Which types of inventions are likely to have the greatest impact in a dynamic environment? : a study of research in motion using patent citations

(1)

Master Thesis

Which types of inventions are likely to have the greatest impact in a dynamic

environment?

A study of Research In Motion using patent citations

Author: Jasper Wolfkamp Student Number: 6067611

Study: Master Business Studies – Strategy Track Supervisor: Dr. R.M. Singh

(2)

2

Abstract

Creating inventions has become very important for firms in order to gain and sustain a competitive advantage. This research has focused on the different types of inventions a firm can create and has examined which types of inventions are likely to have the greatest impact. The findings of this study suggest that firms should use new and extraindustry

knowledge and different technologies to create high impact inventions. These kinds of knowledge and technologies should be gathered by external search. This research has found that inventions that are based on new, extraindustry knowledge gathered by external search and from international sources, and that use different technologies that have not been used before by the company, have the greatest impact in a dynamic environment. This study used patent data of Research In Motion (RIM). RIM was once one of the leading companies in its industry and now has trouble coming up with the right type of inventions. It is therefore interesting to look at the patent data of RIM to determine if they created the right type of inventions. Because patent data provides a lot of information about inventions, it is very useful for this research.

(4)

4

Introduction

The essential point in strategy is to figure out why particular firms have survived and whether their survival can be sustained (Geus, 2002). As research has shown, the life expectancy of a multinational firm from the Fortune 500 is on average between 40 and 50 years, and there are just a few firms that have survived for a longer period of time (Geus, 2002). But why do some firms fail while others do not?

‘‘It is not the strongest of the species that survive, nor the most intelligent, but the one that is most responsive to change.’’

Charles Darwin

As Charles Darwin once said, survival is about the ability to response to change. Because the business environment has evolved into a very dynamic one over the last decades, the way firms gain and sustain a competitive advantage has changed greatly. Recent research has shown that to establish and sustain a competitive advantage today, firms must try to generate innovations in an effective way (Fabrizio, 2009). Firms can generate innovations in different ways. They can either use their exploitative capacity to generate incremental innovations, or they can use their explorative capacity to generate radical innovations (Atuahene-Gima, 2005; March, 1991). A firm must create the right type of inventions to increase its likelihood of survival.

Something that is closely related to innovations is the concept of inventions. The difference between an invention and an innovation is that an invention refers more directly to the creation and formulation of an idea, whereas an innovation is more about the

(5)

5

are also different ways to generate inventions. The types of inventions usually differ in the type of knowledge that is used. Firms can develop their inventions based on existing or new knowledge, internal or external knowledge, and old or recent knowledge (Katila & Ahuja, 2002; Katila, 2002). The type of knowledge used for inventions is of major importance for the value of that invention. Firms need to focus on the right type of inventions, as it becomes more and more difficult to keep up with competition in a dynamic environment. Therefore, this research will look at the different types of inventions and will try to find out which type has the greatest impact in a dynamic environment.

To find an answer to the research question, this study will use patent data from Research In Motion. Patent data is seen as a very rich source of data to study inventions, as it contains information about the patented invention as well as the technological area of that invention (Jaffe & Trajtenberg, 2002b). It also gives an idea of the type of knowledge used within the original patent. Because this research focuses on the type of knowledge used in inventions, patent data is very suitable.

Research In Motion is used as a case in this study because it is a good example of a company that has been very innovative, but now is not able to keep up with competitors. Research In Motion (RIM, and better known as Blackberry) once virtually created the smartphone market, with its pioneering Blackberry phones with wireless email. However, the company has faced many problems during the years since Apple released the Iphone in 2007. Even though the company has been innovative during these years, and came up with some inventions, it was not able to come up with high-impact or breakthrough inventions. Therefore, it is interesting to look at the patent history of RIM to see which types of inventions they used and what the impact of these inventions were. The results of this

(6)

6

research could be very useful for managers and companies, because knowing what types of inventions have the most impact can help them to gain and sustain a competitive advantage.

This research will start with a review of the existing literature in section 1. In section 2 the hypotheses will be stated and in section 3 the methodology will be explained. The findings of the analyses will then be presented in section 4 and further discussed in section 5. After some implications and suggestions for further research are given, this research will end with a short conclusion.

(7)

7

Section 1 - Literature Review

1.1 Innovation

For a long time economists saw firm differences as a cause of general economic differences. From a neoclassical perspective, they were more focused on the general

economic performance of an industry or nation rather than on the firm itself (Nelson, 1991). In contrast, Schumpeter put forth a strong general challenge to the effect that innovation ought to be the center of economic analysis. From this evolutionary point of view,

management researchers see firm differences much more as exploring new, potentially better, ways of doing things (Nelson, 1991). They argue that organizational differences, especially differences in the ability to generate and gain from innovation, are the source of durable and not easily imitable differences among firms. With the emergence of the

knowledge economy, intense global competition, and considerable technological advances, innovation has become even more central to competitiveness (Lawson & Samson, 2001). Especially in a very dynamic environment such as industries are facing today, and in an era of unrelenting competition, innovation has become a priority for corporations, institutions, and nations (Buderi, 1999). To emphasize the importance of innovation, Fabrizio (2009) says that establishing and sustaining a competitive advantage depends upon effectively developing internal knowledge, utilizing external knowledge, and exploiting knowledge to generate innovations. The type of knowledge that is used often leads to a specific type of innovation.

Lawson & Samson (2001) explain innovation as the mechanism by which organizations produce new products, processes, and systems required for adapting to changing markets, technologies, and modes of competition. Innovation can be divided into roughly three categories. The first type is radical innovation, which consists of discontinuous

(8)

8

events that are the result of deliberate research and development activities (Propris, 2002). Radical innovations occur through major changes that make great improvements. These improvements require competencies or skills that are different from those the incumbent has (O’Reilly & Tuschman, 2008), and are based on a new technology, or combination of technologies, induced commercially to meet a user or market need (Propris, 2002). Radical innovation is often based on new and distant knowledge where search behaviors involve a conscious effort to move away from current organizational routines and knowledge bases (Katila & Ahuja, 2002).

The second type of innovation is incremental innovation. In contrast to radical innovations, incremental innovations are innovations in which an existing product or service is made better, faster or cheaper (O’Reilly III & Tushman, 2008). Although these

improvements may be difficult or expensive, they draw on an existing set of competencies and proceed along a known trajectory. Incremental innovations often take the form of smaller enhancements around major radical innovations, such as the form of design improvements, learning by doing and learning by using (Propris, 2002). This type of

innovation relies more on local and existing knowledge, where organizations search locally and use knowledge that is closely related to preexisting knowledge bases (Katila & Ahuja, 2002). Finally, innovation also occurs through seemingly minor improvements in which existing technologies or components are integrated to dramatically enhance the

performance of existing products or services. This type of innovation is called architectural innovation.

The main differences between these types of innovation are the skills and

competencies required and the type of knowledge used for a particular innovation. In this paper the focus will be mainly on the type of knowledge used for innovations, linking it to

(9)

9

the difference between radical and incremental innovations and the explorative and

exploitative capacity of firms. Therefore, the next section of this paper states the importance of radical and incremental innovation in the process of exploration and exploitation.

Following that, the different types of knowledge are discussed.

1.2 Innovation with Exploration and Exploitation

Incremental innovations are the result of improvements to existing products based on procedures along a known trajectory, while radical innovations differ more from the existing capabilities of a company and rely on major changes. The features of these types of innovations can be linked to explorative innovation and exploitative innovation. There is not much empirical evidence about the relationship between exploration and radical innovation, and between exploitation and incremental innovation, but a study of Atuahene-Gima (2005) claims that exploration is positively related to radical innovation and exploitation is

positively related to incremental innovation. Their findings suggest that an exploratory invention requires more information processing and exposure to a variety of technological knowledge domains. This technological knowledge base facilitates complex problem solving, the generation of new ideas, and novel combinations that enhance the development of exploratory innovative competences. In contrast, with incremental innovation complexity and knowledge scope are less important, because this type of innovation requires minor changes to existing knowledge resources in the organization (Atuahene-Gima, 2005).

Another way to look at the relationships between radical innovation and exploration, and incremental innovation and exploitation, is to focus on the knowledge used and the notion of search. Units that engage in exploratory innovation pursue new and distant knowledge (Jansen, Van Den Bosch, & Volberda, 2006; Katila & Ahuja, 2002). As stated

(10)

10

earlier, radical innovation involves the use of new knowledge and distant search, which can both be linked to the use of new knowledge in the explorative capacity of a firm. In contrast, units pursuing exploitative innovation build on existing knowledge and local search (Katila & Ahuja, 2002; Jansen et al., 2006), which can be linked to incremental innovation as

mentioned in the previous section

In addition, some relate exploration and exploitation directly to innovative outcomes, such as products or services, without looking at the use of knowledge and the notion of search. In such cases, exploration is used as a synonym for radical innovation, and

exploitation as a synonym for incremental innovation (Li, Vanhaverbeke, & Schoenmakers, 2008).

1.3 Exploration and Exploitation

Since research of March (1991) exploration and exploitation have dominated the organizational analysis of technological innovation and have became increasingly important in recent years. March (1991) states that the central concern of studies of adaptive

processes is the relation between the exploration of new possibilities and the exploitation of old certainties. He says that exploration includes things captured by terms such as search, variation, risk taking, experimentation, play, flexibility, discovery, and innovation.

Exploitation, on the other hand, includes things captured by terms such as refinement, efficiency, selection and execution. According to Gupta, Smith, & Shalley (2006) exploitative innovations involve improvements to existing components and build on the existing

technological trajectory, whereas exploratory innovation involves a shift to a different technological trajectory. Along with this, He & Wong (2004) state that exploitative

(11)

product-11

market-domains, and exploratory innovation involves technological innovation aimed at entering new product-market-domains. In addition, O’Reilly III & Tushman (2008) argue that exploitation occurs when firms rely on existing competencies or capabilities to sell to existing customers, and that exploration is about new growth opportunities to adjust to volatile markets.

The large number of studies on exploration and exploitation, spread among different research disciplines, have used different levels of analyses and perspectives (Li et al., 2008). There are studies that focus on the individual level, project level, business level, corporate group level, alliance level and industry level. However, even though there are differences in the interpretation of these constructs, all authors agree on the type of knowledge used in exploration and exploitation. There is consensus on the fact that exploration is the search for new knowledge, technology, competences, markets, or relations, and that exploitation is the further development of existing ones (Li et al., 2008).

1.4 Innovation in a Dynamic Environment

As competition intensifies and the pace of change accelerates, firms need to renew themselves by exploiting existing competencies and exploring new ones (Jansen et al., 2006). Various literatures have argued that organizations need to be ambidextrous and develop exploratory and exploitative innovations simultaneously in different organizational units (Benner & Tushman, 2002; He & Wong, 2004). However, there is also a problem with balancing exploration and exploitation. It is clear that exploration of new alternatives reduces the speed with which existing skills are improved. At the same time, organizational exploitation of current capabilities also reduces exploration of new capabilities (March, 1991). This can result in a short-term bias in organizational adaptation, which is crucial in a

(12)

12

dynamic and fast-changing environment such as industries are facing today.

Dynamic environments are characterized by changes in technologies, variations in customer preferences, and fluctuations in product demand or the supply of material. This type of a fast-changing environment makes current products and services obsolete and requires that new ones be developed (Jansen et al., 2006). To develop new products and services, explorative innovations are needed, and therefore the focus should be more on exploration and radical innovations than on exploitation and incremental ones. Explorative innovations require new types of knowledge that can only acquired by distant and external search. This is especially the case in a dynamic environment. However, some researchers argue that focusing on the one reduces the outcome of the other. They argue that focusing on exploration, including experimenting with new alternatives, reduces the speed at which existing competencies are improved and refined (March, 1991). A failed explorative effort may disrupt successful routines in a firm’s existing domains, without any significant success in the new field to compensate for the loss in existing business (He & Wong, 2004).

Although some researchers argue that there needs to be a good balance between exploration and exploitation (O’Reilly III & Tushman, 2008), the research of Jansen et al. (2006) supports the previous statement that firms should focus more on exploration and radical innovations in a fast changing industry. Their results suggest that organizational units operating in more dynamic environments increase their financial performance by pursuing exploratory and radical innovations. They say that these companies could resist the threat of obsolescence of their competences not only by developing new products and services but also by entering new markets and finding new customers. According to their research, exploitation of existing products, services, and markets appear to have a negative effect on financial performance, and these organizations will fall behind as they try to improve existing

(13)

13

products and services for markets that are in decline (Jansen et al., 2006). A review of the literature leads to the conclusion that the ability to continuously innovate has become one of the most important assets and characteristics of a firm today Previous research has shown that in a dynamic environment, the need for new growth opportunities is high, and the focus on explorative and radical innovations is therefore more important than a focus on

exploitative and incremental ones.

1.5 Research Gap

From the literature, it is clear that innovation has become one of the most important sources of differences between firms, and that the ability to continuously innovate has become one of the most important assets and characteristics of a firm. Innovation is needed to come up with new ideas and solutions and to gain and sustain a competitive advantage (Buderi, 1999; Lawson & Samson, 2001; Malerba & Orsenigo, 1996). If a firm is not able to innovate continuously, there is a low chance of survival for that firm. Also, a great deal of research has focused on the different types of innovations. Firms can either be making incremental innovations by focusing on exploiting knowledge, or making radical innovations by focusing on exploring new knowledge. These different types of innovations are related to the explorative and exploitative capacity of a firm, which is a result of the adaptive process (March, 1991). They argue that incremental and exploitative innovations can be made in a static environment, but thatt the focus must be on radical and explorative innovations in a fast-changing environment (Buderi, 1999). Here, the need for new growth opportunities is high, so the focus on explorative innovation is more important than on exploitative

innovation.

(14)

14

somewhat confusing, and is used less often in daily businesses. Because the main focus of this research will be on the types of inventions, which can certainly be linked to the different types of innovations as explained above, it is important to know the difference between the concepts. The concept of an invention is about new means, achieving some function not obvious beforehand to someone skilled in prior art (Kline & Rosenberg, 1986). When an invention is not derived from an existing model or idea, or it has a completely new function, it may be defined as a radical breakthrough. An invention differs from an innovation in that inventions refer more directly to the creation of the idea or method itself whereas

innovations refer more to the use of that new idea or method (Hao, 2012). An explanation that makes the difference more clear is the commercial application or adoption of each concept. Here, inventions are about how the new ideas or products are formulated, while innovations are more about the application of these inventions into products and practices for the market. Because this research will focus more on the creation of new ideas or

products from the very start, and not on their application into the market, inventions will be used from now on.

The literature has stated the general importance of innovations, which are the result of the commercialization of inventions. In every context and environment, there is a specific need for a certain type of invention. The types of inventions (which can be linked to the different types of innovations as explained above) and especially the knowledge used for these inventions are of particular importance for the value of a firm. There is less evidence about the relation between the type of invention and its impact. Therefore, this research will explore the impact of the different types of inventions using Research In Motion as a case study. It will focus on the type of knowledge used in their inventions and how this

(15)

15

question:

"Which types of inventions are likely to have the greatest impact in a dynamic environment?"

(16)

16

Section 2 - Hypotheses Section

2.1 The Use of Knowledge for Inventions

As can be concluded from the previous sections, the main difference between radical and incremental inventions is the type of knowledge used. In general, knowledge is the key economic asset that drives long-term economic performance, especially in a knowledge economy such as companies are in today. The different types of knowledge, either existing or new, affect the type of inventions of a company (Katila & Ahuja, 2002; Rosenkopf & Nerkar, 2001) and are related to the explorative and exploitative capacity of a firm. Units pursuing exploitative inventions build on existing knowledge, whereas units that engage in exploratory inventions pursue new knowledge1_{(Jansen et al., 2006; Katila & Ahuja, 2002).}

However, whether a company uses existing or new knowledge for their inventions depends on the type of search. Katila & Ahuja (2002) and Katila (2002) state that there are two dimensions of search; search depth and search scope. Search depth is the degree to which existing knowledge is reused or exploited (Katila & Ahuja, 2002). Some firms may use existing knowledge repeatedly, while others may use it only once. This difference in search depth leads to differences in the ability of a firm to come up with new solutions. There are some positive effects of an increase in the depth of search, such as reducing the likelihood of errors, making search more predictable and increasing the understanding of the knowledge. However, literature identifies at least two negative effect of the increase of search depth, which at some point exceed the benefits of it (Katila & Ahuja, 2002). The first negative effect of depth is a decrease in returns when building on the same knowledge, because there is a limit in the intrinsic performance of knowledge. Second, depth can make an organization

1

Note: New knowledge refers to the use of external knowledge gathered by new sources instead of existing ones. It does not refer to any time-considered aspects.

(17)

17

rigid by reusing existing knowledge, which can lead eventually to a decrease in product output.

Besides search depth, the search efforts of a firm can also be characterized in terms of search scope. Search scope is about how companies search to gather knowledge, which can be done in two different ways. Either a company can participate in local and internal search or it can participate in distant and external search. With local and internal search the firm searches for its own previously created knowledge that is closely related to their preexisting knowledge base. Here, the firm focuses on knowledge and technologies that are used before. When a company uses distant and external knowledge it moves away from current organizational routines and knowledge base and seeks extraindustry knowledge (Katila & Ahuja, 2002; Katila, 2002). Search efforts with high scope affect product inventions positively. Companies with a high level of search scope will have a more enriched knowledge pool with more distinctive variations of knowledge. Also, a high level of search scope enables firms to improve possibilities for finding new combinations of useful information by adding new elements to the set (Katila & Ahuja, 2002).

By focusing on local search, a firm uses internal and existing knowledge to create incremental inventions. With the use of internal and existing knowledge, firms become more expert in their current domain. Rosenkopf & Nerkar (2001) call this a "first-order

competence", which could be a distinctive competence if it is superior to competition. However, prior research suggests that greater levels of reliance on existing knowledge and on own prior developments (or "first-order competence") is associated with inventions that are less relevant and therefore a hallmark of obsolescence (Sørensen & Stuart, 2000). As first-order competence is related to less relevant inventions, the focus of the company should be on "second-order competence": the ability of a firm to create new knowledge

(18)

18

through recombination of knowledge across boundaries (Rosenkopf & Nerkar, 2001). The theories described above suggest that companies should decrease their search depth, increase their search scope, and focus on their second-order competences to gather new knowledge for their inventions. Therefore, based on the existing literature, we predict that companies that use new knowledge for their inventions, using distant and external search, will create inventions that have a higher impact. This prediction will be tested using the following hypothesis:

Hypothesis 1: The greater the use of new knowledge gathered by external search, the greater the impact of a firm’s invention.

2.2 The Use of Technology

Companies need to use new knowledge within their inventions, but they also need to search for different and new technologies to find new solutions. The use of new

technologies depends strongly on the search for new knowledge, as the technologies used by an organization are based on the type of knowledge they search for. As mentioned before, companies are more likely to use local and internal search, and more often use knowledge that already exists (Katila & Ahuja, 2002; Katila, 2002). The main objective of a company is to create reliable and profitable inventions, which are more easily created by the exploitation of existing knowledge (Sørensen & Stuart, 2000). Following these theories and the logic of evolutionary theory, it is more likely that a firm will also focus their inventions more around a specific technological position in which it uses previously accumulated knowledge and experience.

(19)

19

However, as stated in the literature, in a dynamic environment, focusing on exploiting existing knowledge will not result in high performance (Jansen et al., 2006). In such a dynamic environment as we have today, companies need to focus on exploration and radical inventions. By using similar technologies for new solutions, no breakthrough

inventions can be made and the chance of survival will be low. Therefore, looking at the technological field of an invention, we predict that companies with a background that is technologically close or similar to the technological field they are already in, will have inventions with less impact. This results in the prediction that technological distance will influence the impact of an invention:

Hypothesis 2: The greater the distance between the technology used for an invention and the current technological domain of the firm, the greater the impact of a firm’s invention.

2.3 The Age of Knowledge

As the previous hypotheses predict that the use of new external knowledge and new external technologies will influence the impact of a firm's invention, there is also another aspect of knowledge that needs to be mentioned. The age of knowledge is an important aspect in the creation of inventions as well, and it can affect the innovative output of a firm (Katila, 2002). However, in the existing literature there is not always a broad consensus on this point.

Some researchers argue that firms should focus on older and established knowledge for their inventions for several reasons. First, because inventions are uncertain, knowledge that has existed for a longer time is usually considered more legitimate, reliable, elegant,

(20)

20

and robust (Katila, 2002). Because established knowledge is more reliable, a firm might be able to learn from previous knowledge. Second, older knowledge is seen as more beneficial than recent knowledge because there is a decreased threat of retaliation. Using recent knowledge will more often be seen as an attack against the creators of that knowledge, as these ideas are seen as more valuable. A retaliatory response to building on older knowledge will be much less likely (Katila, 2002). A third argument that supports the use of old

knowledge is that older knowledge is often more difficult to access and build on, so it creates a sense of uniqueness. Research states that all these qualities together will increase the chances of successful inventions.

In contrast, the literature also states that firms should build on the most recent technological foundations to enhance inventions. They say that if knowledge ages, it

becomes obsolete and will no longer match the demands of the current environment (Katila & Ahuja, 2002). This is supported by the view of Sørensen & Stuart (2000), as they state that aging of knowledge leads to rigidity and therefore lower innovative output by the firm. Instead, using recent knowledge could help a firm to adapt to the fast-changing environment and to maintain organization-environment fit (Katila, 2002). It is stated that recent

knowledge can enhance a firm’s ability to expand into new technological areas, and that with recent knowledge firms are better able to predict the nature of future technological advances. Knowledge that is currently available is in the best interest of the firm because it is the best representative of the emerged alternatives (Nerkar, 2003). Another advantage of the use of recent knowledge is that there are reduced search costs, because recently created knowledge is easier to find and is therefore more likely to lead to rewards (Katila & Ahuja, 2002).

(21)

21

knowledge. But is it necessary to choose one or the other? According to Nerkar (2003), the process of knowledge creation can be seen as a path-dependent evolutionary process that involves recombining knowledge spread over time. Many modern technologies are in fact already combinations of ideas discovered during different time period. Therefore, he argues that both current and historical knowledge matter for the future creation of recent

knowledge.

In addition, Katila (2002) found that the age of knowledge has different effects on inventions, based on where that knowledge comes from. As explained in the first part of this section, there is a difference between internal knowledge: knowledge that comes from within the industry (intraindustry knowledge); and external knowledge: knowledge that comes from outside the industry (extraindustry knowledge) or from science based sources, such as universities (Katila, 2002). Even though the most innovative patents are at the forefront of knowledge, coming from a close science link (Buderi, 1999), there could be some problems with the use of the newest available external knowledge. Because recent knowledge coming from science-based resources is often very expensive to access, difficult to interpret ,and not easy to use, it is better to use old external or extraindustry knowledge. However, in contrast, using old internal or intraindustry knowledge can create a number of problems. First of all, when interpretations of events or knowledge are called forth long after their original creation, the original knowledge can become obscured (Katila, 2002). Second, when a firm uses old internal knowledge it is less likely that it will introduce new products or create inventions. Therefore, when using internal or intraindustry knowledge, the

knowledge should preferably be recent. Here it is important to note the assumption about internal knowledge that old internal knowledge is often already used by the firm itself or by others.

(22)

22

As can be concluded from the literature, the age of knowledge has several effects on innovative output. Some researchers argue for the use of recent knowledge, and others for the use of old knowledge or a combination of both. However, Katila (2002) states that old internal/intraindustry knowledge hurts inventions, whereas old external/extraindustry knowledge promotes the impact of inventions. Therefore, based on the literature and previous research, we predict that the age of knowledge influences the impact of a firm’s inventions, based on the idea of search:

Hypothesis 3: Old external knowledge will have a positive effect on the impact of inventions, whereas old internal knowledge will have a negative effect on the impact of inventions.

(23)

23

Section 3 - Methodology

The central purpose of this paper is to determine which inventions are likely to have the greatest impact in a dynamic environment. To find an answer to this question, the study uses Research In Motion (RIM, and better known as Blackberry), a company that was once a leading innovator in its industry and now has trouble surviving in the fast-changing

environment of the smartphone industry. The study focuses on the types of inventions produced by RIM and looks especially at the types of knowledge used in their inventions and how they searched for these kinds of knowledge (Katila & Ahuja, 2002). Before explaining the dataset and sample of this study, it is important to have some background information on Research In Motion. Therefore, first a short overview and history of RIM is given.

3.1 Research In Motion

Research In Motion (RIM) is a multinational telecommunications company. RIM virtually created the smartphone market, and is a company that has always been very innovative. However, even though the company did produce some inventions in recent years, the company is no longer the leader in its industry, and has been outperformed by its

competitors. The company is facing many problems and has burned through almost all of its cash (Gillette, Brady, & Winter, 2013).

Blackberry Limited, formally known as Research In Motion Limited, was founded in 1984 as an electronics and computer science consulting company by Mike Lazaridis and Douglas Fregin. At that time, Lazaridis was an engineering student at the University of Waterloo and Fregin was an engineering student at the University of Windsor. In its first years, the company worked with RAM mobile data and developed products for Mobitex

(24)

24

wireless packet-switched data communications networks, which were often used by military and police forces (“BlackBerry timeline,” n.d.). In 1992, Jim Balsillie, who finished an MBA at the Harvard Business School in 1989, joined RIM as co-CEO and co-former along with

Lazaridis. Together they focused on a breakthrough technology, which was an easy, secure, and effective device that allowed workers to send and receive e-mails while away from the office (Gillette et al., 2013). RIM released the "Interactive Pager", also known as the RIM 900, in 1996, which was the first keyboard-based device of this kind. In 1997, RIM went public on the Toronto Stock Exchange, and introduced the first Blackberry 850 pager in 1999. The first Blackberry was a dramatically slimmed down two-way pager, which could receive and send emails. The introduction of the Blackberry set the stage for future products from the company, such as the Blackberry 957, which was the first Blackberry smartphone. The Blackberry became the indispensable accessory of business executives, heads of state, and Hollywood celebrities and was the pioneer in wireless email. RIM soon began to introduce the devices to the consumer market as well, with the introduction of the Blackberry Pearl, Curve, and Bold. Even though these smartphones were great successes, the introduction of the Iphone in 2007 caused many problems for the company (“Research, no motion,” 2014).

The first models of the Iphone generally lagged behind the Blackberry in sales, as RIM had major advantages in carrier and enterprise support (“BlackBerry timeline,” n.d.).

However, less than a year later Apple had already surpassed RIM in quarterly sales, and continued this growth while RIM began to lose users. This slowing growth prompted the company to undertake a lay-off of 2,000 employees in the summer of 2011. In 2012, Lazaridis and Balsillie resigned as the CEOs of the company, and handed the reins over to executive Thorsten Heins (“BlackBerry timeline,” n.d.). In that same year, the company reported its first net loss in years and announced plans to lay off 5,000 additional

(25)

25

employees. During these years, RIM had lost almost 82% of its market value, as the

company’s market capitalization dropped from $83 billion in 2008 to less than $14 billion in 2011. This is the biggest decline ever seen among communications-equipment providers. The stock price dropped from almost $80 a share in 2009 to less than $8 a share in 2014 (“BlackBerry Ltd.,” n.d.).

3.2 Dataset

To examine which type of knowledge RIM used for their inventions and how they searched for this knowledge, this study uses U.S. granted patent data of RIM from the Delphion Patent Database. Patent data has long been recognized as a very rich and important source of data with which to study inventions and technical changes in the knowledge economy. A patent is a "temporary monopoly awarded to inventors for the commercial use of a newly invented device" (Trajtenberg, Henderson, & Jaffe, 1997). Patent data contains information about the patented invention, including the identity and location of the inventor or his or her employer, the technological area of the invention, and more (Trajtenberg et al., 1997). It is seen as one of the richest sources of data on inventions and provides a great deal of useful information about the type of invention as well. However, the general patent information is not the most useful. It is the patent citations that are the most useful source of information, as they open the possibility of tracing multiple links across inventions, inventors, firms, locations and industries (Jaffe & Trajtenberg, 2002a). Patent citations can provide a notion of "prior art" or prior knowledge that is used within the original patent. This is exactly what is needed for the research of this study. By studying the patent citations of RIM, it is possible to understand what type of knowledge and technology

(26)

26

was used for their inventions and how they searched for these.

However, there are two sets of measures within patent citations. Patent citations consist of forward citations and backward citations, which are each an indicator of different things. In short, forward citations are "citations that are received from other patent

applications and are derived from the relationship between a patent and subsequent

technological developments that built upon it" (Trajtenberg et al., 1997). Backward citations are "citations to previous patent documents and are derived from the relationship between a given patent and the body of knowledge that preceded it" (Trajtenberg et al., 1997). The differences between and usefulness of backward and forward citations will be further discussed at the end of this section with the explanation of the variables.

The data set of this research consists of all patents granted to Research In Motion in the years 2005, 2006, 2007 and 2008. These years were chosen because the mobile telecom industry was changing during that period of time, with the introduction of the Iphone occurring in 2007. During these years RIM was granted 545 patents for the U.S. market. The dataset consists of all these patents, with information on all the backward citations and forward citations. The 545 patents of RIM contain a total of 14,098 backward citations. Also, the number of forward citations of RIM are counted for each year. In total, RIM received 5,999 forward citations during these years, with an average of 11 forward citations per patent.

(27)

27

Table 1 gives an overview of the patents of RIM in the particular time period. All the patents and the information from these patents were downloaded manually from the

Delphion patent database, which contains U.S. Patent and Trademark Office (USPTO) patents.

3.3 Variables

All the variables used in this research are based on the patent data of Research In Motion. Patent data gives a great deal of information about the impact and type of inventions, and therefore is very useful for this research.

3.4 Dependent Variable

Forward citations

The aim of this study is to determine which type of invention is likely to have the greatest impact in a dynamic environment. As explained, this will be measured using the patent data of Research in Motion and especially its patent citations. Over the years, a great deal of research has been done using patent data. Patent data provides unique access to information about the value of inventions (“OECD Patent Statistics Manual,” 2009).

However, it is difficult to find good indicators with which to measure the rate of invention or the value of innovations, which is often a common problem (Trajtenberg et al., 1997). To overcome the limitations of simple patent counts or counts of expert identified inventions (Trajtenberg et al., 1997), this research will use forward citations to measure the impact of an invention. The information provided by forward citations can be exploited to compile indicators of the relative value of patents.

(28)

28

Forward citations are citations received by other patents that indicate that the information in an invention has served as a basis or is used for a future invention (Hall, Thoma, & Torrisi, 2007). Forward citations give a notion of the relationship between a patent and subsequent technological developments that are built upon it, which are called its

descendants. The prior art of the invention cited in patent documents provides useful information about the diffusion of technologies and knowledge used (“OECD Patent Statistics Manual,” 2009). For example, if patent B cites patent A, it implies that patent A represents a piece of previously existing knowledge upon which patent B builds, and over which B cannot have a claim (Jaffe & Trajtenberg, 2002a). Patents with a large number of forward citations have the potential to provide a base for numerous subsequent technical changes and have more impact.

This research will use forward citations to measure the impact of an invention. However, simply counting the forward citations of patents has some limitations that need to be overcome. As stated by Hall et al. (2007) the main difficulty in computing forward

citations is that they appear over time, and sometimes long after the cited patent was filed or granted. The consequence of this is that earlier patents have a lower probability of being cited, as they are "among the market" for a shorter period of time. It is stated that to make sure there is no time bias within the forward citation measure, the "fixed-effects approach" is a useful method to overcome the time bias with forward citations counts.

However, as the time period for the analysis is just four years, the forward citations measure will not be biased in such a significant way that a correction is needed. Therefore, the fixed-effects approach is not used in this study, which could be a limitation, but seems most suitable in this case. The dependent variable will therefore be just a measure of

(29)

29

to note is that the dependent variable is a count variable, so the data will be count data as well. This fact will be important for choosing the right analysis later in this study.

3.5 Independent Variables

The independent variables are all based on the backward citations of the patents of RIM. Backward citations are citations to previous patents. As forward citations say

something about the descendants of a patent and about the diffusion of technologies and knowledge used, backward citations provide information about the antecedents of a patent and give a notion of the relationship between a given patent and the body of knowledge that preceded it (Hall et al., 2007; Trajtenberg et al., 1997). Backward citations can help to track knowledge spillovers in technologies to get a sense of the curve of obsolescence. So, if patent 2 cites patent 1, it implies that patent 1 represents a piece of previously existing knowledge or technology upon which patent 2 builds, and over which patent 2 cannot have a claim. Backward citations are very useful as indicators of the degree of novelty and

originality of an invention and of its knowledge transfer patterns (“OECD Patent Statistics Manual,” 2009; Trajtenberg et al., 1997). As the hypotheses of this research are all based on knowledge transfers and the novelty of knowledge and technology used for inventions, backward citations are good indicators and measures for the independent variables in this research.

Search Scope

The first independent variable of this research is search scope. The theories suggest that companies should search for new knowledge by the use of external and distant search. They have to increase their search scope, which is the notion of exploration of new

(30)

30

knowledge (Katila & Ahuja, 2002). To measure the search scope of a firm, Katila & Ahuja (2002) use the proportion of new citations in a firm’s patent. The proportion of new citations is calculated by looking at the number of previously unused citations in a firm’s list of

citations. This research uses search scope as an independent variable as well, but it is calculated with some different items.

In this research, the independent variable search scope is computed using the

number of self citations by RIM. Self citations are citations made by the same assignee as the one owning the cited patent and may reflect some degree of appropriation of potential spillovers (Jaffe & Trajtenberg, 2002a). So, when RIM owns a patent in which a citation is made to another patent of RIM, it is seen as a self-citation. Self-citations provide a notion of knowledge spillovers, where it is presumed that self-citations represent internal transfers of knowledge. Self-citations are seen as one of the purest forms of internal search, as the firm uses its own existing knowledge for their inventions (Hall, Jaffe, & Trajtenberg, 2001). The first independent variable, measuring the notion of external search, is computed as follows2:

Search scope

The result of search scope will be measured on a scale between 0 and 1. Here a result of 1 means a high search scope, so a high notion of external search. A result of 0 means a low search scope, so a high notion of internal search.

2

As an additional analysis, we also computed this variable by adding the number of citations to close competitors to the number of self-citations. This measures the number of citations that are made outside the industry. To determine the competitors in the industry, this research focused on the top 3 companies with the biggest market share. The most important competitors of RIM in the time period 2005 to 2008 were Nokia Corporations (Finland), Motorola Inc. (US) and Samsung Electronics (South-Korea). The total of citations to close competitors and self-citations is divided by the total number of backward citations. Note that this variable is just used as an additional analysis.

(31)

31

Geographical Search Scope

The previous independent variable measures the notion of external search by looking at the knowledge used within or outside the telecommunication industry. Another way to determine the notion of external search, rather than focusing on the industry, is by looking at the geographical origin of knowledge that is used in inventions. As the development of technological activity indicates that national invention systems have become less distinctive over time, firms have a strong motivation to tap into foreign invention systems to get a wider set of solutions to technological problems (Phene, Fladmoe-Lindquist, & Marsh, 2006). It is expected that knowledge from international sources is useful for inventions due to the differences in the type and content of knowledge available.

To gather new and external knowledge by focusing on the geographical origin of that knowledge, firms need to invest overseas to increase their knowledge diversity. Besides knowledge that is distant from the industry, geographically distant knowledge also provides organizations with the opportunity to make novel inventions (Jaffe & Trajtenberg, 2002a; Phene et al., 2006). However, it is more likely for firms to cite companies that lie

geographically close. As RIM is a Canadian company, it is more likely that it would cite companies that are either in the same country or nearby. Therefore, this research uses Canada and the United States as low geographical scope countries.

To compute the geographical distance of knowledge, the independent variable geographical search scope is computed. The extent of geographical search scope is

measured by calculating the number of citations to companies in Canada and the US, divided by the total number of backward citations of the patent. However, by computing the

variable as just explained, self-citations are not excluded from this variable. Self-citations are highly geographically localized and interact in important ways with geographical localization

(32)

32

(Jaffe, Trajtenberg, & Henderson, 1993). Including self-citations in the variable will create strong localization biases, so this variable will be corrected by the number of self-citations to reduce the geographical localization bias. Therefore, the second independent variable, measuring the notion of external search by looking at the geographical origin of knowledge, is computed as follows:

Geographical Search Scope

Normally, an outcome close to 1 means a high geographical search scope and therefore a high notion of external search, and an outcome close to 0 means a low geographical search scope and therefore a low notion of external search. However, this research hypothesizes that if a company focuses too much on internal and existing knowledge, the inventions of that company will have a lower impact. Therefore, we will use the opposite measure with the independent variables in this research. Here an outcome close to 1 means a low

geographical search scope and an outcome close to 0 means high geographical search scope.

Technological Distance

Hypothesis 3 states that companies need to search for different and new

technologies to find new solutions instead of using existing and old technologies. Knowledge from beyond the firm’s technological domain leads to inventions that have significantly higher impact. Also, companies that explore various types of technologies are more likely to create breakthrough inventions (Ahuja & Lampert, 2001; Rosenkopf & Nerkar, 2001).

However, according to the logic of evolutionary theory, learning, as well as innovation, is seen as a cumulative activity. This suggests that companies are likely to focus more on

(33)

33

existing technologies than on distant and new ones. In other words, it is expected that Research In Motion has applied for patents in industries with a low technological distance and quite similar to the technology they have already used.

In order to compute the technological distance between a firm’s previous patents and the industry it is already in, the patent classification system provided by the USPTO is used. Phene et al. (2006) have used this classification system before to measure

technological distance. The USPTO classification system consists of approximately 400 patent classes and over 120,000 patent subclasses that identify the technological class a patent belongs to (Hall et al., 2001). The main classification of a patent consists of a 3-digit patent class which best represents the technological field of the invention (Hall et al., 2001). The patents are further divided into technological sub-classes, but this research will focus only on the main national classes of the USPTO classification. Because this research focuses on the company Research In Motion, which is active in the telecommunication and mobile telecom business, the main technological classes represented in Table 2 are used.

Table 2 - Classification of USPTO classes of Research In Motion

USPTO Classification – Name Patent Class – 3 digit

Telecommunications 455

Telephonic Communications 379

Multiplex Communications 370

To measure the technological distance of the inventions of RIM, the independent variable is computed as the number of cited patents that were assigned to one of the technology classes similar to the one of RIM (as in Table 1) divided by the total number of backward citations of the patent. It seems that, using this calculation, technological similarity is measured instead of technological distance. However, by recoding the outcome of the

(34)

34

calculation, as measuring similarity and distance are opposite, it will indeed result in a measure of technological distance. Therefore, the thirdindependent variable technological distance will be computed as follows:

Technological Distance

As was explained for the previous variable as well, here an outcome of 1 indicates a low technological distance (similar technology used), whereas an outcome of 0 indicates a great technological distance (more new technology used).

Time Distance

The last independent variable in this research is time distance. Time distance or time lag is "the time difference between the application or grant year of the citing patent, and that of the cited patent" (Hall et al., 2001). Time distance can be used to measure whether patents draw from old or recent technological predecessors. In this research the difference between the application year of a patent of RIM and the year of the backward citations of that patent is used. This research uses the mean of all the time differences between the original patent and the backward citations to compute the average time lag of a patent. However, because the research of Katila (2002) found that there is a difference in the effect of the age of knowledge based on where that knowledge comes from, the time lag will be measured for internal knowledge and external knowledge separately. To see whether a patent uses more internal or external knowledge, we will look at one of the previous

variables, search scope. Search scope measures the notion of internal or external knowledge that is used. To determine if a patent is based more on internal than on external knowledge,

(35)

35

we will use the following rule: when the value of search scope is higher than .50, the patent is based more on internal knowledge. When the value is lower than .50, the patent is based more on external knowledge. The variable time distance will measure to what extent the knowledge used is recent or not. In this research, time distance will be computed as follows:

Time Distance

A high time distance value means that there is a high lag between the patent and its backward citations, so a low extent of recent knowledge is used. A low time distance value means that there is a low lag between the patent and its backward citations, so a high extent of recent knowledge is used. Time distance is measured separately for internal and external knowledge.

3.6 Control Variables

The previous independent variables, which are all based on backward citations, are likely to influence the dependent variable, which is expressed in terms of the number of forward citations. However, there are also other aspects and variables that could influence the number of forward citations. To determine which piece of the dependent variable is only explained by the independent variables, some control variables are used in this study. Control variables are expected to explain a certain proportion of the dependent variable (Field, 2013).

Number of Patents Granted

(36)

36

based on the patents that were granted by the USPTO in the years 2005, 2006, 2007 and 2008. Respectively, there were 27, 132, 148 and 238 patents granted to RIM. It is likely that the number of patents granted is related to the number of forward citations. Therefore, this variable is used as a control variable to determine which proportion of the dependent variable is explained by the number of patents granted.

Number of National Classes

The second control variable is the number of national classes attributed to a patent. As explained before, the USPTO has developed the classification system according to its applicability for different technology areas. The classification system consists of 400 main (3-digit) patent classes and over 120,000 patent subclasses. However, the number of

technological fields a patent belongs to, measured by the number of national classes, can also be used as a proxy for the scope and value of a patent (“OECD Patent Statistics Manual,” 2009). Patents that belong to a greater number of national classes, and therefore cover many technologies, have a greater opportunity to receive a lot of citations. It is likely that the more national classes, the more likely it is that a patent receives a higher number of forward citations. To account for this effect, the number of national classes will be used as a control variable as well.

Number of Claims

The last control variable in this study is the number of claims in a patent. The number of claims appears on the front page of a patent and specifies the novelty and functionality of the invention. The claims of a patent also specify what the invention that is patented does that has never been done before, and where it takes a "claim" on a particular function or

(37)

37

design. The number of claims in a patent may be indicative of the scope and width of the invention, as well as the value of the entire patent (“OECD Patent Statistics Manual,” 2009). As it is an indicator of the value of a patent, it is expected that the number of claims will be related to the independent variable. To take this into account, this variable is used as the third and last control variable.

In Table 3, an overview of all the variables used in this study is given.

Table 3 - Overview Variables

3.7 Statistical Method

In many studies, regression analysis is a common statistical tool for estimating the

relationship between and among variables. It is a straightforward way to analyze data, as the method states whether the relationship between the independent and dependent variables is significant, and also gives an indication of the size of the effect. Most studies on individual patent data use regression analysis as well (Jaffe & Trajtenberg, 2002a). However, the type of regression that needs to be used depends on the type of the variables.

Variable Name Comment Type

Dependent Forward Citations The number of forward citations of a patent. Count Data

Independent

Search Scope Percentage of number of self-citations to all backward citations of _{the company.} Scale (0-1)

Geo Search Scope

Percentage of the number of citations toward Canada or US (corrected for the amount of self-citations) to all backward citations of the company.

Scale (0-1)

Tech. Distance Percentage of the number of citations toward national class 455, _{397, or 370 to all backward citations of the company.} Scale (0-1)

Av. Time Lag

The mean of all the time differences between the application year

of the original patent and the years of its backward citations. Scale

Control

Nr. Patents Granted The total number of patents granted in the particular years. Scale Nr. National Classes The number of national classes a patent is assigned to. Scale Nr. Claims The number of claims made by the company's patent. Scale

(38)

38

As shown in the previous table, the dependent variable of this research is a count variable. A count variable is "a variable that takes on discrete values (0, 1, 2…..) reflecting the number of occurrences of an event in a fixed period of time". As stated earlier, a count variable can only take a positive integer value or zero, because an event or count data cannot occur a negative number of times (Coxe, West, & Aiken, 2009). The literature states that there are two models designed to analyze count data, either the Poisson regression or the Negative Binomial regression. The difference between the models lies mainly in the assumptions that they make about the dependent variable, but there is no real consensus about when to use one or the other. Some authors state that the Poisson regression

assumes that the conditional mean of the dependent variable must equal its variance. When this is not the case, they argue that it is better to use the Negative Binomial regression (Hilbe, 1993). However, others argue that a Negative Binomial regression only needs to be used when there is overdispersion, which is the case when the variance is greater than the conditional mean (Lawless, 1987).

A recent study argues that a lot of researchers commonly select Negative Binomial regression purely because the Poisson model often does not fit with social data. It states that, in addition, both models can be used in many similar cases (Piza, 2012), and that the distribution of the dependent variable could be an indication of the use of one of the models. In Figure 1, the distribution of the dependent variable of this study is illustrated. The distribution of the number of forward citations is positively skewed, which is indicative of a typical Poisson distribution (Coxe et al., 2009). As the patent data in this study is not part of social data, and because the distribution of the dependent variable is similar to a Poisson distribution, this study will use the Poisson regression model to analyze the data.

(39)

39

The Poisson regression is part of "logistic regressions". The reliability of the analysis can be obtained by the large sample size of 436 patents.

(40)

40

Section 4 – Findings

In this section, the findings of the quantitative analysis are presented. First of all, some essential descriptive statistics about the sample are provided. After that, a multi-correlation test is used to check the model for multicollinearity. And finally, the hypotheses are tested with the use of a Poisson regression model.

4.1 Descriptive Statistics

Before the hypotheses are tested with a regression model, it is helpful to understand the general output of the data. Therefore, Table 4 presents some important descriptive statistics about the sample. Looking at the dependent variable, which is the number of forward citations per patent, the maximum value is 84 with a mean of almost 12. This means that a patent of RIM is cited by others at most 84 times and on average 12 times. What also can be seen from Table 4 is that there is quite a high variance, which explains the

skewedness of the distribution, as also illustrated in the distribution of the dependent variable in the previous section.

Looking at the independent variables, the values are on a scale between 0 and 1. It is important to mention again that for the variables geographical search scope and

technological distance, a value of 0 indicates a high geographical scope or high technological distance, whereas a value of 1 indicates a low geographical scope or a low technological distance. For the variable search scope, which is measured by the total number of self-citations of RIM, the values are interpreted in the opposite way. Here, a value of 1 indicates high scope, whereas a value of 0 indicates low scope. If we look, for instance, at

(41)

41

variables is 1.00. This means that for these variables, some patents of RIM are based 100% on knowledge gathered from the same country (in this case Canada or the US) and 100% based on the same and existing technology. For the independent variable search scope, the maximum value is also 1.00. According to the interpretation of this variable, for some patents of RIM knowledge is coming 100% from outside the company.

The mean of the variables is either 0.95, 0.47 or 0.25. This indicates that the patents of RIM in this sample are based, on average, 95% on knowledge from outside the company (search scope), 47% on knowledge from the same country or the US (geographical search scope) and 25% based on the same technology. The last independent variable, average time lag, has a maximum of 28 years with a mean of almost 7 years, which is quite common according to Jaffe & Trajtenberg (2002a). The controls suggest that, on average, RIM makes 18 claims per patent and is assigned to 3.56 national classes.

Table 4 – Descriptive Statistics

4.2 Multicollinearity

In Table 5, the correlation matrix of the data is given. This matrix was computed to check the model for multicollinearity. Multicollinearity exists when there is a strong

(42)

42

correlation between two or more independent variables in a regression model, and it can pose a problem in a regression model with multiple variables (Field, 2013, p. 223). If there is perfect collinearity, as when two variables are perfectly correlated, it becomes impossible to obtain unique estimates of the regression coefficients. By checking the correlation matrix, it is possible to identify possible multicollinearity among the variables. This is the case when the correlation coefficient is above .80 or .90 (Field, 2013, p. 224). Looking at the matrix in Table 5, we don’t see any correlation that exceeds these numbers. Therefore, there is no multicollinearity in this model, so all variables can be used in the regression analysis.

Table 5 – Correlation Matrix

4.3 Regression

Output

As stated in the previous section, this study will use the Poisson regression to analyze the data because the dependent variable is a count variable. The output of the Poisson regression starts with the "goodness of fit" measure, as illustrated in Table 6. Goodness of fit is a list of statistics that indicates how well a model fits the data from which it was generated (Field, 2013). In this case, it gives an indication of whether the Poisson regression is the right model to use for the patent data sample. Looking at Table 6, we can conclude that the model

(43)

43

fits reasonably well because the goodness of fit Chi-Square test is not statistically significant, which means that there are no significant differences within the model (“idre.ucla,” 2014).

Table 6 – Goodness of Fit Model

Value df Value/df

Deviance 5426,851 427 12,709

Scaled Deviance 5426,851 427

Pearson Chi-Square 6783,882 427 15,887 Scaled Pearscon Chi-Square 6783,882 427

Log Likelihood -3474,384

The next output, which we see in Table 7, is the Omnibus test. This statistic tests the model as a whole. Looking at the outcome, we see that the model is statistically significant, with a p value of .000.

Table 7 – Omnibus Test

The last part of the Poisson regression will test how the independent variables influence the dependent variable. This regression model reveals whether the relationship between every variable is significant, if it increases or decreases the outcome of the dependent variable, and it says something about the size of the effect (Field, 2013). However, the outcome of the model needs to be interpreted in a different way from normal regressions, because the Poisson regression test models the "log of the expected count as a function of the predictor variables" (“idre.ucla,” 2014). The Beta of the model, which is the standardized regression coefficient of the Poisson regression, can be interpreted as follows: "for a one unit change in

Likelihood Ratio Chi-Square

df Sig.

(44)

44

the predictor variable, the difference in the logs of expected counts is expected to change by the respective regression coefficient, given the other predictor variables in the model are held constant" (“idre.ucla,” 2014). Here a positive number means a positive effect, whereas a negative number means a negative effect. The values in every model mention the

standardized regression coefficients of each model, and the stars indicate the significance of the relationship. The results of the regression model are presented in Table 8, and contain a two-tailed test.

Table 8 – Results of Poisson Regression

*p < 0.05. **p<0.01.

First, a baseline model is presented with the use of control variables only. Each subsequent model then represents a significant improvement over the baseline model. In Models 2, 3, 4 and 5 each independent variable is tested separately and only with the interaction of the control variables. Here it is possible to see whether a particular

Which types of inventions are likely to have the greatest impact in a dynamic environment? : a study of research in motion using patent citations

Master Thesis