0,4, the correction on the time window becomes five years. Hence, the original sample is reduced from patents that are granted between 1971 and 2006 to patents that are published between 1971 and 2001.

Furthermore, patents with a considerably low number of forward citations can be considered as 'unimportant'. They do not have a high impact on future technologies.

(27)

Therefore, in this research patents with less than 30 forward citations are deleted from the sample.

3.2 Variables

3.2.1. Dependent variable

The dependent variables in this research are radical and incremental innovative capabilities. Organizations possessing incremental innovative capabilities produce innovations that build on prior technologies, whereas organizations possessing radical

innovative capabilities are able to generate innovations with a high degree of novelty. For the classification of radical and incremental innovations, this research makes use of the method introduced by Schoenmakers and Duysters (2010). Schoenmakers and Duysters argue that a patented innovation must satisfy two criteria.

Criterion 1

The first criterion refers to a minimum number of backward citations in order to be a radical innovation. All patents must indicate their prior technological linkage by citing previous patents the new technology is built on (Trajtenberg et al., 1993). Patents that have a sufficiently small number of citations can be considered as innovations that build no further on prior technologies. In other words, the lack of backward patent citations is an indicator of the originality and the creativity of the patented innovation (Trajtenberg et al., 1992). In this research, patents that have no more than three backward references are regarded as radical innovations.

(28)

Criterion 2

If the patent has more than three backward citations, the underlying innovation can still be radical. Novel patents that have citations from other technological domains rather than its own technological domain are considered as an indicator of radical innovations. These "recombinatory" innovations are the result of a recombination of existing knowledge sources from different existing knowledge sources. Therefore, the second criterion to examine whether an innovation is radical is the level of novelty. This can be done by investigating the classification of the focal patent and the classification of the cited patents. The United States Patent and Trademark Office (USPTO) has developed an elaborate classification system for the technologies to which the patented innovations belong. This classification system consists of approximately 400 main (3-digit) patent classes, and over 120,000 patent subclasses (Hall et al., 2001). In assessing the novelty of the innovation, the percentage of the cited patents that have the same national class as the focal patents provides an indication of the degree of

radicality. The lower the percentage of cited patents having the same national class, the less knowledge comes from the patent's technological domain, and the higher the degree of novelty of the patent. To make the two groups of innovations (radical versus incremental) equal in terms of records, the benchmark used to assess whether the innovation is radical is a percentage of 50%. Patents with less than 50% of the cited patents having the same national class, are labeled as indicators of radical innovations. If the percentage is larger than 50%, the focal patent is labeled as an indicator of a incremental innovation.

Criterion 3

A patent must satisfy at least one of the criteria above in order to be categorized as a radical innovation. However, it is not enough to satisfy one of these criteria. The patent must also fulfill the requirements related to the technological impact of the innovation. Generally,

(29)

radical innovations have a larger influence on future technology. Therefore, patents protecting radical innovations will have a larger number of forward patent citations than patents

protecting incremental innovation. So in order to be radical, patents must not only fulfill at least one of the first two criteria, but they must also fulfill the criterion of having a minimum forward citation count. In this research, the minimum forward citation count is 45, which is the median of the sample. Patents that do not fulfill both the backward and forward citation criteria, or the degree of novelty and forward citation criteria, are regarded as incremental innovations.

When a patent fulfills at least one of the first two criteria and the third criterion, the patent is categorized as 'radical'. Otherwise, the patent is categorized as incremental. The distribution of the dependent variable is shown in Figure 1 (Appendix A).

3.2.2. Independent variables Source of technologies

Whether a technology is externally acquired or internally developed is reflected in the source of the patent citations. When an organization generates an innovation by acquiring novel technologies, the patent designed to protect this innovation includes citations from external knowledge sources. However, when an organization generates an innovation by internally developing novel technologies, the patent contains citations that refer to prior innovations patented by the organization itself. These citations, drawing back from the firm's prior patents, are called "self-citations" (Moeen et al., 2013). The innovation's underlying knowledge originates in the firm's own intellectual capital.

Patents applied by alliances can contain citations of patents owned by one of the alliance partners. These patents are also considered as self-citations, since the alliance partner

(30)

transfers (partially) its technological knowledge to the alliance, enabling the alliance

organization to develop the innovation internally rather than acquire it outside of the door. For example, the alliance between Ericsson and General Electric (GE Ericsson) is assumed to have access to the technological knowledge of both organizations.

It is more likely that patents with high numbers of backward citations have in absolute terms more self-citations. To control for this effect of patents, the variable indicating the source of the technology is expressed as a percentage of self-citations with respect to its total number of backward citations. The distribution of the variable 'PercentageSelf-Citations' is shown in Figure 2 (Appendix 2).

Age of technology

The age of a certain technology is determined by measuring the difference between the application date of the focal patent and the application date of the cited patents. When more patents are cited, the time between focal and the most recent citation is used as the age of the technology used for the innovation. The granting date is not taken into account. This is

because of the subjectivity of the examination time (the time between the application date and the granting date). The examination time is highly dependent on the technical experts

examining the patent. Therefore, the granting date is not useful in assessing the age of the technology of an innovation.

The distribution of the variable AgeofTechnology is shown in Figure 3 (Appendix A).

Internal/external human capital

In this research, there are two types of human capital: internal and external human capital. Since each patent contains data about the inventors that have worked on the patented innovation, it can be examined whether these inventors were internalized by the firm that applied for the patent or whether they are contracted. In this research, it is assumed that

(31)

inventors that are located in the same state (or for countries other than the US, the same country) as the applicant are employed internally. Inventors that are located further away from the applicant are considered to be deployed externally. Even when these distant inventors are actually internalized within another subsidiary of the organization, these inventors are

probably not able to enter the knowledge of the applicant due to the difficulties of knowledge transfer between different subsidiaries (Birkinshaw et al., 2008). Due to barriers to knowledge transfers, some subsidiaries within the company might become isolated from the knowledge-sharing activities between different subsidiaries (Birkinshaw et al., 2008). Therefore, it might be hard, for distant inventors, to access the required knowledge about the applicant's

knowledge and to elaborate on the prior technologies. Hence, it is more appropriate to consider the distant inventor as 'external'.

In order to control for the effect of patents with large numbers of inventors having absolutely more external inventors than patents with a small number of inventors, the variable describing the externality of human capital is encoded as the percentage of external inventors. The distribution of the variable 'PercentageExternalInventors' is shown in Figure 4 (Appendix A). From the graph it can be seen that the variable is positively skewed. Most patents have zero percentage of external inventors.

Internal/external attorneys

Similar to the type of human capital, the type of patent attorneys can easily be examined by consulting the patent, because it contains data about the patent attorney that assisted in the application process.

Also for this patent applies that the variable is encoded as the percentage of external attorneys. The distribution of the variable is shown in Figure 5 (Appendix A). As the graph

(32)

shows, the variable is almost binary, with high values for zero and one per cent, but almost no counts for the values between zero and one.

3.3.3. Control variables Age of the patent

An older patent tends to have more forward citations. To adjust for this effect, the patent's age is used as a control variable. The reference date, respect to which the age is measured, is the most recent granting date in the database (2001).

The number of claims

The number of claims defines the extent to which the patent protects the innovation. In other words, it measures the scope of the patent. Logically, a high number of claims is likely to result in a high number of forward citations. To adjust for this effect, the number of claims is inserted as a control variable.

Number patents focal company

The more patents the focal company possesses, the more likely it is that the firm's patents contain self-citations. Therefore, the number of patents of the focal company is included as a control variable.

Reference to science base

This dummy variable set the control for whether the patent contains a reference to the science base. References to science base are referred to as 'other references' in the patent file.

(33)

The value of the variable equals '1' if the patent has such a reference, and it equals '0' if the patent has not.

(34)

4. Results

To test the hypotheses, first, a linear regression test is conducted to test whether there are statistical relationships between the variables. This test can identify dependencies amongst the (dependent and independent) variables. The second test conducted in this research is the binary logistic regression test. Binary logistic regression is a type of regression analysis used for predicting a dependent variable which can have only two categories. In this research, the categorical dependent variable is the character of the innovations, reflecting 'radical

innovative capabilities' and 'incremental innovative capabilities'.

4.1 Correlation

Table 1. shows the results of the linear regression test. At the 0.01 significance level, there is a significant correlation between the percentage of external attorneys and the

percentage of external inventors, between the percentage of external attorneys and percentage self-citations and between the percentage of self-citations and the age of the technology. However, these correlations do not have a high correlation coefficient (r > 0.80), so there is no reason to suspect that the interactions amongst the independent variables will influence the relationship between the independent and dependent variables. The highest correlation is amongst the percentage of external attorneys and the percentage of self-citations. The correlation coefficient is -.249 and is significant at a 1% level.

(35)

Table 1. Mean, Standard Deviation, Skewness and Correlation

N=1951

**. Correlation is significant at 0.01 level (two-tailed). *. Correlation is significant at 0.05 level (two-tailed).

4.2 Binary Logistic Regression

To test the four hypotheses, a stepwise binary logistic regression test is conducted. Logistic regression is used to predict the outcome of categorical (usually dichotomous) dependent variable based on one or more predictor variables. In this case, the categorical dependent variable is binary, which means that the dependent variable can have only two possible outcomes (radical versus incremental). Though the dependent variable binominal, the predictor variables are continuous. Given this difference, logistic regression uses the natural logarithm to transform the binomial cases into a continuous variable. The natural logarithm of

Variable M SD SK 1. 2. 3. 4. 5. 6. 7. 8. 9. 1. Character of Innovative Capabilities 0.26 0.437 1.11 (-) 2. Percentage Self-Citations 0.125 0.191 1.94 -.049* (-) 3. Age of technology 23.70 17.24 4.12 -.021 -.093** (-) 4. Percentage of External Inventors 0.302 0.423 0.86 .035 .048* -.007 (-) 5. Percentage of External Attorneys 0.543 0.493 -0.17 -.049* -.249** .019 -.097** (-) 6. Patent Age 82.77 42.22 1.99 .099** .003 .127** .103** -.125** (-) 7. Reference to Science Base 0.49 0.5 0.04 .031 -.030 -.027 -.020 .033 .001 (-) 8. Number of Patents of Focal Company 1037 1125 0.90 -.059** .218** -.025 .083** .067** -.101** .033 (-) 9. Number of Claims 23.2 19.16 2.91 .068** -.013 -.058* -.030 .011 -.166** .092** .005 (-)

(36)

the dependent variable (the character of the innovative capabilities) is the continuous criterion upon which the linear regression is conducted. Since the predictor variables are not nicely distributed, logistic regression is used rather than discriminant function analysis. Logistic regression makes no assumption about the distributions of the predictor variables.

The first step of the logistic regression describes the null model (the intercept model), which is the model with no predictors. The second step includes only the control variables. The third step includes all the variables, with the independent variables as the predictors for the dependent variable.

3.2.1 Significance model

The fit of the model can be assessed with the pseudo R² values of Cox & Snell and Nagelkerke. Both the model in block 1 (only control variables) and the full model in block 2 has a low R² value, meaning that the model does not fit with the data. However, it can be seen that the addition of the independent variables causes a slight increase in the pseudo R² values. The R² values are shown in Table 2.

Table 2. R² values for statistical model.

Block -2Log Likelihood Cox & Snell R² Nagelkerke R²

1 2185.386 0.019 0.028

2 (full model) 2176.530 0.023 0.035

As in Table 2 can be seen, the -2Log Likelihood (deviance) is very high. This corresponds to the low R² values. However, the deviance of the full model is smaller than the deviance of the null model. The full model has a Chi-square value of 46.338 (df = 8), which is significant at a

(37)

5% level of significance (p = .000 < .05). The Chi-square values for the model with only control variables and the full model are both shown in Table 3.

Table 3. Chi-square test.

Block Chi-square df p-value

1 37.481 4 .000

2 (full model) 46.338 8 .000

Another way to measure the goodness-of-fit is the Hosmer and Lemeshow test. This measure shows also a bad fit between the model and the data. Though, it is clear that the addition of the independent variables causes a decrease in the Chi-square value, meaning that the full model differs less with the data than the model with the control variables has. At a 1% significance level, the differences according to the Hosmer and Lemeshow test are insignificant, meaning that the full model is a good predictor for the data.

Table 4. Hosmer and Lemeshow test.

Block Chi-square df p-value

1 48.571 8 .000

2 (full model) 17.600 8 .024

3.2.2 Intercept model

Table 5. Regression Analysis of the Intercept model.

Regression coëfficient (β)

Standard error Wald

coëfficiënt

Sigma Odds ratio

(38)

The null model does not include any predictors and is therefore not of much interest for this research. The constant of the regression line (β0 = -1.067) is significant at the 1% level

(p < 0.01).

3.2.3 Model with control variables

Table 6. Regression Analysis Model using only control variables.

Regression coëfficient (β) Standard error Wald coëfficiënt

Patent Age 0.005 0.001 20.035 .000 1.005 Reference to Science Base .119 .106 1.263 .261 1.126 Number of Patents of Focal Company .000 .000 4.349 .037 1.000 Number of Claims .010 .003 13.200 .000 1.010 Constant -1.705 .159 114.520 .000 .182

The model with only the control variables points out that the control variables have very low regression coefficients. The dummy variable indicating the reference to science base has the highest regression coefficient (β = 0.119). However, in contrast to the other regression coefficients, the regression coefficient of the dummy variable is not significant at a 5% level (p > 0.05).

(39)

3.2.4 Full model

Table 7. Regression Analysis using control variables and independent variables: Full

model. Regression coëfficient (β) Standard error Wald coëfficiënt

Patent Age 0.005 0.001 16.299 .000 1.005 Reference to Science Base .123 .106 1.354 .245 1.131 Number of Patents of Focal Company .000 .000 2.395 .122 1.000 Number of Claims .010 .003 13.244 .000 1.010 Percentage Self-Citations -.653 .313 4.359 .037 .521 Age of Technology .001 .003 .118 .731 0.521 Percentage of External Inventors .153 .123 1.530 .261 1.165 Percentage of External Attorneys -.217 .112 3.755 .053 .805 Constant -1.576 .192 67.703 .000 .207

The results of the full model, including the independent variables as the predictors of the dependent variable, shows a similar result for the control variables. The results show a high (negative) regression coefficients for the variable reflecting the percentage of self-citations (β = -0.653). This regression coefficient is significant at a 5% level (p < 0.05). Hence, hypothesis 1 is retained, which means that radical innovative capabilities are positively related with the acquisition of external technologies and incremental innovative capabilities are positively related with the internal development of technologies.

The age of the technology seems to be unrelated with the character of the firm's innovative capabilities (β = 0.001). The results from the correlation test (Table 1) show that

(40)

the age of the technology has a significant correlation with control variable 'patent age' (r = 0.127, p < 0.01), which has a very low regression coefficient (β = 0.005, p < 0.01) with the character of the innovative capabilities. This implies that the age of the technologies used to develop an innovation was higher for 'old' patents than for recent granted patents, which can be explained by the fact that the current technological development has a faster rate than a couple of years ago. In this perspective, the age of technologies used to develop innovations is time-dependent and cannot be used as a predictor for the character of the innovation. However, the regression coefficient has a very low significance (σ = 0.731), which implies that no conclusions can be made about the relationship between the age of technology and the character of innovative capabilities, based on the results of this regression test. Hence,

hypothesis 2 is not supported.

Regarding to hypothesis 3, the regression between the percentage of external inventors and the character of the firm's innovative capabilities has a very low significance (β = 0.153, σ = 0.261). Also this variable has a significant (positive) correlation with the control variable 'patent age' (r = 0.107, p < 0.01), which implies that the degree to which inventors are external has decreased over time. Another notable correlation is the correlation between the percentage of external inventors and the number of patents that the focal company possesses (r = 0.079, p < 0.01). Although the correlation coefficient is not high, this correlation implies that firms using external inventors are generating more innovations than firms using internal inventors. Both control variables ('patent age' and 'number of patents focal company') have very small regression coefficients with the character of innovative capabilities (respectively β = 0.005, σ = 0.000 ;β = 0.000, and σ = 0.116).

The regression between the percentage of external attorneys and the character of the innovative capabilities is negative and significant at a 10% level (β = -0.217, p = 0.053 < 0.10). This negative relationship is the inverse of the positive relationship. Therefore,

(41)

hypothesis 4 is not supported. This is contradicting with the negative correlation between the number of self-citations and the percentage of external attorneys (r = -0.249, p < 0.01). This negative correlation implies that the deployment of external attorneys is positively related with the acquisition of technologies. Since the acquisition of technologies is positively related with radical innovative capabilities, the deployment of external attorneys should be also positively related with radical innovative capabilities. In an attempt to explain the relationship between patent attorneys and the firm's innovative capacity, an squared term is added to the variable list in order to test if there is an curvilinear relationship.

3.3 Curvilinear logistic regression

In order to invest the relationship between the deployment of external patent attorneys and the character of the firm's innovative capabilities, a quadratic variable for the external patent attorneys is added to the variable list. From Table 8, it can be seen that the R2 value of the model remains the same when the quadratic variable is included.

Table 8. R² values for statistical model including quadratic variable.

Block -2Log Likelihood Cox & Snell R² Nagelkerke R²

1 2172.271 0.019 0.028

2 (full model) 2163.585 0.023 0.034

3 (incl. quadratic variable) 2163.151 0.023 0.034

The regression test is displayed in Table 9. As can be seen, the quadratic variable expressing the percentage of external patent attorneys shows a very high regression coefficient (β = -0.934), which means that the relationship between the deployment of external patent attorneys and the character of the innovative capabilities is inverted

(42)

curvilinear. Possibly, when the percentage of external attorneys becomes too high, the attorneys cannot link the non-local knowledge with the firm's current knowledge base and hence, they are not able to protect the radical innovation in this context.

Furthermore, the regression coefficient for the linear variable changes drastically from β = -0.225 to β = 0.720, which indicates a positive regression. This implies that the

relationship consists of a linear and a quadratic term. Though, the regression coefficients for both the linear and the quadratic term are insignificant (σ = 0.609 and σ = 0.505), so no conclusions can be made about the relationship between external attorneys and the character of the innovation based on this regression test.

(43)

Table 9. Regression Analysis using control variables and independent variables: Full

model including quadratic variable.

Regression coëfficient (β) Standard error Wald coëfficiënt

Patent Age 0.005 0.001 16.677 .000 1.005 Reference to Science Base .125 .106 1.393 .238 1.133 Number of Patents of Focal Company .000 .000 2.142 .143 1.000 Number of Claims .009 .003 12.469 .000 1.009 Percentage Self-Citations -.651 .313 4.324 .038 .522 Age of Technology .001 .003 .131 .718 1.001 Percentage of External Inventors .155 .123 1.585 .208 1.168 Percentage of External Attorneys .720 1.408 .261 .609 2.054 Percentage of External Inventors (Quadratic) -.934 1.400 .445 .505 .393 Constant -1.592 .193 67.912 .000 .204

Radical versus incremental innovations in the U.S. telecom industry : a study drawing on patent data.