How leader’s transformational leadership affect team members’
propensity to innovate with IT in virtual teams
Thijs Wijnmaalen
a,∗University of Groningen, Faculty of Economics and Business, P.O. Box 800, Groningen 9700 AV, The Netherlands
Dr. U. Yeliz Eseryel
c,, G. Balau
1,a
Student MSc BA Strategic Innovation Management
b
Master Thesis Supervisor
c
Master Thesis Second Supervisor
Abstract
This study investigates whether transformational leadership can influence people to be innovative with IT in the context of virtual teams. Firms work increasingly with information technologies and have to keep up with advancements in technology in order to survive in the global market. With the advent of new technology-enabled organizational forms, people work increasingly in virtual teams to accom- plish their tasks. Especially for firms that employ people working in such teams, it is beneficial to have employees that are innovative with IT. The degree to which people are innovating with IT is subject to considerable differences. A questionnaire was sent to 300 firms in the Netherlands that use virtual teams. The findings did not reveal a significant relationship between the use of transformational leader- ship and team member’s innovativeness with IT, which is not in line with expectations and in contrast with prior similar studies. From a theoretical perspective, the findings helps to understand that the nature of the relationship between transformational leadership and innovativeness is contingent with the research and business context.. From a practical perspective, the findings may help firms that want their employees to excel in their innovativeness with IT, to not depend too much on transformational leadership, as it is shown that the effectiveness of this leadership style varies with the situation.
Keywords: Virtual Teams, Transformational Leadership, Information Technology, Innovativeness with IT
∗
Tel.: +31 63374 4426
∗∗
Tel.: +31 50363 5159 Tel.: +31 50363 3453
Email addresses: m.h.wijnmaalen@student.rug.nl (Thijs Wijnmaalen), yeliz@eseryel.com (Dr. U. Yeliz Eseryel),
g.balau@rug.nl (G. Balau)
1. Introduction
With the advent of new technology-enabled organizational forms, people work increasingly in vir- tual teams to accomplish their tasks (O’Hara-Devereaux and Johansen, 1994). A virtual team is a group of individuals who work across time, space and organizational boundaries with links strengthened by webs of communication technology. People collaborate with others who are geographically dispersed in order to survive and excel in the global market (Lipnack, 1997). These virtual teams benefit greatly from information technologies (also referred to as IT) and the rapid rise in usage of the internet in mak- ing this way of working possible (Townsend, DeMarie and Hendrickson, 1998). Recently, it is argued that the role of leadership within those teams might be better viewed as a distributive effort opposed to a task of a single person (Zigurs, 2003). Teams in which leadership is seen as a distributive effort are called ‘self-managed (virtual) teams’ (Shrednick, Stutt and Weiss, 1992). Moreover, recent evidence show that high performing self-managed virtual teams displayed significantly more leadership behav- iors over time compared to their low performing counterparts (Carte, Chidambaram and Becker, 2006).
More than face-to-face teams, also known as co-located teams, virtual teams function by virtue of in- formation technology. Information technology is defined as the application of computers and telecom- munications equipment to store, retrieve, transmit and manipulate data (Daintith, 2009). Continual in- formation technology innovation is essential for businesses to remain competitive, and can be viewed both from an internal as well as an external viewpoint. First, it enables an organization for swift or- ganizational responses to changing environmental demands (Brown and Eisenhardt, 1997). Second, the effectiveness of a virtual team depends on the ability and willingness of the team members to use information technology, and innovate with new technologies to become more productive (Wang, Li and Hsieh, 2011). Researchers in the field of IS gradually move away from the traditional static, discrete view of IT to models highlighting the dynamic, bottom-up process where individual-level IT use behav- iors and interactions collaboratively create collective-level IT use (Barki, Titah and Boffo, 2007). Given the significance of bottom-up IT use, IS research does not have accumulated rich, robust, empirical findings regarding the mechanisms of these processes (Nan, 2011).
Earlier work shows that the selection and use of IT by ‘leaders’ influences followers attitude in several aspects. Furthermore, Todd, McKeen and Gallupe states that “(...) the perception exists that a successful IS (information systems) professional blends technical knowledge with a sound under- standing of the business while commanding effective interpersonal skills” (Todd et al., 1995, pp. 1–2).
As Börekçi (2009) discovered, leader’s IT usage influences follower’s positive work attitudes, such as
loyalty and hard work. Lewis, Agarwal and Sambamurthy (2003) looked at the pre-adoption stage and
suggest that beliefs about technology use is influenced, apart from individual factors such as personal
innovativeness and self-efficacy, by top management commitment.
Apart from the selection of IT by leaders, the type of leadership style may as well influence the extent to which they motivate subordinates, or team members, to be innovative with IT. Following the work of Burns (1978) on leadership styles, ‘transformational leadership’ most closely resembles the current dynamic, bottom-up processes of IT innovation where individuals are encouraged to go beyond their self-interests for the good of the group (Hater and Bass, 1988).
Summarizing, past literature shows that on the one hand, the use of IT by leaders influences people’s behavior and on the other hand, that people have different attitudes and behaviors toward the use of personal IT. And as people work increasingly in virtual teams, it is interesting to conduct research in this context (Bell and Kozlowski, 2002; Powell, Piccoli and Ives, 2004). Past researchers have exam- ined the role of transformational leadership in various contexts, e.g., organizational change (Eisenbach, Watson and Pillai, 1999), organizational and unit performance (Bass, Avolio, Jung, Berson et al., 2003;
Lim and Ployhart, 2004), organizational culture (Bass and Avolio, 1993), creativity (Shin and Zhou, 2003), education (Leithwood and Jantzi, 2000; Hallinger, 2003) and knowledge sharing (Bryant, 2003).
But apart from some preliminary findings on enhancing organizational innovation (Jung, Chow and Wu, 2003), I could not find existing research that specifically looks at transformational leadership in combination with personal innovativeness in the domain of IT.
Hence, the research gap this study addresses identifies the role of transformational leadership within the context of a virtual team. Also, the study explicitly identifies personal creativity and efficacy with respect to IT: why do some team members go above and beyond and come up with innovative ideas that helps the team perform better? The research question for this study is: how does a leader’s trans- formational leadership affects team members’ propensity to innovate with IT?
This paper is structured as follows. First an overview of existing research relevant for the various topics that this study relies upon is given: post-adoption IT innovativeness, transformational leadership and virtual teams. This argument will result in several hypotheses that will be tested by empirical data which is gathered by means of a questionnaire, as will be described in the methodology section. Finally, implications for research and practice are given.
2. Literature review and hypotheses
2.1. Research context: virtual teams
As stated in the introduction, with the advent of new technology-enabled organizational forms,
people work increasingly in virtual teams to accomplish their tasks (O’Hara-Devereaux and Johansen,
1994). This new organizational form is a relatively new field of research, and getting attention because
working in a virtual team imposes challenges and difficulties, which people often try to overcome using innovative IT. This research is carried out in a virtual team context.
A virtual team is an evolutionary form of a network organization (Miles and Snow, 1986) enabled by advances in ICT (Davidow and Malone, 1992; Jarvenpaa and Ives, 1994). Virtual teams promise the flexibility, lower costs, and improved resource utilization necessary to meet the ever-changing task re- quirements in highly turbulent and dynamic global business environments. (Jarvenpaa and Leidner, 1998; Mowshowitz, 1997; Snow, Snell, Davison and Hambrick, 1996) The term ‘virtual’ implies perme- able interfaces and boundaries and the team can both form and dissolve rapidly based on needs in the market. Furthermore, virtual teams operate independent of time, space and culture. (Kristof, Brown, Sims, Smith et al., 1995) Recent studies on virtual teams look on a wide array of characteristics, and in most cases study antecedents for successful virtual teams, each looking at different stages in the virtual team’s lifetime (inputs, processes, outputs) (Powell et al., 2004). Although most of the research directions highlighted in this review are not very relevant for this study—in this paper I use virtual teams only as a contextual variable and do not study virtual teams in itself—the recent developments around technical expertise and training need attention. First, the team benefits in terms of both satis- faction and performance from a good technical expertise of team members and the ability to cope with technical problems (Kayworth and Leidner, 2000; Van Ryssen and Godar, 2000). Second, virtual teams that are composed of members with diverse technology skills may experience conflict when members are unable to resolve differences and compromise on the use of a specific skill during task completion (Sarker and Sahay, 2002). When team members thus have the (IT) skills needed to complete the task or even innovate with IT, differences in opinion about which technology to use (e.g., using Oracle or MS Access; using Dropbox or Google Drive) can hinder team performance.
Despite the increasing use of virtual teams (McDonough III, Kahnb and Barczaka, 2001), they are no silver bullet for better team performance: creating and sustaining a coherent connection among distributed individuals occupying a shared electronic space presents a major challenge (Schultze and Orlikowski, 2001), but in this research context these teams are interesting because they cannot exists without the extensive use of IT and therefore its performance depends very much on the effective and innovative use of it.
2.2. Innovativeness with IT
There are many theories about adopting new technology by people (Davis, 1985; Venkatesh and
Davis, 2000). Several models have been proposed to conceptualize the complex behavioral and social
process by which individuals adopt new information technologies, beginning with the theory of Rogers
(2010, first ed. 1962). In a work environment, information technology makes up a large part of daily
routines and people’s effectiveness depends on the ability to work with technologies. In this study however, it is more useful to move a step further and look at the antecedents of post-adoption behavior of information technology, like Ahuja and Thatcher (2005) with their theory of trying. Understanding post-adoption behavior has emerged as an important issue in information systems (IS) research lately (Saeed and Abdinnour-Helm, 2008; Kim, 2009). Saga and Zmud (1993) refer to this stage as the infusion stage: after initial learning and acceptance decisions, employees try to innovate with IT in order to meet existing (but unmet) needs.
There are several related constructs developed in this field that each take a different perspective.
First, there is the construct trying to innovate with IT, based on the theory of trying (TT) introduced in the seminal article of Bagozzi and Warshaw (1990), which was later implemented in a framework that explains a user’s goal of finding new uses of existing workplace information technologies (Bagozzi, Davis and Warshaw, 1992; Ahuja and Thatcher, 2005). Second, Davis (1989) defined intent to use IT as the strength of a person’s intention to use IT. This is operationalized as an attitude that varies with beliefs about a specific technology. Third, the intent to explore IT is developed by Nambisan, Agarwal and Tanniru (1999) and is defined as a user’s willingness and purpose to explore a new technology and find potential uses. This is an attitude that is influenced by beliefs about IT.
These constructs all touch upon the construct that will be used in this study. In this study however,
more weight is given to creativity and curiosity in innovating with IT, rather than immutable traits and
intentions to use IT, as the aforementioned theories do. Furthermore, as an intention or attempt to use
an IT may not be the best predictor of usage behavior in the post-adoptive context (Jasperson, Carter
and Zmud, 2005; Kim and Malhotra, 2005). Some people just use the technologies that are given to
them, but others come up with ideas to either use the technology in more innovative ways or propose
to use some other innovative technology in order to help the team do a better job. Agarwal and Prasad
(1998) serve this definition best with the construct personal innovativeness with IT. The concept of
personal innovativeness is grounded in theory of innovation diffusion research such as (Rogers, 2010)
but more specifically in marketing (e.g., Midgley and Dowling, 1978; Flynn and Goldsmith, 1993). Fol-
lowing these theories, people are characterized as “innovative” if they are early to adopt an innovation
and the methods are a means to separate the market in innovators and non-innovators. Following the
importance to conceptually and operationally draw a distinction between global innovativeness and
domain specific innovativeness (Flynn and Goldsmith, 1993), Agarwal and Prasad (1998) define PI in
the domain of information technology, henceforth PIIT. This construct is defined as the willingness of
an individual to try out any new information technology. Global innovativeness exhibits low predic-
tive power whereas domain-specific innovativeness, on the other hand, is posited to exhibit significant
influence on behaviors within a narrow domain of activity. Moreover, this innovativeness has been suggested to be measured directly via self-report (Flynn and Goldsmith, 1993). Leaders ICT usages influence on followers positive work attitudes via perceived leader-follower relations To include the creative and innovative behaviors, a recent study by Wang et al. (2011) extended the construct PIIT by introducing a construct called (propensity to) Innovate with IT (IwIT). IwIT embodies the generation and implementation of individual users’ creative ideas in the form of IT usage behaviors. Specifically, the concept of IwIT describes a user’s applying IT in novel ways to support his or her task performance (Wang et al., 2011).
Selection and use of IT positively impacts the team performance and thus the team outcomes (Robert and Dennis, 2005). Summarizing, it can be concluded that creativity with IT is beneficial for the firm.
The next question is whether or not team leaders are able to influence and encourage this creative post-implementation usage behavior using transformational leadership.
2.3. Transformational IT Leadership
Given that the innovative use of IT is positively linked to team outcomes, it is useful to investigate how team leaders can effectively influence this behavior. In the literature, two broad types of leadership styles are distinguished in the seminal works of Burns (1978) and Bass (1985): transactional leadership and transformational leadership (Bycio, Hackett and Allen, 1995). These ideas of organizational man- agement developed by Burns (1978) were applied by Bass (1985).
Transactional leadership involves motivating and directing people (team members) primarily through appealing to their own self-interest: the theory is founded on the idea that leader-follower relations are based on a series of exchanges or implicit bargains between leaders and followers (Hartog, Muijen and Koopman, 1997). Followers receive certain valued outcomes (e.g. wages, prestige) when they act ac- cording to their leader’s wishes (Hartog et al., 1997).
Transformational leadership, on the other hand, tries to inspire people to do more than expected; it involves shifts in the needs, beliefs, and the values of followers (Kuhnert and Lewis, 1987). Through a strong personal identification with the leader, joining in a shared vision of the future, transformational leaders broaden and elevate the interests of followers, generate awareness and acceptance among the followers of the purposes and mission of the group and motivate followers to go beyond their self- interests for the good of the group (Hater and Bass, 1988). Transformational leaders
attempt and succeed in raising colleagues, subordinates, followers, clients, or constituencies
to a greater awareness about the issues of consequence. This heightening of awareness re-
quires a leader with vision, self confidence, and inner strength to argue successfully for what
he [sic] sees is right or good, not for what is popular or is acceptable according to established wisdom of the time (Bass, 1985, p. 17)
For IT innovativeness, this type of leadership enables people to be more innovative and go beyond the basic tasks their leaders expect them to finish. Furthermore, transformational leadership is more effective at creating and sharing knowledge at the individual and group levels, while transactional leadership is more effective at exploiting the knowledge at the organizational level (Bryant, 2003).
2.4. Transformational IT Leadership and Innovativeness with IT
Now that the characteristics of transformational leadership are clear, it is useful to look at the current state of IT in organizations and observe how this type of leadership can motivate people to be innovative with IT.
As stated in the introduction, over the last decades IT is shifting from a administrative role in the background of organizations to a more strategic one and a source of competitive advantage (Henderson and Venkatraman, 1993). To leverage its potential and business value, it is important to understand what influences the usage and degree of innovativeness with IT for employees. Armstrong and Sam- bamurthy (1999) looked at the role of (leadership of) senior managers on IT assimilation, and find that the intensity of the relationship between CIO’s interactions with the top management team and their level of IT and business knowledge is much stronger in firms that articulate a transformational IT vi- sion (Armstrong and Sambamurthy, 1999). These are interesting insights from a firm-level perspective, and they concentrate on senior business executives responsible for key business or functional areas.
However, the actual work is being done in teams at a much lower level in the organization which have their own local leaders, albeit facilitated by IT from the firms’ management. In practice, IT use is of- ten enacted through self-orchestrated interaction among users and technologies rather than dictated by policies or top-level managerial intentions (Barki et al., 2007). Research is moving in the direction of bottom-up IT use (Nan, 2011).
Organizational innovation and creativity is beneficial for firms (Woodman, Sawyer and Griffin,
1993). If employees exhibit innovative behavior, firms can derive a competitive advantage from it (Hult,
Hurley and Knight, 2004). In the context of a virtual team, firms benefit especially from innovation in
the domain of IT because virtual teams rely largely on IT to function (Majchrzak, Rice, Malhotra, King
and Ba, 2000; Griffith, Sawyer and Neale, 2003). Subsequently, firms have an interest in understanding
how to motivate people to exhibit this creative, innovative behavior in this domain. Previous research
has looked at the role of transformational leadership in enhancing organizational innovation. Jung
et al. (2003) found support for a direct and positive relationship between transformational leadership
and organizational innovation, while Gumusluoglu and Ilsev (2009) also considered creative, innova- tive behavior on the individual level. Contrary to these findings, Jaskyte (2004, p. 162) did not find this direct relationship. Transformational leadership is also linked to performance: Howell and Avo- lio (1993) found that transformational leadership is associated with a higher internal locus of control and significantly and positively predicted business-unit performance, while transactional leadership is negatively related to business-unit performance.
It is clear that a positive relationship between transformational leadership and innovative behavior, on various organizational levels can be found. However, it is not clear inasmuch these findings can be generalized in other contexts; i.e. none of these previous findings explicitly considered a virtual team context nor the way team members used IT in possibly novel ways.
Based on this gap in the literature and the potential for firms for innovative, bottom-up use of IT, I propose these hypotheses.
Hypotheses 1 (for TFL
IST, …, TFL
FAG)
Leader’s transformational IT leadership regarding «dimension»
1, will be positively related to team members’ tendency to display Innovativeness with IT.
This first hypothesis, is divided into six separate sub-hypothesis, given the fact that transformational leadership is composed of six dimensions. Each of these distinct dimensions should have a positive relation toward user’s Innovativeness with IT.
2.4.1. IT self-efficacy
The hypothesized relationship between transformational leadership and the team member’s Inno- vativeness with IT is possibly moderated by personal factors. In the extant literature on technology adoption and user interaction with technology, personal innovativeness and self-efficacy (Compeau and Higgins, 1995) have conventionally been used as behavioral intention determinants (Kwon, Choi and Kim, 2007, e.g., Table 1). Related studies by Hu, Clark and Ma; Ong, Lai and Wang; Vijayasarathy have shown that computer (IT) self-efficacy is positively related to perceived ease-of-use (Hu et al., 2003; Ong et al., 2004), perceived usefulness (Ong et al., 2004) and intention to use (Hu et al., 2003;
Vijayasarathy, 2004). Hence, people who already have high computer self-efficacy might also tend to display innovativeness with IT compared to people with low computer self-efficacy, who in turn might benefit relatively more from transformational leadership.
1
‘Intellectual Stimulation’ (IST), ‘Individual Support’ (ISU), ‘Providing an Appropriate Model’ (PAM), ‘Identifying and Artic-
ulating a Vision’ (AV), ‘High Performance Expectation’ (HPE) and ‘Fostering the Acceptance of Group Goals’ (FAG)
Therefore, in this study IT self-efficacy is used as a moderating variable between the relationship of the two constructs just explained.
Self-efficacy was first introduced by Bandura (1977) grounded in social learning theory (SLT), and is later applied to a variety of disciplines, such as IT. With the advancement of IT, research concen- trated on self-efficacy in these areas (Compeau and Higgins, 1995). Compeau and Higgins (1995) define computer (IT) self-efficacy as “a judgment of one’s capability to use a computer”. Marakas, Mun and Johnson (1998) extend this model by giving a more detailed and isolated explication of the construct.
The construct used by Compeau and Higgins (1995) is what Marakas et al. (1998) refers to as general computer (IT) self-efficacy; as the latter authors identify the former construct as being multilevel. They define general computer self-efficacy as “an individual’s judgment of efficacy across multiple computer application domains” (Marakas et al., 1998).
Self-efficacy should not be confused with self-leadership which is conceptualized as a “comprehen- sive self-influence perspective that concerns leading oneself toward performance of naturally motivat- ing tasks as well as managing oneself to do work that must be done but is not naturally motivating”
(Manz, 1986). Self-leadership in this study is less relevant because transformational leaders should be able to convince both people with high and low levels of self-leadership to do tasks that might not be intrinsically motivating, but nonetheless useful.
Self-efficacy reflects not only an individual’s perception of his or her ability to perform a particular task based on past performance or experience but also forms a critical influence on future intentions.
Self-efficacy is conceptualized as a trait: a relatively stable descriptor of individuals that is invariant across situational considerations. However, other scholars see it as an attitude. [ref] There is a number of related constructs in this field that are sometimes used interchangeably causing confusion, such as computer anxiety, computer attitudes, computer self efficacy and computer experience. A number of large literature reviews and meta analyses have tried to assess the relationships between these con- structs, looking at various aspects from the constructs such as gender effects, i.e. the tendency that females have on average more negative attitudes toward computers than males (Rozell and Gardner III, 2000; Chua, Chen and Wong, 1999; Whitley, 1997). Beckers and Schmidt (2001, 2003) treat computer self efficacy as part of computer anxiety. Levine and Donitsa-Schmidt argue that computer self effi- cacy and computer anxiety are essentially the same thing (Levine and Donitsa-Schmidt, 1998). In a more recent longitudinal study, Marakas, Johnson and Clay conclude that the IT self-efficacy construct is related to, but conceptually different from these other behavioral constructs commonly found in IS research (Marakas et al., 2007).
The original task-specific model derived from Bandura has served as a basis for new task-specific
models, for example to measure the self-efficacy of people in using the Internet (Durndell and Haag, 2002; Hsu and Chiu, 2004), and exploring the difference between general IT self-efficacy and task- specific IT self-efficacy (Marakas et al., 1998; Agarwal, Sambamurthy and Stair, 2000). Because there is no single standardized measure of self-efficacy that is appropriate for all studies, Vispoel and Chen advise researchers to develop new, or to significantly revise and revalidate, existing measures for each study (Vispoel and Chen, 1990). For this study however, where I do not concentrate on a specific technology, there is no choice than to use a general model by Compeau and Higgins or the version by Marakas et al.. The possibility to use the construct in this generic way—acting as a product of a weighted collection of all CSEs accumulated over time—is acknowledged by Bandura (2006) although no empirical evidence clearly establishing the true relationship between the generic and specific forms has yet appeared in the literature (Marakas et al., 2007).
As stated in the beginning of this section, IT self-efficacy influences peoples use of IT can there- fore be of influence in the main hypothesized relationship between transformational leadership and Innovativeness with IT of team members.
Hypothesis 2
The relation between leader’s transformational IT leadership and team members’ tendency to display Innovativeness with IT is moderated by IT self-efficacy.
This yields our research model in figure 1.
Figure 1: Research model
Virtual team member’s innovativeness with IT Virtual team leader’s
transformational leadership
H1
H2 IT self-efficacy (CSE)
3. Data and Methods
Since this study is in the field of information systems (IS), the quest for relevant literature to build
the literature review always began in a search in the top eight journals in this field—the so-called ‘basket
Table 1: Basket of eight
European Journal of Information Systems Information Systems Journal
Information Systems Research Journal of AIS
Journal of Information Technology Journal of MIS
Journal of Strategic Information Systems MIS Quarterly
of eight’—as constructed by the Senior Scholars Consortium of Association for Information Systems.
These top journals are listed in Table 1.
Guided by papers from these top journals, adjacent, more in-depth or more recent literature on a topic was found following references or via searches in the scientific databases of EBSCOhost, Sci- enceDirect and Google Scholar. As a novice in this field, this method ensured both that I worked with high quality research without having to know names of authoritative authors in the field, and served as a starting point to identify papers that developed the constructs that I use in this study. Search terms in- cluded “transformational leadership”, “self-efficacy”, “virtual teams”, “leadership styles”, “technology adoption”, among others. Most searches were entered in different databases to prevent missing impor- tant papers that might not be indexed. As the most influential people within a research field were recog- nized—by the number of citations their papers received, or because they coined a new term—searches started by looking at follow-up papers that cited this initial work.
3.1. Survey Data
Data were drawn from the electronic distribution of a questionnaire during April 2013 and May 2013 to firms in the Netherlands that use virtual teams. Collection of the data was performed by a group of four students (myself included) all working on their master’ thesis in parallel, having roughly the same topic.
A list of firms were collected from data from the Chamber of Commerce, lists of ‘innovative’ firms in the Netherlands and personal contacts within firms of which we knew worked with virtual teams.
For each construct, used in one or more of the theses of the students, we identified a relevant measures in the literature and included the instrument in the questionnaire. The questionnaire could not be too extensive, as it would increase the risk of a lower response rate. Besides data drawn from this questionnaire, there was no additional data gathered.
Several general questions were asked, such as whether or not the firm used virtual teams and if
the person filling out the questionnaire currently participated or had participated in the past in such a
team. Furthermore, concepts that might not be known or might cause confusion, were explained in the questionnaire. In this thesis, the names of the participating firms are not disclosed because identifying participating firms is both not relevant for the research and firms might not want to leak sensitive infor- mation to the public due to privacy concerns. There were no interviews or other qualitative methods used for this study because there is already qualitative research done on the several separate constructs and in this study the developed hypotheses could be tested better with a high number of participants.
This was also the best way to get good quality data in a limited amount of time, because of the tight schedule in which the research had to be carried out.
We contacted 300 people working for firms in the Netherlands, or people working for Dutch firms overseas—as is often the case when working in a virtual team. In most cases, only one person per firm filled out the questionnaire. To increase the response rate, every firm was contacted by phone first with subsequent calls or e-mails a week after we contacted them. From past experiences we knew that just mailing a questionnaire would yield a very low response rate. If a firm indicated to cooperate but had not filled out the questionnaire after a week, a reminder was sent via e-mail. Responses with missing data as well as doubtful or contradictory answers that could not be clarified by follow-up telephone calls were removed from the sample. A total of 108 valid responses (in fact, a total of 166 respondents started the survey, but only this number completely filled out all the questions) were collected from the questionnaire, yielding a response rate of 36 %.
In Table 2 descriptive statistics are listed that provide insights in the sample. Both the distribution of age and the ratio of male to female are balanced, having only slightly more men than women. From the 106 of 108 people who filled in their age, their average age is 38 with a minimum of 22 and a maximum of 66 having a standard deviation of 11 years. It can be drawn from the table, that most respondents entered their demographic data and industry in which they work.
3.2. Variables
The variables from the conceptual model are listed in Table 3 along with measures used, which were
all derived from existing literature. They have all been empirically tested in previous studies, in order
to ensure construct validity. The questionnaire was built around these measures and are mostly seven-
point Likert scale, reaching from 1 (strongly disagree) to 7 (strongly agree). The number of questions
for each construct is also listed in the table. Besides offering the questionnaire in English, all questions
were also translated to Dutch, as this increased the response rate since the majority of the participants
had Dutch as their native language. The original questions that we adapted from the sources mentioned,
can be found in Appendix A. Please also note that I only list the variables relevant for this study, the
questionnaire we sent out contained additional questions for constructs used by my colleague students,
Table 2: Descriptive statistics of sample
Industry Frequency Percent
General business services 10 9.3
Transport 3 2.8
Bank and insurance 3 2.8
Governmental or law organizations 3 2.8
Paper and print 1 .9
Culture, sports and leisure 1 .9 Fuel, plastics, chemical industry 1 .9
Education 15 13.9
Construction 1 .9
Fast Moving Consumer Goods 10 9.3
Development or NGOs 2 1.9
Retail and wholesale 3 2.8
Health care 16 14.8
Agriculture, forestry and fishing 1 .9
Other, services 5 4.6
Hospitality 1 .9
IT industry 14 13.0
Other 16 14.8
(missing) 2 1.9
Total 108 100.0
Sex
Male 61 56.5
Female 47 43.5
Total 108 100.0
Table 3: Variables and measures
Variable Measure No. of Questions
Dependent
Team member’s Innovativeness with IT
Agarwal and Prasad (1998);
Ahuja and Thatcher (2005)
4 + 2 Independent
Transformational IT Leadership Podsakoff et al. (1990, 1996) 24 Moderator
IT Self-efficacy Compeau and Higgins (1995) 10
for their theses.
The firms that filled out the questionnaire all use or have used virtual teams to some extent, because this research explicitly is about leadership and IT in virtual teams. It would have been interesting to empirically test the relationship both in light of virtual teams compared to co-located teams, but due to time constraints this was unfortunately not feasible.
3.2.1. Dependent Variable
The dependent variable that we are considering is Innovativeness with IT. This construct is based on the relatively new construct of Wang et al. (2011), innovative with IT and describes a user’s applying IT in novel ways to support his or her task performance and is based on the theory of Ahuja and Thatcher (2005). It is used in conjunction with the highly related Personal innovativeness with IT (Agarwal and Prasad, 1998) that is defined as a user’s goal of finding new ways of using existing IT and operational- ized as a goal that is influenced by beliefs about the context or personal ability (Ahuja and Thatcher, 2005). The construct refers to the post-implementation stage, in which a users’ familiarity with the in- stalled IT enables them to partake in innovative use that probably could not be identified at the initial acceptance stage (Jasperson et al., 2005).
It is closely related to the moderating variable IT self-efficacy, also known as computer self-efficacy (CSE), which examines the role of individuals’ beliefs about their abilities to competently use computers (Compeau and Higgins, 1995; Marakas et al., 1998; Bandura, 1977). Both are individual differences regarding IT use. However, in this study a distinction is made between these constructs to proof that those are indeed two distinct constructs. Whereas IT self-efficacy is often discussed in relation to the adoption stage of new IT and narrowly defined, ITI explicitly considers post-implementation usage behavior that puts new ideas into action and is influenced by theory of creativity (Wang et al., 2011;
Bagozzi et al., 1992).
3.2.2. Independent Variable
For the independent variable ‘transformational leadership’ we use the Multifactor Leadership Ques- tionnaire that was build by Bass and Avolio (Bass and Avolio, 1995, 1997). As this instrument only considers general transformational leadership behavior and hence is not specifically tailored to iden- tify IT-leadership, it was extended by such factors using the instrument developed by Bassellier and Benbasat.
3.2.3. Preliminary Data Analysis
For each construct used in this study I performed reliability tests on the data gathered from the ques-
tionnaire in order to test the scale’s internal consistency (Nunnally, 2010). When assessing reliability
Table 4: Inter-item correlation matrix for Innovativeness with IT
PIIT1 PIIT2 PIIT3 PIIT4 IWIT1 IWIT2 PIIT1 1.000
PIIT2 .621 1.000
PIIT3 .453 .543 1.000
PIIT4 .622 .733 .524 1.000
IWIT1 .382 .514 .362 .519 1.000
IWIT2 .463 .473 .468 .543 .733 1.000
and validity, a composite reliability of .70 of Cronbach’s alpha or greater is considered acceptable for research (Fornell and Larcker, 1981). The loadings of the scale Innovativeness with IT, that we com- posed ourselves of four items from Agarwal and Prasad (1998) and two from Ahuja and Thatcher (2005) are listed in Table 4. The high Cronbach’s alpha for this scale suggests a very good internal consistency reliability and correspondents with values reported by the original authors, respectively .84 and .78/.87 (male/female) for PIIT and IWIT (Agarwal and Prasad, 1998; Ahuja and Thatcher, 2005). The item PIIT3 scores relatively low compared to the other items, hence may be a candidate for removal. Removing this item however, would only slightly decrease the final Cronbach’s alpha for this construct to .864.
In fact, removing any of the other items would decrease the value. To further test the reliability of this new scale, an inter-item correlation matrix is displayed in Table 4 to see how the combined items score against each other. As there are no negative values in the inter-item matrix, indicating that the items are measuring the same underlying characteristic and looking at these statistics we can safely assume that the construct is reliable.
The items in the scale for IT self-efficacy are separated in a “yes” or “no” response combined with an confidence level expressed in a 7 point Likert-scale, see Appendix Appendix A. In order to perform statistical tests on this scale, I transformed the scale to a 8 point Likert-scale thereby removing the “yes”
responses while leaving the confidence levels intact. For every respondent who answered “no”, I scored
them a 1 on the remaining 8 point Likert-scale
2. Unfortunately, this method causes the scale to be less
normal than would otherwise be the case, thereby somewhat violating the normality assumption of the
statistical test used. Some respondents, less than 5%, did not understand the question and selected
either “yes” or “no” without providing confidence levels. These were filled in with moderate values.
Table 5: Total Variance Explained with Eigenvalues > .8 for Components
Innovativeness with IT
Component Total % of Variance Cumulative %
1 3.67 61.21 61.21
2 .84 13.94 75.15
4. Results
4.1. Principal Component Analysis
I performed principal component analysis
3(PCA) to get an empirical summary of the data set (Tabach- nick and Fidell, 2007) and to identify components and subgroups of items in the variables.
Using PCA it is also possible to check whether the components identified by the original authors of the scales also emerge in this empirical data. Although Tabachnick and Fidell in their review suggest a sample size of at least 300 cases, Stevens (1996) weakens this requirement and suggests that the sample size have been reducing over the years as more research has been done on the topic. Having large marker loadings, as is the case, also permits a lower sample size to be sufficient (Tabachnick and Fidell, 2007, p. 613).
The suitability of data was assessed prior to performing PCA. The Kaiser-Meyer-Olkin value was .81, .91, .93, for ITI, ITSE and TFL respectively, exceeding the recommended value of .6 and Bartlett’s Test of Sphericity was significant for all variables, supporting the factorability of the correlation matrix.
For Innnovativeness with IT, 61% of the variance can be explained by just a single component, while 75% can be explained by adding a second component, see Table 6 for the PCA of this variable. To aid in the interpretation of these two components, Oblimin rotation was performed. The rotated solution revealed the presence of a simple structure (Thurstone, 1947), with both components showing a num- ber of strong loadings and all variables loading substantially on only a single component. Following Kaiser’s criterion to split components based on the components with eigenvalues of 1 or higher, a single component sufficed in this case. However, as this variable is composed of two distinct instruments, it makes sense to also look at the second component which has an eigenvalue of .84. Analysis reveals that two components indeed correspond with the two scales that were used together to create this variable,
2
I am grateful to Mr van der Bij who provided us with this advice in doing statistical analyses on this non-standard type of scale
3
This parametric technique can in principle only be performed on continuous data. Since Likert scales are used in our
questionnaire, this produces ordinal data despite being made up of numbers. However, if certain assumptions about skewness,
number of categories, etc. are met, it is possible to find true parameter values in these techniques with Likert scale data (Lubke and
Muthén, 2004). Furthermore, in the research field this study is performed, management and information systems, it is common
practice to treat Likert scales as interval data in statistical analyses.
Table 6: Component and Pattern Matrix for PCA with Oblimin Rotation with Kaiser Normalization of Two Factors Solution of Innovativeness with IT
Component Coefficients Pattern Coefficients
Item 1. 2. 1. 2.
PIIT1 .852 .883
PIIT2 .839 .858
PIIT4 . 782 . 489 . 795
PIIT3 . 759 −.342 . 738
IWIT1 . 748 . 566 . 959
IWIT2 .704 .883
Note: Only loadings higher than .3 are displayed.
four questions from Personal Innovativeness with IT (Agarwal and Prasad, 1998) and two from Innova- tive with IT (Ahuja and Thatcher, 2005). This indicates that despite the fact that the two components are very similar—see the Components Coefficients column in Table 6—it is still possible to extract two distinct components, which therefore slightly supports the theory in Wang et al. (2011). As a result of this minor difference in components, the low eigenvalue of the second component and the way the research was set up, the distinction is not being used in the further analyses.
The IT self-efficacy variable is an existing tested and validated scale composed of 10 items. There is no need to perform factor analysis on this variable, however to exclude the possibility that people filled out the items in a atypical or highly deviating manner a factor analysis was performed. This results in a explained variance of 59.8 by a single component with an eigenvalue of 5.98. There was no unusual data found.
For Transformational Leadership, an initial PCA revealed a component matrix with only three com- ponents, with almost all items loading above .5 on the first component, half of them loading slightly above .3 on the second component and only 3 or 4 items loading above .3 on the last two components.
However, for this construct a closed approach should be followed. Since I used an existing—tested and
validated—instrument which explicitly contains six dimensions and for which I proposed six distinct
hypotheses, I performed a factor analysis with a fixed amount of six dimensions to see whether or not
those six dimensions emerged in the loadings. If those dimensions did not emerged in the data, I had
no choice but to continue using the scale as a whole thereby losing its richness. The results of the factor
analysis, can be seen in Table 7. Half of the six dimensions continue to be based on three items or more,
while the other three dimensions are now based on a single item after factor analysis.
Table 7: Rotated Component Matrix for PCA with Varimax Rotation with Kaiser Normalization of Six Dimensions Solution of Transformational Leadership
Item Pattern Coefficients
1. 2. 3. 4. 5. 6.
TFL
FAG4.862 TFL
FAG3.813 TFL
FAG1.791 TFL
FAG2.741
TFL
HPE2.843
TFL
HPE3.797
TFL
HPE1.773
TFL
IST3.878
TFL
IST4.799
TFL
IST5.780
TFL
PAM3.913
TFL
ISU1a.954
TFL
AV1.810
Note: Only loadings higher than .4 are displayed.
a.) Originally a reversed item.
Table 8: Means, Standard Deviations and Correlations of Model Variables
a,bexcluding Moderation Ef- fects
Correlations
Mean S.D. TFL
FAGTFL
HPETFL
ISTTFL
PAMTFL
ISUTFL
AVITI ITSE TFL
FAG4.67 1.43 . 939
TFL
HPE4.79 1.36 . 562
**. 902
TFL
IST4.53 1.31 . 512
**. 543
**. 893
TFL
PAM4.17 1.45 .471
**.309
**.360
**1.000
TFL
ISU4.47 1.78 .405
**.072 .169 .154 1.000
TFL
AV4.53 1.53 .482
**.483
**.439
**.390
**.219
*1.000
ITI 4.10 1.32 . 134 . 126 . 127 . 029 −.092 . 127 . 871
ITSE 4.93 1.94 . 208
*. 078 . 157 . 144 . 002 . 054 . 053 . 923
Notes:
∗p < .05,
∗∗p < .01
a. See for a list of all abbreviations, footnote 1.
b. Diagonal figures are Cronbach’s alpha values for composite scales.
Note: For three dimensions of TFL, a Cronbach’s alpha value of 1 is displayed because as a result of the factor analysis, the
variable is based on a single item.
4.2. Correlations
The correlation between the variables are listed in Table 8. Using this first correlation matrix no significant correlation can be found between the independent and dependent variables, because the correlations are too small (Cohen, 1988). Therefore, at this point it cannot be stated that people who ex- perience transformational leadership in their team, have more or less Innovativeness with IT. A possible explanation for this disappointing result could of course be that there is no relation between these two variables, in any way. Another plausible explanation for this result is that one or more of the underlying assumptions to conduct the Pearson correlation method are violated. For example, the method requires the variables to be normally distributed with all of them having the same variance, a property known as homogeneity of variance. The variables should share a common covariance matrix Σ
i= Σ
j, ∀i, j, even the highest value of the Pearson product-moment correlation coefficient in this sample, .134, share only 1.8 % of their variance (0.134
2≈ 0.018).
In earlier versions of this thesis, splitting the sample by sex and running the correlations once more, revealed rather surprising results with coefficients largely positive for males and negative for females.
However, as the sample became larger, this effect diminished.
4.3. Hierarchical Multiple Regression
To statistically test the conceptual model, I perform hierarchical multiple regression, because there is a continuous dependent variable and a number of independent variables. In order to perform multiple regression, some assumptions must be met with respect to the sample: assumptions about sample size, multicollinearity and singularity, outliers and normality. The sample size is large enough to perform multiple regression, using the formula N > 50 + 8m where m is the number of independent variables (Tabachnick and Fidell, 2007, p. 123). The highest variance inflation factor (VIF) among the indepen- dent variables (the six TFL dimensions) was 2.8, which is well below the generally accepted cut-off of 10, indicating multicollinearity. The other requirements were tested by inspecting plots of the sam- ple, such as the normal probability plot (p-p) and scatterplot. These plots had ‘normal’ shapes, which indicated that assumptions about normality and outliers were met as well.
In testing the hypothesized relationship between the independent variables (the dimensions of trans- formational leadership) and dependent variable (innovativeness with IT), I control for age and sex. The results of this regression analysis are shown in Table 9.
The first model only includes control variables Sex and Age; the second model includes the inde-
pendent variables (Y
′= b
0+ b
1TFL
1+ ... + b
6TFL
6+ ϵ) and tests the direct effect; the last model also
includes the interaction effects when the moderator ITSE is included in the model. We can derive from
Table 9: Results of Hierarchical Regression Analysis with and without Moderator Influence for Depen- dent Variable Innovativeness with IT
Coefficient Estimate β Variables Model 1 Model 2 Model 3
Age −.14 −.13 −.14
Sex −.31
**−.31
*−.31
**TFL
HPE.05 .10
TFL
IST−.02 −.01
TFL
PAM−.03 −.00
TFL
FAG.13 .09
TFL
ISU−.09 −.11
TFL
AV. 06 . 06
TFL
HPE× ITSE
a. 18
TFL
IST× ITSE . 21
TFL
PAM× ITSE . 09
TFL
FAG× ITSE .15
TFL
ISU× ITSE .05
TFL
AV× ITSE −.47
**F 5.22
**1.66 1.87
R .31 .35 .48
R
2.09 .12 .23
∆ R
2. 03 . 11
Notes:
∗p < .05,
∗∗p < .01
a. ITSE = IT self-efficacy
Figure 2: Interaction effect of IT self-efficacy on relationship between Transformational Leadership (di- mension ‘Articulating a Vision’) and Innovativeness with IT
Transformational Leadership: art. a vision
2.00 .00
-2.00 -4.00
Innovativeness with IT
7.00
6.00
5.00
4.00
3.00
2.00
1.00
High Moderate Low High Moderate Low IT Self- efficacy
Low : R2 Linear = 0.007 Moderate: R2 Linear = 0.054
High: R2 Linear = 0.006