A Scandal’s Effect on Knowledge Transfer
Within Organizational Networks
A Study of the Enron Email Corpus
K. Alderin
A Scandal’s Affect on Knowledge
Trans-fer Within Organizational Networks
A Study of the Enron Email Corpus
By Kristina Alderin
Under supervision of:
Dr. Nathan Betancourt
A thesis submitted in partial satisfaction of the requirements for the degree of Master of Science in Business Administration program.
Amsterdam Business School - University of Amsterdam June 2017
STATEMENT OF ORIGINALITY
This document is written by Kristina Alderin who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creat-ing it.
The Faculty of Economics and Business is responsible solely for the supervision of com-pletion of the work, not for the contents.
ABSTRACT
The current literature on the concept of knowledge transfer is extensive, yet little at-tention has been given to the concept in a negative setting. The aim of this paper is to fill this research gap and explore how a scandal affects the absolute level of knowledge transfers and the absolute amount of weak and strong ties within an organ-ization network. By using the Enron email corpus the knowledge transfer network within Enron’s organization is plotted before and after the company’s accounting scandal. The knowledge transfers are divided into explicit and tacit knowledge by us-ing the concept of strength of ties from the social network analysis. An ordinary least square regression and a difference in difference regression showcase the change in the absolute level of knowledge transfers and the absolute amount of strong and weak ties from before to during the Enron scandal. The result shows that there is significant in-crease of the absolute level of knowledge transfers of approximately 576 new knowledge transfers from before to after the scandal at Enron. Additionally, there is a significant increase in the amount of strong and weak ties respectively from before to during the scandal. Another discovery conclude that the percentage of strong ties, rel-ative to the absolute amount of ties, decreased significantly. Controversy, the percent-age of weak ties significantly increased during the scandal. Shedding a light on the implication of a scandal on knowledge transfers within an organizational network brings new information relevant for practice and theory.
Keywords: Enron, knowledge transfers, network, scandal, strength of ties, knowledge
CONTENTS
Abstract ...
1. Introduction ... 7
2. Literature review ... 10
3.1 Knowledge ... 11
3.2 Knowledge management and knowledge transfer ... 14
3.3 The strength of ties ... 17
3.5 Scandal response ... 20
4. Hypotheses ... 22
4.1 Knowledge transfer ... 22
4.2 Strength of ties ... 24
5. Research Methodology ... 26
5.1 The Enron scandal ... 27
5.2 Research construction ... 28 5.3 Explanation of variables ... 29 5.3.1 Dependent variables ... 30 5.3.2 Independent variable ... 33 5.3.3 Control variables ... 33 5.4 Multicollinearity ... 37 5.5 Regression ... 41
5.5.1 Ordinary least square regression ... 41
5.5.2 Difference-in-difference regression ... 44
5.6 Goodness of fit ... 46
5.7 Regression considerations ... 48
6. Data Sources and Sample Construction ... 48
6.1 Data gathering ... 49
6.1.1 Data gathering from other databases ... 51
6.2 Data selection ... 51
6.2.1 Phase I – Pre-processing ... 51
6.2.2 Phase II – Exclusion of duplicate emails ... 52
6.2.3 Phase IIIA – Exclusion of unanswered emails ... 52
6.2.4 Phase IIIB – Exclusion of emails sent to more than one person ... 53
6.2.5 Phase IV – Exclusion of emails without specific email addresses ... 53
6.3 Data selection for additional variables ... 53
6.5 Descriptive statistics ... 55
6.5.1 Selection process overview ... 55
6.5.2 Network statistics ... 56
6.5.3 Mapped out network ... 57
6.5.4 General descriptive statistics ... 59
6.5.5 Diff-in-diff statistics... 62
7. Results and Empirical Analysis ... 63
7.1 OLS regression results ... 63
7.1.1 Knowledge transfer ... 64
7.1.2 Percentage strong ties ... 68
7.2 Difference-in-difference regression results ... 71
7.2.1 Knowledge transfer ... 71
7.2.2 Strong ties ... 74
8. Discussion ... 77
8.1 OLS regression discussion – Knowledge transfer ... 77
8.2 OLS regression discussion – Percentage strong ties ... 80
8.3 Difference-in-difference regression discussion ... 82
9. Robustness check ... 84
9.1 Econometric issues ... 84
9.1.1 Multicollinearity ... 84
9.1.2 Heteroscedasticity ... 84
9.1.3 Endogeneity ... 85
9.2 Stock price as a substitute for the scandal ... 86
9.2.1 Interpretation of robustness check regressions ... 93
10. Conclusion, limitations and recommendations ... 93
10.1 Conclusion ... 93
10.2 Limitations ... 95
10.3 Recommendations ... 96
11. References ... 97
12. Appendix ... 101
1. INTRODUCTION
Over the last few years there has been an upsurge of interest among scholars on the importance of knowledge transfer within an organization. This is mainly because knowledge has become recognized as a crucial strategic advantage in the pursuit of organizational performance (Stephen Tallman, Mark Jenkins, Nick Henry and Steven Pinch, 2004; Linda Argote and Paul Ingram, 2000). The implication for the organiza-tion is that the tradiorganiza-tional asset classes, such as capital, land and labor, has been shift-ed to the knowlshift-edge that resides within individuals in the organization (Linda Argote and Paul Ingram, 2000). Over the passed two decades there has therefore been a surge in the focus of organizations on employee knowledge.
Commonly, knowledge is divided into two categories, explicit (exact and ob-servable) knowledge and tacit (sensitive and complex) knowledge (Polanyi, 1966). Both knowledge categories are essential to an organization in the pursuit of perfor-mance, but are not sufficient on their own as a guarantee to achieve organizational performance. Rather, knowledge must be transferred within the organization to realise the competitive advantage it holds and thus increase performance of the organization. While effective knowledge transfer is an important antecedent of improved work practices, better decision-making and the effective transfer of experience and best practices within the organization, its absence can damage the workforce’s skill acqui-sition and ultimately impede the organization’s productivity (Sharon F. Matusik and
Charles W. L. Hill, 1998). It is therefore crucial for the organization and its
manage-ment to understand the concept of knowledge transfer and the factors affecting it. The most popular view on the topic of knowledge transfer is that of knowledge management. Several studies focus on knowledge management as a way of explaining how knowledge is transferred between individuals and organizations
(Nonaka and Takeuchi, 1995; Koraeus, 2008; Grant, 1996) and how knowledge trans-fer can act as a basis for competitive advantage (Argote and Ingram, 2000). Knowledge transfer, as a concept, is thus developed with a long-term orientation, where firms use knowledge transfers to develop competitive advantage over time. Although the concept of knowledge transfer originates from the theory of knowledge management, there are extensive researches on the topic from a variety of perspec-tives. One such perspective is that of Social Network Analysis (SNA).
Numerous studies within SNA focus on the strength of ties as a method to ex-plain which node (individual), with what tie (relationship), that better can facilitate knowledge transfers. One of the more popular studies is that of Granovetter (1973) that concerns the strength of weak ties in knowledge flow. He finds that: 1) Without weak ties acting as bridges, cliques (clusters) composed of strong ties are likely to be cut off from the knowledge flow and thus interrupt the flow across the entire network. 2) Weak ties, as they are distance and infrequent relationships, are efficient for the transfer of novel knowledge as they create bridges between groups of individuals in an organization (Granovetter, 1973). In parallel, strong ties are likely to result in re-dundant knowledge as they consist of individuals that know each other (Granovetter, 1973). Since then, his findings have been confirmed and his thoughts redeveloped in various studies within SNA (Granovetter, 1974, 1982; Hansen, 1999; Fritsch and Kauffeld-Monz, 2008; Wei, Zheng and Zhang, 2011). Consequently, a few years of research within this subject confirm that strong ties better facilitate the transfer of complex and sensitive knowledge (tacit) while weak ties better facilitate the transfer of novel and diverse knowledge (explicit). Hence, in order for the organization to have an effective and efficient knowledge transfer that leads to organizational
perfor-When considering prior research on knowledge transfer, both within knowledge management and SNA, a crucial but forgotten aspect is that of the impact of a scandal, despite the influence is might have on the absolute levels of knowledge transfer and the amount of strong and weak ties within the network of the organiza-tion. Although the majority of organizations take precautions to prevent or minimize the consequences of such an occurrence, the importance of maintaining an efficient knowledge transfer has regularly been overlooked (Gundel, 2005). The fundamental problem is that a scandal is often associated with a sense of urgency and threat, whereas knowledge transfer originally was set out to develop organizations for the distant future and provide competitive advantage (Nonaka and Takeuchi, 1995). Be-cause of this discrepancy, knowledge management practises, having a long-term ori-entation, do not fit with the short-term orientation of a scandal. It is thus questionable how the concept of knowledge transfer as studied in the theory of knowledge man-agement fit into that of a scandal. There is a clear lack of academic studies within this area and this gap must be filled in order to promote a better understanding of how a scandal affects an organization’s knowledge transfer.
This study attempts to extend existing theory on knowledge transfer in the light of an organizational scandal, using the strength of ties to separate between dif-ferent types of knowledge, explicit and tacit. To do this I examine the Enron email corpus that was originally made public by the Federal Energy Regulatory Commis-sion (FERC) during their investigation of the Enron scandal in 2001. The dataset con-tains emails from former Enron employees, during the period 1999-2002. Therefore, I am able to study the development of the knowledge transfers within Enron’s organi-zational network by examining the absolute amount of strong versus weak ties within their organizational network over time. Hence, I will be able to determine how a
scandal affects the absolute level of knowledge transfers and the absolute amount of strong and weak ties within an organizational network.
This paper makes four contributions to the existing literature. First, it contrib-utes to the SNA theory by extending the existing literature by analysing the affect of a scandal on strength of ties within an organizational network. Thus it contributes to a better understanding of how ties inside a network shifts in pressing circumstances and highlights the importance of each of them in such an event. Second, it sheds light on how strong and weak ties influence the knowledge transfers within the organization and thereby contributes to a richer understanding of how a social network is used to conduct knowledge transfers. Third, it extends the literature within the field of knowledge management by displaying how strong and weak ties are used as conduits for different types of knowledge in short-term oriented environments. Fourth, it en-riches the research on scandal response by proposing knowledge transfer as an over-looked but crucial concept within that field of research.
I begin this paper by an overview of the previous literature related to the study, then I develop and present the hypotheses, next I discuss the methodology of the study, thereafter I present the data gathering, then I present the result, after I dis-cuss the robustness in the light of relevant robustness checks, seventh the disdis-cussion is developed and lastly the conclusion of the study is drawn.
3. LITERATURE REVIEW
The following section discusses the main insights in the existing literature on knowledge transfer and tie strength. First, I introduce the concept of knowledge, which I link to the theory of knowledge management. Then I develop the concept of
concept of strength of ties. Then I examine the previous studies within the strength of ties literature and determine what kind of knowledge weak and strong ties are more/less suitable to transfer. Lastly, I develop the concept of knowledge transfer in scandal response, and present the literature gap and research question.
3.1 Knowledge
In the post-industrial society it is evident that knowledge is the most valuable resource to organizations of all kinds (Prahalad and Hamel, 1990; Nonaka, 1991; Drucker, 1993; Nonaka and Takeuchi, 1995; Grant, 1996). Prahalad and Hamel (1990) discuss the importance of core competences as the firm’s competitive advantage and that the-se core competencies must be nurtured and protected in order for the knowledge not to fade. Nonaka (1991) studies the knowledge-creating company in the light of busi-ness environment uncertainty, and concluded that the one source of lasting competi-tive advantage is knowledge. Drucker (1993) continues to build on these thoughts in his study about the post-capitalist society and how the shifts and transformations led to changes in the way businesses do business. He observes that for the emerging knowledge-based economy the individual was central and the knowledge, rather than resources for production, was to become the main resource (Drucker, 1993). Grant (1996) studies the coordination mechanisms through which firms integrate the knowledge of their employees. Various other scholars underscore the importance of knowledge as the main competitive advantage of the firm. Furthermore, knowledge as the main competitive advantage gains further strength by the strong rise of profes-sional service firms and strategic houses like PwC, KPMG, McKinsey, Bain and Company and Deloitte. In these companies the employee’s knowledge is the key company resource, as professional service firms to a large extent offer their clients
advice rather than a given product or service (Marc Baaij, 2014). For those compa-nies, and others in various industries, the root of competitive advantage has shifted from tangible resources, such as land, labor and capital to the intangible resource knowledge (Kogut and Zander, 1992). This increase is based on uncertainty following major events, such as the globalization of industries, increased deregulations, rise of information technology (IT) and privatization (Robin Fincham 1999). As Nonaka (1991) puts it (pp. 162): “In an economy where the only certainty is uncertainty, the
one source of lasting competitive advantage is knowledge”.
Although knowledge is recognized as this crucial company resource that of-fers a basis for competitive advantage it is an ambiguous concept and usually very difficult to define. Nonetheless, past research commonly differentiate between two different types of knowledge, explicit and tacit. Explicit knowledge is exact, observa-ble, codify-able and transferable through a systematic language, e.g. a routine or manual (Polanyi, 1966; Nonaka and Takeuchi, 1995). Tacit knowledge is hidden in people’s experience, knowing and skills and is the knowledge individuals use to make sense of things (Polanyi, 1966). These two knowledge categories are well known and imply that tacit and implicit knowledge are two different knowledge categories with different features (Sajjad M. Jasimuddin, Jonathan H. Klein and Con Connell, 2005). This dichotomy is widely accepted in the majority of literature on organizational knowledge (Sajjad M. Jasimuddin, Jonathan H. Klein and Con Connell, 2005).
In various studies Polanyi develops the discrepancy between the two knowledge categories (Polanyi, 1958, 1996). Polanyi (1996) argues that explicit knowledge in its codified form is easy to acquire and quick to exploit, as such this type of knowledge does not create a competitive advantage for the organization.
Con-organization as a basis for competitive advantage. Nonaka and Kannon (1998) reflects this view by arguing that explicit knowledge is tangible knowledge that can be codi-fied and tacit knowledge is intangible knowledge that individuals possess. As such tacit knowledge cannot be separated from the individual it is embodied in and it is therefore extremely difficult and costly to transfer, increasing its value as a competi-tive advantage (Polanyi 1996). On the contrary, explicit knowledge can be written down, making it easy and cheap to transfer, decreasing its value as a competitive ad-vantage (Polanyi, 1996; Roberts, 2000). While tacit knowledge is an essential basis for competitive advantage to a modern organization, explicit knowledge also plays a fundamental role in organizational survival. Not only does explicit knowledge have the ability to transfer fast and wide throughout the organization, it also offers a possi-bility to keep knowledge within the organization as an employee leaves. Because the employee turnover is high in today’s employable generation, explicit knowledge is crucial in order to retain a specific employee’s knowledge within the organization once this person leave, e.g. through manuals, routines etc. Hall and Andriana (2003) supports this view by pointing out that explicit knowledge is safe internally because it is articulated, codified and available to the entire organization. Additionally, intellec-tual property rights usually protect this knowledge from being exploited by other firms (Hall and Andriana, 2003). On the contrary, an organization cannot store tacit knowledge by other means than retaining the employee who possess it inside the or-ganization. Boiral (2002) recognises this issue and argues that employee turnover translate into losses of tacit knowledge. Additionally, intellectual property rights do not provide any protection for such knowledge (Hall and Andriana, 2003). Moreover, Szulanski (1996) states that individuals are reluctant to share tacit knowledge as it comes with both power and status. Because tacit knowledge is internally vulnerable
many organizations invest heavily in IT systems to preserve the explicit knowledge within the organization (Sbarcea, 1998). However, explicit knowledge is vulnerable to imitation and can thus lose its value. Hence, and organization that focuses on solely one of the two knowledge categories might lose the value of the knowledge it choose to retain. This implies that both explicit and tacit knowledge are crucial for the organ-ization in terms of success (performance). Yet, an organorgan-ization that possesses both explicit and tacit knowledge is not guaranteed a competitive advantage. Explicit and tacit knowledge is necessary for the creation of competitive advantages and thus or-ganizational performance, but not sufficient on its own. Epple, Argote and Murphy (1996) and Galbraith (1990), confirm that knowledge transfer between units within an organization contributes to the performance of firms within the manufacturing sector. Similar studies have been conducted in the service sector and prove the important function of knowledge transfer in creating organizational performance (Baum and In-gram, 1998; Darr, Argote and Epple, 1995). Hence, the key ingredient to ensure that both forms of knowledge are brought to bear on the problems and opportunities that the organization is facing is the distribution, or transfer, of both knowledge categories between the individuals inside the organization.
3.2 Knowledge management and knowledge transfer
The theory of knowledge management arose in the 1990s from the theory of organiza-tional learning. In its time organizaorganiza-tional learning was the idea that an organization contained knowledge in its structures, policies and standards, and by varying these factors an organization could learn and respond to its continuous changing environ-ment. Over time organizational learning became divided into side-tracks, where
Takeuchi invented the field in 1995. They study how Japanese companies translate tacit to explicit knowledge and used it to gain competitive advantages in processes, products and services (Nonaka and Takeuchi, 1995). In that study they write about knowledge management as a learning process and develop the concept of knowledge transfer (Nonaka and Takeuchi, 1995). Nonaka and Takeuchi (2005) set out this con-cept of knowledge transfer as a root of competitive advantage that an organization should build over time. Hence, the view of knowledge transfer within knowledge management has a long-term orientation in which knowledge transfer prepare the or-ganization for the future by shaping its competitive advantage over time. Several scholars reflect the thoughts of Nonaka and Takeuchi. Alavi and Leidner (2001) re-view and analyse knowledge management literature in different fields in order to identify the role of information technology in the process of the creation and transfer of knowledge within organizations. Kogut and Zander (1992) examine the claim that the multinational corporation arises due to its superior efficiency in transferring knowledge across borders. Szulanski (1996) examine the barriers to effective knowledge transfer and best practices within the firm. Argote and Ingram (2000) claim that organizations can obtain a competitive advantage by transferring knowledge internally while preventing knowledge from transferring externally. In ac-cordance with Nonaka and Takeuchi (2005) these scholars discuss knowledge transfer as a long-term investment to build a competitive advantage within the organization.
While scholars have developed more insight into the concept of knowledge transfer, a specific definition is still lacking (Lin, Hung, Wu, and Lin, 2002; Alvesson and Kärreman, 2001). For the purpose of this paper I have chosen to use the defini-tion provided by David Schwartz and Dov Te’eni (2011) in the Encyclopedia of Knowledge Management: “Knowledge transfer is a process by which knowledge,
ide-as and experience moves from the sender of the knowledge to the receiver of that knowledge”. This implies that knowledge in itself does not have to be novel in order
to be transferred between nodes in a network, which was initially proposed by Nona-ka and Takeuchi (1995). Rather knowledge transfer takes place as knowledge goes from the sender of the knowledge to the receiver of the knowledge regardless if the receiver or sender was new or familiar with the knowledge to begin with.
Knowing that both explicit and tacit knowledge can transfer between individ-uals regardless of the aspect of new or experienced knowledge, raises an interesting question: How can this knowledge be transferred between individuals? In the previous section on knowledge it was mentioned that explicit knowledge, for instance, could be written down in a manual or turned into a routine. While tacit knowledge resides within an individual and is increasingly difficult to transfer compare to explicit knowledge. However, common to both knowledge categories is that one individual must interact with another individual’s knowledge in order to retain the knowledge. Consequently, one can argue that knowledge transfer takes place when one individual interact with another individual’s knowledge. In contrast, knowledge cannot be trans-ferred from person A to person B if there is no interaction between these two individ-ual’s knowledge. This holds for both knowledge categories and implies that the trans-fer of knowledge is done through ties between two individuals. The concept of ties acting as conduits for knowledge transfer has been further developed in the literature on SNA. In the light of the SNA theory the concept of ties are known as the strength of ties, where weak and strong ties act as conduits for knowledge transfer.
3.3 The strength of ties
Coming from the social science, SNA is a structural approach that is based on the study of interaction and relationships linking social actors (Freeman, 2004). Since Moreno originated the concept in 1934, several scholars, such as Warner (1937), Laumann and Knoke (1989), Wellman (1981), Lorraine and White (1971), Granovet-ter (1973), Krackhardt (1992) and White (1992), study various aspects within the field. For the purpose of this study I will focus on the previous literature within SNA that emphasize on knowledge transfer and its relation to the concept of strength of ties, while leaving other areas open for exploration in other research papers.
One of the most cited definitions of tie strength is Granovetter’s (1973). He defines tie strength as “…a combination of the amount of time, the emotional
intensi-ty, the intimacy (mutual confiding), and the reciprocal services which characterize the tie” (Granovetter, 1973, pp. 1361). However, multiple other definitions of tie strength
exist. Homans (1950) measures tie strength in terms of frequency, where a high fre-quency between two individuals indicate a strong tie; Granovetter (1974) and Lin et al. (1981) also propose to define strong ties as the more frequent ones; Erickson et al. (1978), Granovetter (1974), Lin and Dumin (1982) and Murray et al. (1981) use indi-cation of “closeness” to describe tie strength, thus close friends have strong ties, while acquaintances or friends have weak ties; Friedkin (1980) focuses on mutual acknowledgement where a strong tie must be acknowledged by both parties; Lastly, Alba and Kadushin (1976) say that the overlap of organizational membership indicate a strong tie. With various existing explanation for the definition of tie strength it is evident that the concept is highly contested. As a result, I chose to adopt the definition proposed by Granovetter (1974) and Lin et al. (1981), and used by, among others,
Homans (1950) and Benassi et al. (1999). Thus, tie strength is defined as the
frequen-cy of interaction between two nodes (individuals) in a network.
With his article “The Strength of Weak Ties” Granovetter is the main contrib-utor to the concept of tie strength and its link to knowledge transfer. Granovetter’s finding is the first notion that tie strength could be used gainfully to understand knowledge flow among social actors (Granovetter, 1973). He argues that groups con-sisting of strong ties tend to group together into cliques and relied on weak ties in or-der for knowledge to flow between their group and other groups (Granovetter, 1973). Moreover, he finds that strong ties tend to hold redundant knowledge, as they interact often and thus develop similarities, while weak ties are more likely to contribute with novel knowledge (Granovetter, 1973). This is initially confirmed in his 1974 study of weak ties in job seeking efforts, in which he finds that job seekers are more likely to learn about new job openings through acquaintances than close friends (Granovetter 1974). And then once again in his 1982 article "The Strength of Weak Ties: A Net-work Theory Revisited”, in which he finds that individuals with few weak ties will not received knowledge from distant parts of the system but rather be constrained to the knowledge of their close friends (Granovetter, 1982). Other authors confirm Granovetter’s research on the strength of weak ties; Andrea Kavanaugh, Debbie Reese, John Carroll and Mary Rosson, 2004, show that people with weak ties across groups have a higher level of involvement in their communities, interest in the society and work better together than groups without weak ties among individuals. Valery Yakubovich (2005) also confirm Granovetter’s study by showing that a worker in the Russian labor market is more likely to get a job through one of his/her weak ties ra-ther than through a strong tie.
While Granovetter and his followers make a strong case for the significant role of weak ties in efficient knowledge flow there has been several scholars with de-parting results. For instance, Fritsch and Kauffeld-Monz (2007) study the impact of network structure on knowledge transfer and find that strong ties are more beneficial for the transfer of knowledge. Hansen (1999) examines the role of weak ties in shar-ing knowledge across organizational subunits and finds that neither weak nor strong relationships between units lead to efficient knowledge sharing between them. Addi-tionally, he sees that strong ties have a better effect when the knowledge is highly complex, whereas weak ties have a better effect when the knowledge is not complex. Marsden and Campbell (2012) find that weak ties facilitate the transfer of novel knowledge, while strong ties act as a dependable resource of social or emotional sup-port. Hansen (1999) studies the role of weak ties in sharing knowledge across the or-ganization, for which he finds that weak ties help individuals to search for useful knowledge in other units, but limit the transfer of complex knowledge. Wei, Zheng and Zhang (2011) find that the distance and structural equivalence between the seeker of knowledge and the source of knowledge would determine how much knowledge is transferred between them. Although there is a considerable debate about the ad-vantages of a weak versus strong tie, the majority of the past research show that strong ties act better as transmitters for complex and sensitive knowledge, whereas weak ties better transfer novel and diverse knowledge. This implies that strong ties act as better transmitters for tacit knowledge as this knowledge is complex and sensitive, and thus hard to transfer. While weak ties act better as transmitters for explicit knowledge as this knowledge is novel and diverse, and thus easier to transfer. Hence, each tie within the network transfers knowledge, regardless of it is a weak or strong tie. The main difference lies in the fact that strong ties more efficiently transfer tacit
knowledge while weak ties more efficiently transfer explicit knowledge. Controver-sially, the phenomenon of knowledge transfer is not to be treated separately from the phenomenon of strength of ties. Rather ties are the conduits through which knowledge transfer; hence necessary for knowledge to transfer between individuals. Indeed, the phenomenon of strength of ties confirms this while adding an important separation between which type of knowledge that is transferred through which type of tie.
The concept of ties as conduits for explicit and tacit knowledge transfers in the SNA theory do not focus on knowledge transfer as a competitive advantage, but ra-ther focus on which type of tie is better suited to transfer tacit or explicit knowledge. Thus, the concept of knowledge transfers in the SNA theory adopts the long-term ori-entation that it was originated with. And while the past literature on tie strength and knowledge transfer is quite extensive it, just like the concept of knowledge transfer within the field of knowledge management, has not been examined in a short-term oriented environment, like a scandal. Consequently there is a clear lack of empirical research on the affects of a scandal on the absolute amount of ties within the organiza-tional network and thus on the affects on the absolute levels of knowledge transfers.
3.5 Scandal Response
A company scandal can take many shapes and cause different degrees of damage to involved stakeholders. But whether the scandal is political, financial or legal, all com-panies involved must respond effectively and with communication towards external and internal stakeholders. In our modern society a scandal has a negative clang and most individuals has come across one, whether in politics, business or our daily lives. Although the various definitions of the word are broad and the impact on stakeholders
gressions which becomes known to others and are sufficiently serious to elicit a pub-lic response” (John B. Thompson, 2013, pp.4). A transgression refers to a violation of
the law, a command or a certain duty. As such, the definition of a scandal implies that any action or event that goes against a stated law, command or duty, which then be-comes known to others and is serious enough to enact a response by the public, can be termed as a scandal. However, it is important to note that not all scandals have the same impact, but rather vary in their strength and potential damage.
Regardless of the impact of the scandal prior literature argue that successful knowledge transfer depend on the absorptive capacity and degree of complexity of the knowledge being transferred (Zahra and George, 2002; Lane et al., 2001). Yet, Martin and Salomon (2003) argue that in a turbulent environment, such as a scandal, a more social aspect is critical to efficiently transfer knowledge. Davenport and Prusak (2000) and Shin, Holden, and Schmidt (2001) add to this argument by highlighting the importance of allowing experts to talk to each other in order to promote effective knowledge transfer by the creation of both strong and weak ties. While this might work very well in the average organization, it might seem like a more difficult task to accomplish in a stressed short-term decision-making setting, such as a scandal. On one hand, one can argue that while an organization can reach an effective knowledge transfer by establishing both weak and strong ties in an organization, the short-term orientation of a scandal impedes this creation of ties. This is especially evident as pre-vious literatures argue that knowledge transfer takes a long-term orientation whereas a scandal takes a short-term orientation. Thus, it is questionable if the concept of knowledge transfer developed in the knowledge management theory can be adopted in the face of a scandal. On the other hand, the situation might also be reversed. A scandal might increase the creation of ties and thus increase knowledge transfers by
stressing the importance having the organizational competitive advantage, knowledge, protected and well utilized to better regain back the organization’s strength. This con-cept is more evident in today’s business environment because of the emphasis on hu-man knowledge as the main organizational resource for competitive advantage.
Recent researches recognize the discrepancy between these two views and suggest that a more comprehensive understanding of the influence of a scandal on knowledge transfer is needed (Koraeus, 2008; Riggs 2014). Gaining that understand-ing is the first step in ensurunderstand-ing that your organization can respond effectively to a scandal. Consequently, this paper studies how a scandal affects the absolute levels of knowledge transfers and the absolute amount of strong and weak ties within an organ-izational network. The research question that this paper aims to answer is:
What effect does a scandal have on the absolute levels of knowledge transfer and the absolute amount of strong and weak ties within the organizational network?
4. HYPOTHESES
The aim of this study is to analyse the impact of the Enron scandal on the absolute level of knowledge transfers and the absolute amount of strong and the absolute amount of weak ties within the organizational network. In order to achieve this pur-pose three different hypotheses are formulated.
4.1 Knowledge transfer
The literature review presents two competing argument for the effect of the Enron scandal on the absolute level of knowledge transfers in the organizational network.
1) Decrease in the absolute level of knowledge transfers: The theory of knowledge management as set out by Nonaka and Takeuchi (1995) adopts a long-term view of knowledge transfer. Because the older and current literature on knowledge transfer has its origin from the theory of knowledge manage-ment it is highly likely that the long-term orientation of knowledge transfer has continued. Consequently it is questionable if knowledge transfers would be considered in a short-term orientation. Because a scandal create a sense of urgency and adopts a short-term orientation, this would be an area where the concept of knowledge transfer is not applicable. Hence, the long-term orienta-tion of knowledge transfer as in theory, is not compatible with a scandal. As a result, the absolute level of knowledge transfers within an organization under-going a scandal will suffer, due to a lack of focus on the long-term, resulting in a decline in the absolute level of knowledge transfers during such an event. 2) Increase in the absolute level of knowledge transfers: For the past two decades
authors such as Stephen Tallman, Mark Jenkins, Nick Henry and Steven Pinch (2004), Linda Argote and Paul Ingram (2000), Prahalad and Hamel (1990), Nonaka (1991), Drucker (1993), Nonaka and Takeuchi (1995) and Grant (1996) argue that the knowledge residing inside of employees is the main competitive advantage of companies. Various business leaders recognize this research and now seek to further enhance their organizations by developing in-house development programs e.g. trainee programs, workshop sessions, highly regarded incentives and other kinds of employee support. That knowledge is the most important resource is confirmed by the strong rise in professional service firms, where the firm “sells” the knowledge of its employees to its cli-ents (Marc Baaij, 2014). Because the view on knowledge has changed the past
decade and today adopts a central role inside organizations, the majority of or-ganizations ensure that organizational structures and processes emphasize the transfer of knowledge between employees. It is therefore highly likely that the shift in focus from tangible resources to the intangible resource, knowledge, outweigh the discrepancy between the original long-term orientation of knowledge transfer and the short-term orientation of a scandal. Particularly because knowledge is a recognized competitive advantage of firms, and the concept of knowledge transfer, like anything else, progress over time. I can therefore draw the conclusion that it is likely that a highly dynamic business environment, which we have today, is automatically transforming the long-term orientation of knowledge transfer into a more short-long-term and flexible concept. Additionally, a competitive resource like knowledge must work in a short-term setting in order to utilize this resource in short-term oriented pro-jects as well. Hence, knowledge transfer can overcome the contradictions of a long-term versus a short-term orientation and the absolute levels of knowledge transfers will increase during a scandal.
Hypothesis 1: In response to the Enron scandal the absolute level of knowledge
transfers increases.
4.2 Strength of ties
Several authors have proven how ties, in terms of their strength, ensure a successful transfer of knowledge. As such ties are conduits for the transfer of knowledge be-tween individuals. In their absence knowledge transfer would be either problematic or
not take place at all. In the light of a scandal and the arguments for hypothesis 1 (H1) there are three possible outcomes of the strength of ties.
1) The absolute amount of weak ties and strong ties increases. From H1 it is
de-termined that the absolute levels of knowledge transfers increases during a scandal. Consequently, it is not possible that the absolute number of weak ties and strong ties both decreases. Because a decrease in both types of ties would imply a decrease in the absolute level of knowledge transfers. Hence, this ar-gument continues to build on H1. It argues that in the light of a scandal the or-ganization is able to overcome the discrepancy between the long-term and short-term and the absolute amount of weak and strong ties increases.
2) The absolute amount of weak ties decreases and the absolute amount of strong ties increases. Previous research concludes that strong ties are better
transmit-ters of sensitive and complex knowledge (tacit knowledge), while weak ties are better transmitters of novel and diverse knowledge (explicit knowledge). It is therefore highly likely that the two ties do not react to the scandal in a simi-lar way. During a scandal it is possible that little new knowledge will be de-veloped within the organization as it focuses on solving the scandal it faces. On the contrary it is possible that more sensitive knowledge e.g. regarding the scandal, is transferred and discussed within the organizational network. Thus it implies that the absolute number of weak ties decreases and the absolute number of strong ties increases.
3) The absolute amount of weak ties increases and the absolute amount of strong ties decreases. Involving the discrepancy between the long-term orientation of
knowledge transfer in the theory of knowledge management and the short-term orientation of a scandal, one could argue that the first knowledge to be
transferred in the light of a scandal would be explicit knowledge. This is be-cause explicit knowledge is easier to transfer since it demands a shorter time-horizon to be transferred, than tacit knowledge that is more complex. Because weak ties are more applicable in a short-term orientation, like a scandal, it is highly likely that these ties increase as the organization tries to utilise its most valuable resource, knowledge. On the contrary, because strong ties are more applicable in a long-term orientation it is highly likely that these ties decrease as the organization faces a short-term oriented environment, like a scandal. Hence, it is not likely that 2) holds. This is strengthen by the logical argument that the employees within the organization are likely to hold any sensitive knowledge regarding the scandal to themselves in order to prevent information from leaking internally to externally. Thus, the information travelling within the organization is more likely to be news regarding the scandal that are less sensitive and therefore pass through weak ties. Hence, the hypotheses are:
Hypothesis 2: In response to a scandal, the absolute amount of tacit knowledge
transfers (strong ties) decreases.
Hypothesis 3: In response to a scandal, the absolute amount of explicit knowledge
transfers (weak ties) increases.
5. RESEARCH METHODOLOGY
The methodology section is constructed as follow. First, I explain the used research design. Second, I display the different variables used in the study. Lastly, I justify the
use of the regression in the study and how it links to the different variables used. Sec-tion 6, which follow, describe the dataset used to test the hypotheses.
5.1 The Enron scandal
The Enron scandal is widely known as one of the biggest audit failures in history. Re-vealed in 2001, the scandal forced the Enron Corporation into bankruptcy. In its fall the American government also enforced the dissolution of Arthur Andersen, one at the time “Big five” companies. A part from being the biggest audit failures in history the Enron Corporation is well-known for its rapid collapse.
Founded in 1985 in Hudson, Texas, the Enron Corporation grew to become one of the biggest energy companies in the world. In the late 1990s Enron became “Americas most innovative company” and was highly praised for its business. The company focused on electricity, natural gas, communication, pulp and paper, and en-ergy trading within odd new markets such as weather futures and broadcast time for advertisers (Li, 2010). Between 1995 and 2000 Enron’s revenues grew with approxi-mately $91 billion and early in 2001 the company hit its peak, however, it did not take too long before the company’s way to the top was discovered to be darker than any-one could have imagined (Li, 2010).
After years of accumulating debt in order to grow, Enron’s management reached a point where they feared that the company’s excessive debt would become public knowledge and thus impact the company’s stock price negatively. In order to avoid that situation Enron decided to turn to Arthur Andersen. With Arthur Ander-sen’s expertise in auditing and accounting Enron managed to use accounting loop-holes to hid the company’s debt and portray an unrealistic picture of the company’s financials to its current and potential shareholders. However, this was not a long
last-ing solution as analysts quickly started to question Enron’s balance sheet. In 2001 it was revealed that the company had hidden away excessive debt by the use of account-ing fraud with its accomplice Arthur Andersen. In less than a year Enron’s share price plummeted from $87,64 per share (Sep-00) to less than $1 per share (Nov-01) (Li, 2010). Frustrated and angry over the misconduct of the company the shareholders of Enron answered by filing a $40 billion lawsuit. This was the end for the giant energy corporation and Enron filed for bankruptcy the 2nd of December 2001, taking with it Arthur Andersen (Hardin, J. S. and Sarkis, G., 2015). In its demise the company left the only real email dataset that is publicly available. Over time this has become known as the Enron email corpus.
Revisiting the definition of a scandal in the literature review section, in the case of Enron, hiding away debt is a violation of the company’s duty towards its shareholders. Additionally, the accounting fraud is a violation against the law. After a while this became known to the public and caused an immediate public response. Since the Enron scandal match the definition outlined above it is clear that it can be termed a scandal. Additionally, for Enron the scandal had a significant impact since it lead the company to go bankrupt. Therefore I draw the conclusion that this scandal is the furthest up the scale of impact of scandals. Other scandals might have negative impacts but ultimately not lead to bankruptcy of the company itself.
5.2 Research construction
This study examines the Enron scandal’s effect on the absolute levels of knowledge transfers and the absolute amount of strong and weak ties. Below is a construct of the framework used to find the answer to the posed research question of this paper.
Framework I
BEFORE THE ENRON SCANDAL
This framework depicts the research construct of this paper. It indicates that the period before the scandal is be-tween Q1 1999 and Q4 2001. Then it shows how the construct moves towards identifying a knowledge transfer, then towards calculating the absolute level of knowledge transfers, next it depict the calculation of the frequency and indicate that this will be used to find the absolute amount of strong and weak ties.
Framework II
DURING THE ENRON SCANDAL
This framework depicts the research construct of this paper. It indicates that the period after the scandal is between Q4 2001 and Q3 2002. Then it shows how the construct moves towards identifying a knowledge transfer, then towards calculating the absolute level of knowledge transfers, next it depict the calculation of the frequency and indicate that this will be used to find the absolute amount of strong and weak ties.
5.3 Explanation of variables
This section covers the variables used in the study to test the hypothesis. Included are the dependent, independent and control variables.
Before the Enron scandal (Q1 1999 – Q3 2001) Absolut level of knowledge transfers in the network Absolute amount of strong ties Absolute amount of weak ties Frequency of email exchange between indi-viduals at En-ron Email ex-change be-tween A and B leads to suc-cessful knowledge transfers During the Enron scandal (Q4 2001 – Q3 2002) Absolut level of knowledge transfer in the network Absolute amount of strong ties Absolute amount of weak ties Frequency of email exchange between indi-viduals at En-ron Email ex-change be-tween A and B leads to suc-cessful knowledge transfers
5.3.1 Dependent variables
Knowledge transfer
The absolute level of knowledge transfers within the organizational network is de-pendent on the scandal and thus viewed as a dede-pendent variable in this study. A suc-cessful knowledge transfer takes place when individual A sends a message to individ-ual B, and B sends a response back to A (i.e. follow up on the email sent from A). The follow-up is a critical condition since this indicates that B actually receives the knowledge sent from A and thus the knowledge is successfully transferred (frame-work 2). Note that this view differentiate from the theoretical view of knowledge transfer, the theoretical view treats knowledge transfer as the movement of infor-mation from the sender to the receiver. This study takes a stricter view and assumes that knowledge transfer occurs when the receiver of an email responds to the sender of an email and the subject of the email remains the same.
It is important to note that knowledge is treated as raw information throughout this paper and that no references had been made to the content of any email.
Framework III
KNOWLEDGE TRANSFER
This framework depicts a successful knowledge transfer between a sender and receiver. First the Sender (A) trans-fers en email to the Receiver (B), B then follow up on the email by transferring back a response to the email by A; hence a successful knowledge transfer is created. The arrows represent the direction of the email.
Sender (A) Receiver (B)
Percentage strong ties
The percentage of strong ties in relation to the total amount of ties is used as the se-cond dependent variable as the study examines if the amount of strong and weak ties changes during the scandal.
It has previously been established that this study use frequency as a definition of tie strength. A part from being validated in previous literature this definition fills another important function as it perfectly match the requirements of the Enron email corpus used in this study. By measuring the strength of each tie in terms of frequency I am able to study the rate of email exchanged between two individuals within the network and thus detangle their tie strength. The frequency is measured over time (months) in order to examine how the strength of ties shifted from before the scandal to during the scandal. The frequency is determined according to the equation below:
(1) !!!!! !!!!! !!!!! !!"!!
!! = 𝐹!!
In which F1T1 represent the frequency of knowledge transfers between two
us-ers in time 1. F2T1 and F3T1 represent the frequency of knowledge transfers of other
users in the network in time 1 and FnT1 represent that the addition continues until user
n, which is the last frequency of knowledge transfers between two users in time 1. nT1
represent the number of users in the network in time 1 and FT1 represent the frequency
in time 1. The frequency according to strong and weak ties is then divided according to the following:
(2) 𝐹𝑟𝑒𝑞𝑢𝑒𝑛𝑐𝑦 𝑜𝑓 𝑡ℎ𝑒 𝑡𝑖𝑒 ≤ 𝑓𝑟𝑒𝑞𝑢𝑒𝑛𝑐𝑦 → 𝑤𝑒𝑎𝑘 𝑡𝑖𝑒 (3) 𝐹𝑟𝑒𝑞𝑢𝑒𝑛𝑐𝑦 𝑜𝑓 𝑡ℎ𝑒 𝑡𝑖𝑒 > 𝐹𝑟𝑒𝑞𝑢𝑒𝑛𝑐𝑦 → 𝑠𝑡𝑟𝑜𝑛𝑔 𝑡𝑖𝑒
The equation and conditions above are used to determine which tie that is strong and which tie that is weak in the organizational network. Once it is established which tie that transfers which kind of knowledge (explicit or tacit) the percentage of strong ties is calculated. This is then used as the dependent variable.
(4) !"#$%& !"#!!
!"#$% !"#$%& !" !"#$!! = % 𝑠𝑡𝑟𝑜𝑛𝑔 𝑡𝑖𝑒𝑠!!
By using the percentage of strong ties as the dependent variable I avoid the necessity to test the percentage of weak ties as well. Instead the regression can be run with solely the percentage of strong ties as the dependent variable. If the regressions result is significantly positive, it indicates that the percentage of strong ties increases. If the result of the regression is significantly negative, it signifies that the percentage of strong ties decreases; hence, the number of weak ties increases. By such, solely one regression is needed to test H2 and H3.
Number of strong ties
The dependent variable number of strong ties is solely used for one of the difference-in-difference (diff-in-diff) regressions. The number of strong ties represents the abso-lute amount of strong ties that exists in the organizational network. This variable is calculated as described above (see: percentage strong ties), and the values are gath-ered over time per month.
5.3.2 Independent variable
Scandal
The scandal that occurred at Enron in Oct-01 is the independent variable in the study, as it affects both the dependent variables above. Thus, I use the scandal to test H1, H2 and H3. The independent variable shifts between no scandal (0) and scandal (1). A variable taking on the value of 0 or 1 is known as a dummy variable. The scandal starts in Oct-01 and this is thereby the first month assigned 1. There are 32 time peri-ods before the scandal and 9 time periperi-ods of the scandal. These time periperi-ods falls monthly and not yearly in order to utilize as many observations as possible.
5.3.3 Control variables
The purpose of the control variables is to filter out any effect they could have on the
independent variable, i.e. there is a possibility that each of the control variables could
affect the dependent variables in the organizational network. Including the control
variables in the regression will separate this effect. Below follows the control
varia-bles used in the study:
Debt ratio
The debt ratio is the organization’s debt to assets ratio. It indicates the percentage of
an organization’s assets that is provided via long-term and short-term debt. It is a
good indicator to see how heavily the company rely on debt to finance its assets and is
commonly used as a control variable in several studies (El Ghoul et al. 2016). A
high-er debt ratio indicates that the company has a highhigh-er degree of levhigh-erage and is thus
more likely to be a subject of financial distress. Financial distress in turn could
e.g. if the company is closer to financial distress it might focus less on providing
in-ternal communication ways to its employees. Indeed this can potentially affect the
dependent variables. Hence this variable is included as a control variable. In order to
avoid any skewness the debt ratio is turned into a logarithm.
(5) ln!"!" !"#$ !"#$!!! !!!"# !"#$ !"#$!!
!"#$% !""#$"!! = 𝐷𝑒𝑏𝑡 𝑟𝑎𝑡𝑖𝑜!!
Return on assets (ROA)
ROA is used as a control variable in various studies (El Ghoul et al. 2016). The ROA
indicates how the company is doing in terms of performance relative to its total assets.
As such, it is a strong measure of a company’s performance. The higher ROA the
more common it is for the organization to be viewed positively by investors. If
inves-tors view the company positively it is possible that it will affect the dependent
varia-bles. Hence, ROA is included as a control variable for all the dependent variavaria-bles.
(6) !"# !"#$%& (!"##)!!
!"#$% !""#$"!! = 𝑅𝑒𝑡𝑢𝑟𝑛 𝑜𝑛 𝑎𝑠𝑠𝑒𝑡𝑠!!
Employees
Employees represent the number of employees in the organization. In line with
multi-ple other research papers, this is a control variable that is good for analysing the size
of an organization. The number of employees might inflict on the absolute level of
knowledge transfers and the strong ties within the organizational network, because the
in-Market Cap
Enron’s stock price and total shares outstanding per quarter and year was used to
cal-culate the market cap of the company. The market cap is commonly used as a
meas-urement of the size of a company and used as a control variable in various studies (El
Ghoul et al. 2016). The market cap was turned into a logarithm in order to avoid
skewness. The size might inflict on both knowledge transfers and the strong ties
with-in the organizational network because it with-indicates how big the company is, how many
employees the organization can sustain and so on.
(7) ln(𝑆𝑡𝑜𝑐𝑘 𝑝𝑟𝑖𝑐𝑒!!∗ 𝑇𝑜𝑡𝑎𝑙 𝑠ℎ𝑎𝑟𝑒𝑠 𝑜𝑢𝑡𝑠𝑡𝑎𝑛𝑑𝑖𝑛𝑔!!) = 𝑀𝑎𝑟𝑘𝑒𝑡 𝑐𝑎𝑝!!
Labor
Labor is calculated on the basis of revenue per employee. The variable indicates the efficiency of the employees. The higher its value, the more efficient the employees of a company are in terms of generating revenues. Labor may inflict on both the absolute level of knowledge transfers and the strong ties within the organizational network. It is possible that the efficiency of an employee determine how many knowledge trans-fers and strong and weak ties that employee has. In line with market cap, labor is turned into a logarithm in order to avoid skewness.
(8) ln !"#"$%"&!!
!"#$%&''(!! = 𝐿𝑎𝑏𝑜𝑟!!
Total emails sent
Total emails sent is the total number of emails sent within Enron, excluding those emails that contain “RE:” in the subject line. RE indicates that there is a direct reply
to an email. The emails containing “RE:” in the subject line are removed because they are already accounted for in the dependent variable knowledge transfer. Hence, the total emails sent represent the total amount of emails circulating in the Enron net-work. This control variable might increase or decrease the absolute level of knowledge transfers and strong or weak ties within the organizational network be-cause it might affect the number of emails that are responded to. Logically, if there are 10 emails sent in one period those emails are more likely to be responded to than in a period where 10.000 emails are sent. This control variable was turned into a loga-rithm in order to avoid any skewness in the regression.
Profitability
The profitability was calculated on the basis of operating net income (loss) and sales. This measure is consistent with multiple other studies. Profitability, or more common-ly referred to as the net profit margin, is used as a measurement of the organization’s profitability and shows which percentage of the selling price that is turned into profit. Hence, the higher its value, the better it is for a company’s financial position. With a high profitability the organization might be more inclined to spend money on com-puter equipment, increasing the amount of email sent and thus possible affect the de-pendent variables.
(9) !"#$%&'() !"# !"#$%& (!"##)!!
!"#$%!! = 𝑃𝑟𝑜𝑓𝑖𝑡𝑎𝑏𝑖𝑙𝑖𝑡𝑦!!
Total unique users
net-as a control variable in order to determine how large the network wnet-as initially. The total amount of unique users might inflict on the dependent variables because a larger amount of total unique users implies a smaller percentage chance that the two users, exchanging an email, would have done so previously. As such the total unique users represent the size of the network. This variable is in line with several social studies that argue that the size of a group inflict with its performance, it is therefore likely that this holds for networks as well. This variable was turned into a logarithm in order to avoid skewness.
Table I
SUMMARY VARIABLES
This table represent all variables included in the study. First are the dependent variables, knowledge transfer, per-centage strong ties and number of strong ties. Second, is the independent variable scandal. Third, are the control variables, debt ratio, ROA, employees, market cap, labor, total emails sent, profitability and total unique users.
Dependent variables Knowledge transfer Percentage strong ties
Number of strong ties
Independent variable Scandal
Control variables Log Debt ratio
ROA
Employees
Log Market Cap
Log Labor
Log Total emails sent
Profitability
Log Total unique users
5.4 Multicollinearity
Multicollinearity is a test that shows the degree of correlation between the variables outlined above. This test is done before the regressions are conducted in order to avoid distortion of the result due to possible strong correlations. The best way to
iden-tify multicollinearity is through a correlation matrix. StatPlus, Linear correlation (Pearson), is used to create Table II. I define a moderate correlation as less than -0,55 and more than 0,55, and a strong correlation as less than -0,75 and more than 0,75. Correlations above -0,55 and below 0,55 are viewed as weak correlations.
Table II
CORRELATION MATRIX
This table display the correlation matrix of the control variables used in the study. Light grey background indicates a strong correlation between variables. 5 moderate correlations exist and 4 strong correlations exist.
ROA Log Debt ratio Log Labor Log Total unique users Log Total emails sent Profitability Log Market Cap Log Em-ployees
ROA 1
Log Debt ratio 0,46 1
Log Labor 0,44 0,59 1
Log Total unique users -0,44 -0,39 -0,27 1
Log Total emails sent -0,43 -0,42 -0,49 0,99 1
Profitability 0,93 0,73 0,72 -0,66 -0,66 1
Log Market Cap 0,67 0,10 0,08 -0,19 -0,19 0,53 1
Log Employees 0,31 -0,35 -0,83 0,27 0,26 0,10 0,85 1
As viewed in Table II there are several moderate and strong correlations between the chosen control variables. Because a strong correlation can possibly bias the regres-sion, market cap, employees and profitability has been excluded. Despite the correla-tion between labor and debt ratio these variables will be tested in the same models. These variables do not stem from the same underlying values, thus there is a possibil-ity that the correlation is a coincidence. Additionally, the correlation is at 58%, indi-cating that there are 42% that is not explained as a correlation between them. Hence, both variables are kept and tried in the same models. Total emails sent and total
ity that the total emails sent will increase (or decrease) as well. Thus the variables were expected to strongly correlate. Both variables are included in the regression but in different models in order to avoid this correlation.
The new correlation matrix, excluding the eliminated variables and including the study’s independent and dependent variables, is displayed below (Table III). The table indicates that there is a moderate correlation between the independent variable scandal and the dependent variable knowledge transfer. Because this is the relation-ship the study is set out to examine this suits as a first indicator that the scandal af-fects the absolute level of knowledge transfers within the organizational network. However, causality cannot be confirmed yet.
Additionally, there is also a moderate correlation between the scandal and ROA. Because the scandal is a dummy variable it is highly likely that the patters of it moves similar to the pattern as the ROA of the organization. This would then explain the moderate correlation between the variables. However, should there be a relation-ship this will be showed in the regression models. There is yet another moderate cor-relation between labor and debt ratio that is expected due to the result in the previous correlation matrix. The correlation lies at the same level, 58%, and the two variables will still be considered in the same models. There is a strong correlation between the total emails sent and the total unique users, this is expected due to its correlation in Table II. These variables will both be tested but in different regression models. More-over, there are two moderate correlations between knowledge transfer and total unique users and knowledge transfer and total emails sent. These correlations are in-dicators that both total unique users and total emails sent may affect the absolute lev-els of knowledge transfers within the organizational network. If this is the case will be shown in the regression results, hence the causality is not confirmed yet.
Table III
CORRELATION MATRIX WITH CHOOSEN VARIABLES
This table display the correlation matrix of the independent, dependent and control variables that are used in the regression models. A light grey background indicates a strong correlation between variables. The correlation be-tween the dependent variables and other variables are solely here to be observed and not interpreted. 5 moderate and 1 strong correlations exist.
ROA Log Debt ratio Log Labor Log Total unique users Log Total emails sent % Strong ties Knowledge transfer Scandal
ROA 1,
Log Debt ratio 0,46 1,
Log Labor 0,44 0,59 1,
Log Total unique users -0,44 -0,39 -0,27 1,
Log Total emails sent -0,43 -0,42 -0,49 0,99 1,
% Strong ties 0,01 -0,10 -0,04 0,17 0,17 1,
Knowledge transfer -0,55 -0,55 -0,55 0,60 0,62 -0,12 1,
Scandal -0,63 -0,24 -0,21 0,35 0,35 -0,21 0,73 1,
For each regression model the pattern is the same apart from a shift of the dependent variable, i.e. both dependent variables is not included in any model. This implies that each model is run twice, but with a different dependent variable each time (X). More control variables are added to the models as they progress.
Table IV
REGRESSION MODELS
This table represents the regression models used for the dependent variables percentage strong ties and knowledge transfer. An X indicates that the variable will be used in the model presented. An (X) indicates that the variable shifts in each model, i.e. no model include two (X) when it is run. The dependent variables are stated with the in-dependent variable scandal and the others control variables.
Model I Model II Model III Model IV Model V Model VI
% Strong ties (X) (X) (X) (X) (X) (X)
Knowledge transfer (X) (X) (X) (X) (X) (X)
Scandal X X X X X X
Log Total emails sent X X X X
ROA X X X X X
Log Efficiency X X X X
