A Shock to the System: The Structural Implications of Enterprise System Technology

Glen Desson Murphy MBus BBus QUTSchool of Management

Faculty of BusinessQueensland University of Technology

Being a report on an investigation submitted as a partial requirement for the degree of Doctor of Philosophy

2006

i

Keywords

Enterprise systems; job design; job characteristics model; social network analysis; organisational change; user acceptance.

ii

Abstract

The last two decades have seen an increasing sophistication in the type of information systems employed by organizations. In particular we have seen the emergence of enterprise systems technology – advanced information technology specifically designed to integrate the vast majority of an organization’s processes and data flows. As the characteristics of ES technology have encroached beyond individual user domains and have become integrated throughout organizations, user acceptance issues have also broadened beyond the individual unit of analysis. At the same time numerous examples can be found both in the trade press and academic literature of organizations wishing to use enterprise systems as a primary driver of widespread organizational change and restructuring.

A fundamental premise of this study is that while it may be intuitively appealing to consider technology as a primary catalyst for organizational change, it neglects to acknowledge the presence of what is referred to as the “duality of structure” (Giddens, 1993). Duality of structure proponents contend that while IT system protocols may to a certain extent determine individual action, human agency can also determine the extent to which the technology is incorporated into everyday operations. The failure of past research to acknowledge the role of individual action and the influence of social context in determining IT usage is considered to be a significant oversight (DeSanctis & Poole, 1994).

Underpinned by the theory of structuration and its notion of duality, a theory of user acceptance is put forward capable of clarifying the process by which users evaluate and react to enterprise systems technology. The thesis reports on an empirical investigation into the relationship between three representations of structure within an organization: the characteristics of ES technology; job design; and social networks. The capacity of ES technology to alter the structural elements of both job design and social networks, and hence form user’s attitudes and behavior towards the system, is the fundamental theoretical premise of the thesis. As such this represents a clear step forward in understanding the implications of ES technology for both users and organizational structure.

Using a longitudinal embedded single case design, this study examines the user acceptance and structural implications of introducing an ES into a large public sector educational institution. A social network and job design perspective was adopted to offer fresh insight into the dynamics of employee reaction to the introduction of ES technology. Five hypotheses support the job design component of the thesis. It was

iii

argued that given the inherent design elements of ES technology, along with the specific intent of the system’s introduction, that users would both anticipate and perceive a decrease in job characteristics following an ES implementation. Further, that the positive relationship between job change and user acceptance would be moderated by the amount of system usage reported by users. Users with a greater exposure to the system were hypothesized to have a far stronger relationship between job change and acceptance than low users. The ramifications of perceived or actual changes to embedded resource exchange networks and subsequent employee reactions to those changes were also considered. Essentially social networks were argued to play a dual role in the user acceptance process, one being a conduit for the facilitation and transfer of user attitudes towards new systems, the other acting as a catalyst for attitude formation towards new systems.

Overall the findings only partially supported four of the eight hypotheses put forward. While users were seen to anticipate an “across the board” decrease in job characteristics at Time 1 following the introduction of an ES, perceived changes in job characteristics at Time 2 were dependant on user hierarchy and the extent of system usage. Those high in formal authority reported an increase in job enrichment following the system’s introduction, while those low in formal authority reported a decrease in overall job enrichment. Usage was also seen to moderate the relationship between job change and user acceptance. At Time 1 low users reported a positive relationship between anticipated changes in meaningfulness and user acceptance. Conversely at Time 1 high users reported a negative relationship between anticipated skill variety levels at Time 2 and user acceptance. Only one job characteristic reported a relationship between usage and user acceptance. Low users reported a positive relationship between changes in task identity and user acceptance. A post-hoc profile of the usage categories indicated that high users were more likely to be a lower hierarchical position than low users.

The positive relationship reported by low users at Time 1 and Time 2 was explained by both the nature of the system, as well as the type and quantity of information received by low users. As senior members of the organization they were considered more likely to receive information that highlighted its attributes in the context of their job roles. The inherent design of ES technology, along with the specific intent it was being introduced, facilitated largely management orientated objectives. Therefore it is unsurprising that low users anticipating an increase in experienced meaningfulness following the introduction of a system that enhanced their job role reported corresponding acceptance levels. In contrast, the negative relationship between anticipated levels of skill variety at Time 2 and perceived ease of use was explained by the affinity that high users were

iv

likely to have with the old system. To high users with a high degree of proficiency associated with a redundant skill set, increased skill variety only represented a steeper learning curve and an increased pressure to adapt to the new system.

The network component of the study also produced mixed results. Of the two networks that were measured over time, only one supported the hypothesized increase in both advice and resource exchange networks over time. Post-hoc analyses indicated that two of the four groups exhibited network change consistent with the hypothesized relationship. Anecdotal reports suggested that contextual elements such as geographical location and managerial policy at a localized level determined the nature of the change for the remaining two groups. The results failed to support the relationship between network change and user acceptance. However, a weak but significant negative relationship between the measure of network efficiency and user acceptance was found. In simple terms users developing an increasingly redundant set of contacts reported higher levels of user acceptance.

In sum, the thesis represents a contribution to enterprise systems, user acceptance and social network literatures. In the first instance the research validates the call by Orlikowski & Iacono (2001) to readily acknowledge the specific nature of the technology under investigation. Despite the growth and saturation of enterprise system types, comparatively little research has been undertaken to examine the user and organizational issues surrounding their implementation. This research has demonstrated the capacity for the inherent design elements of ES technology to have differential effects in terms of job design for different user classifications. This and other findings represent a step forward in understanding the structural and user acceptance implications of this technology, while sign-pointing a number of promising future research avenues.

The job design results, and to a lesser extent the network efficiency results, demonstrate the effect of social context on user acceptance. As such they provide further insight regarding the potential determinants of user acceptance beyond the individual unit of analysis. The findings also indicate an increasing need for user acceptance research to stretch beyond the transitory, short term measures of user acceptance such as perceived ease of use, usefulness, training and computer efficacy.

Finally the thesis contributes to a small, but growing literature examining the role of social networks in the process of organizational change. In particular this thesis has considered in detail, the attitudinal and behavioral consequences of artificially altering established patterns of interaction. As such the study highlights the need to better understand the role of networks not only in the case of facilitating change, but the effect of network change in terms of change intervention success.

v

Table of ContentsChapter One — Introduction ..................................................................................1

1.0 Introduction ......................................................................................................11.1 Background to the Research ..............................................................................21.2 Research Problem and Research Questions ......................................................51.3 Justification of the Research ..............................................................................8

1.3.1 User Acceptance and Social Context ......................................................81.3.2 Information Technology as Structure ......................................................91.3.3 Social Network Analysis and User Acceptance ......................................91.3.4 Implications of Network Change ..........................................................10

1.4 Research Design ..............................................................................................101.5 Key Definitions ............................................................................................... 11

1.5.1 Enterprise Systems Information Technology ........................................121.5.2 Social Network Analysis .......................................................................161.5.3 Structuration Theory and Information Systems Research ....................17

1.6 Outline of the Thesis .......................................................................................191.7 Conclusion .......................................................................................................21

Chapter Two — The Structural Implications of Introducing Enterprise Systems Technology ..............................................................................................................23

2.0 Introduction / Review ......................................................................................232.1 Theoretical Model - Overview ........................................................................242.2 System Driven Changes to Job Design Characteristics ..................................25

2.2.1 A Job Design Approach to Enterprise Systems .....................................262.2.1.1 System Driven Job Characteristics Change ..............................282.2.1.2 System Driven Job Change and User Acceptance ....................35

2.2.2 Summary ...............................................................................................372.3 System Driven Changes to Informal Network Structures ...............................38

2.3.1 Networks and Organizational Change ..................................................402.3.2 Implications of ES Technology Driven Network Shock .......................412.3.3 Actor Centrality ....................................................................................45

2.3.3.1 Central Actors ...........................................................................452.3.3.2 Network Brokers ......................................................................47

2.3.4 Summary ...............................................................................................512.4 User Behavior – Social Networks as a Medium for User Agency ..................52

2.4.1 Attitude Transmission – Brokers and Central Actors ...........................542.4.1.1 Central Players and Network Brokers ......................................55

2.5 Summary .........................................................................................................57

vi

Chapter 3 — Methodology ....................................................................................593.0 Introduction .....................................................................................................593.1 Research Strategy and Design .........................................................................603.2 Research Design ..............................................................................................61

3.2.1 Student Administration System .............................................................623.2.2 Implementation Pilot Sites - City & Country .......................................63

3.3 On-Line Self Administered Questionnaire (SAQ) ..........................................633.3.1. Method Justification ..............................................................................64

3.3.1.1 Advantages of On-line delivery ................................................653.3.2 Method Limitations ...............................................................................663.3.3 Questionnaire Measures ........................................................................67

3.3.3.1 User Acceptance Measures .......................................................673.3.3.2 Job Design Measures ................................................................683.3.3.3 Demographic Variables .............................................................69

3.3.4 Questionnaire Format ............................................................................703.3.4.1 Specific design elements – Web Questionnaire ........................72

3.3.5 Questionnaire Delivery .........................................................................733.3.6 Data Analysis Preparation .....................................................................74

3.4 Social Network Analysis .................................................................................753.4.1 Network Boundary Specification ..........................................................753.4.2 Data Collection Protocol .......................................................................773.4.3. Participant Confidentiality ....................................................................783.4.4 Network Measures ................................................................................793.4.5 Network Data Entry and Storage ..........................................................80

3.5 Summary .........................................................................................................81

Chapter 4 — Social Network Methods, Measures and Application ..................834.0 Introduction .....................................................................................................834.1 Network Data ..................................................................................................84

4.1.1 Data Entry .............................................................................................844.1.2 Network Analysis Software ..................................................................86

4.2 Measuring Network Change ...........................................................................864.2.1 Quadratic Assignment Procedure (QAP) ..............................................874.2.2 Density ..................................................................................................894.2.3 Krackhardt’s External Tie Ratio (E-I index) .........................................91

4.3 Network Change and User Acceptance ...........................................................934.3.1 Network Centrality ................................................................................93

4.3.1.1 Degree Centrality ......................................................................934.3.1.2 Flow Betweenness Centrality ...................................................95

4.3.2 Network Brokers ...................................................................................964.3.2.1 Structural Holes (Redundancy & Constraint) ...........................96

4.3.3 Network Analysis and Multiple Regression .........................................984.4 Summary .......................................................................................................100

vii

Chapter Five — Job Design and ES Technology ..............................................1015.0 Introduction ...................................................................................................1015.1 Sample ..........................................................................................................101

5.1.1 Demographic Summary .....................................................................1025.1.2 Descriptive Statistics and Correlations................................................107

5.2 Job Characteristics Change.............................................................................1075.2.1 Anticipated Job Change at Time 1.......................................................1085.2.2 Perceived Job Change at Time 2..........................................................1095.2.3 Summary..............................................................................................111

5.3 Job Design Change and User Acceptance......................................................1125.3.1 Hierarchical Regression.......................................................................112

5.3.1.1 Composite Score Analyses......................................................1155.3.2 Usage as a Moderator of Job Change and User Acceptance...............117

5.3.2.1 Cross-Sectional Data Time 1 - Anticipated Job Chars............1185.3.2.2 Cross-Sectional Data Time 1 - Perceived Job Chars...............121

5.4 Chapter Summary...........................................................................................124

Chapter Six — Job Design Results: Post–Hoc Analysis & Discussion ..........1276.0 Introduction ...................................................................................................1276.1 Job Characteristics Change ...........................................................................127

6.1.1 Anticipated Job Change at Time 1 ......................................................1286.1.2 Perceived Job Change at Time 2 .........................................................132

6.1.2.1 Perceived Job Change, System Usage And Job Enrichment .1326.1.2.2. The JCM and Contemporary Work Context ..........................134

6.1.3 Summary .............................................................................................1366.2 Job Design Change and User Acceptance .....................................................136

6.2.1 Anticipated Job Change at Time 1 and User Acceptance ...................1366.2.2 Perceived Job Change at Time 2 and User Acceptance ......................138

6.3 Usage As A Moderator Between Job Change And User Acceptance ............1386.3.1 User Profile .........................................................................................1386.3.2 Usage, Anticipated Job Characteristics and User Acceptance ............1406.3.3 Usage, Perceived Job Characteristics and User Acceptance at T2 .....142

6.4 Summary .......................................................................................................144

viii

Chapter Seven — Social Network Change and User Acceptance ...................1477.0 Introduction ...................................................................................................1477.1 Research Context ...........................................................................................147

7.1.1 City Faculty .........................................................................................1487.1.2 Country Faculty ..................................................................................152

7.2 Network Change ............................................................................................1557.2.1 Network Correlation (QAP) ................................................................1567.2.2 Network Density .................................................................................1577.2.3 Direct Ties ...........................................................................................1597.2.4 E-I Index...............................................................................................1647.2.5 Summary - Network Change................................................................166

7.3 Network Centrality and User Acceptance.......................................................1687.4 Chapter Summary............................................................................................172

Chapter Eight — Network Change Post – Hoc Analysis & Discussion ..........1738.0 Introduction ...................................................................................................1738.1 ES Technology and Patterns of Network Change .........................................173

8.1.1 Summary .............................................................................................1778.2 Network Change and Acceptance .................................................................1778.3 Structural Autonomy and Perceived Ease of Use ..........................................1808.4 Conclusion .....................................................................................................182

Chapter Nine — Conclusions ..............................................................................1839.0 Introduction / Review ....................................................................................1839.1 Research Questions and the Overall Research Problem ...............................183

9.1.1 Job Design ...........................................................................................1849.1.2 Social Networks ..................................................................................1879.1.3 Overall Research Problem - Summary ...............................................189

9.2 Theoretical Contribution ..............................................................................1909.2.1 User Acceptance - Technology Acceptance Model .............................1909.2.2 Network Shock – Attitudinal and Behavioral Implications ................1919.2.3 Network Change .................................................................................1929.2.4 Structural Implications of Enterprise Systems Technology ................1929.2.5 Network Analysis and Structuration ...................................................193

9.3 Practical Implications ....................................................................................1949.3.1 Job Design and User Acceptance ........................................................1949.3.2 Social Networks and User Acceptance ...............................................196

9.4 Limitations ....................................................................................................1969.5 Directions for Future Research ....................................................................1989.6 Conclusion .....................................................................................................200

ix

Appendix ...........................................................................................................................201Appendix A.01 .........................................................................................................201Appendix A.02 .........................................................................................................203Appendix A.03 .........................................................................................................209Appendix A.04 .........................................................................................................215Appendix A.05 .........................................................................................................217Appendix A.06 .........................................................................................................233

Reference List....................................................................................................................243

x

List of Tables

Table 3.1 User Acceptance Inter-correlations and Reliabilities (Time 1) .........................66

Table 3.2 User Acceptance Inter-correlations and Reliabilities (Time 2) .........................67

Table 3.3 Job Characteristics Inter-correlations and Reliabilities (Time 1) ......................68

Table 3.4 Job Characteristics Inter-correlations and Reliabilities (Time 2) ......................68

Table 5.1 Pilot Site Participant Demographics at Time 1 and Time 2.............................103

Table 5.2 Pearson’s Product Correlation Matrix for Anticipated and Current Job Characteristics at Time 1 and Time 2........................................................................................104

Table 5.3 Paired Sample t-test for Current and Anticipated Job Characteristics ............108

Table 5.4 Paired Sample t-test for Current Job Characteristics.......................................108

Table 5.5 Means, Standard Deviations and Variance of Job Characteristic Difference Scores ..........................................................................................................................115

Table 5.6 Usage as a Moderator Between Anticipated Meaningfulness and Perceived Ease of Use...............................................................................................................118

Table 5.7 Usage as a Moderator Between Anticipated Skill Variety and Perceived Ease of Use...................................................................................................................119

Table 5.8 Usage as a Moderator Between Task Identity and Perceived Usefulness....122

Table 5.9 Usage as a Moderator Between Task Identity and Perceived Ease of Use...119

Table 6.1 Summary Table of Interaction Effects for Anticipated Job Design at T1 ......138

Table 6.2 Summary Table of Interaction Effects for Perceived Change in Job Design and Acceptance Variables (PU and PEU) ............................................................. .141

Table 7.1 QAP Correlation Scores (City T1 & T2) ........................................................ .153

Table 7.2 QAP Correlation Scores (Country T1 & T2) .................................................. .154

Table 7.3 Bootstrap Paired Sample t-test Of Network Density at T1 & T2 (City). .......155

Table 7.4 Bootstrap Paired Sample t-test Of Network Density at T1 & T2 (Country). ..156

Table 7.5 Paired Sample t-test Comparing IN-Degree Centrality Scores at T1 & T2. ...156

Table 7.6 Change in Participant in-degree ties for City and Country..............................157

Table 7.7 User Acceptance Scores of Country Groups .................................................. .161

Table 7.8 E-I Index of Country and City Advice Networks at T1 and T2 ..................... .162

Table 7.9 E-I Index of Country and City Resource Exchange Networks at T1 and T2. .162

Table 7.10 Perceived Usefulness at T2 as Predicted by Network Change at T2 ............. .167

Table 7.11 Perceived Ease of Use at T2 as Predicted by Network Change at T2. ............168

xi

Table of Figures

Figure 2.1 Job Characteristics Model (Hackman & Oldham, 1976) ..................................26

Figure 2.2 Implications of network shock on resource exchange networks .......................43

Figure 2.3 Example of network brokers (A & B) ...............................................................47

Figure 2.4 Structural holes within a network ......................................................................48

Figure 4.1 Example of a valued, directional matrix and corresponding graph (sociogram) 84

Figure 4.2 UCINET6 Quadratic Assignment procedure (QAP) Output .............................88

Figure 4.3 UCINET6 density output .................................................................................89

Figure 4.4 UCINET6 E-I Index Output ..............................................................................91

Figure 4.5 UCINET6 Degree Centrality Output .................................................................93

Figure 4.6 Freeman’s (1979) Star Network .......................................................................94

Figure 4.7 UCINET6 Flow-Betweenness Output ...............................................................94

Figure 4.8 UCINET6 Structural Holes Output ...................................................................97

Figure 4.9 UCINET6 Vector Regression Output ................................................................98

Figure 6.1 The Differential Effects of an ES implementation on Job Design ................131

Figure 7.1 City Advice Network at Time 1 .......................................................................147

Figure 7.2 City Advice Network at Time 2 .......................................................................147

Figure 7.3 City Resources Network at Time 1 .................................................................148

Figure 7.4 City Resources Network at Time 2 .................................................................148

Figure 7.5 Country Advice Network at Time 1 ................................................................150

Figure 7.6 Country Advice Network at Time 2 ................................................................150

Figure 7.7 Country Resource Network at Time 1 .............................................................151

Figure 7.8 Country Resource Network at Time 2 .............................................................151

Figure 7.9 Country Group A Network (Advice & Resources merged) ............................158

Figure 7.10 Country Group B Network (Advice & Resources merged) ............................159

Figure 7.11 Country Group C Network (Advice & Resources merged) ............................159

Figure 7.12 Country Group D Network (Advice & Resources merged) ............................160

xii

List of Abbreviations

ANOVA Analysis of VarianceAO Administrative OfficerDV Dependent VariableERP Enterprise Resource PlanningES Enterprise SystemsGNS Growth Need StrengthGUI Graphical User InterfaceHLOAR High Level Organizational Analysis ReportHSD Honestly Significant Difference (Tukey’s)HTML Hypertext Markup LanguageINSNA International Network for Social Network AnalysisIT Information TechnologyIV Independent VariableJCM Job Characteristics ModelJDS Job Diagnostic SurveyMPS Motivational Potential ScoreMRP Materials Resource PlanningPEOU Perceived Ease of UsePU Perceived UsefulnessQAP Quadratic Assignment ProcedureQUT Queensland University of TechnologyNET Network Exchange TheorySAQ Self Administered QuestionnaireSNA Social Network AnalysisSNT Social Network TheorySPSS Statistical Package for the Social SciencesTAM Technology Acceptance ModelUCI Unique Code IdentifierVDU Visual Display Unit

xiii

Statement of Original Authorship

The work contained in this thesis has not been previously submitted for a degree or diploma at any other higher education institution. To the best of my knowledge and belief, the thesis contains no material previously published or written by another person except where due reference is made.

Signed:

Date:

xiv

Acknowledgements

A PhD is similar to a black hole, it seems to suck people in towards the centre of a vortex, of which you are the centre. Therefore a lot of thanks are due to a lot of people for getting me this far (whether they liked it or not !).

First I would like to thank my supervisors, for their perseverance over the period. Dr. Artemis Chang, for your energy and commitment to the project, your willingness to push me beyond my comfort level and your tolerance of, at times, a difficult and petulant student ! And to Dr. Kerrie Unsworth - counselor, sage advisor and motivational speaker, you got me through some dark times! Steven Cox also deserves a heartfelt thanks given the amount of work it took to convey the beauty and eloquence that is moderated regression analysis. Undoubtedly this project would not have eventuated without the understanding, commitment, rigor and willingness to help that you all showed.

Thanks also go to colleagues such as Amanda Gudmundsson, Adele Bish and Jack Kegan for helping me see the funny side of it all – you were essential for my sanity ! I would like to also thank my parents again for the sacrifices they made, the opportunities they created, and their continued support. I would like to especially thank my mother Anne, who at times seemed to be the only person who really understood what I was going through. Also, a huge thanks to my sister Hannah, for returning as my “editor in chief”. I can’t believe you enjoy that stuff, but I’m very glad you do !

It would be remiss of me not to also formally acknowledge those users in Erudio that made the study possible. Your willingness to participate and to share your thoughts at a difficult time was truly appreciated. I’m grateful I got a chance to know you as people, and not merely participants in a study.

And finally, thanks to my two girls. Kristine my partner, for enduring this epic and seeing me through; you’ve got me back now – I promise. And to Jordan, my little bear and the most fabulous distraction there ever was - let’s go ride our bikes !

xv

1

Chapter One — Introduction

1.0 Introduction

Enterprise systems (ES) design such as that pioneered by materials resource planning (MRP) and enterprise resource planning (ERP) system types, continue to offer a range of benefits for firms looking to maintain their competitive advantage (Lengnick-Hall, Lengnick-Hall, & Abdinnour-Helm, 2004). In short, enterprise systems are designed to integrate the information and process flows of an organization through a single software package capable of operating across functional and departmental boundaries (Davenport, 1998). The increasing sophistication of ES offered by SAP, Peoplesoft and local vendor MINCOM offer organizations improved networking opportunities, greater integration and exploitation of data, improved resource sharing and “real-time” data access (Chung & Synder, 2000; O’Mahony & Barley, 1999). The costs associated with this system type are also widely accepted, particularly in the case of implementation failure with tangible costs reported in the millions of dollars for any one organization (Hall, 2002). In a departure from what might be expected, observers of such failures tend to lay the blame not at the technology, but at the measures taken to ensure its implementation and acceptance by users (Mitchell & Zmud, 1999).

The degree to which employees accept and utilize new information systems has long concerned both practitioners and researchers in the IT field (e.g. Robey, 1979; Swanson, 1982). However the peculiarities and specific core design elements of ES require a more comprehensive approach to understanding what determines a user’s degree of system acceptance. Previous research has tended to focus on individual antecedents of user acceptance. For example, support has been found for self efficacy (Hong, Thong, Wong & Tam, 2001; Venkatesh & Davis, 1996), pre-training and training environment interventions (Venkatesh, Speier & Morris, 2002), computer anxiety and intrinsic motivation (Venkatesh, 2000) as determinants of user acceptance. The frequent coupling of task and process re-design with ES implementations (either through deliberate intent or as an unintended consequence of system protocols) however requires a broader examination of social and structural elements as possible determinants of user attitude.

The thesis reports on an attempt to investigate the structural implications of introducing ES technology into an organization. In doing so it attempts to recognize the specific nature of ES and their characteristics when being introduced into organizations (Orlikowski & Iaconno, 2001). It also investigates the dual role played by structural

2

properties such as job design and social network configurations when introducing ES technology. Following the lead of authors such as Orlikowski (1991, 1992; 2000), Griffith (1999) as well as Poole and DeSanctis (2004) the thesis examines the reciprocal nature of technology design, structural properties and user acceptance. While ES technology may demonstrate the capacity to induce organizational change it is also suggested that existing organizational institutions may determine the permanence, extent and magnitude of the change experienced by the organization and consequently the degree of acceptance reported by users.

The aim of this introductory chapter is to provide a clear overview and summary of the thesis presented here. Section 1.1 below outlines in more detail the background and context of the research. This is followed by the research questions posed within the thesis including the broad, overall research question along with the more specific questions that have driven the development of the hypotheses presented in Chapter Two. The other main components of the chapter include a review of the methodology undertaken in the study, its practical and theoretical contributions along with its limitations. A number of key definitions are also provided before moving onto the theoretical component presented in Chapter Two.

1.1 Background to the Research

A number of developments within the field of information systems research provide the background to the research contained within the thesis. Three aspects in particular have driven the research questions and the design of the study: One, the increased prevalence of ES technology and their peculiar design characteristics as the predominant technology type being adopted by organizations over the last decade; two, the acknowledged importance of improved theory in IS research and the emergence of structuration theory as a well regarded theoretical framework for IS research; and three, the need to better understand the antecedents of user acceptance, particularly the role of a user’s social and contextual environment in determining acceptance. Each of these three areas will now be discussed in more detail.

This research has chosen to investigate the effects of Enterprise Systems (ES) technology for two reasons. The first is related to the ubiquitous nature of the ES technology and its potential as a significant catalyst for change within organizations (Davenport, 2000). The second is in response to calls for a greater acknowledgement of “technology as artifact” (Orlikowski & Iacono, 2001). Respected commentators in the field have stressed the need to recognize the peculiar aspects of various technologies

3

and to consider user, design and implementation issues in light of those characteristics. Similar notions have been echoed in the change literature, criticizing the tendency to discuss “change” in generic terms rather than considering the types of change being investigated (Rafferty & Griffin, 2001). There are a number of reasons why it is important to consider the unique characteristics of ES technology when investigating the structural implications of this technology type. These are outlined briefly below. For a more detailed discussion refer to Section 1.5.1 later in the chapter.

Enterprise systems (ES) technology currently represent one of the largest growing sectors in the IT market (Lengnick-Hall et al., 2004; Martin, 1998). Forecast expenditure on these system types had been estimated to be (US) $50 billion by 2004 (Bingi, Sharma & Godla, 1999). Until recently the cost and design of ES technology had limited it to large organizations, however recent evidence suggests that ES manufacturers have shifted their focus to also include small – medium enterprises (Oliver & Romm, 2002). Organizations intending to adopt ES technology typically do so for a number of reasons. Principally these reasons revolve around efficiency, responsiveness and control, the representations of which manifest themselves in changes to work processes, information and data flows (Davenport, 1998; Koch, 2001; O’Mahoney & Barley, 1999). A number of organizations have experienced difficulties in realizing the intended benefits of adopting ES technology (Bingi et al., 1999). Unexpected outcomes experienced by firms include implementation delays and increased costs, reduced productivity, poor data management and lost customers, all of which have been linked to some degree to the manner in which users have approached and used the new technology (Battacherjee & Hirchhiem, 1997; Chung & Snyder, 2000; Martin, 1998). However, in spite of the significant financial and organizational investment these systems represent to organizations, comparatively little research has been carried out to examine the user acceptance and organizational outcomes of ES technology (Klaus, Roseman & Gable, 2000; Oliver & Romm, 2002; O’Mahoney & Barley, 1999). An intended aim of the thesis is to further this field of inquiry.

The second area underpinning the research is the need for theoretically driven, rigorous studies in the IS/IT field. The degree to which users accept and utilize new IS has long concerned both practitioners and researchers within the IS/IT field (e.g. Robey, 1979; Swanson, 1982). Unfortunately research within the IS/IT field has also been typified as methodologically suspect (Boudreau, Gefen & Straub, 2001), has tended to lack contextual and technological specificity (Liker, Haddad & Karlin, 1999; Orlikowski & Barley, 2001), lacks sufficient theoretical grounding (Robey & Zmud, 1992) and in general, lacks relevance (Benbasat & Zmud, 1999). Consequently, at every stage of the process the researcher has attempted to address these concerns in research design

4

and methodology, particularly in the area of theory development. A key element of this approach was the adoption of structuration theory as a heuristic framework to discuss the relationship between ES technology and elements of organizational structure such as job design and social networks.

In attempting to explore the structural implications of enterprise system (ES) technology it was considered appropriate to adopt what is termed a structuralist approach (Giddens, 1984). Over the last decade structuration theory has gained increasing acceptance as a recognized and advocated framework for IT/IS research (Poole & DeSanctis, 2004). Initiated by Orlikowski and colleagues, the value of adopting a structuration perspective when carrying out IS research has steadily gained acceptance in the IS field (e.g. DeSanctis & Poole, 1994; Griffith, 1999; Orlikowski, 1991; 1992; Orlikowski & Robey, 1991; Poole & DeSanctis, 2004). In broad terms structuration theory argues that human behavior is determined by a “duality of structure” (Giddens, 1984). The duality of structure is the product of a series of inter-dependent interactions between social system and structure. Structure represents the collective institutionalized properties, rules and resources that mediate everyday activities. Social systems are viewed as sets of interactions concerned with the concrete practices of individuals operating within the set of institutionalized rules and resources, that are structural properties. The regular, patterned nature of these interactions between actors, within existing structural elements that work to reinforce or legitimize those very same structures is termed the “duality of structure”. Structuration theory as it has been adopted within the IS/IT literature has argued that technology should be considered as a representation of structure within an organization (DeSanctis & Poole, 1994). They argue that IT in its many forms is capable of acting as a form of social structure, providing users with both resources and rules that govern and enable behavior. Applying a similar logic this study goes further, suggesting that user’s job characteristics and social networks are also valid representations of structure and the interaction of the three may help to explain user’s reactions to ES technology. A more detailed discussion of structuration theory as it relates to IS/IT research is contained in Section 1.5.3 below.

The emergence of structuration theory as a unifying approach to the study of IS acceptance within organizations has coincided with the need to better understand and identify the antecedents of user acceptance. More specifically a number of commentators have acknowledged the need to better understand the social and contextual elements that may determine user attitudes to new IT/IS (Green, 1998; Karahanna & Straub, 1999; Venkatash & Davis, 2000). User acceptance models such as Davis’ (1989) Technology Acceptance Model in their current form struggle to measure the impact of social variables on user attitudes and their subsequent behaviors (Mathieson, 1991).

5

This is particularly relevant when introducing ES technology where the social and political ramifications can be significant (Huber, 1990; Koch, 2001). This, along with the need to better understand the structural implications of ES technology has provided a stimulus for the research contained within the thesis. The next section will outline in some detail the nature of the research problem being investigated and the associated research questions.

1.2 Research Problem and Research Questions

As outlined in Section 1.1 above, three elements underpin the research reported in the thesis: the ubiquitous nature of ES technology in the marketplace; the need to better understand the social and contextual determinants of user acceptance; and the emergence of structuration theory as a valid framework for IS/IT research. A consideration of all three elements has resulted in the broad research question “what are the structural implications of introducing enterprise systems technology into an organization?”. The thinking behind this question is outlined below.

Overall the research contained within the thesis considers the unique characteristics of ES technology and its effect on other structural components within the organization. An assumption is made that organizations represent an evolution of patterned interactions reflected in the presence of mutually reinforcing and adapting structures of rules, norms, routines, shared values and dependence relationships. A position is taken whereby the social systems and structures required for the process of structuration are conceptualized here as an organization’s job design, ES technology and social networks.

The core characteristics of ES technology include improving process flows, data integration and the enforcement of “best practice” protocols (Davenport, 2000; Oliver & Romm, 2002). These and other ancillary aspects of ES technology all have the potential to impact on structures present within organizations (see Section 1.5.1 below for a more comprehensive treatment of ES’s structural properties). The research seeks to identify whether the introduction of ES technology represents a disruption to established, or at best incrementally changing patterns of interaction. Therefore the broad research question being asked is “what are the structural implications of introducing enterprise systems technology into an organization?”.

6

In keeping with the structuralist approach “structures” are considered to be both in-tangible and tangible institutions or conventions within a user’s context that simultaneously enable and constrain their actions. Specifically the thesis considers the structural properties of user’s job design and their social network configurations. While other equally valid sources of structure may be identified these two have been chosen on the basis of previous research demonstrating strong links between employee attitude and behavior as well as their inherent link to ES design and function (Burkhardt, 1994; Grant & Uruthirapathy, 2002).

Job Design and Social Networks

Job characteristics are one of the more salient sources of structure within an organization (Torraco, 2005). Job tasks, responsibilities and functions provide a structure and a framework for employees to operate within. Job design to a large extent can be shown to determine employee communication patterns, access to resources and a place within the social and formal hierarchy of the organization (Poole, Seibold & McPhee, 1986). A sizable and consistent body of work has also linked job characteristics and their design to employee attitudes such as job satisfaction and morale (e.g. Boonzaier, Ficker & Rust, 2001; Fried & Ferris, 1987). The process oriented design of ES and their permeation throughout the organization function has the potential to reignite concerns based around technology driven job-deskilling and task simplification to ensure system effectiveness (Pruijt, 1998; Tolsby, 2000). However the relationship between job design and ES technologies is yet to be comprehensively understood (Grant & Uruthirapathy, 2002; Robey, Ross & Boudreau, 2002). Therefore given the highly integrated nature of organizational function and ES design it will be useful to consider the impact of ES on user’s job characteristics.

In addition to the potential for ES technologies to alter job characteristics, debate also surrounds its capacity to facilitate workplace democracy, and empowerment (Grey & Mitev, 1995). A number of ES advocates and observers have indicated a range of potential job enrichment and empowerment possibilities (Davenport, 1998; Lengnick-Hall et al., 2004; Tansley, Newell & Williams, 2001). Commonly cited ES characteristics such as improved access to information, increasing levels of information available for analysis and improved communication between organizational functions are considered capable of enriching users jobs and enabling workplace democracy initiatives. However, others argue the empowerment or enrichment potential of ES is often tempered by the tendency for systems to reinforce or establish control structures desired by management (Hall, 2002; Koch, 2001). Moreover, recent research has demonstrated the capability of ES to allow greater monitoring of users and the ability to enforce compliance (Sia, Tang, Soh & Boh, 2002; Tansley et al., 2001). Given the

7

strong theoretical and empirical links between job enrichment and employee attitudes such as job satisfaction and internal work motivation (Boonzaier et al., 2001) the research above gives rise to the question of user acceptance. Two in particular are considered here. One, does ES technology alter job design, and if so in what way; and two, do these changes have an impact on user acceptance levels ?

Therefore the first two research questions are presented below.

Research Question 1: Will ES technology alter the structural elements based around existing job characteristics?

Research Question 2: Is there a relationship between perceived changes in job design and user acceptance following the introduction of ES technology?

In addition to job design, other, less tangible sources of structure are evident within an organization that arguably better reflect Gidden’s (1984) original conception of structuration. In particular, an employee’s informal or social networks can be seen to be simultaneously a source of structure and social action within an organization. Again, when considering the intended function and design of ES technology it is reasonable to consider whether the introduction of this technology type would alter a user’s social networks. One of the defining characteristics of ES technology is the manner in which it can alter the flow of information and allow greater access to data (Davenport, 1998). Changing or redistributing access to information resources may result in incomplete networks, shifting balances of power and possible negative impacts on existing social networks (O’Mahoney & Barley, 1999). As such the following question is posed:

Research Question 3: Will ES technology alter the structural elements based around existing social network configurations?

A logical extension of the question posed above is to consider whether perceived or actual changes to social networks are linked to user attitudes towards the ES technology. A small but compelling group of studies have indicated the ability of social networks to facilitate and transmit attitudes concerning new IT/IS (Fulk, 1993; Meyer, 1994; Schmitz & Fulk, 1991). Moreover, a recently emerging body of work examines not only the role that networks play in facilitating organizational change (e.g. McGrath & Krackhardt, 2003; Tenaski & Chesmore, 2003) but the consequences of organizational change on networks (e.g. Burkhardt & Brass, 1990; Burkhardt, 1994; Shah, 2000; Susskind, Miller & Johnson, 1998). Both bodies of work raise questions

8

regarding the role networks play when introducing ES technology, particularly if that technology threatens to alter established patterns of interaction. Therefore questions exist as to whether user reaction to ES technology is best predicted not by the salient characteristics of the system, but the extent to which they disrupt the social systems and other structural modalities inherent in the structuration process. Consequently the final research question presented below is:

Research Question 4: Is there a relationship between perceived changes in social networks and user acceptance following the introduction of ES technology?

1.3 Justification of the Research

As can be seen from the discussion above, the study contained within this thesis attempts to address a number of areas in both the network and user acceptance literature. However the research offers a number of broader contributions to the fields of IT/IS research and social network analysis.

1.3.1 User Acceptance and Social Context

One of the few empirically and theoretically robust models of user acceptance to emerge over the last decade is Davis’ (1989) Technology Acceptance Model (TAM). Since its inception TAM has received empirical support in a variety of contexts and over a wide range of technologies (Karahanna & Straub, 1999). Efforts have also been undertaken to identify the antecedents of TAM’s two key constructs, Perceived Usefulness (PU) and Perceived Ease of Use (PEOU); such as user training and experience (Igbaria, Guimaraes & Davis, 1995), intrinsic involvement and motivation (Jackson, Chow & Leitch, 1997; Venkatash, 1999) and perceived user resources (Mathieson, Peacock & Chin, 2001). However, in its current state TAM research has also fielded criticism in terms of its failure to adequately identify the antecedents of PU and PEOU (Green, 1998; Karahanna & Straub, 1999; Mathieson, 1991). As indicated in Section 1.0 existing research has tended to focus on individual determinants of user acceptance. Recent commentators have indicated the need to adopt a wider focus, investigating the social and contextual elements that may effect user acceptance (Karahanna & Straub, 1999). While there have been some attempts at redress (Venkatash & Davis, 2000; Karahanna & Straub, 1999) the issue is yet to be investigated comprehensively. As suggested by Karahanna & Straub (1999) social network analysis, as used in this study, provides researchers with a greater understanding of how structural context can effect

9

user acceptance. Investigating the effects of ES technology on both job design and social networks will represent a significant step forward in understanding the role of social context as an antecedent of both user attitude and behavior.

1.3.2 Information Technology as Structure

In its original form structuration theory rejects the idea of structure as being tangible or concrete, arguing that its existence is determined purely by human perception (Giddens, 1984). Authors such as Orlikowski (1992; 2000) and DeSanctis & Poole (1994) have challenged this position, arguing strongly for the role of technology as a set of structural properties to be recognized. Recently Poole and DeSanctis (2004, 213-14) outlined a set of requirements for researchers wishing to examine structuration in an IS/IT context. Among them were the need to identify an array of relevant structures; to clarify the relationship between the structures; to develop a description of how the social system works and a description of how the social context is reproduced; to identify the redefining and reproduction of structures; and to undertake a critical enquiry into the dynamics of power underlying the structuration process. Undertaking a longitudinal approach considering the characteristics of ES technology, job design and social networks will demonstrate the extent to which ES technology can determine human action, but similarly be interpreted by users in a way that reflects the existing social context. Therefore the study looks to further validate the role of technology as a structural modality as defined by those who have adopted structuration theory in IS/IT research (Poole & DeSanctis, 2004).

1.3.3 Social Network Analysis and User Acceptance

Another important contribution of the thesis lies in its application of network analysis concepts to the issue of user acceptance in an ES technology context (See Section 1.5.2 below for an introduction to network analysis). Theoretically and methodologically social network analysis provides a strong complement to researchers adopting a structuralist approach. Pursuing such an approach offers user acceptance researchers a number of advantages. First, a far greater degree of attention can be paid to a user’s social context and the impact it may have on user acceptance outcomes (Karahanna & Straub, 1999). More specifically, adopting a network analysis approach lends a methodological and theoretical specificity to IT research adopting a structuration perspective. Structuration theory by its nature fails to offer researchers directional guidance and is more akin to a theoretical ideology or philosophy than a prescriptive, hypothesis generating model (Poole & DeSanctis, 2004). As social network

10

configurations represent both structures and social action the “duality of structure” can be closely observed, particularly when adopting a longitudinal approach. Therefore adopting a network analysis approach allows researchers greater opportunities to both measure and observe the dynamic process of structuration.

1.3.4 Implications of Network Change

Finally, the study is considered valuable in its attempt to further understand the impact of formal change initiatives on informal network configurations. To date much of the research employing a social network perspective has tended to focus on the consequences of network structure (Brass, Galaskiewicz, Greve & Tsai, 2004). As a result few scholars have investigated how networks change over time, or indeed the implications of network change (Borgatti & Foster, 2003). Along with other recent attempts (Shah, 2000; Susskind, Miller & Johnson, 1998) this research represents a departure from this tradition, considering both the antecedents and the nature of network change. In doing so it also builds upon and complements previous work such as Brass & Burkhardt (1990-1994) and Barley (1990) in considering patterns of change following the introduction of specific types of technology. This will not only allow a better understanding of how networks can facilitate organizational change (Stevenson et al., 2003), but will also help to understand user’s reactions and behaviors in response to network change.

1.4 Research Design

In brief, the research design employed here can be summarized as a longitudinal case study using multiple methods to capture data from a single case organization. The organization in question was Erudio (a pseudonym), a large tertiary education institution. Comprised of fifteen quasi-independent learning centers Erudio claims to be the largest provider of post-secondary education in the state with over 200’000 international and domestic students enrolled each year. Service delivery provided by Erudio ranges from conventional classroom instruction to on-line learning, distance education, on-site training and long term traineeships. The data were collected from two of the centers that had been designated as pilot sites for the implementation of the system. Two principle methods of data collection were employed: 1) a longitudinal on-line survey distributed to all users throughout the organization, and 2) a series of structured interviews aimed at collecting social network data from a narrower, pre-determined set of users.

11

A number of factors were considered when deciding to adopt this particular research design. A longitudinal research design was considered better able to capture the extent to which structural change occurs following the introduction of an ES. Barley (1990) and more recently Poole & DeSanctis (2004) highlight the necessity for structuration researchers to adopt research designs capable of identifying both structural and social change.

The primary research tool employed to collect data in this study was an on-line self-administered questionnaire. The use of an on-line survey was identified as suitable for the target audience and able to overcome a number of geographical and financial constraints. However, a number of design and deployment considerations had to be made in order for the tool to successfully capture the data. These considerations are reported in more detail in Chapter Three. Also outlined in Chapter Three, social network analysis is principally concerned with establishing patterns of interaction between actors. Data collected for network analysis are necessarily dyadic in nature, requiring the collection of data that is non-anonymous and able to capture the nature and type of relations between actors. The specific and sensitive nature of the data meant that a series of structured interviews carried out with a defined set of actors were the most appropriate method to use.

In order to test the hypotheses relevant for job design change a variety of analyses were employed including correlational analysis, ANOVA’s, paired sample t-tests and moderated hierarchical regression. A variety of analyses were also used to test the extent and nature of network change. Quadratic Assignment Procedure (QAP) routines, network density IN-degree scores and E-I indices were all employed during the study. While regression techniques were also used to test some of the social network hypotheses, the specialized nature of network data required the use of a number of non-parametric techniques to derive appropriate confidence intervals.

1.5 Key Definitions

This section of the chapter clarifies a number of key concepts dealt within the thesis. Contained below are definitions of Enterprise Systems (ES) technology, social network analysis and a brief discussion of structuration theory as it relates to the IT/IS research field.

12

1.5.1 Enterprise Systems Information Technology

In essence the thesis concerns itself with technological change within one case organization. The technology in question is a type of software package evolved from early Materials Requirements Planning (MRP) and Manufacturing Requirements Planning (MRPII) packages into what have been termed Enterprise Resource Planning (ERP) and “mega-packages” (Davenport, 1996). While ERP is a commonly used term it has received criticism for being confusing, lacking relevance and unreflective of the technology’s characteristics or wider capabilities (Klaus, Rosemann & Gable, 2000). Consequently in an attempt to reflect the contemporary nature of these systems, the term Enterprise System (ES) is used throughout the thesis. The nature of what is precisely meant by enterprise systems technology is defined below.

Klaus et al. (2000) provide a comprehensive review of enterprise based system typologies. While stressing that at a grass roots level ES are ultimately customized to perform within “industry best practice” they highlight three core elements of any enterprise based software package. First, at its most basic ES technology is defined as a standard software package that can operate across a number of different hardware platforms, typically tailored to various industry sectors such as mining and education. Second, they take the position that ES technology is seen to be the key element of an infrastructure that delivers a solution to business. Primarily this solution is achieved by offering an “underlying integrated database that stores master and transactional data in a consistent way and with controlled redundancy” (p143). Third, the software is designed to support multiple functions within an organization such as human resources, production, sales, logistics and so on. This element is possibly the most visible from an operational and functional perspective. In contrast to stand-alone productivity software, ES technology provides the infrastructure and capacity to integrate data and process flows within and between organizational functions. Following a process orientated view of organizations these system types aim to “seamlessly” support the core processes of the business and administrative functions. This forcibly promotes a greater level of inter-dependence between organizational functions both from a procedural and data communication perspective. Other capabilities cited by Klaus et al. (2000) include: Greater access for users to functional areas and greater navigation through different abstraction levels; organizational transaction processing across geopolitical boundaries; and multiple transaction processing in real time.

13

Davenport, a long time commentator in the enterprise systems debate, adds to the framework suggesting that in principle the characteristics listed above allow organizations advantages in four key areas (Davenport, 2000).

Information scope: Refers to the sheer range of information that ES’s can effectively deal with.

Transaction automation: Allows the performance of real-time sequential transactions without specific instruction by the user. Transactions are able to be carried out within and without the organization if so configured.

Process management: The ability of the system to take system data and to translate it into business processes. Assists organizations in achieving a rigor in their process management by encouraging adherence to “best practice”.

Knowledge management: Considered by Davenport as the most sophisticated ES capability, knowledge management and refers to the ability of ES to be a strategic management tool. Representations of this include: increased access to data; increased availability of data for decision making; facilitating the management of multiple data sources, and; increased data analysis capabilities.

Not all share the entirely positive assessment of ES technology’s characteristics however. Even Davenport (1998, 122) notes the overly deterministic nature of this technology type: “one major criticism of enterprise systems is their tendency to impose their own logic or business process on companies”. Six years on this characteristics remains a key element of ES design, with Lengnick-Hall et al. (2004, 314) observing that “idiosyncratic ways of doing business are not supported by ERP systems”. As a result organizations are typically required to engage in a series of structural changes aimed at fitting the requirements of the system (Chan & Land, 1999; Davenport, 2000). Numerous examples are to be found in both the professional and academic press documenting such changes (e.g. Battacherjee & Hirschheim, 1997; Hall, 2002; Paper, Tingey & Mok, 2003; Songini, 2004).

14

The organizational ramifications of such a design element are compounded by the mandatory nature of the system (Brown, Massey Montoya-Weiss & Burkman, 2002).ES technology usage is largely mandatory due to their integral linkages with task and job functions (Wright, 2001). Consequently, ES characteristics contrast heavily with information systems of the past, which lacked integration with other systems and functions of the organization, offered reduced decision making capability due to limited data resources and allowed information only to key organizational actors or units (Bhattacherjee & Hirschheim, 1997; Brown & Jones, 1998). Others argue that the inherent characteristics of ES technology enable management to reinforce or establish desired control structures (Hall, 2002; Koch, 2001). Moreover, recent research has demonstrated the capability of ES to allow greater monitoring of users and the ability to enforce compliance (Sia et al., 2002; Tansley et al., 2001).

In summary, this brief outline highlights five key elements of systems defined as enterprise systems technology. One, they provide an integrated software solution for all key functional components of the organization; two, where possible routine, systematic transactions are automated; three, in line with their integrated functionality the systems are designed around process flows rather than functional or departmental structures; four, these system types afford organizations significantly increased opportunities for data management and storage; and five, the introduction of ES technology often coincides with, or requires organizations to undertake some form of significant structural change.

Case organization – HAL Student Administration System

Klaus et al. (2000) advocate ES technology as being highly customizable to each enterprise context. As such they caution that ES systems can only be discussed in generic terms prior to implementation. The extent to which these systems can (or more importantly, are) customized is cause for debate (Pollock & Cornford, 2004; Sia et al., 2002) however it is considered valuable to examine the system implemented in the case examined here. The aims of the case organization in relation to the ES being implemented were consistent with the generic capabilities of ES technology reviewed above. The information reported below was obtained from a variety of sources. The primary sources of information were a series of reports generated for the implementation project management team. Other sources of information included vendor and implementation consultancy press releases. Due to the confidentiality agreement in place with the participating organization however, these cannot be referenced in the usual manner. The organization will be referred to using the pseudonym Erudio.

15

The introduction of the system was touted as one that allowed a number of operational and efficiency advantages. Primarily it was communicated as a tool to enable improved student enrolment through on-line enrolment and student administration. Prior to this decision each of Erudio’s 15 Centres used an in-house package which over time had been customized to reflect the particular requirements and requests of each Centre’s users. Despite its apparent flexibility and customizability the old system suffered from dated software and a basic infrastructure that was over a decade old. Built around a commercially available enterprise system, the new system aimed to standardize processes across the 15 Erudio Centers, allow developments such as on-line enrolments and other benefits such as cross functional data sharing, access and reporting. Both internal reports and publicly available communications reflected this aim.

The HAL project creates strategic infrastructure, delivering a Student Administration System with the capability to deliver timely, relevant information to both community and government stakeholders, in an open e-business framework (internal newsletter).

[the system will] improve Erudio’s client services and system operation, increase the accuracy of reporting data and produce significant administrative savings (internal newsletter).

At a deeper level however the introduction of the system represented some fundamental changes to the manner in which the organization operated. These aims or objectives were not as publicly conveyed and were only identified in senior management reports. The system was described to users as a “vanilla” system, indicating that little or no customization was available. For example, the organization was required to change its entire vocabulary in relation to its structure, courses and offerings. For example, “Course” became “Program”, “Faculty” became “Group”, “Module / Competency” became “Course” and so on. The superficiality of language change belied the more strategic changes reflected in the system’s introduction. For example, a consistent theme was the desire to obtain an information system that was centrally controlled at a corporate, rather than at a faculty level (HLOAR, 2002, 6). In keeping with the autonomous nature of each Centre, users had previously been able to develop on-site, Centre specific functionality. Unsurprisingly this led to inconsistencies in reporting and work processes across Erudio as a whole. This further reinforced the notion that each Centre operated independently and not as part of the larger Erudio organization as desired by management and powerful external stakeholders. It was anticipated that the new system would allow greater control and standardization of organizational process.

16

The aim of HAL is to replace the locally designed, centre specific legacy system with one that is centrally controlled (HLOAR, 2002, 6).

“Users may no longer ‘own’ standalone systems and may need to sacrifice user convenience provided by custom built databases for the benefits of greater data integrity, reporting flexibility and process consistency. (HLOAR, 2002, 7).

Other documents alluded to the magnitude of the change that may occur with the introduction of the new system, but failed to provide specifics.

“Implementing HAL is more than just a technical solution. It is about achieving business success by ensuring maximum alignment between the ERUDIO people, process, policy and the system. The implementation of HAL will drive a fundamental shift in the way the organization manages student administration services and ERUDIO staff must be aligned with the new paradigm for maximum benefits to be achieved” (Change Management Strategy, 2003, 11).

It could assist in the development of an Erudio culture, rather than a [centre] culture (HLOAR, 2002, 7).

It is these deeper, more fundamental, structural changes that the thesis is concerned with. In particular do ES technologies such as the one described above provide an “occasion for structuring” (Barley, 1990). More specifically, are ES technologies a catalyst for job design and network change, and further, are these changes a determinant of user attitude and behavior? Chapter Two will expand on this further before the remainder of the thesis reports the results of the study aimed at answering this question.

1.5.2 Social Network Analysis

While the study of social networks has long been the domain of sociologists, the utility of the approach has received an increasing level of support among organizational theorists (Borgatti and Foster, 2003). In simple terms a social network can be defined as a set of nodes and the set of ties representing some relationship, or lack of, between the nodes (Brass, 1995). Alternatively, it has been referred to as an identified set of relationships or patterns of interaction between two or more actors, in this instance, users within an organization (Scott, 2000). The principle unit of analysis when examining social networks is the nature of the relationship between each actor node,

17

not the actors themselves (Brass, 1995). The strength of the relationship, the structural position and legitimacy of each actor, combined with the extent of their connection to other actors are all considered key determinants of attitude and behavior (Podolny & Page, 1998). A variety of networks can exist at any one time, with previous researchers having examined friendship, influence, advice, support, resource transfer and power networks (Brass & Burkhardt, 1993; Ibarra, 1993).

Social network’s capacity to link the micro and macro elements of organizational research allows a unique perspective to be gained, one not offered by more conventional forms of organizational research (Tichy, Tushman & Fombrun, 1979). A key assumption among social network theorists is that the social context surrounding an individual is the primary source of reference for the interpretation and development of perceptions regarding expected behaviors and relationships with others (Meyer, 1994). As organizations are essentially socially constructed artifacts, social network theory (SNT) provides a mechanism to understand the social composition and context of an organization. Consequently, this allows researchers to better understand variations in individual attitudes and behavior (Nohria, 1992), to identify and understand the role of sub-groups within an organization (Fombrun, 1982) and the structure of power distributions within organizations (Brass, 1984; Ibarra, 1993; Tichy et al., 1979).

1.5.3 Structuration Theory and Information Systems Research

At the core of the structuralist approach are the three components of structure, social system and the “duality of structure” (Giddens, 1999). Social systems are viewed as sets of interactions concerned with the concrete practices of individuals operating within the set of institutionalized rules and resources, referred to as structure. In brief, structure represents the collective institutionalized properties, rules and resources that mediate their everyday activities. However it is the regular, patterned nature of these interactions between actors, within existing structural elements that work to reinforce or legitimize those very same structures. Mouzelis (1989) recounts Gidden’s analogous use of language to help clarify this recursive, inter-dependant relationship. Humans communicate using language, which is bounded by conventions, grammar and vocabulary (resources). In communicating to each other humans further institutionalize the conventions of language by their use. It is this recursive interaction of social system and structure that is termed the “duality of structure”. Together the three elements of duality, structure and social system provide the foundations for the process of structuration - the conditions governing the continuity or transformation of structures and therefore the reproduction of systems (Mouzelis, 1989). Importantly

18

structuration recognizes that while structures constrain the actions of actors, previous action has also determined the arrangement of those structures and can alter (albeit incrementally) conditions in the future.

A strict application of Gidden’s theory when considering the role of technology in organizational structures can be problematic. Structures as conceptualized by Giddens (1984) are tangible only in the sense that they are perceived by humans to exist and are enacted in their behavior. The recursive nature of both structure and social action appears to prevent the separate consideration of elements such as technology as sources of structure (Poole & DeSanctis, 2004). Recently authors such as Orlikowski (1992; 2000) and DeSanctis & Poole (1994) have argued strongly for the role of technology as a set of structural properties to be recognized. Most notable has been the development of DeSanctis and Poole’s (1994) Adaptive Structuration Theory (AST). AST argues for the capacity of IT to impose yet another set of structural attributes or properties on users. In particular their theory addresses in detail the characteristics and features of the technology by defining a system types’ “spirit” as well as its structural properties.

DeSanctis and Poole (1994) define a technology’s structural properties as the specific rules, protocols, capabilities and resources provided as part of the system’s inherent design and construction. Importantly AST makes explicit the capacity of contemporary advanced IT systems to offer a wide variety of capabilities and configurations. The other, less tangible component of technology is what DeSanctis & Poole (1994) describe as a technology’s spirit. The spirit of a system reflects the manner in which people are expected to interact and interpret the system’s function and intent where these elements may not be specifically communicated. While spirit can reflect a normative structure of system use and intent, the spirit of a technology is determined largely by how it is presented to, and perceived by users. The interpretive nature of spirit is important in that the features of a technology can be appropriated by users faithfully as intended by design, function or intent. Technology can also be appropriated in a way that is described as ironic (Poole & DeSanctis, 1990). For example, while the faithful appropriation of a phone book would be to use it as a data storage and retrieval mechanism, most of us just use them to raise the height of our computer monitors.

AST proposes the manner in which actors faithfully or ironically adapt the structural elements of IT and appropriate the system into their existing social structures accounts for variations in IT implementation success. Consistent with Giddens’ conception of structuration it is the recursive interaction between the structural aspects of the ES and existing social structures that determine the extent to which the characteristics of the system are appropriated in a faithful or ironic nature. DeSanctis and Poole

19

(1994) argue a key strength of AST is its ability to account for the structurational potential of technology while also acknowledging the characteristics of the technology and its effect on users as called for by authors such as Orlikowski and Iacono (2001). Variants of the structuralist approach such as AST have enjoyed an increasing amount of acceptance within the IT research field (Poole & DeSanctis, 2004).

Therefore, while it may be intuitively appealing to consider technology as a primary catalyst for organizational change it neglects to consider “duality of structure” and the process of structuration. Applying structuration theory to the role played by technology in organizations avoids an overly deterministic approach while at the same time acknowledging the presence and effect of the IT artifact (DeSanctis & Poole, 1994; Orlikowski & Iacono, 2001). IT researchers adopting a structuralist stance argue that while ES system protocols may to a certain extent determine individual action, human agency also determines the extent to which technology is incorporated into everyday operations (Orlikowski, 1992). Applying a similar logic, job design characteristics and social networks are also considered capable of being considered representative of structure and social action within an organization. The failure of past research to acknowledge the role of individual action and the influence of social context represented by this approach when considering the acceptance and use of IT is considered to be a significant oversight (DeSantis & Poole, 1994).

It should be noted however that the research contained within this thesis does not claim to formally test variants of structuration theory such as AST. Instead, the approach has been adopted as a useful heuristic framework to discuss the recursive interaction between the structural properties of ES characteristics, job design and informal networks and its impact on user acceptance. While the theoretical component of the thesis will consider the effect of social action in response to structural change, the testing of these propositions lies outside the empirical scope of the thesis.

1.6 Outline of the Thesis

The thesis is divided into nine chapters of which this chapter is one. Chapter Two establishes the theoretical grounding upon which the remainder of the research is based. An overall conceptual model and eight supporting hypotheses are presented throughout the chapter. Divided into two major sections, Chapter Two first considers the relationship between ES and job design. The second component of the chapter investigates the role of social networks in determining and facilitating the transfer of attitude among users.

20

Chapter Three outlines in some detail the methodological aspects of the study reported within the thesis. The context and particulars of the research site are first covered, giving the reader an insight into the data collected and later tested in Chapters Five and Seven. Chapter Three defends the overall strategy employed and also justifies the use of data collection methods such as on-line surveys and highly structured interviews. The chapter also reviews the measures used in each of the data collection tools and justifies their inclusion.

Chapter Four is aimed at providing a brief introduction to social network analysis and its methods. While network analysis has been an established field of enquiry for over a century it remains an emerging paradigm in the areas of organizational and IS research (Borgatti & Foster, 2003). Consequently it is likely that some readers may be unfamiliar with the unique nature of network analysis, data capture and its analytical capabilities. This chapter aims to provide the reader with an understanding of the methods and analysis used in the study and reported in Chapter Seven.

Chapter Five reports the results of the job design component of the study. Three basic sets of analysis and their findings are provided. One, a series of paired sample t-tests are reported examining whether users anticipate or perceive a change in job characteristics over time. Two, a series of hierarchical regressions are reported, testing the relationship between anticipated change in job characteristics and user acceptance. Three, a series of moderated regressions testing the hypothesized interaction effect between system usage, perceived job change and user acceptance are reported.

Chapter Six provides a discussion of the job design results reported in Chapter Five.

Chapter Seven reports the results of the network component of the study. Two major sets of analyses were conducted. One, a series of analyses aimed at ascertaining whether network change did occur, and if so to examine the nature of that change. Two, the findings from a series of hierarchical regressions testing the effect of change in centrality or brokerage on user acceptance.

Chapter Eight provides a discussion of the network results reported in Chapter Seven.

Chapter Nine concludes the thesis, providing a review and a discussion of the results as they relate to the theory put forward in Chapter Two. The significance and contribution of the findings are also raised both in relation to IS practice and theory. Finally, the chapter puts forward some suggestions for future research, addressing some of the current study’s limitations and building on some of its findings.

21

1.7 Conclusion

Chapter One has provided an introduction and summary of the thesis contents. The primary aim of the chapter was to provide an introduction to the rationale, aims and structure of the thesis. The background to the research and associated research questions were presented, along with an explanation of the contribution the study expects to make to the fields of IS/IT research and social network analysis. A review of the overall research strategy was previewed, outlining the data collection methods and analysis techniques that were employed.

Chapter Two outlines a series of arguments based around the capacity of ES technology to change the structural properties based around user’s job design and social networks. Influenced by the notion of structuration it is argued that while ES technology may represent the potential for significant structural change, user responses to those changes via their social networks may determine the degree of acceptance demonstrated by users. Consequently the extent to which the system is successfully adapted into their everyday structures and social systems is dependant on user action within existing structures and systems.

Chapter Two also presents a series of hypotheses designed to underpin the research questions. In the interests of clarity the chapter is split into two major sections. The first section investigates the relationship between ES technology characteristics and job enrichment. The second section concerns the relationship between the disruption of network structures and user acceptance.

22

23

Chapter Two — The Structural Implications of

Introducing Enterprise Systems Technology

2.0 Introduction / Review

Chapter One provided a summary of the background, aims and justification of the research undertaken in this thesis. Chapter One also provided a preview of the key theoretical and methodological tenets underpinning the research design. In doing so the introduction highlighted the need to better understand the implications of introducing enterprise systems (ES) technology. A brief review of the literature highlighted the need to better understand the social and structural implications of this technology type and to investigate further the effect of social context on user acceptance of ES technology. The intent of Chapter Two is to further elucidate on these elements, providing the theoretical basis upon which the research design and methodology is derived.

Chapter Two presents the model to be tested, outlines the variables under consideration and puts forward a number of supporting hypotheses. As discussed the thesis is principally concerned with the structural implications of introducing contemporary IT into organizations (defined here as Enterprise Systems). To this end the research identifies two structural elements present within an organization – elements of a user’s job design and their social networks. These are examined in terms of their relationship with a third structural element, ES technology. The potential consequences for both job design and pre-existing network configurations following the instruction of ES technology are considered in detail. The chapter is divided into four major components. The first provides a brief overview of the theory discussing the two comparative approaches to examining ES technology and structural change in the context of structuration theory. The second examines in detail the relationship between the structural properties of job design and ES technology characteristics. The third uses social network theory to examine the relationship between user interaction patterns (social networks) and ES technology. The fourth component concludes the discussion, considering the potential for social networks to be used as a mechanism for user action as users influence others as to the relative merits of ES technology via their network connections. While this final component falls outside the empirical scope of the thesis it is considered useful to include this component in order to provide a comprehensive theoretical piece of value to both the user acceptance and social network literatures. Chapter Two concludes by summarizing the contents of the chapter and reiterating the main arguments represented in the model.

24

2.1 Theoretical Model - Overview

The value of adopting a structuralist approach has already been established in Chapter One. A principal contribution of the thesis lies in its application of social network and job design theory within a structuralist framework to better understand the antecedents of user acceptance.

Drawing on social network and job design theory a new approach to ES technology acceptance is offered that implicitly acknowledges the process of structuration in organizations. It is argued that changes resulting from the introduction of an ES stimulate a number of responses from users. In contrast to many examinations of user acceptance strong emphasis is placed on the structurational processes as reflected by informal network exchange mechanisms in determining how the system is perceived. While users may make their own judgments of the system’s utility based on job and task related criteria, they may also be subject to influence within their informal networks as to the systems worth. The tension existing between individual assessments and the normative pressures of social system and structure culminate in a range of behavioral responses towards the system.

In broad terms the thesis is concerned with the potential for ES technology to stimulate change in two representations of structure within an organization. One, changes to a user’s job design characteristics and two, changes to informal resource exchange networks. Section 2.3 below draws on both user acceptance and job characteristics theory to explain user reactions to system driven job and task changes. Overall it is argued that ES technology can have a negative effect on the enrichment capabilities of a user’s job and thus have subsequent effects on user acceptance. However the degree of system usage reported by a user may moderate the extent to which their job characteristics are altered and hence form their attitudes towards the system.

Section 2.4 then presents an argument to suggest that the inherent design elements and system protocols associated with ES types may result in perceived or actual changes to a user’s social networks, especially those relating to advice and resource exchange. A set of hypotheses are put forward to suggest that the degree of user acceptance demonstrated by users is moderated by their position within existing network configurations and the amount of perceived control over resource flows afforded to them by the new system.

25

Finally, in Section 2.5 attention shifts to the to the role played by social networks as another structural property capable of determining and facilitating user behavior. The ability of the model to depict the process of structuration by which users evaluate, assess and react to ES technology via network analysis concepts and methodology represents a clear step forward in understanding user acceptance.

Therefore this chapter concerns itself with three fundamental examples of structure and social action within an organizational context. The first are the structural properties associated with the characteristics of ES technology (See Section 1.5.1 for a brief review). Second, the model is concerned with the structural properties of job design and job characteristics. Finally the role of informal social networks and their ability to reflect both structure and social action is considered. In short the chapter represents an attempt to determine whether the interdependent relationship between the structural attributes represented by ES technology characteristics, along with the existing structural and social attributes represented by job design characteristics and social networks do at least in part determine user acceptance levels. The remainder of the chapter is devoted to presenting the theoretical basis of this argument and the related hypotheses.

2.2 System Driven Changes to Job Design Characteristics

Structure as defined by structuration theory involves the rules, resources and guidelines for social action (Giddens, 1999). One approach to empirically measure or conceptualize such structures is to examine job design characteristics as a representation of structure within organizational contexts (Poole, Seibold & McPhee, 1986). This section of Chapter Two investigates the potential of ES technology to not only add another set of constraining and enabling structural properties, but to alter existing structural properties represented by a user’s job characteristics. The implications of systems driven job change, or even the threat of such change is considered in light of user acceptance issues. In short, the question “will ES technology alter the structural elements based around existing job characteristics and will it effect user acceptance?” is considered.

Two parallel sets of hypotheses are put forward, reflecting the longitudinal nature of the research. Previous user acceptance and satisfaction research has indicated the effect of user anticipation on acceptance outcomes (Szajna & Scamell, 1993; Ryker, Nath & Henson, 1997). Fishbien and Azjen’s (1975) Theory of Reasoned Behavior upon which the TAM model is derived also acknowledges the influence of perception on behavior. Consequently one set of hypotheses is concerned with users anticipating ES

26

driven job change prior to its introduction, while the other is concerned with perceived changes over time after implementation. Section 2.5 will address in greater detail the implications of these attitudes in relation to user behavior and the effect “duality of structure” has on the acceptance, appropriation and usage of ES technology.

2.2.1 A Job Design Approach to Enterprise Systems

Vendors and industry advocates highlight the potential of Enterprise Systems (ES) to facilitate broader organizational initiatives such as workplace democracy, empowerment and job enrichment (Tansley, Newell and Williams, 2001). It is proposed that ES characteristics can increase the scope of a position outside the traditional functional boundaries present with isolated legacy systems; facilitate the potential devolvement of information and decision making to line supervisor and workers; and allow users to manipulate multiple screens and system tasks thanks to “Windows based” user interface design1 (Sia et al., 2002). If ES technology does in fact deliver benefits such as workplace democracy and job enrichment one would reasonably expect to see relatively high levels of user acceptance following the system’s introduction.

However, Sia et al. (2002) cite a contra-argument suggesting that ES technology can also facilitate greater visibility and monitoring of employees, therefore increasing managerial prerogative and control. Hall (2002) concurs, arguing that ES technology is regularly associated with downsizing and furthers management control, job deskilling and the intensification of work. The frequent coupling of process re-design with ES implementations (either by deliberate intent or as an unintended consequence of system protocols) requires a broader examination of system driven job design change as a possible determinant of user acceptance levels (Hall, 2002; Oliver & Romm, 2002). Section 2.3.2 below investigates this idea further, offering up a number of hypotheses based around Hackman and Oldham’s (1976) job characteristics model.

Hackman and Oldham’s (1976) job characteristics model (JCM) is one of the key theoretical approaches to job design developed over the last three decades (Parker, Wall & Cordery, 2001). Briefly, the JCM holds that a number of individual work outcomes, such as motivation and job satisfaction, are at least to some degree determined by three critical psychological states — meaningfulness of work, felt responsibility and knowledge of job results. Hackman and Oldham (1976) argue that these three states are

1 Technically referred to as a Graphical User Interface (GUI) this relates to the use of screen displays that rely on cursor navigation, drop down menus and “drag ‘n drop” operations for the management of data, files and the entry of information (as opposed to monochrome menu or DOS driven systems).

27

determined by five core job characteristics. The core job dimensions of skill variety, task identity and task significance collectively represent characteristics responsible for felt meaningfulness. The characteristics of autonomy and feedback represent felt responsibility and knowledge of results respectively. The model is now well regarded as a strong predictor of affective outcomes such as job satisfaction and internal work motivation (Parker et al., 2001; Schmidt & Daume, 1993).

Figure 2.1 Job Characteristics Model (Hackman & Oldham, 1976)

Using the JCM as a framework it is possible to discuss the potential of ES to alter task and function responsibilities and hypothesize as to the consequences of those changes. The JCM also provides structuration researchers with a means to both conceptualize and measure a key source of structure within an organization. An individual’s job design can do much to determine their action within an organizational context. Often the job performed by users encourages defined patterns of communication, determines the amount and type of resources available, determines their social standing within the organization, and provides the stimuli for types of social interaction (Pierce, Dunham & Cummings, 1984). Importantly, using the JCM framework takes into account the wider social implications of job design, such as the perceived contribution made by the job incumbent.

28

The aim of this section is to consider the characteristics of ES systems as they pertain to existing job designs. Readers are directed to Chapter One (Section 1.5.1) for additional information concerning the core elements of ES technology. Three broad arguments are presented here in relation to the introduction of an ES technology, its impact on job characteristics and its effect on user acceptance. First, it is suggested that rather than merely adding an alternate or additional range of structural properties, ES technology has the capacity to alter existing structural conditions based around user’s job designs. Second, it is argued that system-initiated changes to job design may determine the degree of user acceptance demonstrated by users. More specifically, it is hypothesized that users will perceive an overall decrease in the enrichment quotient of their jobs, which in turn may be associated with reduced levels of user acceptance. Finally this argument is qualified, indicating that the extent to which system driven change may be linked to user acceptance is dependent on the degree of job related IT usage required of the user.

2.2.1.1 System Driven Job Characteristics Change

Section 1.5.1 attempted to provide an overview of what have been collectively termed Enterprise Systems (ES) technology. In the most generic sense ES technology provide organizations with a single, integrated software platform that aims to streamline the flow of information throughout the organization (Klaus et al., 2000). The increased quantity and quality of reporting data available to organizations is considered a significant source of competitive advantage (Lengnick-Hall et al., 2004). Organizations adopting ES technology exhibit a tendency to alter the manner in which they operate in order to accommodate the process flow orientation of ES design and to maximize the benefits afforded to them by aspects such as transaction automation.

While ES technology has several dominant characteristics, the context within which the system operates needs to be considered to gain a clear understanding of the systems’ aims and functionality (Klaus et al., 2000). This approach is also in keeping with the embedded case study design of the study adopted throughout the thesis.

In essence the organization had three principle aims it hoped to achieve with the introduction of their ES. The first and the most immediate was the replacement of an internally developed and maintained system that was dated in its technology, capabilities and usability. The new system was expected to allow the standardization of processes across the organization and allow new developments such as on-line enrolments, and cross functional data sharing, access and reporting. At a more strategic level the design of the ES allowed an information system that was centrally controlled at a corporate,

29

rather than at a faculty level (HLOAR, 2002, 6). Finally, it was anticipated that the new system would allow greater control and standardization of organizational process as well as driving a range of administrative efficiencies.

With these core elements in mind we now shift our attention to the impact of introducing this range of ES characteristics on the set of structural conditions represented by user’s job characteristics. The first three components of the JCM typically discussed are those of Skill Variety, Task Identity and Task Significance. Collectively, these components are the characteristics responsible for engendering a sense of “experienced meaningfulness at work”. It is suggested that given the aims of the system in this instance, overall users are likely to perceive a reduction in the characteristics relating to experienced meaningfulness following the introduction of the ES.

Meaningfulness

Skill Variety is defined by Hackman & Oldham (1976) as the degree to which a job holder must carry out a variety of different activities and be required to use a range of skills and abilities in performing their job. Intuitively the introduction of a new ES would result in certain jobs demonstrating an increase in skill variety due to a new and varied set of skills related to the use of the new system. Activities such as generating new reports and accessing data from other functional areas would require a new set of skills to be developed. However, these skills would not be limited to simply performing system commands. Advocates of ES point to an increase in skill variety by requiring users to effectively process and utilize newly available data. Improved recall, access and management of data arguably places far greater demands on users in relation to their analytical and decision making skills. A limited amount of research supports the notion that increased computerization may result in increased levels of job satisfaction via increased levels of skill variety (Ghani & Al-Meer, 1989; Zeffane, 1994). It should be noted however, that both Zeffane (1994) and Ghani & Al-Meer (1989) limited their findings to management positions or those already demonstrating high levels of skill variety.

Other questions can be raised about the introduction of a new ES and a corresponding increase in skill variety. One, in many cases the process could be legitimately viewed as a skills replacement, rather than a skill increase. Most ES will replace a legacy system, with users possessing skills based around its proprietary characteristics. With the advent of a new system the majority of users are likely to exchange a non-transferable skill set and proficiency level for a new, equally non-transferable skill set. Two, ES typically support frequent, reoccurring business processes (Klaus et al. 2000). Therefore it is possible that routine, laborious job tasks (but still requiring

30

a certain degree of knowledge and skill) are made obsolete by system transactions. Consequently most users again face the prospect of a redundant skill set due to an ES’s introduction. For example, Haines (1999) documented skill redundancies experienced by users moving from a manual filing process to an electronic data warehouse. Whereas the filing department in the past were responsible for the physical command and control over files, their lack of IT competence meant file management required IT skills possessed by other technical experts. Process automation, combined with the tendency of management not to capitalize on the data distribution potential of ES may result in a situation whereby only those users already in decision making roles may benefit from the new system. Those already lacking in decision making authority or involvement are unlikely to benefit from the increased informational capacities of ES.

In the case reported here the new ES replaced a system that had been developed internally over a 10-15 year period. As part of the in-house development process users were able to request enhancements at a local level. In addition to the customized nature of the system, its interface reflected an older era of IT, with monochrome screens and function-key driven menu systems. In contrast the new system represented a range of innovations such as a “Windows” based user interface with mouse controlled drop-down menus and the capacity for multiple sessions to be running concurrently. As such the majority of users possessed a propriety skill set inherently linked to the peculiarities of the old system. Consequently the portability of user’s existing skills are likely to be severely limited given their close association with the new system. Therefore it is argued that in this instance the majority of users are unlikely to perceive an increase in skill variety as a result of a new ES. Rather, if users perceive any change at all it is suggested that they will perceive a decrease in their skill variety quotient.

Task Identity is the degree to which a job requires completion of a whole, identifiable piece of work – that is doing a job from beginning to end, with visible results attributable to the job holder’s performance. On the face of it few characteristics of an ES lend themselves to positively altering the task identity of any given job. In fact, ES’s reliance on sequential processing and a high degree of process integration may conceivably result in decreased task identity. Because ES adopt a process view of organizations, less emphasis is placed on functional departments, groups or individuals (Davenport, 2000). Rather, task and process responsibilities defer to the cross-functional demands and requirements of the system. System designers tend to place a higher premium on the work carried out by users as a component within the process flow. The frequent coupling of process re-design with ES implementations is widely acknowledged (Hall, 2002). Conventional job design based around functional isolation is discouraged and

31

perceived to be contributing to resource and process inefficiency (Mitchell & Zmud, 1999; Mitev, 1996). A shift towards a process orientation of work design is unlikely to result in users perceiving an increase in task identity - their ability to complete and take responsibility for an entire piece of work. Users are more likely to perceive their contributions as being “dwarfed” by the system and have difficulty separating their job tasks from others who also contribute to the process. Arguably this could result in an inability to assess the completeness of their own contributions. As previously mentioned, “re-engineered” job role changes may also leave users unclear as to what now constitutes a “whole and identifiable piece of work”. In addition, system design elements may have removed aspects of user’s jobs that are now essentially automated, therefore increasing the potential for reduced task identity.

As discussed, the intention of the case organization’s system was to realize administrative efficiencies and increased information control via the technology’s process consistency and standardization capability. It is considered likely that the introduction of such characteristics will result in much of what has been discussed above. Consequently it is expected that overall users are likely to perceive a reduction in task identity following the introduction of the ES.

Task Significance is the perceived impact of the job on the lives or work of other people whether within or without the organization. As discussed, ES rely heavily on the flow-on effects of information throughout an organization. Consequently, the process orientation of ES with elements such as real time processing and single point data entry may impart on some users a greater sense of involvement in the organization and its operations. The knowledge that their component of the work process is vital to ensure effective transaction processing “down the line” could conceivably lead to greater levels of task significance. It is also likely that an ES could place greater demands on a small portion of users in relation to their analytical and decision making skills (Huber, 1990). If perceived in a positive light, changes in expectations concerning increased decision making and the results of improved analysis could result in increased task significance for that particular group. For example, if a user is now making decisions previously made by others, or able to make more meaningful contributions to the development of corporate strategy, it is likely that the user will experience a greater sense of responsibility for their activities.

However, the potential of ES to devolve decision making throughout the organization and promote employee empowerment is often tempered by the tendency for systems to reinforce or establish control structures desired by management (Hall, 2002; Koch, 2001). Therefore, the extent to which system characteristics are enabled and deployed

32

in such a manner to increase workforce empowerment relies heavily on management’s willingness to relinquish control to its members further down the hierarchical chain. This is further exacerbated if those responsible for re-configuring work processes operate within the pre-enterprise context and fail to move beyond the status-quo (Tansley et al., 2001). Consequently, while ES may possess the capability to devolve decision making and encourage greater participation, few organizations capitalize on this opportunity. In fact the introduction of the system is typically used to enforce greater compliance to procedure (Hall, 2002).

A review of the case organization’s ES technology characteristics would appear that consistency, standardization and centralized control are the primary focus of the organization, not the devolution of authority and information. In fact company documents readily acknowledge that users “may need to sacrifice user convenience provided by custom built databases for the benefits of greater data integrity, reporting flexibility and process consistency” (HLOAR, 2002, 7). As a result it is unlikely that the majority of users in this case will perceive an increase in their task significance following the introduction of an ES. Rather, due to measures aimed at securing management control over organizational processes, users are likely to perceive a reduction in their levels of task significance.

H1a: Users will anticipate a reduction in the characteristics relating to experienced meaningfulness of the work carried out by users following the introduction for ES technology.

H1b: Users will perceive a reduction in the characteristics relating to experienced meaningfulness of the work carried out by users following the introduction for ES technology.

Autonomy

The degree of freedom, independence and discretion in scheduling work that the job provides is all encompassed by the term, autonomy. According to the JCM, autonomy is largely responsible for producing experienced responsibility for work outcomes. Much of what was discussed in the previous section is also relevant in relation to the impact of ES on perceived levels of autonomy. The rhetoric surrounding ES would suggest that greater access to information, improved opportunities to utilize and manage data and the capacity to decentralise decision making would result in more autonomous working conditions (Hall, 2002). Clearly any system that allows a user to reduce their dependence on others for the distribution and access to essential data can be seen as increasing the degree of job autonomy. However, as with task significance

33

much of an ES’s capacity to increase autonomy is dependant on those introducing the system into the organization. Only if management actively engage in configuring the system to maximize its potential for increased autonomy will it actually occur. Otherwise users are likely to perceive ES as reducing autonomy due to the mandatory nature of the system and the increased control afforded to management.

A consistent theme within the ES literature is the deterministic nature of this technology type: “one major criticism of enterprise systems is their tendency to impose their own logic or business process on companies” (Davenport, 1998). Unlike stand alone packages (e.g. word-processing) or function specific software, ES are mandatory in nature due to their inherent association with work processes. Researchers have observed increases in workload monitoring and scrutiny, as well as changes in accountability and audit trails of work following an ES implementation (Haines, 1999; Lau, Wong & Law, 2001). Therefore, the highly integrated nature of ES and organizational function ensures that system use is no longer a voluntary act, but a necessary one in order to fulfill job requirements. Whereas in the past employees may have had discretion over how and when a task should be carried out and who they went through (or who they didn’t) system requirements now dictate process flows.

Increased levels of scrutiny and visibility afforded to management by an ES can also effect levels of perceived autonomy. Recent research has demonstrated the capability of ES to allow greater monitoring of users and the ability to enforce compliance (Sia, Tang, Soh & Boh, 2002; Tansley et al., 2001). A small but persuasive body of work has also consistently shown the tendency of ES to reinforce the existing command structures of organizations (Besson & Rowe, 2001; Koch, 2001; Orlikowski, 1991; Tansley et al., 2001). Previous modes of operation may have necessitated “bending the rules” in order to achieve required performance levels, or to circumvent bureaucratic procedure to ensure customer satisfaction (Pollock & Cornford, 2004). In contrast, the “panoptic” characteristics of ES allow management greater ability to verify procedural adherence, therefore reducing the ability of employees to engage in work-arounds, however legitimate (Sia et al., 2002).

Given that the stated objectives of the case organization’s system are to ensure “process consistency”, to “increase the accuracy of reporting data” and stress the need to “ensure maximum alignment between people, process policy and the system” it would appear that increased autonomy for many users is unlikely (various internal documents – see Section 1.5.1). Consequently it is considered likely in this instance, that overall employees will perceive a reduction in their levels of autonomy following the introduction of an ES.

34

H2a: Users will anticipate a reduction in autonomy following the introduction of an enterprise system.

H2b: Users will perceive a reduction in autonomy following the introduction of an enterprise system.

Feedback

Feedback concerns the degree to which carrying out job activities results in clear and direct information about the effectiveness of performance. Sia et al. (2002) outline three characteristics of an ES that may provide employees with greater feedback in their job role: comprehensive system tracking of data and procedures; greater accountability for process transactions; and an improved ability to manage aspects such as cost effectiveness, productivity and efficiency. These characteristics may result in greater visibility to management, employees and colleagues in relation to user performance and fulfillment of job tasks. This increased level of exposure could be translated into higher levels of feedback for employees in the performance of their job role. Earlier exploratory research has reported that while ES can provide feedback in the form of error detection and reporting, this failed to constitute a significant improvement in the overall feedback mechanisms currently present within job roles (Grant & Uruthirapathy, 2000). Participants indicated that while system initiated error detection did increase it represented a minimal component of the feedback elements present within their job. Feedback elements driven by organizational policy or initiated by supervisors were considered more relevant in terms of enhancing the psychological state of “knowledge of results”.

It is highly likely that given the case organization’s focus on administrative efficiencies and process standardization that feedback mechanisms such as job tracking and user accountability measures are included in the system’s design. Consequently the feedback mechanisms incorporated into user’s job designs via the mandatory usage of the ES may provide the opportunity for increased levels of feedback.

H3a: Users will anticipate an increase in feedback levels following the introduction of an enterprise system.

H3b: Users will perceive an increase in feedback levels following the introduction of an enterprise system.

35

In summary, this section argues that the introduction of ES technology has the potential to change existing job design characteristics. However, the argument goes further to suggest that the introduction of an ES may result in a decrease in the majority of characteristics, effectively reducing the overall motivational and enrichment capabilities of user’s jobs. While it may be possible for ES technology to drive empowerment and job enrichment initiatives organizations leveraging the inherent capabilities of ES based around process consistency, process standardization and administration are likely to result in an overall decrease in the enrichment quotient of user’s jobs. It is suggested that overall employees are likely to perceive a reduction in four of the five job characteristics and a reduction in overall job enrichment. The implications of such change in relation to user acceptance are considered in the section below.

2.2.1.2 System Driven Job Change and User Acceptance

The potential conflict between the structural conditions of existing job characteristics and the design elements of ES requires a closer examination of system driven job change as a possible determinant of user acceptance levels (Hall, 2002). It is argued that the introduction of ES technology is likely to result in a perceived decrease in overall job enrichment, resulting in reduced levels of user acceptance.

Under the JCM framework the enrichment or motivating potential of any particular job is typically taken as a multiplicative of each of the five reported job characteristics (See Chapter Three, Section 3.3.3.2 for more detail). The Motivational Potential Score (MPS) reported by the Job Diagnostic Survey (JDS) has received significant support within the job design literature (Parker et al., 2001). Consequently while the JCM emphasizes the critical nature of each of the five characteristics, the MPS allows researchers to gauge the overall potential of any one job to produce the outcomes hypothesized by Hackman and Oldham. Recent reviews of the JCM have noted the ability to predict affective response outcomes such as internal work motivation, general job satisfaction, and growth satisfaction (Boonzaier et al., 2001; Parker et al., 2001). This would appear particularly relevant when attempting to assess user attitudes towards the impact of system driven job change.

When considering technology related job re-design the three characteristics of skill variety, task identity and autonomy appear to be consistently related to outcomes such as job satisfaction and motivation. Early case study research by Preece (1987) documented decreases in employee motivation resulting from job de-skilling following the introduction of new printing technology. More recently Wall et al. (1990)

36

investigated the job design implications of a computer controlled manufacturing technology (characterized by improved levels of autonomy, task identity, and skill variety). Engineers using the new technology demonstrated higher levels of intrinsic job satisfaction and reported reduced levels of perceived work pressure. In terms of information technology Ghani and Al-Meer (1989) identified a relationship between the use of personal computers (PC’s) and job satisfaction. Interestingly their findings indicated that the results were particularly strong if the incumbent’s job was characterized by high skill variety, task identity, autonomy and feedback. This result was supported by more recent research reporting an interaction effect between autonomy and variety as it relates to satisfaction (Dodd & Ganster, 1996). High variety jobs with increased levels of autonomy were better able to predict satisfaction than were low variety jobs. Similar results were found for performance as an outcome variable. The link between system characteristics offering users a greater sense of control, the capability to reduce routineness or increase variety, and the outcomes of motivation and commitment to the IT system, has also been confirmed more recently by Gill (1996). Finally, Zeffane (1994) found support for skill variety, as well as a weak negative relationship between computer dependence and job satisfaction.

Based on the brief review above it would appear that systems driven changes enabling greater levels of autonomy, skill variety, task identity and feedback are likely to result in greater levels of user job satisfaction. If so, then the reverse may also be true, that systems inhibiting or reducing job characteristics such as autonomy and skill variety may result in reduced levels of job satisfaction. It is considered likely that if a newly introduced ES is perceived to be responsible for a decrease in user’s affective responses to job design (such as satisfaction and motivation) then user acceptance of that intervention is also likely to be low. Users perceiving or experiencing a decrease in job characteristic ratings leading to a reduction on MPS overall and attributing that decrease to the introduction of an ES are likely to view the system in negative terms.

H4a: Users will report a positive relationship between anticipated job characteristics change and levels of user acceptance at Time 1.

H4b: Users will report a positive relationship between perceived job characteristics change and levels of user acceptance at Time 2.

While intuitively appealing the argument presented above lacks a degree of specificity. A possible criticism is its “one size fits all” approach, failing to consider a number of possible moderating variables that may effect the relationship. One variable capable of moderating the degree of system driven job change experienced by users is the

37

degree of system usage required by the user in their job role. For example, Davis, Cox & Beale (1991) investigated the prevalence of repetitive VDU work over four job categories - data entry operators, administrators, programmers and supervisors. Only data entry operators demonstrated significantly lower levels of job satisfaction and a corresponding lack of control over their work. Similarly, Zeffane (1994) identified a negative relationship between the extent of computerization and the degree of dependence on computers reported by non-managerial employees. Therefore the extent to which users report a positive relationship between system driven job change and acceptance depends on the extent to which they are required to use the system. For example, senior roles typically involve greater levels of decision making, analysis, and interpersonal roles incorporating elements such as leadership, performance management and various communication and liaison roles (Pavett & Lau, 1983). Hence the degree to which an individual holding a senior position is beholden to the use of information technology is greatly reduced. While an ES may remain an essential component of the job role it is less likely to define the core elements of the position and therefore less likely to alter significantly those core elements. In such cases it is considered unlikely that system driven job change would have a significant impact on user acceptance levels. In contrast, users such as Davis et al.’s (1991) data entry operators, are more likely have a higher degree of interaction with the system as part of their job role. High users are more likely to have a higher component of their role determined by system protocols or process flows, and consequently, are likely to experience a higher degree of job change following an ES implementation. As such it is expected that high users will report a stronger relationship between job design and user acceptance following the implementation of ES technology.

H5: The degree of system usage required by the user will moderate the relationship between system driven job change and user acceptance, such that high users will report a stronger positive relationship between job change and user acceptance than low users.

2.2.2 Summary

Overall Section 2.2 indicated the intention of the researcher to investigate the structural implications of introducing ES technology into an organization. In particular two structural components of an organization were considered likely to be affected by the introduction of an ES – user job designs and social network configurations. Section 2.3 outlined a series of hypotheses based around the notion of system-initiated job design. In summary, it was argued that the characteristics of ES technology, along

38

with the manner in which they are typically introduced into organizations, are likely to result in a perceived change in job characteristics by users. Further, it was suggested that a perceived change in job characteristics would lead to a corresponding change in user acceptance. However the argument was qualified somewhat, indicating that the degree of system-driven job change experienced by users would be moderated by aspects such as job-related system usage.

The job design implications of ES technology represent a useful starting point in considering the structural implications of ES technology (Poole & DeSanctis, 2004). It is open to criticism however, particularly when considering user acceptance from a structurational approach. First, it tends to represent an overly deterministic approach to the effect of technology on user action, behavior and roles played within an organizational setting. Second, it is especially weak in its ability to predict activities undertaken by users in response to anticipated or perceived changes in job design. As it stands the approach struggles to acknowledge the capacity of actors to determine their own work context and the manner in which ES technology may be incorporated into that context, regardless of its mandatory nature (DeSanctis & Poole, 1994) .

To address some of the acknowledged weaknesses the part played by informal network configurations is also considered. Particular attention is paid to the ability of social networks to play a dual role during the introduction of ES technology – which the job design approach in isolation cannot provide. The remainder of the chapter will discuss the potential for social networks to act as occasions for attitude formation as well as the capacity of networks to act as mechanisms for attitude transmission between users. However, before Section 2.5 examines the role of network interactions as mechanisms for attitude transfer the chapter will also address some of the network implications of ES technology. Section 2.4 below considers the potential of ES technology to act as a catalyst for “network shock” and the ensuing consequences in terms of user attitude.

2.3 System Driven Changes to Informal Network Structures

A key contribution of the thesis lies in its application of network analysis concepts to the issue of user acceptance in an ES technology context (See Chapter One, Section 1.5.3 for an introduction to network analysis). Pursuing such an approach offers user acceptance and structuration researchers a number of advantages. First, a far greater degree of attention can be paid to a user’s social context and the impact it may have on user acceptance outcomes (Karahanna & Straub, 1999). Structuration theory by its

39

nature fails to offer researchers directional guidance and is more akin to a theoretical ideology or philosophy than a prescriptive, hypothesis-generating model (Poole & DeSanctis, 2004). As social network configurations represent both structures and social action, the “duality of structure” can be closely observed, particularly when adopting a longitudinal approach. Therefore adopting a network analysis approach allows researchers to both measure and observe the dynamic process of structuration. More specifically, adopting a network analysis approach lends a methodological and theoretical specificity to IT research adopting a structuration perspective.

Typically social networks measured within an organizational setting are concerned with aspects such as communication, friendship, resource exchange and influence. Networks of this type present users with a context capable of determining their behavior, access to resources, opportunities and hindrances in how they go about their daily activities (Nohria, 1992). However, consistent with structuration theory the actions of employees also serve to reinforce or alter the structural nature of social networks over time. Therefore while job characteristics represent one set of defining structures within an organization, another less tangible but equally powerful set can be found in the social networks of employees.

It is considered that network structures have the potential to play a dual role in determining user attitudes towards newly introduced ES technology – in facilitating the formation and transmission of attitudes of users toward newly introduced ES. On one hand it will be argued that under certain conditions system driven changes to existing network configurations may help form attitudes towards the new technology. On the other it is considered that network configurations and the actions of users within those networks will help to transmit attitudes between users about the new technology.

This portion of Chapter Two considers the potential for ES driven network change to act as a catalyst for attitude formation. The main thrust of this section is devoted to discussing the ramifications of users perceiving or experiencing changes in their informal networks, and their subsequent reactions to those changes. More specifically, it is argued that the degree of support demonstrated for ES technology may depend on the position or location of users embedded within resource exchange networks. Two network positions are considered in particular, those of central actors within a network and those identified as boundary spanners between networks.

40

2.3.1 Networks and Organizational Change

Exploring the structural characteristics of an organization via a social network approach has steadily gained legitimacy within the management and organizational change literature (Borgatti & Foster, 2003). Recent publications have also shown an increasing interest in the potential for network analysis to further understand the change process (e.g. Stevenson, 2003). However, due to its relative youth as a methodological and theoretical approach only a limited number of researchers have applied a social network approach to the study of organizational change. As yet few attempts have been made to address the role of relational networks in the facilitation and deployment of organizational change interventions (Khan et al., 2003).

McGrath and Krackhardt (2003) recently put forward their “strong ties” theory of organizational change. They argue that a high degree of intra-organizational interdependence and interaction results in higher levels of change acceptance. The rationale behind the model is that multiple strong ties across organizations assist in the transfer and sharing of information aiding a reduction in both conflict levels and the presence of factions acting to influence the change process. The existence of dense strong ties across organization sub-groups also helps to engender a greater sense of trust between members, particularly if those strong ties are based around friendship. A small amount of research tends to support the strong ties model. Krackhardt and Stern (1988) and Tenaski & Chesmore (2003) both noted that the presence of strong, dense ties within and between organizational units facilitated change implementation success. Tenaski & Chesmore (2003) reported that increased density within change recipient groups and between implementation units resulted in increased on-time implementation and usage. Krackhardt & Stern (1988) found the presence of multiple strong friendship ties helped organizations deal with crises more effectively and in a more timely fashion. Others such as Brown and Quarter (1994) highlight the tendency of employees to trust and rely on social referents for information regarding the implications of a change process. Their results also indicated that in periods of uncertainty, decision-making is made via a process of identifying with network opinion leaders and comparing information with other network actors. At a more pragmatic level, Mohrman et al. (2003) advocate planned networks of multi-level and multi-disciplinary members to facilitate change activities. They also note the success of change activities depends heavily on the presence or creation of networks able to direct the exchange and sharing of knowledge.

41

While the work reviewed above highlights the valuable role played by networks in facilitating change, less emphasis has been placed on the implications of network disruption arising from organizational change. There continues to be a lack of attention paid to the structural implications of organizational change initiatives and their impact on the established social networks of employees. In the context of this study it is also important to consider the type of change being experienced (Rafferty & Griffin, 2001). This is no more obvious than in studies investigating technological change. A number of commonly cited and more recent studies all place a reduced emphasis on the characteristics and the design of the technology in relation to the outcomes of the change process (e.g. Aydin & Rice, 1991; Burkhardt & Brass, 1990; Burkhardt, 1994; Hislop, Newell, Scarbrough & Swan, 2000; Macri, Tagliaventi & Bertolotti, 2001; Rice & Aydin, 1991; Robertson, 1989). However, as recently argued by Orlikowski and Iacono (2001) this hinders the development of IT theory capable of reflecting the increasingly dynamic technologies operating within organizations and prevents a greater understanding of IT phenomena. The lack of attention paid to the type of technological change being experienced, as well as the network implications of the change, highlights the need to examine whether user acceptance may be linked to the degree of disruption caused by ES technology to user’s social networks.

2.3.2 Implications of ES Technology Driven Network Shock

Earlier research has alluded to the need to pay greater attention to the implications of network disruption or shock (e.g. Burkhardt & Brass, 1990; Burkhardt, 1994; Mohrman et al., 2003). Network shock is defined here as forced or non-emergent changes to established patterns of interaction. Typical sources of network shock may include: Job re-design activities such as multi-skilling, job sharing and job rotation (Krackhardt & Porter, 1986); downsizing and departmental restructuring (Shah, 2000); training and development activities; and the introduction of new technology (Burkhardt, 1994). All of these change types, inadvertently or by design, may force actors to alter their established and regular patterns of interaction within their work context.

Research investigating technologically-based change provide excellent examples of technology-induced network shock. For example, Barley (1990) and Haines (1999) document the challenges faced by previously powerful employees following the introduction of new technologies. In each of the case examples their lack of IT competence meant they were either bypassed for information or they themselves had to seek previously unsolicited advice. Burkhardt and Brass’ (Brass & Burkhardt, 1993; Burkhardt, 1994; Burkhardt & Brass, 1990) longitudinal study is another of the few to recognize that change activities do (at least in the short term) alter network

42

configurations. For example, Burkhardt and Brass (1990) identified changes in advice network characteristics, namely measures of centrality and power following the introduction of a health information system. Advice network changes were largely attributed to employee system proficiency and knowledge as team members became dependant on those regarded as system experts for advice and assistance. Questions arising from this study centered around the permanence of network disruption and the long-term effects of forced network change. Barley (1990) shared similar thoughts, declaring the introduction of IS/IT as “occasions for structuring”. He also noted that more work was required to ascertain whether old patterns of interactions reasserted themselves or were permanently altered following the introduction of the technology. While the examples reviewed here do not consider ES technology per say, they are still considered useful to highlight the need to investigate what impact this contemporary and prevalent technology type may have on informal network configurations.

ES technology is considered particularly capable of occasioning far more permanent changes to social network interactions. As reiterated throughout the thesis, the potential for increasingly integrated ES technology to change work processes, communication patterns, and reporting relationships are all areas of particular interest to organizational researchers (Chung & Synder, 2000; O’Mahony & Barley, 1999). ES technology is widely touted for its capacity to decentralise information and access to data (see Section 1.5.1). Essentially this means previously unattainable information maybe available instantaneously to a far wider range of organizational constituencies. Wider access to critical data and resources has the potential to redistribute decision making concentrations, with changes to the amount and type of individuals involved in decision making as well as a decrease in both face-to-face decision making and the number of hierarchical levels required for organizational decision making (Huber, 1990). The highly integrated nature of ES technology and organizational function also ensures that system use is no longer a voluntary act, but a necessary one in order to fulfill job requirements (Brown, Massey, Montoya-Weiss & Burkman, 2002). Researchers have noted changes such as increases in lateral (horizontal and inter-group) communication, increases in workload monitoring and scrutiny, shifts in resource and information providers, as well as changes in accountability and audit trails of work following the introduction of ES technology (Haines, 1999; Lau, Wong & Law, 2001). Therefore due to ES’s capacity to alter the rate, quantity and quality of available information, they have significant ramifications for social networks based around information and resource transfer (Huber, 1990). It is highly possible that in order to achieve the information and data flows required by ES technology the manner in which people communicate, access information and exchange resources may be fundamentally altered. These changes will by default alter the social networks constructed around resource exchange and dependence.

43

As discussed on several occasions, a key aim of the case organization’s ES implementation was to achieve data accuracy, integrity and consistency. The system’s introduction also represented a move away from locally “owned” systems to one that reflected organizational, rather than faculty processes. Other ES characteristics such as password controls and transaction automation allowed greater control over transactions that represented a conflict of interest if currently performed by one individual or group (Sia et al., 2002). Inherent capabilities such as these therefore improve data and process integrity, the intention of the case examined here. It is suggested that as the flow of data throughout the organization shifts away from local repositories to a series of organizational wide process flows it is likely that the degree of resource exchange between users will effectively increase.

H6a: Social networks based around resource exchange will increase in density following the introduction of ES technology into an organization.

It is also suggested that the introduction of new technology also stimulates the need for users to increase their advice networks following the introduction of IS/IT such as ES technology. Prior research has already identified the tendency of employees to interact with those considered early adopters of technology (Burkhardt & Brass, 1990). As discussed in relation to job design, the introduction of ES technology invariably results in users needing to develop new skills based around the systems functionality. When faced with the need to use new technology and develop new skills it would seem likely that users would attempt to maximize their sources of advice and assistance. Therefore it is proposed that following the introduction of an ES, users will increase their advice networks in an attempt to reduce their initial knowledge and skill gap.

H6b: Social networks based around advice sharing will increase in density following the introduction of an ES technology into an organization.

The consequence of these changes discussed above is a scenario whereby established patterns of relational exchange are discontinued. In their place new, alternate pathways are created and new patterns of resource exchange emerge. This is represented below in Figure 2.4 as a simple diagram with lines indicating an exchange relationship between “actors”. The dashed lines in the second network diagram represent new pathways of resource exchange, effectively bypassing previously powerful actors A, B, C and D (see over).

44

Figure 2.2 Implications of Network Shock on Resource Exchange Networks

Network exchange theory (NET) is considered useful to frame the argument presented here. Supporters of NET concern themselves with structural sources of power, identifying positions, resources and network interactions that determine power distributions (Walker, Thye, Simpson, Lovaglia, Willer & Markovsky, 2000). Actors maintain their source of power based on their ability to prolong the dependant relationship existing between themselves and other actors. If alternate access to the source of dependency occurs this reduces the strength of the power relationship between the two actors. Applying NET would suggest that network shock acts as a catalyst for changes in dependence relationships among users. It is conceivable that levels of change acceptance are linked to the degree of change, either perceived or experienced by employees in their dependence relationships. Those peripheral individuals previously dependant on others for information and resources, but now having access to alternate sources due to the changes, will most likely endorse its introduction. Previously powerful actors, due to their control of essential information and resources, will in all probability see the proposed change as a threat to their vested interests and consequently resist its implementation.

With the aim of furthering the above argument we shift our attention to the consequences of network shock on the attitudes of employees towards the introduction of an ES. It is considered likely that the degree to which an employee supports or resists an ES is determined partially by the degree to which their existing network configurations are disrupted. The potential for ES to radically alter the political arena can be perceived as a source of opportunity for some actors and a source of considerable threat to others. Schien (1992; cited in Schwarz & Brock, 1998) cautions that even a perceived loss of control over information or knowledge will stimulate attempts to subvert the process in order to secure a hold over current sources of power. Consequently the key argument put forward here is that the degree to which an ES is endorsed or resisted by

45

users depends heavily on the net effect of those changes to their networks of resource dependence. The next section lends some specificity to this argument, proposing a set of hypotheses based around a user’s position within network structures.

2.3.3 Actor Centrality

An obvious concern for researchers and IT practitioners is to identify users potentially susceptible to changes associated with exogenous network shock such as the introduction of an ES. Previous research examining the structural attributes of networks has identified two network properties considered likely to be affected by forced network change – actors with high centrality and those identified as boundary spanners.

2.3.3.1 Central Actors

An essential element of network research has been the concept of centrality and its relationship to power in networks (e.g. Brass, 1984; Ibarra & Andrews, 1993). Centrality can be defined as the extent to which it reflects an actor’s position relative to the multiple exchanges characterizing the system as a whole (Ibarra, 1993). The degree of centrality possessed by any node within a network is indicative of its activity level (Scott, 2000). A node with a high degree of centrality will experience a high degree of interaction with other nodes in the network (Brass, 1995). Typically this node will act as a conduit, facilitating the transfer, distribution and dissemination of resources such as information, ideas and attitudes between nodes (Brass, 1995). Consistent with resource exchange theory, this control over the distribution of resources affords highly central actors increased levels of informal power within an organization (Brass & Burkhardt, 1993). Centrality has other significant advantages, including an increased awareness of an organization’s political environment and agendas, as well as understanding the political implications of organizational change initiatives (Ibarra, 1993). While centrality as a general concept is easily understood a number of theoretical and methodological alternatives have been offered up as sophisticated extensions of the basic idea. The most commonly discussed distinctions are degree, betweenness and closeness-centrality (Freeman, 1979). Others include, information (Stevenson & Zelen, 1989), flow-betweenness (Freeman, Borgatti & White, 1991), power (Bonacich, 1987) and eigenvector centrality (Bonacich, 1972).

46

Burkhardt and Brass (1993) suggest that examining power from a resource dependence perspective requires researchers to consider centrality in terms of actor alternatives, access and control. The discussion below highlights the value of considering both Degree centrality and a more recent refinement of its theoretical underpinnings, Bonacich’s (1987) “power” centrality in their capacity to measure a user’s alternatives, access and control over resources within a network. Both degree centrality and Bonacich’s power measure have been used previously to measure links between social influence (Ibarra & Andrews, 1993) and the exercise of power within organizational contexts (Brass, 1984; Burkhardt & Brass, 1990). Those wanting a more comprehensive review of other centrality measures are directed to Wassermann and Faust (1994).

Degree centrality concerns itself with the number of direct ties flowing to and from an actor. In terms of resource exchange, actors with high out-degree centrality possess influence in their network due to their rate of resource distribution. Conversely, those identified as having high in-degree centrality possess a relative amount of resource independence as they have access to a number of alternative suppliers. Those with equally high in and out-degrees would be considered the most powerful actors in a network due to their access to and control of resource distribution (Brass, 1984). Methodologically the degree centrality measure is useful as it takes account of directional relations, essential when considering the pattern of exchange relationships within a network.

While one of the most commonly discussed measures of centrality and useful as a conceptual springboard, Freeman’s (1979) measure has significant weaknesses. One of the most obvious is its inability to account for the influence of indirect links in an exchange network. Bonacich (1987) argued that only considering direct links ignores the wider network context and fails to acknowledge the “true” nature of exchange relations within networks. In doing so he introduced the notion of dependence between network exchanges into the measurement of centrality. He proposed that high degrees of centrality were only indicative of power within a network if those you were connected to were relatively isolated. In this scenario Freeman’s degree centrality appears to hold up, with the central actors possessing a high degree of control over resource exchange between peripheral actors. However, if connected to actors that are also well connected, an actor’s command over the flow of essential resources is greatly diminished. Therefore degree centrality must take into account the wider pattern of exchanges and accept the maxim that “all actors are not created equal”. In doing so Bonacich’s measure appears to hold more faithfully to the network approach by considering the structural constraints within which actors operate. In simple terms, calculating Bonacich’s (1987) power measure involves applying negative weightings

47

to connections with well connected actors, and positive weights to connections with actors with few other connections. Using this measure allows researchers to better identify those that may be threatened or affected by network shock. Using a power measure of degree centrality also reduces the risk of over estimating the role played by some actors in the exchange of resources throughout a network.

ES technology properties capable of altering established patterns of interaction are likely to be viewed by high “power” actors with some suspicion. Central actors perceiving a system instigated threat to their networks are likely to demonstrate reduced support for the system. In fact, some may even use the power and influence afforded to them by their central positions to influence those around them to also boycott the introduction of the change (Brass, 1984; Ibarra, 1993). For example, Cavaye and Christiansen (1996) describe a group’s subversion of an IT implementation over a seven year period. Similarly, Brown and Jones (1998) found that the actions of one influential group within an organization led to the eventual failure of a ES implementation. Perceiving the new system as a threat to its privileged relationship with other key network actors, the group engaged in a range of activities designed to discredit the usefulness of the system and highlight its perceived failings. However, to assume a “knee jerk” reaction of change resistance from central actors would be wrong. In some cases ES may assist in re-enforcing levels of dependence between actors (Koch, 2001; Macri et al., 2001). In cases where central actors perceive an opportunity to consolidate their exchange networks they may embark on a campaign aimed at encouraging support for the new system, effectively becoming informal “change champions” (Howell & Higgins, 1990). From this then, the following is proposed:

H7: There will be a positive relationship between change in network centrality and user acceptance.

2.3.3.2 Network Brokers

The other network property likely to be effected by an ES triggered network shock is that of the broker or cut-point. In a strict graph-theoretical sense a cut point is an actor that allows a connection between two or more components of a network (Scott, 2000). Actors identified as cut-points typically act as boundary spanners, linking otherwise isolated sub-networks to each other due to their relationship with members from each component. The relationship between two-cut points is referred to as a bridge (Wasserman & Faust, 1994). Removal of a cut-point or the dissolution of a bridge may result in a disconnected network and prevent the transmission of communications or

48

resources between various network components. Figure 2.3 below shows a network connected by cut points A and B. Together their relationship (e.g. between two production teams) forms a cut-set.

Figure 2.3 Example of Network Brokers (A & B)

In a similar fashion to that of a central actor the presence of a cut-set affords either cut-point actor a degree of control over the flow of resources throughout a network. A cut-point can act as a filter, choosing when, how and for “what in return” resources are released (Burt, 1992). Their ability to obtain resources unavailable to other actors within their network again sets up a pattern of dependence by virtue of their position. At the very least cut-points have greater access to valued resources such as information and advice due to their wider degree of network contacts (Burt, 1999). Isolated subgroups with few external ties have a reduced ability to influence change processes threatening their interests (Hislop et al., 2000). However, unlike the example provided in Figure 2.5 few organizational networks provide such clearly defined relationships. As is the case with centrality there are a number of alternative approaches to identifying actors acting in the capacity of brokers. In an attempt to better understand the specific nature of the brokerage role two approaches to the measurement of brokers are discussed below.

One approach is to examine another measure of centrality. While the ideas of centrality and cut-points are conceptually different, methodologically the two share commonalities. For example, one approach to measuring the extent to which an actor acts as a gatekeeper or broker is to use Freeman, Borgatti and White’s (1991) measure of flow-betweenness. Flow-betweenness is a refinement of Freeman’s (1979) original betweenness measure of centrality. Unlike degree centrality discussed previously, betweenness measures the extent to which actors are positioned between actors within a network. Central to the notion of betweenness is the geodesic, the shortest path

49

between two actors (Scott, 2000). Betweenness measures the extent to which actors are present on the geodesic between other pairs of actors. It follows that for an actor to demonstrate high betweenness they must be represented on a large number of geodesics, therefore affecting the flow of resources between a large number of actors.

The assumption made by betweenness (that actors will always choose the shortest path between each other) led to the development of Freeman et al.’s (1991) flow-betweenness measure. Flow-betweenness makes the alternate, possibly more realistic assumption that actors will use all available pathways open to them, relative to the length of the path. This is particularly evident in a scenario where choosing the geodesic may be politically un-advantageous or involve dealing with actors perceived as being unwilling to supply a requested resource. The other strength of flow-betweenness over previous measures is its ability to handle valued data, which is useful when attempting to measure the strength of dependence relationships (Freeman et al., 1991). Therefore using the flow betweenness measure helps to avoid the constrictions of Freeman’s (1979) measure. Clearly there are a range of reasons as to why actors would not chose the shortest (i.e. quickest) way to contact each other. Flow-betweenness accounts for the apparently irrational behavior of actors while identifying those actors that by virtue of their location along various paths, can be considered as resource brokers.

An alternate approach to identifying brokers gaining a large degree of attention from network, social capital and entrepreneurship researchers is Burt’s (1992) concept of structural holes. Instead of focusing on the degree of connectedness between network actors, Burt and colleagues instead consider the extent to which actors link other disconnected actors (individuals or groups) Structural holes are observed when gaps exist between disconnected or separated actors (Burt, 1992). For example, in Figure 2.4 below it can be seen that actor A fills structural holes existing between group X, Y and Z, however other structural holes remain between individual actors and between for example, groups Y and Z.

Figure 2.4 Structural Holes Within a Network

50

In doing so actor A has the opportunity to operate in a broker capacity to facilitate the transfer of resources (information, goods, affect) between the three groups (Burt, 1997). Research investigating the outcomes of structural hole brokers has tended to support the assertions of Burt’s (1992) theory. Recent studies have identified brokers as having greater involvement in innovation and creativity processes (Burt, 2004) enjoying greater management success (Burt, 2000) receiving greater rewards and demonstrating higher levels of achievement (Burt, 2002). Conversely Cummings and Cross (2003) found evidence to suggest that leader’s structural holes were negatively associated with performance.

As can be seen in Figure 2.4, in bridging the structural holes to groups Y and Z, actor A has an advantageous position over other actors within his or her group. Networks with large numbers of structural holes provide opportunities for a small number of actors for improved access to, and greater control of vital information and resources (Burt, 1997). Burt (1999) attributes both informational and control benefits to individuals spanning multiple structural holes. Informational benefits resulting from brokerage roles revolve around improved access, timing and referrals. The ability of brokers to selectively filter and disseminate information between structural holes is also recognized as particularly valuable. Similarly the brokerage role allows actors to favor various constituencies within the network that they connect, whether it be themselves or others depending on the specifics or politic of the situation.

However, a fundamental element of the benefits afforded to structural hole brokers is the notion of resource redundancy. Structural hole theory posits that a brokerage opportunity only exists if in bridging a gap an actor gains access to resources that are different to those they have already have access to. Burt (1997) measures redundancy based on the degree of cohesion present within a network and the amount of structurally equivalent2 actors the potential broker is connected to. Arguably one of the weaknesses of the flow-betweenness approach discussed previously is its failure to take into account the redundancy of resources available to an actor within a network. Therefore adopting a structural hole approach provides an alternate approach to identifying actors in brokerage roles that may be susceptible to changes in resource exchange stimulated by the introduction of ES technology.

2 In simple terms, the extent to which two actors are attributed the same role or position within a network based on their connections to exactly the same actors (Knoke & Kuklinski, 1982). See Scott (2000) and Wasserman & Faust (1994) for a more rigorous discussion of structural equivalence and its properties.

51

For example, the touted ability of ES systems to facilitate greater access to data for decision making (Tansley, Newell & Williams, 2001) represents the potential for a reduction in structural holes. However if as suggested by Koch (2001) the intent of an ES system is to reinforce the status quo and increase management control then it is highly likely that the introduction of an ES will see a strengthening or even an increase in structural holes. The process or transaction automation capability of ES may also result in the increase of structural holes for some users. If aspects of a user’s role are automated it is possible that brokerage relationships established based around that role will dissolve as their participation in the process becomes redundant. It is suggested that the extent to which users perceive an increase or decrease in their structural holes following the introduction of an ES will effect on their attitude towards the system. In support of this argument Susskind, Miller and Johnston (1998) reported a relationship between downsizing survivors structural hole experiences and their support for the change, as well as perceptions of organizational chaos. Importantly they found that a network shock (such as a downsizing initiative) did not have a universal effect. Employees experiencing an increase in structural holes following the change reported higher levels of perceived organizational chaos and higher levels of resistance to change. It is suggested that similar responses may be observed following the introduction of an ES. Therefore based on the material presented above the following hypotheses are offered:

H8: There will be a positive relationship between change in brokerage and user acceptance.

2.3.4 Summary

Section 2.4 discussed the potential for a second fundamental structural change to result from the introduction of ES technology. It was hypothesized that both advice and resource exchange networks would increase in density following the introduction of ES technology. The section also addressed the capacity of network shock to act as a catalyst for the formation of user attitudes towards ES technology. In short, ES typically provide greater access to information and alter protocols concerning the storage and retrieval of data. The introduction of this system type has also been shown to be capable of facilitating a change in control mechanisms or to support a wider political process (Koch, 2001). In either scenario the introduction of such system characteristics can act as a catalyst for changes to the dependence relationships between users. Access to resources has been previously realized as a key source of power within organizations (Pfeffer, 1992). On reviewing network exchange theory, it becomes clear to see that

52

any perceived or actual change to the flow, or opportunity to access essential resources could result in perceived power imbalances and shifts in dependence relationships (Brass, 1984). Therefore, an argument was presented to suggest that levels of user acceptance may be linked to the extent that a user’s resource exchange networks are altered. Those users previously dependant on others for information and resources (but now having access to alternate sources via the new system) will most likely endorse and support its introduction. Previously powerful actors, due to their control of essential information and resources, will in all probability see the introduction of the new system as a threat to their vested interests and consequently, resist the implementation. The argument was taken further to suggest that user attitudes towards the system depend on the structural characteristics of the network within which they were embedded. Users occupying highly central positions, those occupying brokerage roles and those in peripheral relationships were all considered in detail in relation to their possible acceptance levels.

However, networks are seen to play a dual role when considering user acceptance of ES technology. While social network change has the capacity to stimulate attitude formation, existing network configurations also have the capacity to transmit attitudes among users. It is the capacity of social networks to facilitate both attitude formation and transmission that represents the “duality of structure” evident within the structuration process. Section 2.5 explores this idea further, examining from a network perspective the behavioral implications of users affected by ES driven structural changes such as job design or social network change. In doing so the value of adopting a network perspective to examine structuration process is emphasized as it provides an opportunity to observe and measure human agency in response to the introduction of information technology.

2.4 User Behavior – Social Networks as a Medium for User Agency

To this point Chapter Two has only considered the attitudinal consequences of ES initiated structural change. Network shock leading to attitude formation has already been addressed, but it is useful to also consider the role of networks as mechanisms by which users exchange attitudes. In an extension to the theory posed above this next section considers the behavioral consequences of network shock. In doing so the theory acknowledges the second component of structuration theory - social action. The reader should be made aware at this point that this section will not be empirically tested in his thesis. The nature of the data required, along with the increased project size that this

53

component represents, means that it lies outside the empirical scope of the thesis. However, in attempting to present a theoretical piece of work that is both useful and represents an important contribution, it is considered vital to also consider the potential for networks to play another role, as a mechanism of attitude transmission between actors (Whetten, 1989). Addressing this component helps to answer the “so what” question also posed by Whetten (1989) and therefore works towards a theory able to contribute to the wider body of knowledge based around ES technology, social networks and user acceptance.

As a result this final section of Chapter Two intends to consider some of the behavioral implications of user attitudes formed by the events discussed in Section 2.4. While a range of behavioral responses may be considered, especially in the case of user dissatisfaction (Rusbult et al., 1988) the chapter will continue its focus on the role played by informal networks during the introduction of an ES. While established social networks may be altered by the introduction of ES technology, research also suggests that they play a role in determining user attitudes and hence the adoption and absorption of technology into an organization (Burkhardt, 1994; Burkhardt & Brass, 1994). Adopting a network analysis approach allows an opportunity to consider in detail the mechanics of “duality” and the process of structuration. While ES characteristics may represent potentially significant structural change in the form of job change or network shock, the actual degree of change is dependant on the reactions of users towards those changes. The actions of users holding specific network positions are considered in their capacity to affect attitude change throughout the network.

As demonstrated in 2.3 and 2.4 above, the nature of ES implementations often leads to differing attitudes towards the project either at an individual, group or organizational level. It is thought that users perceiving a change in circumstance resulting from the introduction of an ES would attempt to influence other’s perceptions of the new system. At one extreme some users may perceive the system as beneficial and actively endorse its implementation. At the other end of the spectrum users may perceive the system as detrimental and work towards preventing a successful adoption of the system. Importantly, due to the ambiguous and uncertain context created by the system’s introduction, surrounding actors are more likely to be influenced by those around them (Ellis, 1992), particularly by those holding powerful network positions, such as network centrality and brokerage roles.

54

2.4.1 Attitude Transmission – Brokers and Central Actors

A key assumption among social network theorists is that the social context surrounding an individual is a primary source of reference for the interpretation and development of perceptions regarding expected behaviors and relationships with others (Meyer, 1994). The notion that an individual’s social networks are a powerful determinant of attitude has significant implications when considering the behavioral reactions of users perceiving system driven job change or network shock. For example, while users may make their own judgments of the system’s utility based on job and task related criteria, they may also be subject to influence, or indeed influence others within their network as to the system’s worth. If empirically supported this may help to explain the apparent disparity between user acceptance levels and actual user behavior (Szajna, 1996).

One of the most commonly discussed theories used to explain the derivation of attitudes in the context of social networks is Social Information Processing (SIP) theory, initially proposed by Salancik & Pfeffer (1978). Proponents of SIP argue that individuals develop attitudes based on the information available to them at the time of developing the attitude. Further, they argue that a key source of information is an individual’s social environment, providing cues as to how events should be interpreted and evaluated. This is derived from the fundamental premise that “individuals, as adaptive organisms, adapt attitudes, behavior and beliefs to their social context and to the reality of their own past and present behavior and situation” (Salancik & Pfeffer, 1978, 226). Attitude similarity within social contexts is argued to occur in two ways. First, a process of socialization allows actors to interpret informational cues from other actors as to expected attitudes and behaviors within a given setting or towards a specific workplace phenomena (Rice & Aydin, 1991). Second, particularly in ambiguous situations, SIP facilitates the development of beliefs through a shared sense-making experience involving the development of an understanding of the ambiguous event via employee interactions (Chattopadhyay et al., 1999). The second point appears especially relevant given the degree of structural change potentially occurring following the introduction of an ES. It would be fair to describe the introduction of an ES technology as an event possessing a high degree of ambiguity. Given this, the extent to which users will depend on others around them to make sense of system driven changes is highly likely (Chattopadhyay et al., 1999).

55

2.4.1.1 Central Players and Network Brokers

Previous network research has supported SIP theory, showing the tendency of attitudes within network configurations to be similar in relation to newly introduced technology. For example, Aydin and Rice (1991) examined the relationship between “social worlds” and attitudes towards a newly introduced heath care system. Pre-and post-implementation comparisons of attitudes toward the system demonstrated differences in attitude based on occupational and departmental networks. Evidence also suggests that group norms and levels of attraction within groups can influence assessments of usefulness and actual use of newly introduced technology (Fulk, 1993; Schmitz & Fulk, 1991). Importantly, it would appear that these norms develop around network configurations and have demonstrable effects on user behavior (Postmes et al., 2000). Of considerable interest is a limited but compelling group of studies identifying a solid link between informal network position and use of power within organizational networks (Brass, 1984; Brass & Burkhardt, 1993; Ibarra & Andrews, 1993). Continuing the theme set in Section 2.4, this section will now consider the network positions of brokers and central players in relation to their behavioral responses to ES structural change.

Highly central positions (see section 2.3.3 for more detail) have been shown to correspond with perceived levels of influence (Brass, 1984), involvement in administrative and technical innovation (Ibarra & Andrews, 1993) and types of behavioral tactics employed in the exercise of power (Brass & Burkhardt, 1993). It is considered likely that some may use the power and influence afforded to them by their central positions to influence those around them to also boycott the introduction of the change (Brass, 1984; Ibarra, 1993). Several examples can be found within the literature describing the efforts of individuals or groups perceiving IT as a threat to their privileged position, and engaging in a range of activities designed to discredit the usefulness of the system and highlight its perceived failings (e.g. Brown & Jones, 1998; Cavaye & Christiansen, 1996). Others such as Macri, Tagliaventi and Bertolotti (2001) document the cancellation of an entire IT implementation due to fears that its introduction would increase the centrality and power of IT personnel. Ironically the original motivation for the purchase of the system was to reduce dependence on this group. Given the findings reviewed above, it is considered highly probable that central actors will use their position within network structures to influence others regarding the value and nature of the new system preceding or following its introduction.

56

It can be seen that a similar argument can be proposed for those users operating within brokerage roles. Burt (1999) makes a strong case for brokers not just as resource accumulators and disseminators but as opinion leaders within networks. He makes the point that brokers have the ability to influence (or at the very least) transfer opinion between groups of individuals. This is in contrast to highly central individuals who may only have influence within their network or group rather than between groups. It is arguable that those holding brokerage positions have the potential to influence a wider range of users than do those with high degrees of centrality.

Research examining the role of brokers during exogenous shock has tended to support the pivotal role they play in influencing actor’s attitudes towards change interventions. For example, Brown (1995) observed the politically astute machinations of a project team responsible for the introduction of a health management system. By identifying the predominant political actors within the organization the project team selectively chose what information was distributed to those groups. The approach resulted in a distribution of information appealing to the needs of each user while at the same time playing down the changes resulting from the introduction of the system. In contrast, Hislop et al.’s (2000) case study documents the failure of a disaffected group to prevent an ES implementation. They cite an example where functional isolation prevented a previously powerful production team resisting an ES implementation that reduced the source of their power within the organization. Their lack of connections to the actors responsible for decision making stymied their attempts to have their concerns heard and attempts to halt the process. There is also evidence to suggest that previously established brokerage relationships may resist dissolution, even though the original stimulus for the link is discontinued (Seabright et al., 1992). It is possible that brokers with no remaining “legitimate” connections to other users may attempt to sustain these links to not only maintain their source of influence but exercise that influence in relation to how others perceive the system. Susskind et al. (1998) found evidence of previously powerful brokers re-establishing their command over newly created structural holes post-change intervention, but failed to provide an explanation as to how or why.

As discussed, this section of the thesis lies outside the empirical scope of the thesis. As such a proposition, rather than a hypothesis is put forward to reflect the argument presented above.

Proposition 1: Actors in highly central or key brokerage positions will use the power afforded to them by that position to influence other peripheral users regarding their assessment of the system’s value.

57

2.5 Summary

Chapter Two provided the theoretical underpinning for the research carried out and reported throughout the remainder of the thesis. In all, eight hypotheses and one proposition were put forward, the contents of which are summarized below.

Chapter One introduced structuration theory as the over-arching perspective underpinning the research approach. Structuration theory as it applies to IT research concerns the capacity of technology to represent another set of structural properties that recursively constrain and enable user action within an organization. Within this framework attention shifted in this chapter to the relationship between the structural properties of ES technology and two additional sources of structure and social action within an organization – aspects of a user’s job design and informal resource exchange networks. Overall an argument was presented, suggesting that the capacity of ES technology to alter structural elements relating to both job design and exchange networks may stimulate both user attitude and behavior towards the newly introduced technology. The next two sections of the chapter then examined in detail the potential of ES technology to alter both job design characteristics and network configurations.

Section 2.2 used Hackman and Oldham’s (1976) Job Characteristics Model as a theoretical framework to consider ES characteristics in light of their potential to alter a number of job characteristics. It was argued that in the context of the case organization examined here users were likely to perceive a reduction in the job characteristics of skill variety, task identity, task significance and autonomy following the introduction of an ES. Building on this argument it was also suggested that user acceptance was linked to the degree of system driven job change experienced by users. In particular it was argued that users reporting high usage rates would report a stronger relationship between job change and user acceptance than low users.

Section 2.3 highlighted the potential of ES technology to stimulate what was defined as network shock. It was argued that ES characteristics such as process automation, changes in data retrieval and changed access controls can alter established dependence relationships based around resource exchange. System driven changes can represent a serious threat to the established interests of some and an opportunity for others to gain or consolidate power. Consequently, it was hypothesized that those individuals perceiving a positive consequence to system driven changes in social networks would demonstrate a greater degree of user acceptance. Conversely, those individuals perceiving negative consequences would demonstrate reduced levels of user acceptance. In particular it

58

was considered probable that those holding highly central positions or those acting in brokerage roles would be most likely to demonstrate positive or negative attitudes towards the system.

The overall theme of sections 2.2 and 2.3 was that user acceptance levels were dependant on the degree of system driven structural change experienced by users. This apparently deterministic approach may seem contrary to the notion of structuration. However, as suggested by Poole and DeSanctis (2004) for IT/IS researchers to apply causal logic while remaining within a structuration framework requires a degree of determinism to be present. This allows researchers to deconstruct and understand the structuration process. In doing so, as has been done here, the approach allows researchers to explain or at the very least hypothesize about user attitudes and behavior.

Section 2.4 considered a number of behavioral implications of users reacting to ES driven structural change. It was argued that powerful users such as those in highly central or in brokerage positions would influence those around them in order to encourage resistance or endorsement of the system. In doing so the final section of the chapter achieved a number of aims. First, it highlighted the value of adopting a network analysis approach to investigate and measure the dynamic process of structuration. The duality of structure is represented by users undergoing a process of interpretation (as in Section 2.3 and 2.4) and social action as they attempt to either reinforce the status quo or consolidate changes to their structural context by using the structural properties of their social networks. Second, it emphasized the dual role played by social networks when introducing ES technology. While changes to network structures may act as a catalyst for attitude formation (as shown in Section 2.4) existing network structures can also act as conduits for user attitude. In doing so the model overall makes a valuable contribution in its ability to highlight the tension existing between a user’s own appraisal of system driven changes and the weight of influence resulting from network interaction. On one hand users may make their own value judgments of the system’s utility based on job and task related criteria. On the other, users may be subject to influence based on how the system is perceived by others around them. Similarly, they themselves may use their positions within social networks to influence others as to its worth and value, hence furthering their own agendas. It is held that the recursive interaction between ES characteristics and the structural properties represented by network interactions and the outcomes produced as a result of users operating in this context is largely responsible for the degree of acceptance demonstrated for ES.

Chapter Three will outline the data collection procedures and methods undertaken in order to collect data capable of testing the hypotheses set down in this chapter.

59

Chapter 3 — Methodology

3.0 Introduction

Chapter Three provides an outline of the methodological aspects of the research undertaken to test the theoretical propositions put forward in Chapter Two.

Chapter One provided a review of the thesis and its contents, providing a justification for the research and its aims. Emphasis was placed on the lack of attention paid to contextual elements within organizations when considering user acceptance issues. In particular perceived changes to structural elements such as job design and social networks were put forward as possible key determinants of user attitude and behaviors towards newly introduced IT.

Chapter Two presented the theoretical model underpinning the research. Adopting a structuralist approach it was suggested that the acceptance of ES systems, at least in part could be attributed to the amount of ES driven structural change experienced by users. It was suggested that the inherent characteristics of ES systems were capable of disrupting the structural properties associated with job design characteristics and social networks. System usage as a moderating factor was also discussed, as was the dual role played by social networks in both forming and transmitting attitudes towards ES systems. In all eight hypotheses supporting the structural model of user acceptance were put forward.

The chapter is divided into three major sections. The first section outlines the characteristics of the case study research strategy, justifying the research design and data collection methods. A brief outline of the case organization is also provided. Aspects such as organizational history, structure and details of the new IT system are covered. The second section examines in detail each of the data collection methods used in the study. First, the development of a conventional survey instrument and its delivery are presented. In addition to covering areas such as questionnaire design, scale development and sample characteristics, details of the on-line survey design are included. Second, the chapter deals with the network analysis component of the study. Although network analysis has been known to the organizational sciences for over 20 years (Tichy & Fombrun, 1979) only recently has it received considerable attention (Stevenson, 2003). This section aims to provide readers possibly unfamiliar with network analysis with an understanding of the methodology undertaken as part of this thesis. Topics such as participant sampling, interview protocol, sociometric

60

measures, participant confidentiality issues and data storage are all covered in detail. Chapter Four will continue in this theme, examining the peculiar network measures and analysis used to report the results of the study in Chapter Five and Seven.

3.1 Research Strategy and Design

The research strategy employed in this study can be described as an embedded single case design (Yin, 1994). More specifically, it is a longitudinal analysis of a single case organization undergoing the implementation of ES technology.

There are number of reasons as to why the embedded case study approach was considered the most appropriate research design to adopt. First, a case design is most appropriate when the phenomenon under investigation is difficult to separate from its context (Yin, 1994), and the researchers are interested in examining the complex interrelationships between people, organizations and technology (Dube & Pare, 2003). The research questions posed in this thesis are concerned with the degree of structural change initiated by technology and the resulting attitudinal and relational change. Another strong theme throughout the thesis is the failure of ES technology to account for the contextual peculiarities of the organizations they are introduced into. It was therefore necessary to adopt a research design that allowed a range of methods that helped explore the dynamic interaction between ES technology, job design and social networks while taking into account the context within which this was all occurring.

The approach is also well suited to the examination of a phenomenon at several levels of analysis (Yin, 1994), which in this study involved the examination of individual attitudes, dyadic relations and the structural characteristics of groups and departmental networks over time. The longitudinal aspect of the research design was considered particularly important given the nature of the research questions. A longitudinal research design was considered better able to capture the extent to which structural change occurs following the introduction of an ES. Barley (1990) and more recently Poole & DeSanctis (2004) highlight the necessity for structuration researchers to adopt research designs capable of identifying both structural and social change.

Two principle data collection methods were employed in this study. These data sources were supplemented by the analysis of project documentation and semi-structured interviews collected from a select number of participants engaged in the network component of the study. Data relating to the degree of ES driven job change were collected by a survey method. Hypotheses 1 – 5 attempted to ascertain whether

61

the extent of user acceptance reported by users could be attributed to changes in job design characteristics. Job design research has long demonstrated its suitability to survey research. Survey measures such as the Job Diagnostics Survey, have received extensive support in their ability to investigate job design issues (Fried & Ferris, 1989; Teigs, Tetrick & Fried, 1992). Therefore in order to maintain consistency and comparability a survey research design was chosen as the principle method chosen to test hypotheses 1 – 5.

The specific nature of the remaining hypotheses prevented the use of a survey design to capture appropriate data. As outlined in Chapter Two social network analysis is principally concerned with establishing patterns of interaction between actors. Data collected for network analysis is necessarily dyadic in nature, requiring the collection of data that is non-anonymous and able to capture the nature and type of relations between actors. The specific and sensitive nature of the data and meant that a series of structured interviews carried out with a defined set of actors were the most appropriate method to use. The particulars of this data collection method are reviewed later in Section 3.3. Attitudinal data was collected from the participants via a paper version of the on-line survey. The design implications to having both paper and on-line versions are discussed further in Section 3.3.4.

Overall the research design employed can be summarized as an embedded longitudinal case study using multiple methods to capture data from a single case organization. Two principle methods of data collection were employed 1) an on-line survey distributed to all users throughout the organization, and 2) a series of structured interviews aimed at collecting social network data from a narrower, pre-determined set of users. Other ancillary methods of data collection included organizational documentation and a series of semi-structured interviews. The remainder of the chapter will discuss each of these area in detail as well as providing an overview of the research site.

3.2 Research Design

All data collected were from a large tertiary educational institution with campuses located throughout Queensland, Australia. Due to the request of the participating organization to remain anonymous it will be referred to using the pseudonym Erudio. Erudio claims to be the largest provider of post-secondary education in the state with over 200’000 international and domestic students enrolled each year. Service delivery provided by Erudio ranges from conventional classroom instruction to on-line learning, distance education, on-site training and long term traineeships.

62

A predominantly publicly funded institution, Erudio is divided into fifteen separate Centres located throughout the state, each offering their own courses. Due to a number of historical, geographical and political reasons the fifteen Centres have tended to develop rules, operational procedures and work processes specific to their own needs and requirements (HLOAR3, 2002, 7). Consequently, each Centre has been seen to operate “largely as autonomous entities for some years” (HLOAR, 2002, 7). At the time of collecting the data Erudio management had approved the introduction of HAL (also a pseudonym), an ES designed to manage the core functions of the organization such as student administration. The introduction of this system was also seen as a catalyst for a number of changes aimed at establishing consistency between Centres and the establishment of a single set of organizational protocols in relation to student administration (HLOAR, 2002, 3).

3.2.1 Student Administration System

This section provides an overview of the system and the context into which it was introduced. For more information refer to Section 1.5.1 Chapter One.

In June 2003 Erudio was given authorization to begin the configuration and implementation of a new student administration system - HAL. Prior to this decision each Centre used an in-house package which over time had been customized to reflect the particular requirements and requests of each Centre’s users. In keeping with the autonomous nature of each Centre, users were able to develop on-site Centre specific functionality. Unsurprisingly this led to inconsistencies in reporting and work processes across Erudio as a whole. This further reinforced the notion that each Centre operated independently and not part of the larger organization as desired by Erudio senior management and powerful external stakeholders. However, despite its apparent flexibility and customizability the old system suffered from dated software and a basic infrastructure that was over a decade old.

Built around a commercially available enterprise system, its aim was to standardize processes across the 15 Erudio Centers, allow new developments such as on-line enrolments and other benefits such as cross functional data sharing, access and reporting.

3 An internal report produced by Erudio, titled “High Level Organisational Analysis Report”

63

“The HAL project creates a strategic infrastructure delivering a student administration system with the capability to deliver timely, relevant information to both community and government stakeholders, in an open e-business framework” (Internal newsletter, Sept/Oct, 7).

Despite the size and magnitude of changes accompanying HAL’s introduction not all of the organization’s employees were directly affected by the system. The occupation classification most affected were administration staff. Other key occupational classes within the organization, such as teaching and general staff would, in this version of the system, not have any direct interaction with the new system.

3.2.2 Implementation Pilot Sites - City & Country

Due to the size and degree of change associated with the introduction of HAL, two Centres were selected to pilot the system. Following a successful implementation the plan was to roll out HAL to the remaining thirteen Centers 3-6 months after the pilot’s go-live date. Data for the study was collected at both pilot sites, hereafter referred to as City and Country.

The City Centre was one of Erudio’s larger Centres with three campuses located within 7 kilometers of the CBD in Queensland’s capital city, Brisbane. City has over 30’000 students enrolled in eight faculties. Courses offered by City range from graphic arts, to business studies, to applied science.

The Country Centre was one of Erudio’s regional centres with its two primary campuses located approximately 700 – 1000 kilometers from Brisbane. Country serviced regional, coastal and rural Queensland with four campuses and has approximately 25’000 enrolments per year. Course offerings at Country reflect the nature of its location with less depth than that offered by City, concentrating on areas such as business studies, engineering and hospitality. Due to its commitment to rural and isolated areas Country also has a higher proportion of distance learning activities.

3.3 On-Line Self Administered Questionnaire (SAQ)

The aim of the research is to determine the impact of elements collectively referred to as “social context” on user attitudes towards newly introduced IT. Consequently all of the outcome variables under consideration were of an attitudinal or perceptual nature. In addition to measuring user acceptance the study required the measurement

64

of perceived job design change and demographic variables. The method considered most appropriate to achieve this aim was the use of a self administered questionnaire (SAQ). This section justifies the use of this method for the intended purpose, details its essential design elements and in particular, its primary delivery mode – as an on-line survey.

3.3.1. Method Justification

A self-administered questionnaire (SAQ) was selected as one of two research methods used to collect data for the study. An SAQ requires participants to record their responses on a form distributed via conventional paper or electronic media (Fink, 1995). A commonly used method to effectively measure attitudes, beliefs and demographic data (Schmitt & Klimowski, 1991), the SAQ had a number of advantages in the context of this research.

As explained in section 3.2 the target population was one that was geographically dispersed. Even within the urban setting of the City Centre its campuses were located in and around the CBD and outlying suburbs. Therefore the logistical difficulties involved in engaging participants for various research activities reduced the available options. One method widely recognized as being appropriate for a geographically dispersed target sample is the SAQ (Sackett & Larson, 1990).

The characteristics of SAQ’s also mean that they possess a relatively high degree of delivery mode flexibility (Zikmund, 2000). At various stages of the research the survey was required to be distributed via paper, on-line or as an e-mail attachment. The SAQ format allowed multiple distribution methods without altering any of the original measures or research design.

SAQ’s are also relatively quick in providing results (Zikmund, 1997). As the research was longitudinal in design it was vital that a method be selected that effectively captured large amounts of data repeatedly over time. It was also important that in capturing that data repeatedly the demands placed on participants were minimized, SAQ’s allowed this to occur.

Finally, the cost effective nature of the SAQ was also appropriate given the financial constraints placed upon the project (Dillman, 1991). QUT Faculty assistance in the form of web design, HTML programming and web hosting reduced costs commonly associated with on-line survey development (Thomson, Surface, Martin & Sanders, 2003).

65

3.3.1.1 Advantages of On-line delivery

A number of additional advantages were also evident when choosing on-line delivery as the primary method of deployment.

For example, some have cautioned the use of electronic survey distribution due to the risks associated with biased samples (Thompson et al., 2003). Clearly, distributing a questionnaire via electronic means limits a sample to those with access to appropriate technology. However, in this case the target population not only had access to the appropriate technology, but it was the focal element of the study. The research was primarily interested in the perceptions of IT users to a new system. Consequently the researcher felt comfortable in making the assumption that the target population would have access to on-line technology.

The use of HTML code and the use of web design packages such as Dreamweaver also allowed far more latitude in relation to the “look and feel” of the questionnaire than the paper medium did. The use of web pages as survey forms allow forms with more advanced graphic layouts, greater use of color and font, and the inclusion of interactive elements and other visual features (Couper, Traugott & Lamias, 2001; Manfreda, Batagelj & Vehovar, 2002). In the case of the web survey reported here it allowed the design of a form that was both aesthetically pleasing and easy to fill out.

A number of technological enhancements helped encourage greater participant response, reduce the rate of missing data, non-response due to participant fatigue and improve perceptions of respondent anonymity. For example, a multiple login feature and a progressive save function presented throughout the questionnaire allowed participants to respond when they were able.

While SAQ’s as a research method are commonly cited as being relatively cost effective, conventional means of distribution do require significant funds. Printing costs, mail-outs, reminders, pre-paid envelopes as well as the significant costs associated with data entry can all accumulate. These costs cease to be a concern with the use of an on-line survey (Saxon, Garrat, Gilroy & Cairns, 2003). The ability of on-line surveys to automatically record and store data in a format suitable for analysis has other advantages other than reducing data entry costs. In particular automated data storage can eliminate the issue of keying errors when manually entering survey data (Saxon et al., 2003).

66

3.3.2 Method Limitations

While the SAQ is a commonly used research tool and acknowledged as having a number of legitimate benefits, it does possess a number of limitations (Podsakoff & Organ, 1986). Issues such as the risk of common method variance and poor response rates resulting in unrepresentative samples are two limitations relevant in this case.

Common method variance: A widely cited criticism of questionnaire driven research design has to do with common method variance (Williams, Cote & Buckley, 1989). Briefly, common method variance refers to variance resulting when measuring two or more variables from the same source and/or the same data collection method (Spector, 1994). The concern lies in the possibility that any error inherent in measuring one variable will contaminate the other, resulting in over-estimated correlations between variables (Williams, Buckley & Cote, 1989). While some question the extent and degree of the impact common method variance has on research outcomes (e.g. Spector, 1992; Spector, 1994) it is generally agreed that questionnaires should be designed with sources of method variance in mind (Podsakoff & Organ, 1986). Possible sources of method variance in this instance were aspects such as the “consistency motif”. The consistency motif refers to the tendency of respondents to engage in a consistent set of responses to a series of questions. The questionnaire used here required participants to respond to a number of different aspects of the ES implementation. It was considered important that responses were provided as a result of proper consideration rather than any consistency motif. Formatting was used to ensure items were presented in short “blocks”, reducing the risk of respondents developing a response pattern. Where possible items for scales were allocated randomly into sections, and care was taken to ensure duplicate or obviously similar items were removed (Podsakoff et al., 1986).

Response rates: Influential commentators such as Dillman (2000) have suggested that on-line questionnaires represent an exciting new dimension in survey research. However, they also caution that the approach does have a number of potential disadvantages. The most commonly cited problems with web based survey methods concentrate on issues of response rate and sampling. The issue of sample representativeness has already been addressed in Section 3.3.1.1. A number of studies have suggested that electronic surveys tend to demonstrate poorer response rates than paper SAQ’s (e.g. Gunter et al, 2002; Dommeyer & Moriarty, 2000). However many of these studies have examined e-mail attachment based surveys, not web surveys as is the case here. Web based surveys require participants to access web surveys via a hyperlink embedded in an e-mail notification. Web surveys can be hosted on secure servers protected by security firewalls and virus checkers, reducing the reluctance of

67

participants to open unsolicited e-mails containing attachments. The increased risk of participants deleting e-mailed notifications does exist however, and a number of activities were undertaken to increase participant awareness and commitment to the research. For example, as suggested by Dommeyer & Moriarty (2000) potential participants were contacted prior to the survey release and informed about the aims, procedure and format of the study. Information was also provided in the initial e-mail inviting participation. Attached to the e-mail was a short, three slide PowerPoint presentation outlining the aims of the study and encouraging participation (See Appendix A.01). Two follow up e-mails were also sent out to the target population one week apart. Alternate methods of raising participant awareness were also employed, such as using Centre Newsletters and Centre Director memo “directives”.

3.3.3 Questionnaire Measures

The pivotal role of the questionnaire meant the choice of measures were especially important. Where possible previously established scales were used. Established scales tend to possess higher degrees of validity and reliability, and therefore are considered superior to the development of new measures where extensive development is not possible (Hinkin, 1995).

3.3.3.1 User Acceptance Measures

The degree to which users accepted and adopted the HAL system was operationalized using Davis’s (1989) Technology Acceptance Model (TAM) measures. These measures have been used in several user acceptance studies and are considered internally reliable, show good convergent and discriminated properties and have demonstrated high predictive ability (Igbaria et al., 1995; Mathieson, Peacock & Chin, 2001). TAM measures consist of two scales (four items each) measuring perceived usefulness and perceived ease of use. Perceived Usefulness (PU) measures the extent to which the system is perceived to be useful in relation to task requirements and job performance. An example item is “Using HAL in my job increases my productivity”. Perceived Ease of Use (PEU) measures the extent to which users perceive the system to be free from effort and difficulty when using it. An example item is “Interacting with HAL does not require a lot of mental effort”. Both scales demonstrated high internal reliability with Perceived Usefulness (PU) exhibiting Chronbach alpha coefficients exceeding .90 and Perceived Ease of Use (PEU) exceeding .82 over Time 1 and Time 2. Previous research has reported alphas of between .87-.99 for both scales (Brown et al., 2002; Mathieson, Peacock & Chin, 2001; Venkatesh & Davis, 2000).

68

Table 3.1 User Acceptance Inter-correlations and Reliabilities (Time 1)

Time 1 M SD 1 2 3

Perceived usefulness 4.38 1.19 (.910)Perceived ease of use 4.04 1.14 .672** (.862)

Note: N=131. **p<.01 (1-tailed)

Table 3.2 User Acceptance Inter-correlations and Reliabilities (Time 2)

Time 2 M SD 1 2 3

Perceived usefulness 2.31 1.32 (.907)Perceived ease of use 2.41 1.31 .780** (.863)

Note: N=208. **p<.01 (1-tailed)

3.3.3.2 Job Design Measures

In order to measure perceived changes in job design scales from the revised Job Diagnostic Survey were used. The original Job Diagnostic Survey was designed as the empirical tool to validate Hackman & Oldham’s (1975; 1980) Job Characteristics Model (JCM). Empirical testing suggests the revised JDS as better able to support the five factor solution as proposed by the model (Boonzaier, Ficker & Rust, 2001) and requires less items therefore reducing the risk of participant fatigue. However, only the scales measuring the five job characteristics were chosen to include in the study. Measures such as Growth Need Strength (GNS) and job satisfaction were removed as they were not a key focus of the study. Participants were required to rate their job characteristics before and after the introduction of HAL at each data collection round. The inter-item correlations and reliability scores for Hackman and Oldham’s (1979) Job Characteristics model are presented below in Tables 3.3 and 3.4 with the item reliabilities in brackets. The comparatively low reliability scores of the five job characteristics are typical of previous research (e.g. Schmitt & Daume, 1993; Teigs,

69

Tetrick & Fried, 1992) and possibly reflective of the low number of items in each scale (Cortina, 1998). However, it is interesting to note that task significance (T1 - .74; T2 - .59) and autonomy (T1 - .76; T2 - .58) reported noticeably lower scores in Time 2. At this stage it is unclear as to whether this is a methodological or theoretical issue. However, if one considers the well regarded nature of the scales one could be tempted to suggest that the possible uncertainty about job characteristics is an artifact of the changes experienced by users in their job role.

Table 3.3 Job Characteristics Inter-correlations and Reliabilities (Time 1)

Time 1 M SD 1 2 3 4 5 6

Skill Variety 5.23 1.18 (.869)

Task Identity 4.58 1.42 .215** (.739)

Task Significance 5.60 1.05 .603** .229** (.747)

Autonomy 5.04 1.30 .501** .348** .533** (.769)

Feedback 4.59 1.30 .415** .605** .355** .573** (.781)

MPS 24.85 4.91 .663** .673** .711** .783** .811**

Note: N=124. ** statistically significant at p<.01 (1-tailed)

Table 3.4 Job Characteristics Inter-correlations and Reliabilities (Time 2)

Time 2 M S.D 1 2 3 4 5 6

Skill Variety 4.99 1.40 (.795)

Task Identity 4.72 1.43 .191** (.733)

Task Significance 5.66 1.01 .427** .105 (.590)

Autonomy 4.94 1.36 .358** .476** .326** (.579)

Feedback 4.72 1.35 .240** .525** .248** .607** (.675)

MPS 24.84 4.85 .559** .677** .552** .798** .759**

Note: N=197. ** statistically significant at p<.01 (1-tailed)

3.3.3.3 Demographic Variables

Single item questions were also included in the questionnaire to measure a variety of

70

demographic variables such as gender, age, group tenure, organizational tenure, job grade and system usage. All of the above variables were recorded using an ordinal scale. In all cases except for gender and age respondents were provided with five levels of the relevant variable. For example, respondents were able to indicate their length of organizational tenure in any of the following levels: “less than 1”; “1-2”; “3-4”; “5-10”; and “11+”. Age had six categories ranging from “15-19 years” through to “60+ years” increasing in ten year increments.

3.3.4 Questionnaire Format

Fowler (1995) outlines several principles of effective SAQ design, four of which were adopted in the development of the SAQ used in this study. While the primary delivery method was an on-line survey, a paper variant was also administered to those participants chosen for the network study. Therefore due to the multiple delivery solutions employed by the researcher certain design elements were altered to better reflect the nature of the delivery method. For example, the use of color and section spacing was different in the on-line version than in the paper version. These differences will be explored below.

First, due to the “self service” nature of SAQ’s the questionnaire should be self explanatory. In both the paper and electronic versions the questionnaire was prefaced by a brief set of instructions, information on the aims of the study and the contact details of the researcher. The initial briefing also outlined the format of the questionnaire, defined any key terms and return or submission instructions (Fife-Schaw, 1995).

Question Type: Fowler’s second design feature concerns the type of questions used in the questionnaire. Fowler advocates the use of closed format questions in SAQ’s to reduce the risk of vague, incomplete or difficult to code responses. Consequently all measures used a Likert scale response format. Careful editing and piloting procedures ensured items were free of ambiguous terms, leading questions indicating bias, “double barreled” items and technical jargon (Oppenheim, 1992; Weisberg et al., 1996). The length of the items within each scale was also considered. Ensuring the items were simple and easy to understand was particularly important in light of the large number of items to be completed by participants.

Question Form: Fowler’s third principle states that question forms should be kept to a minimum. In the case of the paper version checkboxes were used throughout. The on-line variant used radio buttons to ensure only one response per item could be

71

provided. In both questionnaire formats Likert scales were used for the answering of attitudinal related questions. The demographics section also used checkboxes but required participants to respond in a nominal response format.

Seven point Likert scales were used throughout the questionnaire. A widely used method for the measurement of opinions, Likert scales require participants to respond to a number of statements via a pre-determined set of options (DeVellis, 1991). It was considered vital that the scale points were able to generate sufficient variance in response (Hinkin, 1998). However scales were limited to seven points due to respondent’s difficulty in differentiating more than seven items (Weisberg et al., 1996). Using a seven point scale for all measures also ensured consistency throughout the questionnaire. Care was also taken to label each scale point. Scale points with labels indicating their relative importance on the scale have been demonstrated to significantly improve reliability and validity of measures (Krosnick, 1999). Therefore rather than only providing two descriptions at either end of the scale spectrum, the need to label every point specifically was acknowledged.

Questionnaire Layout: The need to have a clear and uncluttered aspect to the questionnaire underpins principle number four. Each theoretical component being measured by the questionnaire was given a different section. Where possible like elements were grouped into a broad classification. For example, Part 2 of the survey “your work environment” dealt with issues of job design. All sections were numbered and care was taken to ensure items were well spaced out and formatted to fit on one complete page where appropriate (Weisberg et al., 1996).

Item and Scale Sequence: Finally, the order of sequence of topics and items was considered. For example, the decision to end the questionnaire by asking respondents biographical information was based on two considerations. First, demographic data in most cases is cognitively easy for people to provide, requiring relatively little recall effort. Typically participant energy levels tend to lag towards the end of lengthy questionnaires (Fife-Schaw, 1995). Therefore effective questionnaire design would incorporate easier questions such as biographical data, at the rear of the questionnaire. The placement of measures followed this theme throughout the questionnaire with increasingly familiar topics being placed later in the sequence. The questionnaire began with attitudes towards HAL and then to issues of HAL driven changes to job design, before ending with demographic data. The placement of demographic data at the end of the questionnaire had another advantage. People are more likely divulge this type of information after knowing the contents of the survey and the issues being examined (Fife-Schaw, 1995). Therefore the decision was made to ask for participant data after the main body of the questionnaire.

72

3.3.4.1 Specific design elements – Web Questionnaire

When designing the survey instrument the researcher was careful to maintain the principles of effective survey design, regardless of the medium (Dillman, 2000). However, the nature of the on-line survey provided the researcher with a number of alternatives not normally open to designers of paper surveys. One problem encountered by researchers using lengthy, multiple item questionnaires is that of participant abandonment caused by participant fatigue (Roth & BeVier, 1998). Two features were incorporated into the web-survey to address this issue.

The first feature encountered by participants was the requirement to submit a unique log-in password. A common feature of many longitudinal survey studies, the Unique Code Identifier (UCI) requires participants to generate an ID code based around personal information (e.g. “last two letters of your father’s name”; “first two letters of your mother’s maiden name”). Typically, conventional paper surveys use the UCI to track respondents over time without compromising confidentiality. The use of electronic media allowed it to be used for an additional function. Using the UCI as a log-in password allowed participants to return and fill out the questionnaire as time permitted. This allowed participants a degree of flexibility usually offered by paper surveys, but not always by electronic formats. However, ensuring participants could fill out the survey at their convenience required another automated feature. After each set of measures the participant was given the option to save their responses, allowing them to return at a later date and complete the survey, continuing where they left off. In addition to maximizing participant convenience the “log-in & save” functionality also ensured the maximum amount of data was captured. Typically participant fatigue (in either paper or electronic formats) will result in participant non-response, regardless of how much of the questionnaire is filled out. It was considered that the “save option” would ensure at least a partial data capture if participants decided to abandon the survey.

One problem with web based survey design not already addressed are the difficulties associated with maintaining a consistent layout across browser types and PC configurations. Differences in user hardware and software can affect questionnaire length, formatting, the function of some interactive components and the time taken for the survey to load and complete (Couper, 2000). The risk of users possessing limited bandwidth or reduced download capacity drove at least two design elements of the survey. The first concerns the use of progress indicators. Web surveys have the capacity to incorporate increasingly sophisticated progress indicators alerting participants to

73

their progress (Couper et al., 2001). However, rather than use animated or interactive graphics that may increase download times a simple text reminder was used instead. A similar thought process was employed in the decision to use a single HTML page rather than a series of “click though” pages. While “click through” question designs reduce the risk of respondents skipping or missing items, participant abandonment rates tend to be higher (Manfreda et al., 2002). Browser re-fresh rates, software incompatibilities and the difficulty for participants to assess their progress though the questionnaire are all possible reasons for this phenomenon (Couper et al., 2001). Consequently a faster loading single HTML page where participants can scroll through was chosen.

Finally, screen size had to be considered in order to accommodate different screen sizes and browser settings. Typically screen settings can vary from 800x600 up to 1024x768 which may alter the format of the form if not planned for. The form was designed in the smallest size to ensure no significant alterations occurred as a result of screen variations (Frazer & Lawley, 2000).

3.3.5 Questionnaire Delivery

The multiple method research design required the questionnaire to be presented to participants in two distinct ways. Participants engaging in the network component of the research received a paper version of the questionnaire. Instructions were given regarding the filling out of the questionnaire and recovery of the completed document. The “one on one” interaction between researcher and participants reduced the need for a range of activities aimed at ensuring a maximum response rate. Data was collected from participants at two times. The first was three months prior to the system being implemented. The second was 6-7 months after the users had begun using the system.

The characteristics of the organizational wide on-line survey required a completely different approach. Much of the activities carried out followed Dillman’s (1991) guidelines for maximizing participant response rates in survey research. A range of strategies were employed to ensure that an appropriate degree of participant awareness was achieved prior to releasing the on-line survey. The month prior to the survey “go-live” date the researcher attempted to meet as many potential participants as possible through department and team meetings. Attendance at the meetings consisted of briefing staff on the research objectives and the value of participating. Organization newsletters and an organizational wide memo were also used to highlight the aims of the research and the value of supporting the project.

74

Erudio employees were invited to participate via the use of a broadcast e-mail. The e-mail provided a brief explanation of the study and its aims; reassurances of confidentiality; contact details of the research team; and a hyperlink to the web-page address. The e-mail also provided participants with a password to the site. The password was encrypted into the page to reduce the potential for parties outside Erudio entering false or nonsense data. In the event that individuals wished to participate, but wanted an alternative to the on-line survey, instructions were given as to how they could access a paper version of the questionnaire. The survey was hosted for three weeks, with two reminder e-mails sent seven days apart.

3.3.6 Data Analysis Preparation

The questionnaire data collected as part of the network interaction interviews was coded and manually entered by the researcher. The data was entered directly into the Statistical Package for the Social Sciences (SPSS, 2001). The keying was randomly checked for errors and missing data confirmed as missing and not data erroneously left out. Where missing data were identified to be items belonging to multiple item scales the missing value was replaced with the mean score of the observed items (Tabachnick and Fidell, 2001). The reduced sample size and comparatively small data set prompted a decision to retain the remaining items as missing and not remove the subject as is sometimes recommended (Cohen, Cohen, West and Aitken, 2003). Similarly while list-wise deletion of subjects during analyses was an option, the decision was made to use pair-wise deletion to retain the maximum number of subjects within the data set (Cohen et al., 2003). Screening for outliers and assessing the normality of the data was carried out with the use of SPSS’s box-plot and bar-chart routine respectively.

The data captured from the on-line survey component of the study was automatically stored in an MS Excell spreadsheet file. However, the automated process, with the electronic submission process meant no hardcopy of the survey responses existed. This limited the ability of the researcher to verify the responses and check the accuracy of the data entry program. Instead, a series of rigorous tests had to be performed prior to the distribution of the survey. Dummy data (a predetermined set of survey responses) were fed into the on-line survey then compared to the database within which the data was collated automatically. This procedure was carried out several times using different personal computers serviced by different Internet Service Providers (ISP) and with different firewall conditions. Every possible permutation of data entry, saving and submission was conducted and compared with the data stored in the database to ensure the integrity of the final data set.

75

3.4 Social Network Analysis

A key component of this research lies in its attempt to determine the effects of social context on user attitude. Chapter Two explored the value of adopting a social network analysis (SNA) approach in the study of social context and user acceptance. Using SNA afforded the researcher a number of advantages not readily achieved via alternative, more common research methods.

Especially important was the capacity of SNA to link both the micro and macro elements of organizational research. This approach allows researchers to recognize the social nature of individuals and organizations (Brass, 1995) and not limit the research to a consideration of individual antecedents, so typical in user acceptance research (Karahanna & Straub, 1999). Its focus on structure as being the measurement of repeated patterns of behavior also allows a more detailed examination of social structure and composition (Brass & Burkhardt, 1993). In particular, SNA allows variations in actor behavior and attitude to be better understood by knowing the position of actors relative to others within their network structure (Nohria, 1992). For example, conducting network analysis allows researchers to investigate in greater detail the effect of network structure, strength of network interactions and amount of network ties on user attitude. It also allows valuable insight into where power is held within organizational settings (Nohria, 1992; Tichy et al., 1979) and allows researchers to better identify and understand the composition of sub-groups within organizations (Fombrun, 1982).

3.4.1 Network Boundary Specification

A number of decisions have to be made when considering sampling issues in network analysis. Deciding what actors to include in the study depends on a number of factors, both theoretical and practical. From a theoretical standpoint one of the most important decisions concerns boundary specification (Wasserman & Faust, 1994). Often those adopting a network perspective tend to view actors as participants in networks within networks. At some point the researcher has to decide how to bound the network, cutting off any possible interactions beyond that established perimeter.

In order to establish a network boundary that was meaningful, theoretically useful and logistically achievable, a positional approach to participant recruitment was adopted. The positional approach involves sampling from a formally defined group or positions of interest (Scott, 2000). In this instance the researcher was principally interested in interactions between actors identified as future users of the new HAL system. In the

76

main employees identified as HAL users were the organization’s administrative staff and some members of Erudio management. The boundary of the network was further defined by focusing on administrative staff from one specific faculty within both City and Country.

The decision to limit data collection from one defined area in each Centre was based on a number of considerations. Each Centre had a “Business” faculty of a similar size allowing the researcher to conduct “between” and “within group” analyses. By choosing similar organizational groups with similar functions the researcher attempted to reduce the amount of extraneous variables that may account for differences in results between centres. Network research by its nature is open to criticism from being context specific. It was considered that the sampling approach undertaken here would help to dilute any such criticism.

Other, more practical concerns also influenced the researcher’s decision. Both Centres were large organizations in their own right, limiting the ability of the researcher to faithfully capture the entire network from a time and resource viewpoint . The organization itself also placed limits on the access provided to the researcher. Erudio management had concerns regarding the additional disruption caused by research activities on staff already under pressure from the changes required by HAL’s introduction. Consequently, access to the Business faculties was negotiated as a workable compromise for all concerned. While it was considered unfortunate that the entire Centre network could not be captured, previous studies of networks in academic settings have demonstrated faculties to be suitable arbitrary network boundaries (e.g. Stevenson, Bartunek & Borgatti, 2003).

Scott (2000, 53-54) raises a number of issues concerning static approaches to boundary delineation such as by department, work group or organization. Scott makes the point that the nature of human interaction is not in reality arbitrarily bounded and that to ignore ties outside set boundaries may compromise the true nature of the network and its effect on actors within the bounded network. He stresses that network boundaries must be set by identifying the variables under consideration. In the case reported here I was not so interested in the extent of network relations between users, but more to examine in detail changes to a specific set of exchange networks and their affect on user attitude. As such while it was considered important to rigorously measure network interactions within the set boundary, not capturing external interactions was not thought to have a significantly detrimental effect on the study.

77

3.4.2 Data Collection Protocol

Network researchers map interactions between individuals based on self reports, direct observation or archival data (Scott, 2001). To effectively capture all of the information required and to maximise response rates network data was collected from participants using a highly structured interview format (Papa, 1990). Responses to a series of sociometric questions were recorded on an interview sheet. Other methods such as archival data or direct observation were rejected either due to their inability to provide all relevant interactions, or logistical difficulties due to actors situated in multiple locations.

An interview protocol was established to ensure consistency between participants and the information they received. All participants were invited to attend an initial briefing of approximately 15 minutes. Each briefing covered: the broad aims of the research; the commitment required by participants; the characteristics of the research methods; and the protocols instituted to ensure participant anonymity and confidentiality. Potential participants were also made aware of the longitudinal nature of the research and that they would be required to provide data several times over a 12 month period.

Immediately following the briefing participants were given the opportunity to proceed with the network component of the study. The network data collection process followed a highly structured interview format whereby participants were guided through two sociometric questions (See Appendix A.03 for a copy of the interview form). In eliciting responses from participants network researchers have a number of methods at their disposal (Wasserman & Faust, 1994). The method used in this instance is commonly referred to as the roster method (Scott, 2000). Names of all potential HAL users from each faculty were obtained from Faculty management. For ease of participant recall and speed of response the names were then formatted into alphabetical listings based on work group location.

The use of the roster method has a number of advantages when collecting network data. In particular participant recall has been reported as higher using roster methods in comparison to unaided recall (Hammer, 1984). This was thought to be especially relevant when collecting data in a large organization such as Erudio where the potential for multiple interactions existed. Roster methods of data collection are also regarded as easier for participants to use and complete (Papa, 1990). The potential for this method type to improve levels of participation and achieve shorter interview times, resulting in improved response rates and support from the participating organization were all factors in the choice of method.

78

Some have called into question the accuracy of network data collected via questionnaire. Most commonly cited are the collective works of Bernard, Killworth, Kronenfeld and Sailer (cited from Marsden, 1990). Their research questioned the accuracy of reported interactions in particular situations. However, their conclusions have been challenged by more recent research by Freeman, Romney and others (Wasserman & Faust, 1994, 57). Freeman and his colleagues have argued that participant recall is actually biased towards routine, typical interactions thus providing researchers with a truer picture of network structure over time. Other commonly cited problems associated with the roster method such as difficulties in developing name lists for large target populations (Scott, 2000) were alleviated by incorporating the “snowball” sampling technique. In addition to a standard list of current employees spaces were provided at the end of each list for participants to include contacts not already listed. Providing opportunities to nominate alters outside of the listed names helped to avoid some of the disadvantages associated with roster and other fixed list formats (Wasserman & Faust, 1994).

On completion of the network interview participants were given a self administered questionnaire measuring user acceptance and participant demographics (see Section 3.3 for further detail on the design and content of the questionnaire).

3.4.3. Participant Confidentiality

The non-anonymous nature of network research creates additional challenges for researchers. The potential for participant harm or disadvantage as a result of network data misuse highlights the numerous ethical implications for network research (Borgatti & Molina, 2003). All procedures and protocols followed in the conduct of this research were granted expedited approval by the Queensland University of Technology Ethics Committee (QUT reference no. 2747H).

In order to protect participant confidentiality and to engender participant trust a rigorous system of document control was employed. Two code books were established. The first code book had the participant’s name and their Unique Code Identifier (UCI). The UCI was a participant specific code attributable only to them, developed with the use of participant information (See Appendix A.02 for an example). The UCI was provided by the participant on the front of their attitude questionnaire. The second code book contained a list of UCI’s and the corresponding number of the participant’s network questionnaire. All four documents were then stored in separate secure locations after data processing. The use of the code book system and multiple storage sites dramatically reduced the potential for accidental or even deliberate attempts to match

79

network responses to participant details and attitudinal responses. When entering the network data into UCINET6 the “ID Codes” function was used to mask the identity of the participants in the matrix.

All data collection interviews were begun with the signing of a consent form (see Appendix A.04). The consent form was designed to act as a checklist for the researcher to uniformly explain the relevant issues to participants, as a formal understanding of commitment and participation, and as a tangible commitment to the protection of the participant’s confidentiality and anonymity.

3.4.4 Network Measures

The research questions and hypotheses developed around the notion of social networks were primarily concerned with attitude transmission and resource dependence. As reviewed in Chapter Two a number of networks can exist simultaneously in the same environment. While some overlap may occur it is important to recognize that each network may have its own peculiar composition and therefore requires measurement (Ibarra & Andrews, 1993). In the current research project two network types were identified as being relevant and requiring measurement, those of Advice and Resource exchange.

Both network measures required respondents to respond to a single question about their interactions. The commonly accepted practice of single item scales for the measurement of network interactions has not escaped debate (Marsden, 1990). Authors such as Bernard, Killworth, and colleagues (1984; 1982; 1981; 1977) questioned the accuracy of self report interaction data citing poor accuracy levels as compared to comparable observation data. However more recent research has questioned their findings, arguing that respondents tend to report interactions characteristic of routine, typical interactions (Freeman & Romney, 1987; Freeman, Romney & Freeman, 1987: Cited in Marsden, 1990). As these types of interaction patterns are of primary interest to network researchers the practice of self report single item measures has tended to continue. Actions taken to maximise respondent accuracy included the use of previously published items and the use of the roster method to elicit participant recall (see Section 3.3.2).

Network types: One of the key arguments presented in Chapter Two related actor position within resource dependence networks to user attitude. Within most organizations individuals known to possess the ability to dispense advice when

80

required are often regarded as valuable commodities (Cross, Borgatti & Parker, 2001). Often as a consequence of their perceived value, knowledgeable individuals are also acknowledged to possess varying levels of power within organizations (Ibarra, 1993). A similar line of thought accompanies the inclusion and measurement of the “resource exchange” network. As also discussed in Chapter One ES technology typically alters the flow of resources within an organization (Huber, 1990). Therefore it was considered necessary to examine whether resource dependence relationships changed following the introduction of HAL and then to examine the implications in relation to user attitude. Definitions of each resource type were provided to the participants. “Resources” referred to elements such as funds, human resources, time, material supplies or such things as information or reports. “Advice” referred to advice given about work related activities, solutions to problems or guidance on such things as policy and procedure.

Participants were asked to indicate levels of dependence within various resource network types. Tie strength or degree of resource dependence was measured by asking participants to indicate a level of dependence 0 (no interaction); 1 (low); 2 (moderate) 3 (high) where relevant. In order to reduce self report bias the resource networks were measured twice. Participants were required to indicate who they relied on to carry out their job duties and then on a separate form, who relied on them.

“Who interacts with you to obtain the required resources (information, data, advice, materials, approval) they need to carry out their duties/tasks/job?”

“Who do you interact with to obtain the required resources (information, data, advice, materials, approval) you need to carry out your duties/tasks/job?”

3.4.5 Network Data Entry and Storage

The data collected from the network interviews were imputed into a data matrix using UCINET6 (Borgatti, Everett, and Freeman, 2002). Two sets of network data were collected and imputed. In terms of how the data were treated three issues were of critical importance – treatment of directional ties; reported values; and missing data.

One, a decision had to be made whether to retain the directional nature of the data or to symmetrize the ties between actors. Tie direction or matrix symmetry indicates the flow of interaction, whether it be advice, information or communication from

81

actor to actor. The decision to symmetrize the network data or to retain its directional nature should be driven by the research questions and the type of analysis planned (Scott, 2000). For example, a researcher examining friendship networks may assume that when actor A nominates Actor B as a friend that this relationship is reciprocated. Therefore even if Actor B is unable, or fails to nominate A as a friend, the reciprocal nature of the tie is still recorded. However in this case it was considered theoretically appropriate that both sets of network data would be stored asymmetrically as it was important to identify suppliers and receivers of advice and resources in order to test the hypotheses. As such only nominated ties to and from actors were recorded.

“Valued” data was collected for both networks indicating the frequency of the interaction. For the advice and resource dependence networks a level of dependence was indicated by 0 (no interaction); 1 (low); 2 (moderate); and 3 (high).

Reviewing the response data highlighted an issue associated with self report data and response discrepancies between respondents. Scanning the data revealed numerous instances where pairs of actors disagreed with each other in relation to the frequency or intensity of their interactions. A review of the literature failed to identify a consistent approach to dealing with this problem. Baldwin et al. (1997) made an arbitrary decision to record the maximum value indicated between respondents, while Krebs (2004) suggests recording the lowest reported interaction in order to mitigate the incidence of “over reporting”. Others, such as Rice and Aydin (1994) calculate an aggregate level over a nominated time period (e.g. a month). Instead, for the purposes of this study an estimate pooling approach was used to reduce bias and improve respondent accuracy when reporting levels of resource dependence (Borgatti & Cross, 2003). The degree of dependence of i on j was calculated as an average of the extent to which i claimed to be dependant on j and the extent to which j claimed i was dependant on them. This approach is recommended to reduce bias associated with the over-reporting of outgoing ties and the under-reporting of incoming ties.

3.5 Summary

Chapter Three provided a detailed and comprehensive outline of the methodological issues relevant to the case study reported in this thesis. The chapter began by outlining and justifying the research strategy employed to answer the research questions and the hypotheses set out in Chapters One and Two. This was followed with a summary of the research site and context within which the case was conducted. The remainder of the chapter was divided into two distinct sections, one dealing with the development

82

of a self administered questionnaire, and the other with the procedures required to measure various network types within the Erudio faculties. The use of a self administered questionnaire as the primary research tool was justified in light of the research aims. The remainder of the section went on to outline the characteristics of its design. Special attention was paid to the peculiar design elements of the on-line or web version of the SAQ. The second major section of the chapter reviewed the methodological issues surrounding the collection of social network data. The data collection method, interview protocols and sampling strategy were all covered. The sociometric measures used in the study were also reviewed and explained. Before proceeding to the reporting of results in Chapter Five, the next chapter will examine in more detail the methodological and analytical aspects of network analysis. The contents of Chapter Four will help those with limited exposure to network analysis to interpret and critique the results and discussion presented in Chapters Seven and Eight.

83

Chapter 4 — Social Network Methods, Measures and

Application

4.0 Introduction

This short chapter is intended for those readers unfamiliar with the procedural and analytical aspects of network analysis. A brief review of network data entry, management and analysis will be covered. The aim of this chapter is to provide a segue between the more procedural aspects of the data collection reported in Chapter Three and the results of network analysis presented in Chapter Seven. Those familiar with the analysis and interpretation of network data may choose to avoid this chapter and proceed to Chapter Five.

In review, Chapter One set the context for the research, providing a justification for the area of investigation and defining the key concepts. Chapter Two presented a model of structural change investigating the effect of ES characteristics on job design and social networks. Influenced by the notion of structuration it was argued that while ES technology may represent the potential for significant structural change, user responses to those changes via their social networks may determine the degree of acceptance demonstrated by users. Consequently the extent to which the system is successfully adapted into their everyday structures and social systems is dependant on user action within existing structures and systems.

Chapter Three reviewed the process undertaken to gather data capable of testing each of the eight hypotheses. The examination of both user attitudes and network effects required a multiple methods approach to data collection. The design and content of the first method, a self administered questionnaire was adequately covered in Chapter Three. The chapter also covered elements of the second method, network analysis. However, the specific nature of network analysis and its status as an emerging methodology within both the IT and organizational literature, warrant more attention to the specifics of network methodology.

While it is intended that the information contained within this chapter would provide researchers with adequate information to conduct similar research, readers are advised to also consult the following sources. Scott (2000) and Knoke and Kuklinskii (1982) provide a non-technical introduction to a wide range of network concepts, including

84

others not addressed in this study. Wasserman and Faust (1994) is widely regarded as one of the most comprehensive approaches to the study of networks, for both method and analysis.

The structure of this chapter is as follows. First the storage and retrieval of network data will be covered, along with a brief summary of the popular software packages used in the analysis and visualization of network data. From there the chapter will progress through the various types of network measures relevant to the study. The discussion of the network measures will mirror the structure of Chapter Seven, examining analysis relevant to network change before moving on to measures of centrality and brokerage.

4.1 Network Data

4.1.1 Data Entry

The following information has been obtained from Knoke and Kuklinskii (1982), Scott (2000) and Wasserman and Faust (1994). An excellent introductory text designed to be used in conjunction with the SNA package UCINET, is also available electronically at http://faculty.ucr.edu/~hanneman/nettext/.

Data collected for network analysis can take a number of different forms. At its most basic network data can be binary in nature with the presence of a relation between two actors represented by 1 and the absence of a relation 0. However network data can also be valued, representing the increasing or decreasing tie strength of a relation. A further enhancement of the data can be added by indicating tie direction. Network researchers have to make theoretically based decisions as to the symmetric nature of ties, with asymmetric ties indicating the direction of a relation from one actor to another, and not automatically assuming the tie is reciprocated. For example, network relations such as advice networks are typically regarded as asymmetric, whereas communication networks are typically assumed to be symmetric in nature (Burkhardt, 1994; Fulk, Schmitz & Steinfeld, 1990). When required valued, directional data can be transformed into binary format.

In most cases network data are initially stored in matrix form. In the case of relational analysis (as has been undertaken here) a square “case by case” matrix is used. Following convention the direction of a relation is represented as running from a “row” to a

85

“column”. Figure 4.1 below provides an example of a valued, directional network in both matrix and graphic form. Note that the greater strength of tie running from Mel to Kosta is indicated in the matrix with a higher numerical value. Due to the focus on didactic interactions network researchers rarely consider actors as having relations with themselves, hence the absence of data in the diagonal below.

Figure 4.1 Example of a valued, directional matrix and corresponding graph (sociogram)

Richard Kosta MelRichard 0 1 0Kosta 1 0 0Mel 0 2 0

Network researchers also have differing levels of analysis to consider when collecting network data. Network data can be collected to represent a complete; ego-centric; or an ego-centric network with alters (Hanneman, 2001; Knoke & Kuklinski, 1982). The data collected and analyzed in Chapter Seven are considered to be a complete network. As such attempts were made to collect interaction data from every actor within a bounded network (Scott, 2000). Typically researchers will attempt to achieve at least an 80% response rate to ensure an accurate representation of the network (Sparrowe, Liden, Wayne and Kraimer, 2001). Other approaches such as ego-centric approaches are more limited in that the focus is on the individual and the number of ties each actor has (alters). More complex ego-centric approaches (typical of structural hole research) attempt to identify what relationships exist between the reported alters (Hanneman, 2001).

In terms of network notation it is convention to refer to rows using “i” and columns as “j” with the letter a to indicate the value of the cell, resulting in the notation a(i,j) as referring to a network relation. Data collected for this thesis were stored on UCINET6’s propriety spreadsheet editor as a valued, directional network for advice and resource exchange networks. An estimate pooling approach was used to reduce bias and increase respondent accuracy when reporting levels of resource dependence (Borgatti & Cross, 2003). The degree of dependence of i on j was calculated as an average of the extent to which i claimed to be dependant on j and the extent to which j claimed i was dependant on them.

86

4.1.2 Network Analysis Software

The last decade has seen the emergence of numerous software packages able to carry out network analysis routines and perform network visualization. It is not the intent of this section to provide a comprehensive review of all available network analysis software. Rather, the aim is to highlight some common packages, their basic function and their relevance to this research. An extensive list and links to network software can be found on the International Network for Social Network Analysis (INSNA) website under “software” (http://www.sfu.ca/~insna/).

Generalist packages such as UCINET6 (Borgatti, Everrett & Freeman, 2002) and NETMINER (2004) all perform similar degrees of analysis based around relational and structural network analysis. Typical analysis routines can be run for centrality, network cohesion, structural equivalence and various statistical routines such as quadratic assignment procedures (QAP). Programs such as STRUCTURE (Burt, 1991) focus specifically on routines based around structural holes and equivalence. It should be noted that UCINET6 and NETMINER also perform a limited range of structural hole routines as part of their package.

Other packages such as NETDRAW (Borgatti, 2002) Krackplot (Krackhardt, Blythe & McGrath, 2005) and MAGE (Richardson & Presley, 2002) are graphical packages able to provide graphical representations of network data. NETDRAW has a number of useful features and numerous options in terms of layout and network format as well as offering a limited range of analyses. MAGE is particularly useful in its ability to provide a three dimensional representation with the ability to rotate the network around any user defined axis.

The output examples presented in this chapter, as well as the output presented in Chapters Seven are all produced by UCINET6 (Version 6.98). The network diagrams presented throughout this thesis have been produced by NETDRAW (Version 2.18) which is offered as an addition to the UCINET6 package.

4.2 Measuring Network Change

Chapter Two set out two hypotheses concerning the potential for advice and resource exchange networks to change following the introduction of ES technology. Four methods were employed to investigate whether network change had indeed occurred following the introduction of the ES technology. The four were chosen based on their

87

ability to provide additional information regarding the nature of the change being investigated. Three of the four methods used are briefly reviewed including Quadratic Assignment Procedure (QAP) routines, testing network density, and Krackhardt’s E-I index. The fourth approach, In-Degree centrality will be discussed in Section 4.3.

4.2.1 Quadratic Assignment Procedure (QAP)

In attempting to measure what has been defined as network shock (see Section 2.3.2) it is essential to effectively measure the degree of change (if any) occurring within network structures across time. A frequently advocated method of testing the extent to which networks are significantly similar is to use what is termed a Quadratic Assignment Procedure (QAP) (Krackhardt, 1988). The QAP procedure was principally designed to examine whether two NxN matrices are similar beyond that which could be expected based on chance (Meyer, 1994). A QAP takes into account all dyadic information represented in both matrices, comparing each dyadic cell in matrix A with the corresponding dyadic cell in matrix B (Meyer, 1994). The QAP has been shown to produce robust results under conditions where variables are likely to exhibit high levels of autocorrelation (Borgatti et al., 2002). This is discussed further below.

A fundamental difference between network data and more conventional data collected for quantitative statistical analysis needs to be briefly addressed. Tests of statistical significance such as those run for both correlation and Ordinary Least Squares (OLS) regression rely on the assumption that the estimation of standard errors are based around independent observations. Network observations by definition are not independent as is typical in conventional sampling situations. Rather, they are taken to be representative of the population at any given time, as opposed to being a sample of that population (Hanneman, 2001). Therefore, the approximations applicable to independent random samples do not hold, and if used in a network context are likely to result in under-estimations of sampling variability (Martin, 1999). This requires the use of non-parametric tests, such as Quadratic Assignment Procedures (QAP) (Krackhardt, 1988; Martin, 1999). In situations requiring matrix correlation and multiple regression network analysts are able to use Quadratic Assignment Procedures (QAP) to run correlations and Multiple Regression Quadratic Assignment Procedures (MRQAP) to run multiple regressions. The QAP and MRQAP routines are similar to conventional correlation and multiple regression procedures except that they use a randomization / permutation technique to construct significance tests (Borgatti & Cross, 2003).

88

The QAP test is carried out in two stages. First, it examines the correlation between each corresponding cell of the two matrices under examination. However, due to the nature of network data and the potential for high degrees of autocorrelation between variables it performs a random permutation test to establish the standard error for the correlation coefficient (Borgatti et al., 2002). The random permutation test randomly permutes the rows and columns of one matrix onto the other repeatedly. After each permutation the correlation coefficients are re-calculated and compared against the original result. A pseudo-probability measure results, whereby the number of times a random permutation is larger or equal to the original result, is reported. The number of permutations carried out improves the accuracy of the standard error calculations and confidence in the reported p values (Borgatti et al., 2002). UCINET6 allocates a default number of 2500 random permutations, which can be increased to improve estimates of standard error, but increases computational time (Borgatti et al., 2002). An example of the output produced by UCINET6 is presented below.

In the example provided below, two networks of in-degree and out-degree ties of actors within an advice network are compared. As a percentage of the random permutations carried out none produced a correlation as large as .394 (using the (Pearson Correlation score) and therefore fulfils the accepted significance levels of p<.05. In Chapter Seven QAP analyses will be used to compare network structures across time as well as compare advice and resource exchange network structures within each data collection period.

89

Figure 4.2 UCINET6 Quadratic Assignment procedure (QAP) Output

4.2.2 Density

The density measure is used to test hypotheses 6a and 6b, to determine whether advice and resource exchange networks increase following the introduction of ES technology. In simple terms network density refers to the proportion of users in a network that are connected to each other. More formally it is expressed as the proportion of connected points in a graph expressed as a proportion of the maximum lines possible between all points (Scott, 1991). Given this definition it becomes clear that a fully connected graph, whereby every user is connected to each other would be attributed a density value of 1. Conversely a network attributed the value of 0 would indicate nil connections between the actors being observed. Density can be dependant on the size of the network based on this definition as larger networks could be expected to report lower density scores due to the relational nature of the measure (Anderson, Butts & Carley, 1999). Consequently, caution should be taken when evaluating the relative densities of networks and consideration should be taken to consider the impact of network size on the result.

90

The data reported in Chapter Seven are both valued and directional and therefore density was calculated as the total of all values divided by the number of possible ties, giving the average value of those ties (Borgatti, Everett, & Freeman, 2002). An example of this is provided below in Figure 4.3. In this instance the density of one network measured at two times is reported (T1 and T2). The data for both networks is valued with participants reporting a frequency of interaction (1 = low; 2= moderate; 3 = high). In this case at Time 1 the average value per possible tie was .35 at T1 and .57 at T2. It would appear that in this case the amount of interaction in the network has increased over time. It is not possible however to ascertain whether there has been an increase in ties overall, whether some actors have increased more than others or whether the frequency of interaction has increased. In this instance additional tests such as those reviewed below could be used as well as visual confirmations of network graphs to examine the nature of network change (McGrath, Krackhardt, Blythe, 2003). UCINET6 also offers a permutation test to examine whether the difference in density is more or less than could be expected by chance.

Figure 4.3 UCINET6 density output

91

4.2.3 Krackhardt’s External Tie Ratio (E-I index)

Network analysis techniques such as measuring density and undertaking QAP routines are only able to provide limited information about the nature of network change over time. For example, the density measure can only tell researchers whether overall ties have increased or decreased, it is unable to identify what ties have changed and how. Krackhardt and Stern (1988) introduced a measure aimed at identifying the distribution of ties within work groups and between work groups – termed the E-I index. Krackhardt and Stern’s (1988) original work tested the hypothesis that organizations demonstrating a higher number of “cross-organizational” friendship ties would respond more effectively to organizational crises. The pair suggested that friendship ties lead to greater levels of organizational trust, leading to increased willingness to cooperate and therefore adaptation to crisis. The results of six experimental trials indicated support for their hypothesis. More recently McGrath and Krackhart (2003) extended the theory in relation to organizational change success.

In simple terms Krackhardt and Stern’s (1988) E-I index provides a ratio of ties reported outside the immediate workgroup to all those observed. Membership of groups is identified a-priori by the researcher. In Chapter Seven group membership was based on formal workgroup arrangements. The E-I index routine offered by UCINET6 provides a raw E-I score (External – Internal / External + Internal ties), expected score (ratio of maximum and minimum scores given the size of sub-units and density of the network) and a re-scaled index (raw score rescaled to fit within the expected score). E-I scores are provided for the overall network, each nominated group and for each actor. A random permutation significance test of the overall E-I score is also provided. E-I index scores range from -1 to 1 with a positive ratio indicating a tendency for actors to interact with others outside of their nominated groups. A negative score indicates a tendency for internally focused actor interaction and group closure. Therefore Krackhardt and colleagues would suggest that those networks reporting a positive ratio would be better able to respond to change due to their extensive links throughout the network leading to increased cooperation, information and support.

Figure 4.4 over provides an example of the output provided by the E-I index routine. As evidenced, a large amount of information is provided to the researcher. In the example below four groups have been identified a-priori by the researcher. After the density matrix a table is provided with the raw data required for the E-I ratio. Column one tables the frequencies of the observed internal and external ties; column two presents these frequencies as percentages of the total number of ties in the data; column three reports the total number of possible ties given the size of each group; the

92

final column gives the E-I index assuming that all of the observed ties had been evenly distributed across the network and becomes the expected value of the E-I index. This value is then used to derive the re-scaled E-I index that better reflects the size of the groups and the network. A permutation test is then used to determine whether the E-I index reported is greater or smaller than might be expected by chance given the size of the groups. Following the permutation test the output also provides an E-I score for each nominated group as well as an E-I score for each individual actor (not shown).

Figure 4.4 UCINET6 E-I Index Output

93

4.3 Network Change and User Acceptance

This section of Chapter Four will examine in detail each of the four types of network centrality and brokerage measures undertaken in the study. The structure of this section follows closely the structure of Chapter Two as the measures and analysis for the relevant hypotheses will be dealt with in the same order. The final component of this section will deal with the use of regression when using network data. Where relevant the output will refer to analysis carried out on the example data set introduced at the beginning of the chapter (Figure 4.1 reproduced below).

Figure 4.1 Example of a valued, directional matrix and corresponding graph (sociogram)

Richard Kosta MelRichard 0 1 0Kosta 1 0 0Mel 0 2 0

4.3.1 Network Centrality

Hypotheses 8a and 8b required the identification of users within a network that could be considered highly central in their network. The discussion surrounding hypotheses 8a and 8b also considered the value of using various centrality measures in light of the theoretical aims. The specific analysis required by both measures are reviewed below.

4.3.1.1 Degree Centrality

Freeman’s (1979) degree measure is used twice in Chapter Seven. In the first instance it is used to identify changes in direct ties between actors in order to better understand the change in network structure. In the second it is used as one measure of centrality in order to determine any link between changes in network centrality and user acceptance.

The degree centrality routine run by UCINET6 allows researchers to take into account direction of ties, allowing researchers to distinguish resource flows to (in-degree) and from (out-degree) actors at a localized level. The ability to identify peripheral actors with low out-degree scores (and high in-degrees) is also useful. Degree centrality has been used in several previous studies of power and influence in an technological change context (Brass & Burkhardt, 1993) and therefore is considered appropriate to

94

use in this instance. The calculation of out-degree centrality is a simple addition of ties sent and ties received by each actor in the matrix. If the matrix is asymmetric and therefore directional the output will provide both in-degree and out-degree scores. If the data are valued then the degrees (in and out) will consist of the sums of the values of the ties (Borgatti et al., 2002). The output (See Figure 4.5 below) provides raw centrality scores, normalized scores expressed as a percentage of the maximum possible degree, and network centrality measures. Borgatti et al (2002) advise that for valued data (as used in this case) non-normalized scores should be used and the network centrality index should be ignored.

Figure 4.5 UCINET6 Degree Centrality Output

In addition to a range of standard descriptive statistics the reader will also note that UCINET6 also produces a measure of network centrality. Network centrality is a measure of variability displayed by a network as a percentage of that in a network of the same size in a star configuration (Freeman, 1979; Hanneman, 2001). Figure 4.6 over is an example of an idealized star network, of which the percentage of centralisation is 100. The higher the degree of centralisation the more concentrated ties are towards a small number of actors from a relatively large number of peripheral actors. The implications of network centralisation will depend on both the nature of the networks under consideration and the research questions of the researcher.

95

Figure 4.6 Freeman’s (1979) Star Network

4.3.1.2 Flow Betweenness Centrality

Freeman et al.’s (1991) flow-betweenness estimates the proportion of complete pathways between two actors (e.g. X and Z) that occur on the path of a given actor (Y). This proportion for Y is calculated for every other actor in the network. This process is repeated for every other actor, the flow-betweenness scores for each actor are then standardised as a ratio of total paths not containing each actor. An example of the output for the flow-betweenness routine is provided below (Figure 4.7).

Figure 4.7 UCINET6 Flow-Betweenness Output

96

In addition to the “raw” centrality scores notice that UCINET6 also provides normalized scores. Because the size of the raw scores is dependant on both the size and density of the network score results can become unwieldy in some circumstances. UCINET6 standardizes the scores by calculating a ratio of flow betweenness of each actor by the total flow betweenness not involving that actor, this is shown by nFlowBet. Higher scores indicate an actors presence on a larger amount of paths between other actors in the network.

4.3.2 Network Brokers

In addition to the importance placed on highly central users in determining levels of user acceptance Chapter Two also considered the role played by users in brokerage holes. Burt’s (1992) structural hole measures were used to measure the relationship between brokerage and user acceptance.

4.3.2.1 Structural Holes (Redundancy & Constraint)

Burt’s (1992) structural holes approach focuses on those individuals that span gaps between network actors, those considered high in structural autonomy. Actors high in structural autonomy are able to fill a number of structural holes present within a network, and consequently benefit from increased access to resources and information (Burt, 1992).

Burt (1992) specifies two key conditions for the identification of a structural hole. First, it must provide access to a non-redundant resource. The degree of redundancy present within an actors (ego) network is captured by the effective size and efficiency of their ego-network. The effective size of an ego network is established by identifying the number of alters (ties to other actors) minus the average number of ties of those alters. The efficiency of an ego-network is determined by the effective size divided by the number of alters in the ego-network (Borgatti et al., 2002). An efficiency ratio of 1 indicates every contact within the network is non-redundant.

Second, the degree of network constraint evident within an actor’s network must be identified. In the context of the study carried out here constraint is of greater importance as it represents the degree of dependence evident between network members (Burt, 2003).. In simple terms constraint is concerned with the extent to which an actor is dependant on a limited number of contacts within their ego-network (Susskind et al., 1998). More constrained networks span fewer structural holes leaving the actor

97

constrained by their reduced number of contacts. However, an increased number of ties does not guarantee reduced degrees of constraint. In calculating constraint consideration is also placed on the potential for maximizing the returns resulting from the investment made by the ego in the relationship to the alter. Constraint scores vary from 0 (indicating many redundant contacts) to 1 (indicating only one contact). Therefore, actors reporting low constraint scores can be identified as possible brokers within a network (Burt, 2003). Conversely those actors reporting high constraint scores can be identified as users dependant on a limited number of resource providers.

The use of Burt’s (1992) constraint measure is considered an appropriate alternate measure of a user’s information availability. In a recent reassessment of the constraint measure (Burt, 2003) highlights the usefulness of constraint in relation to the diffusion of ideas within an organization. He puts forward evidence to support the argument that those exposed to more sources of information benefit in relation to idea generation and acceptance. In a similar vein Perry-Smith & Shalley (2003) linked individual creativity to the amount of weak ties reported by actors. It is considered that a similar logic can be applied to the amount and quality of information about an ES received by users. Users limited to a reduced number of information sources regarding the characteristics and perceived value of an ES are likely to be more influenced by the sources of that information (Burt, 1997). Measures of constraint experienced by users would be obtained using the Structural Holes routine available in UCINET6.

The results presented in Chapter Seven will report both efficiency and constraint measures. Figure 4.8 over provides an example of the Structural Holes output provided by UCINET6. Notice the brokerage role highlighted by the constraint measure in Kosta (refer to the network graph at the beginning of the section for a visual confirmation of this).

98

Figure 4.8 UCINET6 Structural Holes Output

4.3.3 Network Analysis and Multiple Regression

Each of the four network measures reviewed above were used as the independent variables in a regression analysis with Perceived Usefulness and Perceived Ease of Use as the dependant variables. The unique nature of network data and consequently the measures derived from the data prevented the use of conventional OLS regression techniques to be used. As discussed earlier, the non-independent nature of network data requires the use of non-parametric tests that produce a more robust assessment of the data variability. Instead, researchers have refined a method akin to bootstrapping that uses permutations to estimate the standard errors of a network statistic (Hanneman, 2005). The permutation technique is able to be applied to a range of analysis techniques such as t-tests and the regression of network data. For example, using the permutation technique UCINET6 allows researchers to assess the significance of the r-square and regression coefficients of attribute data collected from network study participants.

UCINET6 allows researchers to conduct simple or multiple regression using network data. UCINET6 performs the routine in two steps. First, a standard multiple regression is performed across the cells of the dependant and independent variables. In the second stage the elements of the dependant variable are randomly permutated and the regression is recalculated. The resulting values of r-square and all of the coefficients

99

are stored then compared to the coefficient reported in step one. This permutation process is repeated as many times as specified by the researcher. The standard errors are calculated as the proportion of random permutations that yield a coefficient as extreme as the one reported in step one (Borgatti et al., 2002). As with the QAP routine increased permutations result in increased computational time but improved confidence in significance levels (Borgatti & Cross, 2003). Krackhardt’s (1987) “High-tech managers” attribute data is used to provide an example of the Regression output provided by UCINET6 (See Figure 4.9 below).

Figure 4.9 UCINET6 Vector Regression Output

The output, while providing the regression coefficients for each independent variable does not provide the associated standard errors as is commonly reported with OLS regression. Instead, the proportion of random trials yielding a coefficient i) as large or larger ii) as small or smaller and iii) as extreme as the original observed value are reported (Borgatti et al., 2002). All statistics reported in Chapter Seven were generated using UCINET6 and unless specified all confidence intervals are reported refer to the results of the permutation tests reviewed here.

100

4.4 Summary

The intent of this brief chapter was to provide a brief overview of the social network measures and analysis techniques used within the thesis. While not intending to be a comprehensive review of network methods and analysis, the chapter did intend to provide readers unfamiliar with network analysis an understanding of the analysis and results presented in Chapter Seven.

A limited amount of attention was paid to the procedural aspects of storing network data and the forms it may take. An increased amount of time was taken to explore the difficulties inherent to the application of conventional statistical approaches to network data. The need to use permutation and randomization techniques such as quadratic assignment procedures was covered, as was the use of similar procedures to carry out simple and multiple regressions.

The chapter also reviewed in detail the network “measures” used to test the hypotheses presented in Chapter Two. The first component reviewed the measures used to determine whether network change had occurred over time and the nature of that change. The remainder of the chapter presented the measures used when considering the links between changes in centrality and brokerage and user acceptance. The next chapter, Chapter Five will present the results of the job design component of the study. Chapter Seven will present the results of the network analysis previewed in this chapter.

101

Chapter Five — Job Design and ES Technology

5.0 Introduction

The previous two chapters outlined the methodology of the thesis. Chapter Three provided an overview of the data collection procedures and the methods used to collect both attitudinal and network data. Chapter Four provided a brief introduction to network analysis techniques, covering a range of issues pertinent in Chapter Seven. Chapter Five presents the results of analysis aimed at testing the job design hypotheses laid down in Chapter Two.

Overall the chapter reports on the results of analyses aimed at investigating the potential for job design change following the introduction of ES technology, and the user acceptance implications of those changes. The chapter is divided into four main sections. The first describes the characteristics of the sample, descriptive statistics and variable correlations. The second reports on analyses testing the impact of ES technology on job design. Paired sample t-tests were used to determine whether perceived job characteristics changed from pre- to post-implementation. The third section presents a series of hierarchical regressions aimed at testing the relationships between system driven job change and user acceptance. Finally, a series of moderated regressions are reported, examining the extent to which system usage moderates the relationship between perceived job change and user acceptance.

5.1 Sample

As outlined in Chapter Three, the data collected for the study were from a large educational institution undergoing an ES technology implementation. Two of the institution’s fifteen education centres had been chosen to pilot the system. It is from these pilot sites (referred to as City and Country) that the data have been drawn.

Data used for the testing of the job design hypotheses were obtained from an on-line self administered questionnaire broadcast to all users at both City and Country sites. The first round of data collection took place three months before the system went live. A total of 145 responses were obtained. The second round of data collection took place four months after the system became operational, and seven months after the first round of data collection. A total of 218 responses were collected from the second round of data collection. A total of 144 responses over the two time periods were obtained.

102

One disadvantage of the on-line distribution method of the survey was the difficulty in determining an exact response rate. Estimations by Erudio staff identified approximately 450 potential users in the City faculty and approximately 250 in the Country faculty accounting for a population of approximately 700 system users. A total of 145 responses were received at Time 1 for an approximate response rate of 20.7%. This level of participation was poorer than expected and a number of reasons were considered possible. User training was only just beginning when the survey went live and had the dual effect of removing users from their daily activities and leaving others to make up the shortfall. Consequently people may have been unable to complete the survey due to increased work pressures, training commitments or may have been reluctant to participate due to a lack of knowledge about the system. A total of 218 responses were received at Time 2 for an approximate response rate of 31.1%. This slightly healthier response rate is possibly reflective of the lack of obstacles present at Time 1. In an attempt to verify the representativeness of the sample the demographics were compared to those collected from an organizational wide employee attitude survey earlier in the year (WERP, 2005).

5.1.1 Demographic Summary

Participants responding at both data collection points had the following characteristics (See Table 5.1 below for details). In keeping with the overall Gender profile of the organization the majority of respondents were female (71%). The Age profile of the respondents also reflected the aging nature of the organization’s population with 45% of respondents over the age of 40. The highest single age category reported however was from the 20-29 age group (28%). The Job Grade of respondents was also reflective of the organizational profile. 37.5% and 19.5% respectively were of an AO2 or AO3 level. The AO2 classification is a lower administrative position with no management responsibility. In the context of this organization an AO2 would be primarily responsible for student administration, data entry, answering student enquiries and other service based roles. An A03 would be likely to adopt supervisory roles as well as other higher order administrative tasks. It is likely that a high proportion of employees classified as AO2 and AO3 would be required to interact with the system on a regular basis. Unsurprisingly the largest proportion of participants were users reporting 3+ hours usage (45%), with users reporting 1-2 hours usage (30.5%) and less than 1 hour usage (24.5%) sharing similar response rates.

A review of the demographic profile provided the researcher with a degree of confidence that the longitudinal participants were representative of the wider population. The demographic variables of Gender and Job Grade in particular exhibited patterns

103

consistent with the organizational profile. It should be noted however that in terms of Age employees between 15-29 years were over represented and those 40 years and over were under-represented. It is possible that the web-based mode of survey delivery was better received by workers in the 20-29 age bracket than those in the 40+ age bracket. It is also plausible that the high number of younger participants is reflective of the high number of participants in junior or low level positions.

Those that did not respond at Time 1, but did respond at Time 2 demonstrated no observable differences in demographic profile except in regard to Job Grade. A greater number of participants classified as AO2 (56.5%) responded at Time 2 than in Time 1.

104

Table 5.1 Pilot Site Participant Demographics at Time 1 and Time 2 (% in brackets)

Time 1 (N=145)

Time 2 (N=215)

Longitudinal participants

(N=144)

2005 Employee

Survey DATA

(N=2477)

Gender Male 58 (45.3) 84 (41) 57 (29.1) 33%c

Female 70 (54.7) 118 (59) 70 (70.9) 66%Missing 17 13 17

Age 15-19 18 (14.3) 5 (2.4) 18 (14) 2.2%c

20-29 35 (27.8) 36 (18.0) 35 (28) 9.7%30-39 27 (21.4) 53 (25.9) 27 (21.5) 19.5%40-49 27 (21.4) 54 (26.8) 26 (21) 31%

50+ 19 (15) 54 (27) 19 (15.5) 30%Missing 19 13 19

Tenure (Years)

>1 9 (7.0) 13 (6) 13 (9.8)1-2 23 (18.0) 34 (17) 19 (14.4)3-4 28 (21.9) 34 (17) 22 (16.7)

5-10 39 (30.5) 46 (23) 28 (21.2)11+ 28 (21.9) 72 (36) 50 (37.9)

Missing 17 16 17

Group Tenure (Years)

>1 13 (10.3) 38 (20) 9 (7)1-2 15 (11.9) 50 (25) 23 (18)3-4 44 (34.9) 48 (24.5) 28 (22)

5-10 50 (39.7) 37 (19) 38 (30)11+ 4 (3.2) 24 (12) 28 (22)

Missing 19 18 17

Job Gradee

AO2 55(25.2) 80 (43) 54 (37.5) 31.6%d

AO3 28(12.8) 49 (27) 28 (19.5) 21.8%AO4 21(9.6) 26 (14) 21 (14.5) 16%AO5 13(6) 10 (6) 13 (9) 10%

AO6+ 7(3.2) 18 (10) 7 (5) 16%Missing 21 32 21

System Usea (Hours)

>1 42 (33) 71 (35) 30 (24.5)1-2 37 (29) 70 (35) 37 (30.5)3+ 49 (38) 60 (30) 55 (45)

Missing 17 14 22

a. Self report measure; b. Number in brackets indicates valid percent; c. Represents percentages of all employees within Erudio such as teaching and maintenance staff (not required to use the system); d. Represents percentages of Administrative and Management classifications only; e. AO2 (Administrative Officer 2) an administrative position, typically appointed after a 12month traineeship. AO6 (Administrative Officer 6) A senior management position within the organisation – typically a Faculty director or equivalent.

105

Tabl

e 5.

2 Pe

arso

n’s P

rodu

ct C

orre

latio

n m

atri

x fo

r ant

icip

ated

and

cur

rent

Job

Cha

ract

eris

tics a

t Tim

e 1

and

Tim

e 2

MSd

12

34

56

78

9

1U

sefu

lnes

s T1

4.39

1.19

2E

ase

of U

se T

14.

041.

140.

67**

3U

sefu

lnes

s T2

2.15

1.31

0.01

0.01

4E

ase

of U

se T

22.

211.

30.

04-0

.03

0.75

**

Tim

e 1

5Sk

ill V

arie

ty5.

231.

160.

050.

18*

0.04

0.04

“Cur

rent

”6

Task

Iden

tity

4.58

1.42

0.17

*0.

130.

130.

040.

21*

7Ta

sk S

igni

fican

ce5.

601.

060.

020.

040.

03-0

.01

0.60

**0.

23**

8A

uton

omy

5.04

1.29

0.17

*0.

25**

0.07

0.06

0.50

**0.

35**

0.53

**9

Fee

dbac

k4.

591.

290.

31**

0.32

**0.

120.

000.

42**

0.60

**0.

35**

0.57

**10

MPS

24.8

54.

910.

17*

0.19

*0.

110.

010.

66**

0.67

**0.

71**

0.78

**0.

81**

Tim

e 1

11Sk

ill V

arie

ty5.

101.

150.

050.

080.

04-0

.03

0.74

**0.

090.

50**

0.32

**0.

24**

“Ant

icip

ated

”12

Task

Iden

tity

4.56

1.31

0.41

**0.

30**

0.15

*0.

10-0

.07

0.60

**0.

060.

140.

30**

13Ta

sk S

igni

fican

ce5.

501.

070.

100.

070.

01-0

.07

0.38

**0.

060.

80**

0.32

**0.

17*

14A

uton

omy

4.87

1.21

0.36

**0.

37**

0.04

0.02

0.22

**0.

15*

0.34

**0.

70**

0.36

**15

Fee

dbac

k4.

461.

160.

49**

0.45

**0.

06-0

.03

0.17

**0.

40**

0.17

**0.

41**

0.74

**16

MPS

24.3

64.

560.

36**

0.31

**0.

08-0

.02

0.36

**0.

39**

0.49

**0.

52**

0.52

**

Tim

e 2

17Sk

ill V

arie

ty4.

991.

430.

080.

02-0

.04

-0.0

1-0

.09

0.06

-0.0

9-0

.16*

-0.1

0“C

urre

nt”

18Ta

sk Id

entit

y4.

721.

440.

040.

070.

060.

090.

060.

03-0

.18*

-0.0

3-0

.04

19Ta

sk S

igni

fican

ce5.

661.

010.

07-0

.03

-0.0

7-0

.08

-0.0

4-0

.10

-0.0

4-0

.18*

-0.1

8*20

Aut

onom

y4.

941.

34-0

.02

0.03

0.03

0.12

0.11

-0.0

3-0

.15*

-0.2

2**

-0.1

221

Fee

dbac

k4.

721.

340.

000.

070.

050.

15*

0.17

*0.

06-0

.16*

-0.1

9*-0

.09

22M

PS24

.84

4.92

0.10

0.11

-0.0

80.

080.

130.

00-0

.11

-0.1

7*-0

.14

N =

131

; **p

<.01

; *p<

.05

106

Tabl

e 5.

2 Pe

arso

n’s P

rodu

ct C

orre

latio

n m

atri

x fo

r ant

icip

ated

and

cur

rent

Job

Cha

ract

eris

tics a

t Tim

e 1

and

Tim

e 2

(Con

tinue

d)

1011

1213

1415

1617

1819

2021

2211

Skill

Var

iety

0.53

*Ti

me

112

Task

Iden

tity

0.35

**0.

14“A

ntic

ipat

ed”

13Ta

sk S

igni

fican

ce0.

49**

0.66

**0.

21**

14A

uton

omy

0.52

**0.

46**

0.37

**0.

50**

15F

eedb

ack

0.57

**0.

35**

0.56

**0.

30**

0.63

**16

MPS

0.71

**0.

68**

0.64

**0.

71**

0.81

**0.

78**

17Sk

ill V

arie

ty-0

.05

0.02

0.11

0.06

-0.0

70.

000.

07Ti

me

218

Task

Iden

tity

-0.0

20.

130.

08-0

.11

0.04

0.10

0.09

0.16

*“C

urre

nt”

19Ta

sk S

igni

fican

ce-0

.06

0.12

0.08

0.11

0.01

0.06

0.16

*0.

44**

0.23

**20

Aut

onom

y-0

.10

0.14

0.04

-0.1

0-0

.11

0.00

-0.0

10.

35**

0.34

**0.

44**

21F

eedb

ack

-0.0

70.

140.

09-0

.11

-0.1

1-0

.03

-0.0

10.

25*

0.51

*0.

30*

0.64

*22

MPS

-0.0

40.

23*

0.13

0.05

0.01

0.05

0.15

0.52

*0.

62*

0.63

*0.

77*

0.77

*N

= 1

31; *

*p<.

01; *

p<.0

5

107

5.1.2 Descriptive Statistics and Correlations

The means, standard deviations and inter-correlations of all variables analyzed in this chapter are presented in Table 5.2 above. As expected the outcome variables of Perceived Usefulness and Perceived Ease of Use were moderately correlated within each data collection point. Across time however the acceptance measures did not correlate with each other.

“Current” job characteristics at Time 1 were moderate to highly correlated with their comparable “anticipated” characteristics recorded at Time 1 (.60 - .80, p< .01). “Current” characteristics at Time 1 showed little correlation with “Current” characteristics at Time 2. In fact only Autonomy (r = -.22, p<.05) demonstrated any significant correlation to itself at Time 1. However, given that the change being experienced by users was not a radical departure from their existing role (as for example in the case of a department restructure) one would have expected some degree of correlation to be present. This aspect will be discussed further in the next section concerning change in job characteristics over time.

5.2 Job Characteristics Change

The initial aim of the job design component was to identify whether users perceived a change in job design following the introduction of an ES. Two elements were considered: One, whether users would anticipate a significant difference in their job characteristics prior to the introduction of the system; and two, whether users would perceive a significant difference in their job characteristics post-implementation. The relevant hypotheses are repeated below.

The first set of hypotheses related to users anticipating changes to their job characteristics prior to the introduction of the system. Data used to test these hypotheses were collected three months prior to the ES implementation.

H1a: Users will anticipate a reduction in the job characteristics responsible for the experienced meaningfulness of the work carried out by users following the introduction for ES technology.

H2a: Users will anticipate a reduction in autonomy following the introduction of an enterprise system.

108

H3a: Users will anticipate an increase in feedback levels following the introduction of an enterprise system.

The second set of hypotheses were related to user’s perceptions of their jobs after the system’s introduction. User’s perceptions of their current job characteristics collected four months after the system went live were compared to user’s perceptions of their job characteristics recorded in the first round of data collection.

H1b: Users will perceive a reduction in the job characteristics responsible for the experienced meaningfulness of the work carried out by users following the introduction for ES technology.

H2b: Users will perceive a reduction in autonomy following the introduction of an enterprise system.

H3b: Users will perceive an increase in feedback levels following the introduction of an enterprise system.

Comparing perceived job characteristics pre- and post-implementation required the use of paired sample t-tests (Cooper & Emory, 1995). Paired sample t-tests are appropriate in cases where two means are drawn from the same sample and need to be tested for any significant difference (Cooper & Emory, 1995). Participants were required to indicate at Time 1 and Time 2 the perceived level of job characteristic as categorized by Hackman & Oldham’s (1979) job characteristics model. At Time 1 they were also required to indicate their Anticipated job characteristic level following the introduction of the system (i.e. what users thought their job would be like following the implementation of the system).

5.2.1 Anticipated Job Change at Time 1

Current job characteristics at Time 1 were compared against Anticipated job characteristics. The results are presented below in Table 5.3. In all five cases the difference between current and anticipated job characteristics is negative, indicating that users anticipated a reduction in these characteristics. However, only four recorded a significant decrease following the system’s introduction, those of Skill Variety, Task Significance, Autonomy and Feedback. Therefore the results provided partial support

109

for hypothesis 1a and confirmed hypothesis 2a. The analysis also failed to support hypothesis 3a, as a significant decrease in feedback was observed, rather than an increase, as was hypothesized.

Table 5.3 Paired Sample t-test for current and anticipated Job Characteristics at Time 1

Ma S.Da t df Sig. (1-tailed)

Skill Variety Anticipated -T1 -.18 .82 2.41 123 .01

Task Identity Anticipated -T1 -.05 1.2 .45 122 .33

Task Significance Anticipated -T1 -.16 .64 2.73 122 .01

Autonomy Anticipated -T1 -.22 .95 2.51 122 .01

Feedback Anticipated -T1 -.14 .87 1.71 121 .05

N= 145; a Paired Differences

5.2.2 Perceived Job Change at Time 2

The second set of analyses ran paired sample t-tests to determine whether a significant difference existed between “current” job characteristics at Time 1 and “current” job characteristics at Time 2. No significant differences were found between any of the five job characteristics over time. Consequently hypotheses 1b, 2b and 3b were not supported by the analyses. The results of the analysis are presented in Table 5.4 below.

Table 5.4 Paired Sample t-test for current Job Characteristics Time 1 & Time 2

Ma S.Da t df Sig. (1-tailed)

Skill Variety T2 – T1 -.23 1.92 1.30 115 .10

Task Identity T2 – T1 .12 1.99 -.66 112 .25

Task Significance T2 – T1 .03 1.49 -.22 112 .41

Autonomy T2 – T1 -.18 2.10 .85 112 .20

Feedback T2 – T1 .16 1.96 -.87 111 .20

N= 131; a Paired Differences

On the basis of these results it would appear that while users anticipated a reduction in at least three job characteristics before the system went live, this concern on average was not realized by users. Instead, the mean scores of job characteristics over time indicated a non-significant change following the introduction of ES technology.

110

However, a review of the data indicated that the relationship between ES characteristics and job change may be more complex than initially argued. A closer examination of the data revealed that different types of users did experience different degrees of job change. As observed previously, while the mean scores did not significantly change there was a lack of correlation between the two sets of scores over time (See Table 5.2 above). In an attempt to identify why this might be the case a series of analyses were undertaken to investigate this further.

Initially users were categorized into those that had increased (n=45), decreased (n=52) or experienced no change (n=19) in job design over time. Hackman and Oldham’s (1976) composite variable of Motivational Potential Score (MPS) was used as an overall measure of job design. As briefly outlined in Chapter Two, MPS represents the overall enrichment capability of a job, reflected by the addition of all five characteristics to form one measure (Boonzaier et al., 2001). In the analysis reported here MPS is used as a parsimonious measure of job design as it reflects the overall pattern of all five characteristics in one measure. As such the use of MPS increases the clarity of the results and reduces the risk of family-wise error posed by replicating the analyses over all five characteristics (Tabachnick & Fidell, 1989).

A series of ANOVA’s were run to compare the characteristics each user category. Significant differences were observed between users’ change in MPS and their level of MPS at Time 1 (F(2, 112) = 36.19, p<.001), System Usage (F(2, 111) = 2.68, p<.10) and Group Tenure (F(2, 111)=3.55, p<.05). No significant differences were identified between MPS change and the demographic variables of Job Grade, Tenure, Age, and Number of Subordinates. Where a significant difference was reported Tukey’s Honestly Significant Difference (HSD) was used to identify exactly where the differences occurred (Berenson & Levine, 1996; Corston & Colman, 2000).

Change in MPS was seen to be differentiated by System Usage. Users reporting no change in MPS reported the lowest usage rates (Musage = 1.78). Those reporting an increase (Musage = 2.56) or decrease (Musage = 2.51) reported similar usage rates but at higher levels than those reporting no change. As predicted usage rate appears to act as a moderator, with low users reporting little or no change in their job characteristics. In contrast, higher usage rates appear to have a relationship (albeit inconsistent) with job change. In particular increased system usage appears to have a polarizing effect on user’s job design with users reporting both increases and decreases in MPS over time. Usage as a potential moderator is investigated in greater detail later in this chapter.

111

Those reporting “no change” were also more likely to report higher levels of group tenure (Mtenure= 3.28) than those reporting an increase (Mtenure = 2.96) or decrease (Mtenure = 2.43) in MPS. However those reporting an increase or a decrease in MPS were not significantly different from each other in regard to group tenure. It is possible that those reporting longer levels of group tenure held positions less dependant on the use of the system (such as group supervisors) and therefore experienced less job change.

Additional analysis revealed significant differences in users reporting MPS change and pre-existing levels of job enrichment prior to the system’s introduction. All three categories reported significantly different levels of MPS at Time 1 to each other. Users decreasing in MPS over time reported significantly higher levels of MPS at Time 1 (Mmps1 = 27.73) than those reporting no change (Mmps1 = 25.23) or those reporting an increase (Mmps1 = 21.04). In an attempt to further understand the nature of the change in MPS over time experience by high users another series of ANOVA’s were run. High users experiencing a decrease in MPS at Time 2 were more likely to be of a lower rank (F(1,82) = 3.23, p<.10) have less subordinates (F(1, 88) = 3.69, p<.01) and report shorter work group tenure (F(1, 94) = 3.90, p<.10) than high users increasing in MPS over time.

Overall these results suggest a move towards parity in enrichment levels between users, with those reporting high levels of enrichment pre-implementation experiencing a reduction in job enrichment levels following the system’s introduction. Conversely those with comparatively low levels of job enrichment experienced an increase in enrichment post-implementation. However hierarchical level appeared to determine the direction of the change in job enrichment experienced by users. Those with less formal power in the organization tended to experience a decrease in job enrichment. Conversely, those higher in formal hierarchy experienced an increase in job enrichment. Possible explanations for these results will be covered later in Chapter Six.

5.2.3 Summary

In summary, the argument that users would universally anticipate and perceive a change in job characteristics following the introduction of ES technology was only partially supported. While users anticipated a decrease in Skill Variety, Task Significance and Autonomy at Time 1 as predicted, users did not anticipate a decrease in Task Identity or an increase in Feedback. At Time 2 the data reported no significant change in the mean scores of job characteristics across time. Further analysis indicated that perceptions of job change were determined by variables such as usage, hierarchy and pre-existing levels of job enrichment. In short, users reporting no change to their job

112

characteristics were more likely to report Low usage and have been a member of their work group the longest. High users were more likely to experience change in their job characteristics, but not uniformly. Those categorized as high users with pre-existing high levels of enrichment reported a decrease in their job’s enrichment quotient at Time 2. Conversely high users with low job enrichment levels at Time 1 reported an increase in job enrichment. The direction of change in job enrichment appeared to be determined by the degree of formal authority reported by the user.

While these results indicate the need to further investigate these results a limited sample size prevented further investigation at this time. The following section will report a series of analyses carried out to determine whether anticipated or perceived job change had a relationship to user acceptance.

5.3 Job Design Change and User Acceptance

In addition to discussing the potential for job design change following the introduction of an ES, Chapter Two also considered the attitudinal consequences of such change occurring. This section of the chapter reports on the analysis carried out to test the hypotheses relevant to that discussion. Section 5.3 is divided into two parts in order to communicate the results effectively. The first part presents a series of hierarchical regressions testing the direct relationship between job design change and user acceptance. The second reports the results of a series of moderated regressions testing System Usage as a hypothesized moderator between job change and user acceptance.

5.3.1 Hierarchical Regression

The aim of this analysis was to test the relationship between job characteristics and user acceptance following the introduction of an ES. Hypotheses 4a and 4b argued that user’s perceptions of system driven job change would determine levels of system acceptance. As in the previous set of analyses, the theory was interested in user’s expectations of job change as well as their perceived job change over time in relation to levels of user acceptance. Hypotheses 4a and 4b argued for a positive relationship existing between job characteristics and user acceptance following the introduction of the new system.

H4a: Users will report a positive relationship between anticipated job characteristics change and levels of user acceptance at Time 1.

113

H4b: Users will report a positive relationship between perceived job characteristics change and levels of user acceptance at Time 2.

Two set of analyses were planned. One, to examine the relationship between “Anticipated” job characteristics and user acceptance at Time 1, prior to the system going live. Two, to examine the relationship between current job characteristics change and user acceptance at Time 2 (post-implementation).

Anticipated Changes in Job Design at Time 1 and User Acceptance

In order to test the relationship between anticipated job characteristics and user acceptance, data collected at Time 1 were used. Participants were asked to consider their “current” job characteristics and then to consider what those characteristics may be like after the system was implemented. Davis’s (1989) TAM variables of Perceived Usefulness and Perceived Ease of Use were used as the dependant variables in two sets of analysis.

In order to test the hypotheses concerning the link between anticipated job characteristics and user acceptance a hierarchical regression technique was used, with the user acceptance variables at Time 1 (Perceived Usefulness; Perceived Ease of Use) as the dependant variables. A similar approach has been used previously to measure the effect of work characteristics on employee attitudes following a change intervention (Axtell & Parker, 2003; Parker, Griffin, Sprigg & Wall, 2002) and is considered preferable to the use of difference scores when comparing the effect of an IV on a DV over time (Edwards, 1994; 2001).

In the first step Current job characteristics were entered to control for existing job characteristic levels and to test for lagged effects (Parker et al., 2003). These were user’s job characteristics as they perceived them prior to the introduction of the ES. In the second step Anticipated job characteristics were entered. These were user’s job characteristics as they anticipated them after the introduction of the ES. Any variance accounted for in the final model is the contribution of anticipated job characteristics on user acceptance (Axtell & Parker, 2003). While causality remains untested, controlling for Current job characteristics allows a stronger test of association between user acceptance and Anticipated job characteristics (Parker et al., 2002). This method is used for the remainder of the analyses reported in this chapter. The directional nature of the hypotheses allowed the use of 1-tailed levels of significance, all results reported from this point onward are reported as such.

114

Two hierarchical multiple regressions were carried out. The effect of anticipated job change on acceptance was measured by regressing all five Anticipated (pre-system) characteristics onto the variable of Usefulness while controlling for perceived Current characteristics. The equation (Adjusted R2 = .36, ∆R2 = .30, F∆R

2 (5,111) = 11.30, p<.001) was significant and reported moderate variance explained. The coefficients indicated that anticipated Autonomy (β=.53, p<.01), Task Identity (β=.36, p<.01) and Skill Variety (β= -.90, p<.001) explained the majority of the variance in relation to Usefulness.

The second regression examined the same independent variables (anticipated job characteristics, controlling for “current” job characteristics) but instead used Ease of Use as the dependent variable. As to be expected given the similarity in distribution of scores for the DV’s, the equation produced a similar result. The five job characteristics were able to explain 29% of the variance in the DV (Adjusted R2 = .39, ∆R2 = .29, F∆R

2(5,111) = 11.74, p<.001). A review of the coefficients identified anticipated Skill Variety (β= -1.11, p<.001), Autonomy (β=.56, p<.01), and Task Significance (β=.51, p<.01) as those explaining the majority of the variance after controlling for those variables at Time 1.

The partial, zero-order and simple bivariate correlations of both equations however indicated the need to interpret the coefficient results with caution (Cohen, Cohen, Aitken & West, 2003). An initial review of the partial and zero-order correlations highlighted the possible presence of suppression variables (Tabachnick & Fidell, 2001). Tabachnick and Fidell (2001) offer two possible indicators of suppressor variables: One, where the simple correlation score of the IV is significantly different from zero and has a different sign to that of the standardised regression weight of the IV; and two, when the absolute value of the simple correlation is substantially smaller than the beta weight of the IV. If either of these conditions are fulfilled then it may indicate the presence of a suppression variable (Tabachnick & Fidell, 1996, 165). Reviewing the results for Perceived Usefulness revealed possible problems with the IV’s of Task Identity (r = .17; β = .32) and Autonomy (r = .17; β = .52). Similarly potential problems with Skill Variety (r = .18; β = -1.10) and Autonomy (r = .25; β = .56) were identified when Perceived Ease of Use was the DV.

Further investigation and an examination of the bivariate correlations suggested that multicollinearity between the job characteristic variables was preventing a reliable computation and estimation of the coefficients (Cohen et al., 2003). A review of the variance inflation factor (VIF) indicated ratios of above 10 for Task Significance and above 7 for all remaining variables bar Task Identity when treating Usefulness and

115

Ease of Use as DV’s. VIF scores of 10 and above have been cited as indicative of multicollinearity problems, however Cohen et al. (2003) suggest this is too high and in fact should be lower. Consequently it is suggested that the results reported here should be interpreted with caution. Additional analyses to better understand the relationship between job characteristics change and user acceptance were carried out. These are reported below.

5.3.1.1 Composite Score Analysis

In attempting to remedy issues of multicollinearity Cohen et al. (2003) suggest three remedies: One, collect more data; two, respecify the model; or three, conduct a range of alternate regression approaches such as Ridge and Principal Components regression. Given the nature of the data collection and the characteristics of the existing data set it was considered that the second option was the most viable. If theoretically appropriate, researchers may choose to combine the correlated measures to produce a composite variable (Cohen et al., 2003). In the case of Hackman and Oldham’s (1979) job characteristics model the composite variable of Meaningfulness already exists. In review, Hackman and Oldham’s (1983) model of job enrichment outlined five elements that if present within a job leads to increasing levels of enrichment through the realization of increased meaningfulness, felt responsibility and knowledge of job outcomes. Meaningfulness is represented by an aggregate of Skill Variety, Task Identity and Task Significance (Boonzaier et al., 2001). As such the three variables of Skill Variety, Task Identity and Task Significance were aggregated to form a new composite variable of Meaningfulness.

Two regressions were run with the same dependant variables and with Meaningfulness, Autonomy and Feedback included as the IV’s. In all cases the effect of the job characteristics on the DV was measured by controlling for the effect of that variable in its “current” form. The regression with the three variables of anticipated Meaningfulness, Autonomy and Feedback while controlling for those “current” variables were able to explain 15% of the variance when attempting to predict Perceived Usefulness (Adjusted R2 = .22, ∆R2 = .15, F∆R

2(3,115) = 7.79, p<.001). A review of the coefficients identified Feedback (β = .46, p<.05) as explaining the majority of the variance over Autonomy (β = .29, ns) and Meaningfulness (β = -.14, ns).

A similar result was observed when the same IV’s were regressed onto the DV of Perceived Ease of Use (Adjusted R2 = .20, ∆R2 = .12, F∆R

2(3,115) = 6.01, p<.001). Again, a review of the coefficients identified Feedback (β = .49, p<.05) as explaining

116

the majority of the variance over Autonomy (β = .36, ns) and Meaningfulness (β = -.34, ns). In both regressions a review of the correlational data and VIF scores failed to identify any multicollinearity or suppression effects being present.

Perceived Changes in Job Design Over Time and User Acceptance

While participants reported an anticipated change in the job characteristics pre-implementation, they failed to report any perceived change in mean scores over time. However, the pattern of the data reported in the correlational analysis along with a review of the descriptive statistics (see Table 5.5 below) indicated that the relationship between perceived job change over time and acceptance should still be tested (with little optimism however).

As in previous analyses the same hierarchical regression technique was employed, however with one difference. In order to control for the effect of user acceptance at Time 1 each of the acceptance variables were entered in as the first step in the model. Controlling for the DV at Time 1 allows the researcher to effectively capture change in the DV at Time 2 (Axtell & Parker, 2003). Two regressions were carried out, each concerned with Perceived Usefulness and Perceived Ease of Use respectively.

The relationship between Perceived Usefulness at Time 2 and job characteristics at Time 2 was first tested. At step one Perceived Usefulness at Time 1 was entered in as a control variable. At step two the three variables of Meaningfulness, Autonomy and Feedback at Time 1 were entered. The final step entered the three IV’s measured at Time 2. A similar process was followed for Perceived Ease of Use. In the resulting equations no significant results were found for either Perceived Usefulness (Adjusted R2 = .02, ∆R2 = .01, F∆R

2(3,103) = .23, ns) or Perceived Ease of Use (Adjusted R2 = .05, ∆R2 = .05, F∆R

2(3,103) = 1.67, ns). Consequently the null for hypothesis 4b was unable to be rejected.

Table 5.5 Means, Standard Deviations and Variance of Job Characteristic difference scores (T1 – T2)

M SD Var...

Skill Variety .23 1.92 3.69Task Identity -.12 2.00 3.99Task Significance -.03 1.50 2.24Autonomy .17 2.10 4.42Feedback -.16 1.96 3.85MPS -.61 3.46 11.95

Note: N= 122; Score was measured on a seven point Likert scale.

117

5.3.2 Usage as a moderator of Job Characteristics Change and User Acceptance

Hypothesis 5 required the testing of System Usage as a moderator of the relationship between job characteristics at Time 2 and acceptance levels. In brief it was argued that those with increasing levels of system usage would report a stronger relationship between job characteristics and user acceptance at Time 2 than those with reduced levels of usage.

H5: The degree of system usage required by the user will moderate the relationship between system driven job change and user acceptance, such that high users will report a stronger positive relationship between job change and user acceptance than low users.

System Usage was a self report measure coded into a discrete variable with two categories determined using a median split process into High (>3 hours per day; n = 63), and Low users (<3 hours per day; n = 69). The discrete variable of Usage was then recoded into dichotomous “dummy” variables (Jaccard & Turrisi, 2003). Consistent with previous analyses four initial sets of analyses were run treating each of the variables of Meaningfulness, Autonomy and Feedback as separate IV’s. In the case of a significant interaction for Meaningfulness regressions were also carried out for the variables of Skill Variety, Task Significance and Task Identity to investigate further the effect of Usage on the Meaningfulness components.

In order to reduce any potential effects of multicollinearity the six independent continuous variables were centred around their respective means (Cohen et al., 2003). In order to test for possible interaction effects both the cross-sectional Time 1 data and the longitudinal data at Time 2 was tested. In all, 12 regression analyses were performed for each IV and both DV’s at each time point using SPSS regression. Each of the five job characteristics and the composite score of Meaningfulness were examined for the presence of Usage as a moderator. As with previous analyses, due to the directional nature of the hypotheses all significance levels are reported at a 1-tailed level of significance. However, the high number of regressions involving two DV’s from the same data set increased the potential for a Type 1 error due to family-wise error (Tabachnick & Fidell 1989). A Bonferroni adjustment was performed to correct for the chance of a false positive. The standard formulae (p/C, where C = the number of tests in a series) was used, requiring the set p<.05 significance level to be adjusted downwards to p<.03 (Maxwell & Delaney, 1990). As such all results reported below reflect the Bonferroni adjustment. The results of the analyses are described below.

118

5.3.2.1 Cross-Sectional Time 1 Data - Anticipated Job Characteristics

In review, at Time 1 participants were required to indicate their perceived “current” job characteristics while still using the old system. They were also required to anticipate the level of those same characteristics after the system had gone live. A test of the mean scores of all five characteristics indicated significant anticipated decreases in Skill Variety, Task Significance, Autonomy and Feedback. A series of hierarchical regressions indicated that anticipated levels of Meaningfulness, Autonomy and Feedback predicted up to 15% of the variance in acceptance scores at Time 1.

The analysis below attempted to determine whether a user’s amount of system use better predicted a relationship between Anticipated job characteristics and user acceptance at Time 1. Significant interaction effects were observed when examining anticipated levels of Meaningfulness, Skill Variety and Task Significance in relation to Perceived Ease of Use and Perceived Usefulness at Time 1. No significant interactions were observed for anticipated levels of Task Identity, Autonomy or Feedback for either dependent variable when considering usage as a potential moderator.

System Usage As A Moderator Between Change In Meaningfulness and User Acceptance

Perceived Ease of Use was the only dependant variable reporting a significant interaction when considering the effect of Usage on the relationship between anticipated Meaningfulness and Perceived Ease of Use (R2 = .13, ∆R2 = .05, F∆R

2(1, 116) = 6.13, p<.01). Simple slope analysis identified significant effects for Low users (.44, p<.001) when testing the relationship between anticipated Meaningfulness and Perceived Ease of Use but not for High users (-.02, p=.46). A graph of the simple slope can be found over in Figure 5.1.

119

Figure 5.1 System Usage as a moderator of the relationship between anticipated Meaningfulness and Perceived Ease of Use at Time 1

Note: Low Users (.44, p<.001) High Users (-.02, p=.46)

Table 5.6 System Usage As A Moderator Of The Relationship Between Anticipated Meaningfulness And Perceived Ease Of Use At Time 1

Step & Predictor 1 2 3 4

1. Time 1 Meaningfulness .20* .02 .04 .01

2. Time 2 Meaningfulness .27* .27* .44***

3. Usage category -.06 -.04

4. Interaction term -.27*

R2 .04 .08 .09 .13Adjusted R2 .03 .07 .06 .10

∆R2 .04* .00 .05*

Note: DV is Perceived Ease of Use; *p<.03; **p<.01; ***p<.001

System Usage As A Moderator Between Change In Skill Variety And User Acceptance

Significant effects for both dependant variables were also observed when examining the effect of Usage on the relationship between anticipated Skill Variety and Perceived Usefulness R2 = .07, ∆R2 = .05, F∆R

2(1, 117) = 5.65, p<.01) and Perceived Ease of Use (R2 = .08, ∆R2 = .04, F∆R

2(1, 117) = 5.14, p<.02) when controlling for current Skill Variety levels. Simple slope analysis failed to identify either category as having a significant effect at p<.03 for Perceived Usefulness (Low use .28, p=.04; High Use -.19, p=.07). Simple slope analysis did identify a significant effect for High users

120

when considering anticipated Skill Variety and Perceived Ease of Use (-.33, p<.03) but not for Low users (.12, p=.24). A graph of the simple slope is presented below in Figure 5.2. Of note is the negative relationship between anticipated Skill Variety and Ease of Use for High users. High users anticipating an increase in Skill Variety are more likely to report reduced perceptions of Ease of Use. In contrast, there appears to be no relationship between anticipated levels of Skill Variety and Ease of Use for Low users. This may be explained by the fact that Low users are perhaps less likely to receive a significant change in Skill Variety following the system’s introduction.

Figure 5.2 System Usage as a moderator of the relationship between anticipated Skill Variety and Perceived Ease of Use at Time 1

Note: Low Users (.12, p=.24) High Users (-.33, p<.03)

Table 5.7 System Usage As A Moderator Of The Relationship Between Anticipated Skill Variety And Perceived Ease Of Use At Time 1

Step & Predictor 1 2 3 4

1. Time 1 Skill Variety .19* .26* .26* .172. Time 2 Skill Variety -.10 -.01 -.33*3. Usage category -.01 -.024. Interaction term -.26*

R2 .04 .04 .04 .08Adjusted R2 .03 .03 .02 .05

∆R2 .00 .00 .04*

Note: DV is Perceived Ease of Use; *p<.03; **p<.01; ***p<.001

121

While a significant interaction was observed for the effect of Usage on anticipated Task Significance (R2 = .05, ∆R2 = .05, F∆R

2(1, 117) = 6.13, p<.01) simple slope analysis failed to identify any significant effects for either user category.

In summary, the results suggest prior to the introduction of an ES low users are more likely to associate anticipated changes in meaningfulness with acceptance following the introduction of an ES, than high users. High users were observed to have a significant negative relationship between anticipated Skill Variety and user acceptance. In contrast to the hypothesized aims of Hackman and Oldham’s (1976) job characteristics model high users are more likely to report lower acceptance levels the larger the anticipated increase in Skill Variety. Again the implications of this finding will be discussed in Chapter Six. The next section below reports the results of moderated regressions testing the presence of usage as a moderator between perceived job characteristics at Time 2 and user acceptance at Time 2.

5.3.2.2 Cross-Sectional Data – Perceived job characteristics at Time 1 and Time 2

For this data set participant’s “current” job characteristics scores at Time 1 and Time 2 were used. In review, on average no significant differences in job characteristics levels were observed over time. Twelve models were tested for the effect of Usage as a moderator between job characteristics change and the two dependent variables (Perceived Usefulness and Perceived Ease of Use). The two usage categories were again used to reflect High (>3hrs) and Low users (<3hrs). See Table 5.10 and 5.11 below for a summary table of the results.

Only one job characteristic supported Usage as a moderator of the relationship between job characteristics and acceptance at Time 2. As to be expected given the increased correlation of the two DV’s at Time 2, the results were similar for both. Usage reported a significant but weak effect for Task Identity at Time 2 when predicting Usefulness at Time 2 (R2 = .07, ∆R2 = .05, F∆R

2(1, 106) = 5.11, p<.02) and Ease of Use at Time 2 (R2 = .10, ∆R2 = .08, F∆R

2(1, 108) = 9.27, p<.001) when controlling for each DV at Time 1 and Task Identity at Time 1. Simple slope analysis identified a significant effect for Low users (<3hrs) but not for High users in both cases. See Figures 5.3 and 5.4 below for a plot of the simple slopes. As such, while the result represents an interesting finding it fails to support hypothesis 5 in that low users have reported a stronger relationship between Task Identity levels at Time 2 than high users.

122

In response to the results reported earlier (see Section 5.2.2) it was considered necessary to control for any possible effects hierarchy may have on the relationship between job change, usage and user acceptance. The ordinal variable of “Job band” was used to control for any direct effects. An interaction term calculated between “job band” and Task Identity at Time 2 was used to control for any possible effects of an interaction between hierarchy, Task Identity and both user acceptance variables. The results indicated that even when controlling for possible direct and interaction effects explained by hierarchy, the interaction terms for both equations for Low users continued to be significant. While a relationship between usage and hierarchy may exist when considering job change over time (as was indicated in Section 5.2.2), the presence of hierarchy as a possible confounding variable has been discounted. As such the results allow the researcher greater confidence that usage does moderate the relationship between Task Identity at Time 2 and user acceptance.

Figure 5.3 System Usage as a moderator of the relationship between perceived change in Task Identity and Perceived Usefulness at Time 2

Note: Low (β=.31, p<.02) High (β=-.13, ns)

123

Table 5.8 System Usage as a Moderator Of The Relationship Between Task Identity And Perceived Usefulness at Time 2 While Controlling for Task Identity at Time 1

Step & Predictor 1 2 3 4 5

1. Time 1 Perceived Usefulness .05 .03 .03 .03 .062. Time 1 Job Characteristics

Task Identity .11 .11 .11 .133. Time 2 Job Characteristics

Task Identity .11 .11 .31*4. Usage category .00 -.015. Interaction term -.30*

R2 .00 .02 .03 .03 .07Adjusted R2 .00 .00 .00 -.01 .03

∆R2 .00 .01 .01 .00 .05*

Note: DV is Perceived Usefulness; *p<.03; **p<.01; ***p<.001

Figure 5.4 System Usage as a moderator of the relationship between perceived change in Task Identity and Perceived Ease of Use at Time 2

Note: Low (β=.40, p<.001) High (β=-.19, ns)

124

Table 5.9 System Usage as a Moderator Of The Relationship Between Task Identity And Perceived Ease of Use at Time 2 While Controlling for Task Identity at Time 1

Step & Predictor 1 2 3 4 5

1. Time 1 Perceived Ease of Use -.03 -.03 -.03 -.03 .012. Time 1 Job Characteristics

Task Identity .01 .01 .01 .043. Time 2 Job Characteristics

Task Identity .13 .13 .40**4. Usage category .02 .015. Interaction term -.40**

R2 .00 .00 .02 .02 .10Adjusted R2 .00 .00 .00 .00 .06

∆R2 .00 .02 .00 .08**

Note DV is Perceived Ease of Use; *p<.03; **p<.01; ***p<.001

5.4 Chapter Summary

Chapter Five reported the results of analysis testing hypotheses 1 – 5 set down in Chapter Two. In summary it was hypothesized that users would both anticipate (1a-3a) and experience (1b-3b) a change in job characteristics following the implementation of ES technology. A positive relationship between job change and user acceptance was also hypothesized (4a and 4b). Finally System Usage was argued to be a moderator of the relationship between job change and acceptance (5).

Analyses using the Time 1 data found partial support for Hypotheses 1a, 2a, 4 and 5. Consistent with user expectation research (Ryker, Nath & Henson, 1997; Szajna & Scamell, 1993) participants anticipated a decrease in Skill Variety, Task Significance, Autonomy and Feedback prior to the system going-live at Time 1.

Analyses concerned with job change over time and user acceptance found partial support for hypothesis 5, but not for hypotheses 1b, 2b, 3b or 4. Paired sample t-tests comparing mean differences in job levels found no significant difference in job characteristics across time. A closer examination of the data indicated that increases or decreases in job enrichment were linked to both system usage and pre-existing levels of job enrichment. On average Low users reported no change in job enrichment, however High users reported both increases and decreases in job enrichment. Further analysis

125

indicated a change in job enrichment was consistent with hierarchical structure. Users increasingly higher up the organizational ladder tended to increase their enrichment levels, while those in lower ranking positions tended to report a decrease in their job enrichment.

More targeted analysis confirmed System Usage as a moderator for both acceptance measures at Time 1 and Time 2. At Time 1 Low users were seen to report a positive relationship between anticipated job meaningfulness and user acceptance levels at Time 1. In contrast High users reported a negative relationship between anticipated Skill Variety and perceived ease of use at Time 1. At Time 2 Low users reported a significant relationship between Task Identity at Time 2 and both acceptance variables at Time 2 after controlling for levels of the DV and IV’s at Time 1 as well as hierarchy as a possible confound.

Regression analysis testing direct effects between job characteristics at Time 2 further supported the hypothesis that anticipated job levels had a significant relationship with user acceptance when controlling for current job characteristics. Anticipated levels of Feedback reported the strongest relationship to both Usefulness and Ease of Use at Time 1. Similar analyses failed to identify a relationship between perceived job levels at Time 2 and user acceptance at Time 2.

The next chapter will discuss these results in light of the hypothesized relationships, examining referring to previous research and contextual elements that may help to better explain the pattern of results.

126

127

Chapter Six — Job Design Results: Post–Hoc

Analysis & Discussion

6.0 Introduction

The previous chapter presented the results of the analysis for the job design component of the study. This chapter will discuss each of the findings in light of the hypotheses they were testing. The structure of the chapter will mirror that of Chapter Five, beginning with a discussion of job change over time. The hierarchical regression results will then be addressed, before concluding with a discussion of the moderated regression results.

Where appropriate the findings will be considered in light of the research context, attempting to provide clearer understanding of what was found. In doing so supplementary sources of information will be used. Documentation such as Erudio implementation and change management reports will be used, as well as user reactions and reflections obtained through a series of semi-structured interviews (See Appendix A.05 for interview notes). It should be noted that the supplementary information is not offered up as an additional data set but as an attempt to contextualize the findings and therefore better explain what was reported.

6.1 Job Characteristics Change

“Less work, less challenging, less responsibility and “no say” in the work that is being done” (user #7).

This quote was offered by a particularly disgruntled user who felt that the password access controls, the institution of “new roles” and limited training had resulted in a de-enriched job, and was particularly vocal in their condemnation of the system (HAL, a pseudonym). Hypotheses 1, 2 and 3 were formulated to test whether a representative proportion of users shared similar views. The study was designed to measure user’s expectations as well as any perceived change that may have occurred over time. In the broader context of the study this allowed the researcher to examine whether ES technology altered or replaced the set of structural properties represented by existing job design elements. It was argued that users would both anticipate and perceive a

128

decrease in the job characteristics of Skill Variety, Task Identity, Task Significance and Autonomy following the introduction of ES technology. It was also hypothesized that users would anticipate and perceive an increase in Feedback.

6.1.1 Anticipated Job Change at Time 1

In terms of predicting user’s expectations of job change following the introduction of an ES, the following hypotheses were put forward.

H1a: Users will anticipate a reduction in the job characteristics responsible for the experienced meaningfulness of the work carried out by users following the introduction for ES technology.

H2a: Users will anticipate a reduction in autonomy following the introduction of an enterprise system.

H3a: Users will anticipate an increase in feedback levels following the introduction of an enterprise system.

A series of paired sample t-tests partially supported hypothesis 1a, fully supported hypothesis 2a but failed to support hypothesis 3a. Users were seen to anticipate a significant decrease in Skill Variety, Task Significance and Autonomy. However, they failed to anticipate a decrease in Task Identity or an increase in Feedback. The results here further our understanding of how ES may affect work context and of how users anticipate those changes. The results also provide support for the argument that perceptions concerning the mandatory and automated nature of ES can lead to an anticipated reduction in key job characteristics. This is discussed in detail below.

Skill Variety, Autonomy and Task Significance

Prior to the introduction of the system users reported an anticipated decrease in four of the five job characteristics measured in the study. Skill Variety, Autonomy, Task Significance and Feedback were all identified as characteristics that may decrease following HAL’s implementation. This is of particular interest as the two elements of Skill Variety and Autonomy have received strong support within the literature as predictors of attitudinal outcomes (Dodd & Ganster, 1996).

129

The reader is cautioned however that the result for Skill Variety, Task Significance and Autonomy may be an artifact of the JCM constructs being measured. The three variables reported moderate – strong correlations with each other at each time point (.66 -.46). Existing JCM research suggests that participants may fail to differentiate between the three variables, and that together they may actually measure a single construct tapping into a single domain (Fried & Ferris, 1987). While causal attribution is difficult one can see the potentially high degree of interrelationship between the three. Levels of skill variety are likely to effect the amount of autonomy present or required with a job (Dodd & Ganster, 1996). Arguably, increasing levels of autonomy and skill variety could lead to increasing levels of task significance (or vice-versa). As a result it may be difficult to identify what aspect of ES technology users perceived as contributing to a decrease in these three characteristics.

Chapter Two predicted a reduction in Skill Variety due to aspects such as process automation and the redundancy of an existing and non-transferable skill set. Similarly the mandatory nature of the system, the increased controls and visibility of user compliance were considered likely to result in perceptions of reduced autonomy. It is possible that users feared a reduction in their existing skill set developed over an extensive period. On average 43% of users surveyed in this study had been in their current workgroup for over 5 years. Entrenched users may have accumulated increased levels of Skill Variety, Autonomy and Task Significance over time due to their knowledge and/or control of the current system and associated business processes. If job characteristics such as these are heavily linked to a proprietary skill set about to be made redundant, it is unsurprising that users would anticipate a decrease in their characteristics.

Arguably Task Significance is one of the least obvious job characteristics to be affected by ES technology. It is possible that the concerted shift away from task and role based job functions to a work process mode of operation may go some way to explaining user’s expectations of a decrease in both Task Significance and Autonomy. Erudio users were informed of the shift towards “roles” required by “system managed workflows” (HLOAR, 2002, 8). It is possible that the heavy emphasis placed on the vital importance of process over function had the unintended effect of alienating employees. Felt ownership of their role and job responsibility may be subsumed by the ES and its “system managed workflows”. This feeling of displacement may have been further compounded by lack of ownership of the system itself. In the case reported here, HAL replaced a highly customised legacy system that had evolved over a decade in response to user’s needs. Anecdotally, while many users recognised the dated nature of the legacy system they also expressed an affinity with it and the benefits of being

130

able to request enhancements at a localised level. Ironically it was this attribute that led to a range of procedural inconsistencies across the organisation and the decision to adopt an ES.

Task Identity and Feedback

The failure of the analyses to support the Feedback and Task Identity hypotheses at Time 1 however are particularly surprising, especially given the extent to which the need for system compliance is typically stressed prior to an ES implementation (Oliver & Oliver, 2002). A common theme among ES technologies is their integrated nature and sequential treatment of data (Lengnick-Hall et al., 2004). Arguably those users more aware of the data process and their role within it would report increased Task Identity. Similarly, given the potential for system integrity to be compromised by poor user protocol, one would have considered that feedback mechanisms and developing an increased awareness of sequential processing would have been salient in user training. However a follow-up interview with a trainer indicated that this wasn’t the case. Users had limited exposure to their particular functions and little else, describing the training as “more of an overview” (user #18). It is possible that due to the lack of information users failed to realize the potential impact the system may have on both Feedback and Task Identity. Instead, the potential for the system to impact on their skills, autonomy and overall contribution to the organization were the areas most apparent to users.

Context Factors

Discussions with users indicated that the lack of information provided may have compounded fears of a decrease in job enrichment following HAL’s introduction. As mentioned previously, several users made reference to the lack of training and briefing about the system prior to go-live. One user (#5) in particular complained that “if people are only told about “big changes” then they’ll anticipate them and be unsure [of what those changes are]. If we had been told that you’re doing the same role/task, just slightly differently it would have allayed our fears” (user #5). This comment highlights the importance of clear and frank communication to users about anticipated changes (Miller, Johnson & Grau, 1994). A “need to know” information strategy runs the risk of speculation and rumour which may unwittingly provide the basis for resistance or reduce user’s openness to change (Brown & Quarter, 1994; Miller et al., 1994). Of course the extent to which anticipated changes are communicated to users is dependant on how well informed the organisation is about those changes. There

131

is evidence to suggest that Erudio management did not have a clear picture of the changes that were (or were not) about to occur. Official documents present conflicting messages. The following quotes indicate a continuum of opinion on the degree of planned change (emphasis added).

“Job change is not an intended outcome of the HAL project” (HLOAR4, 2002, 8).

“The introduction of HAL will encourage and/or require some level of standardization of work practices within and across the groups” (HLOAR, 2002, 7).

“Implementing HAL is more than just a technical solution. It is about achieving business success by ensuring maximum alignment between the ERUDIO people, process, policy and the system. The implementation of HAL will drive a fundamental shift in the way the organization manages student administration services and ERUDIO staff must be aligned with the new paradigm for maximum benefits to be achieved” (Change Management Strategy, 2003, 11).

Given the degree of ambiguity regarding the extent of job change evident at a project-management level, it is unsurprising that a similar (or increased) degree of uncertainty existed among users prior to the system’s implementation.

In summary, this component of the study found support for the idea that users would anticipate a decrease in job characteristics prior to the implementation of an ES. Considering the strong support that the JCM has with attitudinal and to a lesser degree behavioral outcomes (Fried & Ferris, 1987), these results may be of interest to IS/IT practitioners. Post-hoc reports suggested that improved organizational communication about the planned changes may have reduced anticipation levels. Further, organizational awareness and readiness for the planned changes were also identified as possible contributory factors in the quality and quantity of information provided. Previous research supports the link between user expectations and the amount of realistic and credible information available pre-implementation (Miller et al., 1994; Ryker et al., 1997; Szajna & Scamell, 1993). Consequently the use of techniques such

4 HLOAR – High Level Organisational Analysis Report: A report presented to the project steering committee detailing the existing state of the organization and the possible impact and issues surrounding the system’s implementation.

132

as network analysis to identify and recruit informal opinion leaders as credible sources of information may be an effective strategy to reduce anticipated decreases in job enrichment prior to an ES implementation (Borgatti, 2003; Ellis, 1992).

6.1.2 Perceived Job Change at Time 2

The study also measured the extent to which users perceived a change in job characteristics following the introduction of ES technology. Paired sample t-tests comparing job characteristics mean scores over time found no support for the hypothesis that users would report a significant change in job characteristics over time. There are a number of possible explanations as to why the data failed to support the hypotheses 1b, 2b and 3b. Two of the most plausible are discussed below – the presence of moderating variables in the form of enrichment levels at Time 1 and usage rate, and the ability of Hackman & Oldham’s (1979) job characteristics model to capture elements of contemporary job design.

6.1.2.1 Perceived Job Change, System Usage And Job Enrichment

Analysis of results from both correlational analysis and ANOVA routines highlighted a relationship between system usage, pre-existing levels of job enrichment and job change over time. Comparing the characteristics users “increasing”, “decreasing” or reporting “no change” in job enrichment highlighted a difference between high and low users. Low users were more likely to report no change in their levels of job enrichment at Time 2. In contrast high users were more likely to experience change in their job characteristics, but not uniformly. Those categorised as high users with pre-existing high levels of enrichment at Time 1 reported a decrease in their job’s enrichment quotient at Time 2. Conversely high users with reduced job enrichment levels at Time 1 reported an increase in job enrichment at Time 2. Additional post-hoc analyses yielded greater insight into the particulars of users increasing or decreasing in MPS over time. High users experiencing a decrease in job characteristics at Time 2 were more likely to be of a lower rank, have less subordinates, and report shorter work group tenure than high users increasing in job characteristics following the ES implementation. Therefore those with less formal power in the organization tended to experience a decrease in job enrichment. Conversely those senior in the formal hierarchy experienced an increase in job enrichment. These results are summarized graphically in Figure 6.1 over.

133

Figure 6.1 The Differential Effects of an ES implementation on Job Design

No changeIncrease in

MPSDecrease in

MPS

No change Low MPS at T1

Decrease in MPS

Low MPS atT1

Decrease in MPS

System usage rate

Change in MPS at Time 2

MPS at Time 1

Rank

Low useHigh use

Note: MPS (Motivational Potential Score) is used as a measure of overall job enrichment

At this point it is unclear why hierarchically senior users would have reduced levels of MPS at Time 1. It is however possible to speculate as to why those with more formal authority in the organization increased in overall job enrichment. A review of the literature pertaining to the aims, intents and characteristics of ES technology would suggest the primary benefits of an ES are typically discussed in terms of strategic or management related outcomes (Lengnick-Hall et al., 2004). Other broader IS/IT research has also identified a relationship between increased computerization within management roles and job satisfaction (Ghani & Al-Meer, 1989; Zeffane, 1994). The key aims of the ES implementation were to standardize work possesses, to enable the organization to centrally control the flow of information and data, improve the accuracy of reporting and to drive administrative efficiencies (see Section 1.5.1). Therefore the aim of the system’s introduction reflects much of what would be considered valuable by those in senior, or management roles. Consequently the new tools and altered work context afforded to these users may have resulted in an increase their overall job enrichment.

It is also possible that the new system disenfranchised those with a degree of informal authority over the work process. The long standing nature of the existing legacy system may have allowed proficient users to increase aspects of their role such as Task Significance, Task Identity, Autonomy and Skill Variety over time. Their knowledge

134

of the system and its relationship to work processes may have allowed them to establish a degree of influence outside the formal boundaries of their role. Consequently any level of job enrichment associated with their legacy system proficiency would have evaporated with the introduction of the ES.

Finally the results are worth considering in relation to the notion of structuration and the effect of a new structural modality (ES technology) on an existing structural modality (job design). Prior to the system’s introduction users anticipated the new structure to have a negative impact on existing job design elements. However the pattern of change indicates that existing structural properties based around hierarchy determined the manner in which the structural properties of ES technology were incorporated into the structural fabric of the organization. The results highlight the limitations of adopting a deterministic viewpoint when considering the effect of technology on organizational function.

In summary, it appears ES design characteristics have a bipolar effect on those who use the system on a regular basis in terms of job enrichment. In the case reported here an inverse relationship was identified between formal authority and changes in job enrichment over time. Those high in formal authority tended to report an increase their job enrichment over time, while those low in formal authority decreased in enrichment over time. In an attempt to better understand the pattern of results reported here it is suggested that future research investigate role orientation as a potential moderator of the relationship between ES technology and job design.

6.1.2.2. The Job Characteristics Model and Contemporary Work Context

The failure of users to uniformly report any perceived change contrasts with comments received from users 3-4 months after the Time 2 data collection. Almost universally users talked about the manner in which their job had become narrow and restrictive in the duties they were required to undertake (user #2;4;7;9;11;12; 14;17;19). Senior employees referred to terms such as “segregation” and a “culture of compartmentalization” to describe the way that user’s job functions were now organized. Of the 19 users and management interviewed nine explicitly referred to this outcome of the implementation. One manager (#2) at Time 2 stated emphatically that the HAL “roles” were a mismatch with previous job descriptions and presented a list of HAL role statements that ordinarily would be grouped into larger task descriptions. They indicated that in several instances HAL roles (often with varying permission levels) cut across previously established jobs. The implication was that people (due to

135

their current permission level) were unable to do the job they used to and now relied on others to carry out certain steps. Similar examples were provided by users in both pilot sites, highlighted by this quote.

“You can’t finish a job in a complete process. The system now requires the task to be completed in stages that frequently requires another part of the system to verify each stage or to provide a required piece of information etc. This can take three or four days to complete a task rather than all in one hit like I used to be able to do” (user #9)

At least three senior users commented that task segregation appeared to be changing the career profile of ERUDIO employees, from one of a multi-skilled generalist (Administrative Officer 1, AO2 etc.) to specialist roles based around system requirements and password controls. This appeared to be reinforced at City by what one user referred to as “a mentality of you don’t get the training unless you do that job” (#4)5. Users voiced concerns regarding their ability to achieve promotion and development opportunities, as they were effectively digitally handcuffed to their role (#4;11;14;19). There is evidence to suggest that users at other sites also shared these concerns (#14). Interestingly, the narrowing and specialization of job roles was identified by the network results, confirming the utility of social network analysis in IT implementation research. The value of adopting a social network approach to the study of technology adoption will be explored further in the final chapter.

If as prevalent as the above discussion would have us believe, changes such as those described should have been reflected in terms of a perceived reduction in Skill Variety, Task Identity and Autonomy and perhaps to a lesser degree, Task Significance. The comments highlight the possibility that Hackman and Oldham’s (1976) job characteristics model is ill-equipped to measure such changes or to capture elements of contemporary job design (Parker et al., 2001). Traditional departmental and organizational boundaries continue to become increasingly blurred with increased use of project teams, strategic partnerships and networked organizations (Symon, 2000; Torraco, 2005). Consequently the restricted nature of the JCM struggles to capture the dynamics of contemporary work contexts (Parker et al., 2001; Torraco, 2005). For example, users talked about feeling a loss of control over business processes and how this was a major source of frustration (users # 9; 11). Instead, on average users perceived no significant change in the JCM’s five measures of job design. The JCM does not explicitly measure constructs such as control (Wall, Corbett, Martin, Clegg, & Jackson,

5 This however was not consistent across all locations as will be highlighted in the network component of the discussion.

136

1990) and therefore this loss of control may have not been reflected in the results. Future research would benefit from alternate, broader approaches to the measurement of contemporary work design, incorporating aspects such as user’s emotional demands (Parker et al., 2001) and a focus beyond the individual level of analysis (Torraco, 2005).

6.1.3 Summary

The initial phase of the study was concerned with whether users anticipated and perceived a change in job characteristics following the implementation of ES technology. At Time 1 prior to the implementation users reported an anticipated decrease in Skill Variety, Task Significance and Autonomy. The results indicated that users may have equated the loss of skills and knowledge relating to the existing system to decreases in these three job characteristics. Post-hoc discussions with users indicated that improved communication of clear and credible information prior to go-live may have reduced users concerns regarding decreases in job characteristics.

The study also examined the perceptions of users over time, comparing job enrichment levels pre- and post-implementation. The results identified ES technology as having a bipolar effect on users, depending on their amount of system usage. Whereas Low users tended to report no change over time, High users reported both increases and decreases in job enrichment. Further analysis identified an inverse relationship between formal authority and changes in job enrichment over time. Those high in formal authority tended to report an increase in their job enrichment over time while those low in formal authority decreased in enrichment over time. It was suggested that future research investigate role orientation as a potential moderator of the relationship between ES technology and job design. Finally, it was suggested that the job design measures may have struggled to reflect the nature of contemporary work design, particularly in light of the tight coupling of ES technology and work tasks.

6.2 Job Design Change and User Acceptance

The second research question presented in Chapter One was concerned with the extent to which changes in job design were related to user acceptance. Two hypotheses were developed to better explore answers to this question.

H4a: Users will report a positive relationship between anticipated job characteristics change and levels of user acceptance at Time 1.

137

H4b: Users will report a positive relationship between perceived job characteristics change and levels of user acceptance at Time 2.

6.2.1 Anticipated Job Change at Time 1 and User Acceptance

Hierarchical regression was used to measure the strength in the relationship between job change and user acceptance following the introduction of HAL, the new ES. Two regressions were carried out, one for each user acceptance variable (Usefulness; Ease of Use).

Both regressions were run with anticipated Meaningfulness, Autonomy and Feedback regressed onto the acceptance variables while controlling for “current” levels of Meaningfulness, Autonomy and Feedback. Both reported a significant result with weak effects (Usefulness ∆R2 = .15, p<.001; Ease of Use ∆R2 = .12, p<.001). When controlling for all other variables in the model only feedback reported a significant coefficient (Usefulness β = .46, p<.05; Ease of Use β = .49, p<.05). Therefore while Meaningfulness and Autonomy may account for some variance it appears that anticipated levels of Feedback is the primary driver of user acceptance when considering these variables.

Hypothesis 3a argued that users would anticipate an increase in feedback due to ES technology’s comprehensive tracking of data and procedure and an improved audit trail for user’s transactions, as well as improved error detection and reporting (Grant & Uruthirapathy, 2000; Sia et al., 2002). Based on the theory and findings supporting the JCM it was also argued that users would report a positive relationship in feedback levels following an ES implementation and user acceptance. While the results here support the above arguments it is unclear as to why feedback in particular reported the strongest effect. A series of post-hoc analyses were undertaken, but failed to identify any plausible explanations for the nature of the relationship between anticipated change in feedback and user acceptance. Discussions with users also failed to uncover any possible reasons for the result. Therefore at this point it is unclear as to why feedback reported the strongest effect when controlling for all other variables. A greater amount of work is required to identify whether ES technology effects different categories of users, and if so to identify the particulars of those user groups.

138

6.2.2 Perceived Job Change at Time 2 and User Acceptance

A series of hierarchical regressions failed to identify the presence of a direct relationship between job design levels at Time 2 and user acceptance at Time 2. Drawing on the job change results reported earlier it would appear that user’s experiences of job change following an ES implementation are moderated by factors such as usage, hierarchy and pre-existing levels of job enrichment. As a result the failure to report a direct relationship between job change and user acceptance is not surprising. As will be discussed in the next and final section of this chapter usage did report significant effects when tested as a moderator between job design change and user acceptance.

6.3 Usage As A Moderator Of The Relationship Between Job Change And User Acceptance

A series of hierarchical regressions were carried out to determine whether usage would moderate the relationship between perceived job change and user acceptance. Users were categorized as either high (>3hrs a day) or low users (<3hrs a day).

Consistent with the analysis throughout the study, usage as a moderator between anticipated job characteristics and acceptance, while controlling for current job characteristics at time 1 were first examined. A second set of regressions were then run to examine usage as a moderator of the relationship between job characteristics at Time 2 while controlling for the effects of both acceptance levels and job characteristics levels at Time 2. This approach allows a higher degree of confidence in examining the contribution of change in job characteristics to levels of user acceptance at Time 2 (Axtell & Parker, 2003). Interaction effects were identified at both Time 1 and 2 for both acceptance variables. At Time 1 interaction effects were reported for both usage categories in relation to Meaningfulness and Skill Variety. At Time 2 only one interaction was identified between Usage and Task Identity.

6.3.1 User Profile

A post-hoc user profile was developed better understand the results of the moderated regressions. An initial profile was formed using demographic data collected for the study. This was then refined based on discussions with users, observations and official documentation detailing the various role requirements of users.

139

An initial test of the demographic variables indicated that low and high users were significantly different to each other in relation to job band (F(4,109) = 7.57, p<.000). Job bands AO46 (MUsage = 2.357), AO5 (MUsage = 1.88) and AO6+ (MUsage = 1.31) were more likely to be low users. As might be expected AO2’s (MUsage = 2.98) and AO3’s (MUsage = 3.12) were more likely to be high users. It should be noted that the job band of AO4 represents a mid point in the usage continuum and was not significantly different to the AO5 or AO2 job band. Interestingly AO3’s tended to report the highest levels of usage. Discussions with users and Erudio management provided additional information to build a detailed profile of each usage category. These profiles are presented below.

Low users: Low users were identified as those that had limited direct exposure to the previous legacy system or the new ES. It is also likely they were senior within the organization, holding a management or teaching role rather than a technical or purely administrative position. Consequently Low users may have experienced very little disruption in regard to their core job functions as a result of the ES implementation. While this group may have had little direct involvement with the system, these users were considered to be the primary benefactors of the benefits afforded to the organization by HAL. As reviewed in Chapter One, a primary aim of the system was to ensure data integrity, process standardization and greater reporting flexibility. These attributes are far more likely to be valued by management rather than the rank and file employee (Lengnick-Hall et al., 2004).

High users: High users were largely comprised of users within an AO2 or an AO3 job band. AO2’s were considered to be those directly responsible for engaging with the system on a regular basis. It is assumed that system use forms a large component of their job role pre- and post-implementation. It is highly possible that these users have a high degree of skills based around the replaced legacy system, but may have also recognized the need for a new system based on the old system’s dated nature. AO3’s were identified as those primarily holding administrative or supervisory roles within the organization. However, it is thought that the primary reason for system involvement was either report generation

6 A continuum of administrative hierarchy: A02 – a low level position, concerned with data entry and processing; A06 – a senior management position in this case responsible or the running of Faculties and other high level duties7 A self report measure captured as an ordinal variable on a four point scale

140

or management of those responsible for data entry and maintenance. It is possible that access to various parts of the system previously enjoyed by these users to examine data and reports were no longer available under HAL. Consequently these users now require assistance to access information or lack the frequency of system interaction to learn and/or retain the new system knowledge.

6.3.2 Usage As A Moderator Of Anticipated Job Characteristics and User Acceptance

Table 6.1 presents a summary of the moderated regressions testing the effects of usage as a moderator between anticipated job characteristics levels and acceptance at Time 2, while controlling for “current” job characteristics at Time 1. In simple terms the results indicated that low users are more likely to be effected by anticipated changes in Meaningfulness than high users. Low users are more likely to report an increase in Perceived Ease of Use the higher the anticipated job meaningfulness following the system’s introduction. High users reported a negative relationship between anticipated levels of skill variety at Time 2 and Perceived Ease of Use. These results are discussed briefly below.

Table 6.1 Summary Table of Interaction Effects for Anticipated Job Design at Time 1

Usage Category Perceived Usefulness Perceived Ease of Use

Anticipated levels of Job Enrichment

Low --- Meaningfulness (+ve)High --- SV (-ve)

Meaningfulness

Low users reported a significant positive relationship between anticipated job meaningfulness and Perceived Ease of Use following the system’s introduction at Time 1. In review, meaningfulness is a composite variable reflective of skill variety, task significance and task identity. Users with jobs high in meaningfulness are more likely to have a higher degree of skill variety, are likely to perceive their job as being important to the organization and be more able to complete a job from start to finish. In simple terms the more job meaningfulness low users anticipate after the system

141

goes “live”, the more likely they are to report higher acceptance levels. Low users appear to make a closer association between job meaningfulness and the system being introduced than high users.

A review of the user profile indicates that low users are more likely to be senior members of the organization. Given what has been discussed throughout the chapter (e.g. see section 6.1.2.1) it is not surprising that low users report a stronger association between anticipated job enrichment levels after the system’s introduction and levels of user acceptance. As discussed, there is evidence that management were the primary benefactors of the system’s introduction. It is also possible that given their position they were privy to additional information and briefings highlighting the positive nature of the system from an organizational perspective. Therefore low users, as senior members of the organization were both better informed and “sold” on a system that was being introduced primarily to facilitate the aims of their job roles and objectives. Consequently access to greater amounts of information allowed a better informed decision as to the potential outcomes of the implementation.

In contrast high users, being hierarchically lower may have been exposed to much less information about the aims, benefits and particulars of the system. A common emerging theme when discussing the implementation with users was the lack of training received. Both management (#1, 6) and front-line users (# 4, 5, 9, 14) commented that the majority of users were exposed only to the system’s most basic functions. One user alleged that only “about an hour” of training was received (user #9). It is plausible that if users received minimal training then knowledge of the system’s impact on their work context would be negligible. As such high users would be unlikely to report a relationship between anticipated levels of job meaningfulness and acceptance, as was the case here.

Skill Variety

High users did report a significant negative relationship between anticipated post-implementation skill variety and Perceived Ease of Use. In short High users anticipating an increase in skill variety are more likely they are to report a reduction in their perceptions of how easy the system will be to use. The results also conflict with the basic job enrichment argument that as Skill Variety increases so do attitudinal measures such as acceptance (Boonzaier et al., 2001). Instead, high users reported reduced levels of acceptance following an anticipated increase in Skill Variety.

142

One could speculate at this point that High users would have a greater amount of skill and proficiency invested in the current system and may resent the loss of those skills. Anecdotal evidence suggests that many users were unfamiliar with the Graphic User Interface (GUI) based software and more familiar with DOS / menu based operating environments. The researcher was told of several instances where trainers had to instruct users in basic navigation techniques before moving onto system characteristics training. Borrowing from the resource dependence argument outlined in Chapter Two, it is possible that a high degree of power and influence was afforded to these users due to their system proficiency. Loss of old skills may accompany an anticipated loss of power and hence a reduction in user acceptance levels. Post hoc discussions with users indicated support for this assertion. Several lower ranking (but higher usage) users observed that “everyone is on the same playing field” (user #11)and that with [the previous system] “I knew where to go, I knew people to get help and if you scratch my back I’ll scratch yours”, all my contacts are gone ‘cause its all different” (User #7).

On the other hand, low users failed to report a relationship between skill variety and either dependant variable. It is suggested that low users have a much lower proportion of their skill variety invested in the organization’s information technology. As a result they are less likely to associate job characteristics such as skill variety with the introduction of a new ES.

6.3.3 Usage As A Moderator Of Perceived Job Characteristics and User Acceptance at Time 2

A similar process was carried out to test for the presence of Usage as a moderator between perceived job change and user acceptance at Time 2. As well as controlling for job characteristics at Time 1, levels of user acceptance at Time 1 were also controlled for in order to better capture change in the DV over time (Parker et al. 2002). As discussed earlier, analyses investigating job change over time indicated a possible relationship between Usage and hierarchy. Consequently hierarchy as a possible confounding variable was also controlled for in the equations.

Task Identity was the only characteristic to report a significant interaction at Time 2. A significant positive effect was found for Low users (<3hrs) when considering levels of Task Identity post-implementation for both acceptance variables at Time 2. Post-hoc interviews provided anecdotal evidence to support the Task Identity results at Time 2, but failed to provide any additional information regarding the absence of additional job characteristic interactions. Therefore at this point it is unclear why Task Identity was the only characteristic to report a significant interaction.

143

Table 6.2 Summary Table of Interaction Effects for Perceived Change in Job Design and Acceptance Variables (PU and PEU)

Usage Category Perceived Usefulness Perceived Ease of

Use

Time 2Low TI (+ve) TI (+ve)High --- ---

Task Identity is the degree to which a job requires completion of a whole, identifiable piece of work – that is, doing a job from beginning to end, with visible results attributable to the job holder’s performance (Hackman & Oldham, 1980). Task Identity is a fundamental aspect to consider when introducing a system designed around “process flows” and transaction automation (See Chapter One, section 1.5.1 for a review). As discussed in Chapter Two, the importance placed upon users completing a piece of work from beginning to end is subsumed by a greater focus on the user’s contribution to the data flow throughout the organization. As such the introduction of an ES can have significant implications for user’s Task Identity. As discussed, interviews with users indicated a preoccupation with Task Identity in regard to the career, training and productivity implications flowing from ES driven task and role segregation. Moreover, users and management alike referred to the lack of understanding in relation to process flows and the sequential processing nature of the system (Users # 2, 4, 9, 14, 16, 18). Given the tight coupling of ES technology and business processes it would appear reasonable that low users may not fully understand the changes being introduced. Not being exposed to the new processes on a frequent basis may make it difficult for low users to develop an awareness of the now workflows and processes. For example, one system trainer felt that the lack of supervisor participation in the initial stages of the project led to a “reduced appreciation and understanding about what was going on” (user #18). Therefore it is likely that low users not familiar with the new processes would report a reduction in Task Identity and consequently a lower level of acceptance.

Conversely, those users able to use new tools and functionality may experience an increase in Task Identity. Aspects such as report generation, data analysis and control over work processes may increase the Task Identity of some roles. Andreau & Boone (2002) concluded that the closer the match between the system and the needs of the user, the higher the potential for increased Task Identity. Consequently, ES technology in its capacity to collate, analyse and provide a high degree of reporting flexibility would allow some low users a greater degree of control over their work environment. In addition, the aim of the ES in this case was predominantly based around process

144

standardisation and the centralisation of control of the organisations’ IT resources. This capacity may have allowed Low users responsible for these areas greater levels of Task Identity in thier role. Therefore the system and its outputs may be of vital importance to some users, regardless of how little they use it.

An aspect more difficult to explain however is the failure of High users to report a relationship between changes in Task Identity and user acceptance. Worth noting perhaps is the increase in the mean scores of Task Identity for High users (T1 4.5; T2 4.94, p<.05). It is suggested perhaps that High users are more acutely aware of the sequential processing nature of the system and as a consequence are more aware of business processes and their essential part in maintaining the data flow. The process and automatic transaction oriented design of ES technology is similar to that of a production line. When an error is made at any one point the process is disrupted either by error or by being suspended. ES technology also allows management to identify the source of the problem due to security protocols and improved monitoring capability (Sia et al., 2001). As a result High users are likely to identify quickly process flows to and from their job role. Given this, one would expect High users to report a positive relationship between changes in Task Identity and user acceptance. At this point however it is unclear as to why High users failed to report a positive relationship between a change in any of the five job characteristics and user acceptance.

6.4 Summary

This chapter provided a discussion and a post-hoc review of the findings reported in Chapter Five. Similar to the results chapter the discussion was divided into three sections dealing with job design change; job change and acceptance; and the effect of Usage as a moderator of the relationship between post-implementation job characteristics and user acceptance.

Prior to the introduction of the system, the results indicated that users overall anticipated a decrease in job characteristics. Discussion centered around the possibility that reduced levels of information about the system may have resulted in a general pessimism regarding the system’s impact on job design. At Time 2 the results indicated that job change was dependant on aspects such as formal hierarchy and system usage. High users with increased levels of formal authority appeared to increase in overall job enrichment. Lower ranking High users appeared to decrease in overall job enrichment following the ES implementation. It was suggested that future research investigate the role orientation as a moderator of the relationship between job change and ES

145

technology. Questions concerning the ability of Hackman and Oldham’s (1979) Job Characteristics Model to adequately capture elements of contemporary work design were also raised.

The findings of the moderated regressions were also discussed. It was suggested that the positive relationship reported by low users in relation to anticipated Meaningfulness and user acceptance was in part due to their likelihood of holding senior positions. Low users higher in the organizational hierarchy were thought to have received more information about the relative benefits of the system. As a result, they were able to make a more considered judgement regarding the possible positive or negative outcomes of the system. The negative relationship between anticipated Skill Variety and acceptance for high users was attributed to the anticipated loss of one skill set, while having to rapidly develop a new skill set during a time of high stress. It was also suggested that the negative relationship was perhaps reflective of an anticipated loss of proficiency and influence due to the process of skills transfer. At Time 2 the positive relationship between Task Identity and Perceived Ease of Use was explained by ES technology’s tendency to place a reduced emphasis on the task identity of individual roles. Low users able to utilize the tools offered by ES technology may experience an increase in task identity. However those that are still coming to terms with the system’s functions may experience a decrease due to the change in process flows and changes to aspects of their job that are system related.

The next chapter, Chapter Seven will report the results of the network component of the study. Chapter Eight will again provide a discussion and post-hoc analyses of the network results. Chapter Nine will conclude the thesis.

146

147

Chapter Seven — Social Network Change and User

Acceptance

7.0 Introduction

The thesis considers three representations of structure within an organization, ES technology, job design and social networks. This second results chapter reports the results of the analysis testing the extent to which social networks change following the introduction of ES technology and if so, whether this determines user acceptance levels. User acceptance was again measured using Davis’s (1989) variables of Perceived Usefulness and Perceived Ease of Use. As was covered in Chapter Four, the method used to collect social network data required an alternate approach to the one taken to collect the job design data. Data were collected from the case organization Erudio, but from two specific sites within the organization. Network boundaries were drawn around two faculties from within each of the two pilot sites, City and Country.

The chapter is divided into three sections. The first outlines the context of each site, detailing the characteristics and the demographic particulars of each network. The second details the analyses relating to the measurement of network change. Finally the chapter reports on a series of hierarchical regressions aimed at determining the extent of the relationship between network change and user acceptance. Chapter Eight will discuss the results reported here.

7.1 Research Context

As briefly outlined in Chapter Three (section 3.2) the data used in this study were collected from an educational organization implementing an ES into two separate faculties, City and Country. The two networks examined below represent the respective “Business School” within each Faculty. The unique characteristics of each are dealt with over.

148

7.1.1 City Faculty

The City faculty was responsible for providing Business Administration and Technology courses to students and operated as an external training provider to industry. The allocation of administration services and support to teaching personnel was reflective of a de-centralized model of operation. Approximately 25 administrative staff were assigned to the faculty to administer student enrolments, coordination of classes and the issuing of results. These staff reported to a Faculty head. The administrators and management staff of this school represent the participants of the City site. The administrative staff were essentially divided into four teams representing separate aspects of the faculty’s operational requirements. Three of the teams were housed in one building on different floors, with the other located in a separate office location.

The dyadic nature of network analysis requires researchers to achieve as close to a 100% response rate as possible. While not ideal, response rates of 80% are commonly regarded as representing a network to some degree of accuracy (Sparrowe, Liden, Wayne & Kraimer, 2001). Response rates for the City network were strong at both Time 1 and Time 2. Of 26 possible participants within the bounded network, 24 participated at Time 1. 24 participants also responded at Time 2. 21 participants reported at both data collection points giving an overall response rate of 81%. The majority of City participants were Female (81%), classed as AO3 or lower (70%) and had been in the organization longer than 5 years (66%). 56.8% of participants were over the age of 40, with 38% between 40-49, with 38% younger than 39 years old. In comparison to Erudio overall, participants were over represented in terms of gender and those holding administrative positions, but were representative in terms of age.

Graphical representations of the City network structures are presented below. The diagrams indicate ties between network actors in relation to the flow of Advice and Resources. In order to improve interpretability, the graphs represent directional data dichotomized at >=1. In doing so the graphs only show ties representing low, moderate or high levels of dependence. Ties valued less than 1 are considered too weak to be of interest when visually representing the networks. The nodes indicate participants within the network, colored to indicate work-group.

It can be seen that despite the physical separation of the groups, interaction is not solely determined by proximity. Even though three of the four groups were housed in the same building, they were all located on separate floors. Despite the spacial dispersion of the groups interaction between different group members remains high. Note also that despite the relative density of the Advice and Resource networks at

149

Time 1, the amount of ties in both network types visibly increase at Time 2. Also notable is the increased density of the Advice network in comparison to the Resource exchange network. Appendix A.06 contains network graphs for each group present within the network. A range of summary statistics are also provided such as density, degree of centralization, degree of reciprocity, E-I index and the group mean score of the dependant variables of Usefulness and Ease of Use.

150

Figure 7.01 — City Advice Network at Time 1

Figure 7.02 — City Advice Network at Time 2

151

Figure 7.03 — City Resources Network at Time 1

Figure 7.04 — City Resources Network at Time 2

152

7.1.2 Country Faculty

A similar group of participants were chosen from the second pilot site (Country). The networks examined were that of an administrative group responsible for the provision of services to the “Business and Administration” school within the Country faculty. However in contrast to the City faculty the administrative staff at Country operated under a centralized mode of operation. Administrative staff reported to an administrative manager (rather than School or Faculty management) and were “shared” among a limited number of schools. In addition to this, the geographically dispersed nature of the Country faculty required administrators to be located at each major teaching facility. This resulted in four teams located approximately 4-6 hours drive from each other. While some management members did travel between campuses the majority of communication between administrative staff occurred via telephone or e-mail.

Response rates at the Country site were lower than that of the City site. The decentralized structure of the Country site, along with the geographical dispersion of the participants was a contributory factor in the failure to achieve higher response rates. 31 out of a nominated 39 participants agreed to participate at Time 1 resulting in a participation rate of 79.5%. At Time 2 only 36 of the original 39 network members were available for participation. One participant had left the organization and another two were on recreational leave. A total of 31 participants completed both network questionnaires, effectively resulting in a participation rate of 79.5%. Four network members declined to participate over Time 1 and Time 2. All four non-respondents held management roles within the organization and declined due to time pressures, lack of involvement in the system or for unspecified reasons. Country participants were similar to City in terms of Age (46% over the age of 40), Gender (82% Female) and Job Grade (70% AO3 or lower) but over represented in terms of AO3’s (a senior administrative position; 28%).

Network graphs are again provided to help conceptualize the structure of the Country network structures. The nodes are again colored, but in this case indicate different locations, rather than specific work groups. Note that in contrast to the City networks, Country network interactions are predominantly based around locale. As with the City networks, Country’s Resource exchange network is sparser than the Advice network. The network graphs also clearly show a number of Country actors establishing ties outside their immediate geographical location at Time 2, primarily in relation to Advice and to a lesser extent, Resources. It is also possible to note an increase in density within a number of the Country locations, particularly RED group (advice, resources) and BURGUNDY (advice) at Time 2. Appendix A.06 also contains network graphs

153

for each group present within the network. A range of summary statistics are also provided such as density, degree of centralization, degree of reciprocity, E-I index and the group mean score of the dependant variables of Usefulness and Ease of Use.

Figure 7.05 — Country Advice Network at Time 1

Figure 7.06 — Country Advice Network at Time 2

154

Figure 7.07 — Country Resource Network at Time 1

Figure 7.08 — Country Resource Network at Time 2

155

7.2 Network Change

In Chapter Two a number of hypotheses were put forward concerning the network implications of introducing ES technology into an organization. The first concerned the potential for ES technology to act as a catalyst for network shock. It was argued that the peculiar characteristics of ES technology had the capacity to fundamentally change the manner in which users interacted within an organization. While one might expect an increase in advice following an IS implementation (Burkhardt & Brass, 1990) it was also suggested that ES technology may precipitate a far more permanent change in the way which users exchanged resources. In particular it was suggested that ES technology was more likely to reduce the centralization of resources due in part by their process driven design elements. As such the following hypotheses were put forward.

H6a: Social networks based around resource exchange will increase in density following the introduction of ES technology into an organization.

H6b: Social networks based around advice sharing will increase in density following the introduction of an ES technology into an organization.

Procedures used to examine whether structural change had occurred were adopted from Burkhardt & Brass (1990), Shah (2000), Stevenson, Bartunek and Borgatti (2003) and Totterdell, Wall, Holman, Diamond and Epitropaki (2004). Each analysis type was adopted based on their ability to present an additional or alternative perspective of the network change being examined. First, a series of quadratic assignment procedures (QAP) were run to examine the degree of correlation within and between network structures at Time 1 and Time 2. Second, the overall density of both the Advice and Resource dependence networks were compared within and between data collection points. Third, a series of t-tests were carried out to identify whether there were any significant differences in actor degree centrality scores. Finally, Krackhardt and Stern’s (1988) E-I ratio was used to examine the extent to which a-priori selected groups within each network altered their within-group and between-group ties. The details are provided below.

156

7.2.1 Network Correlation (QAP)

The null of hypotheses 6a and 6b is that no change in network structure would be observed over time. In this instance both network structures would present as being similar, not different to each other following the introduction of ES technology. One method of comparing the extent to which network configurations are similar is to use Hubert’s (1985) Quadratic Assignment Procedure (QAP). QAP procedures allow researchers to measure the degree to which two NxN matrices are similar beyond what might be expected by chance (Krackhardt, 1988). The QAP test is carried out in two stages. First it examines the correlation between each corresponding cell of the two matrices under examination. Due to the nature of network data and the potential for high degrees of autocorrelation between variables, it performs a random permutation test to establish the standard error for the correlation coefficient (Borgatti et al., 2002). The permutation test randomly permutes the rows and columns of one matrix onto the other a number of times. After each permutation the correlation coefficients are re-calculated and compared against the original result. A pseudo-probability measure results whereby the number of times a random permutation is larger or equal to the original result is reported. The number of permutations carried out improves the accuracy of the standard error calculations and confidence in the reported p values (Borgatti et al., 2002).

The QAP routine provided by UCINET6 provides a range of summary statistics to describe the relationship between the two matrices. In this instance Pearson’s r is reported. Both network structures (Advice, Resources) were compared within each time period and across each time period. Both City and Country sites demonstrated similar patterns of correlation within and between data collection points (See Tables 7.01 and 7.02 respectively). City reported higher correlation scores (.742-.834) than did Country (.662-.778). In contrast, both sites recorded lower correlations across Time 1 and Time 2. City (.287 - .437) recorded lower scores than Country (.398-.515).

Table 7.01 QAP Correlation Scores (City Time 1 & Time 2)

1 2 3 4

Advice T1

Resources .742**

Advice T2 .394** .369**

Resources .366** .354** .759**

Note: **p<.01; *p<.05

157

Table 7.02 QAP Correlation Scores (Country T1 & T2)

1 2 3 4

Advice T1

Resources .663**

Advice T2 .515** .398**

Resources .423** .462** .715**

Note: **p<.01; *p<.05

In summary, Advice and Resource networks at both sites correlated highly with each other at each data collection point but reported a far weaker association when compared across time. If following the introduction of the system either network structure had remained stable, one could expect high degrees of correlation between Time 1 and Time 2. The QAP correlation scores indicated that this was not the case. The nature of the analysis does not allow one to report either Advice or Resource networks as being significantly different across time. The QAP test does however provide support for the idea that both the Advice and Resource network structures have significantly reduced in the extent to which they are similar.

While the QAP routine is able to statistically determine the extent to which interaction patterns are similar, it is unable to provide data on the nature of any change present. For example, the analysis is unable to determine whether network ties increased or decreased over time. Reported below are a number of tests better able to document the nature of the change reported by the network data.

7.2.2 Network Density

A method better able to indicate the nature of the change is to compare network density over time (Burkhardt & Brass, 1990). Density is one of the more straightforward network measures able to give an indication of the overall distribution of ties within a network (Scott, 2000). In this case the data were both valued and directional and therefore density was calculated as the total of all values divided by the number of possible ties, giving the average value of those ties (Borgatti, Everett, & Freeman, 2002). The figures reported in Table 7.03 and 7.04 are the density results for both network types in City and Country sites respectively.

158

Overall the density of both site’s networks exhibited a similar pattern of change in density. Both reported an increase in Advice and Resource density at Time 2. The City network reported a significant change in Resource density (24.8% increase in average tie value) and Advice density (19.5% increase in average tie value) over time. Overall the Country networks reported a similar trend but were not significant at a p<.05 level. The decentralised, sparser nature of the Country network was thought to have contributed to this result (Rice & Richards, 1985). An examination of work group density within the Country network indicated that two out of the four work groups had significantly increased in density, while the other two remained stable (See Appendix A.06 for network graphs of each site’s work group and summary statistics).

A review of network graphs presented above by NETDRAW for both City and Country networks provide a visual confirmation of the increase in density, particularly for the City network (See Figures 7.01 - 7.04 above). As predicted actors appear to establish new sources of advice following the introduction of the new system. Resource exchange networks also appear to have changed, increasing in density over time. An argument was made in Chapter Two regarding the propensity for ES technology to reduce the concentration of resource allocation and dissemination. This result would appear to provide support for the idea that ES systems with their focus on process flow are more likely to force a more equitable distribution of resources throughout an organization rather than encourage the centralisation of resources at any one point or individual.

Table 7.03 Bootstrap Paired Sample t-test Of Network Density At Time 1 And Time 2 (City)

M SD t Sig.Advice network density at T1 .41 .74 --- ---Advice network density at T2 .61* .64 2.24 .02

Resources network density at T1 .29 .82 --- ---Resources network density at T2 .54** .78 2.73 .01

Note: Density significantly different at Time 2 **p<.01; *p<.05

159

Table 7.04 Bootstrap Paired Sample t-test Of Network Density At Time 1 And Time 2 (Country)

M SD t Sig.Advice network density at T1 .19 .57 --- ---Advice network density at T2 .22 .50 1.06 .14

Resources network density at T1 .14 .61 --- ---Resources network density at T2 .17 .54 1.01 .15

Note: Density significantly different at Time 2 **p<.01; *p<.05

7.2.3 Direct Ties

While density provides an overall summary of tie distribution, examining the change in individual ties can give a clearer picture of the structural change taking place (Shah, 2000; Totterdell et al., 2004). A series of paired t-tests were conducted to assess whether participant’s network ties had changed over time. UCINET6 was used to calculate each participant’s IN-degree centrality score. IN-degree centrality measures the number of direct ties to an actor from other actors in the network (Freeman, 1979). In this case high degrees of “IN-degree” ties would indicate high levels of access to advice and reduced dependence on a small number of resource providers. An increase in IN-degree ties at Time 2 would suggest a reduced dependence on an existing number of providers at Time 1. IN-degree, rather than OUT-degree measures were chosen due to their reduced exposure to self report bias. Table 7.05 shows that overall employees increased their ties at Time 2.

Table 7.05 Paired Sample t-test Comparing IN-Degree Centrality Scores at Time 1 and Time 2

Time 1 Time 2M SD M SD df t Cohen’s d

City Advice 10.34 5.64 15.23 6.82 25 3.35* .80

Resources 7.29 4.58 13.5 8.50 25 3.27* .93Country

Advice 7.23 3.88 8.47 4.94 38 1.42 .28Resources 5.42 3.89 6.55 4.28 38 1.31 .28

Note: * p<.05 indicates a significant difference in ties at Time 2

160

City participants recorded statistically significant differences, across time in Advice (t(25) = -3.35, p< .05) and Resources (t(25) = -3.27, p<.05). On average City users increased from 10 to 15 advice ties after the implementation and doubled their resource ties from 7 to 14. A closer examination of the individuals decreasing in centrality identified one as the Faculty head, one had been transferred to a teaching position and the other two were from one group within City (coded as “Group 4” see Appendix A.06 for a network profile). Overall there was no discernible pattern as to why these particular City respondents may have decreased IN-degree ties.

While Country reported a similar trend to City, the difference was not significant for either Country network. On average Country users increased their direct advice and resource connections by only one tie. An examination of in-degree tie change scores further indicates a different pattern of change for each site. Whereas a sizable number of City participants (73-80%) increased their centrality scores (See Table 7.06 below) only 51-64% of Country participants did the same. In addition, 33-44% of Country participants actually decreased in centrality, with the rest remaining stable.

Table 7.06 Change in Participant In-degree ties for City and Country at Time 2

Increase Decrease No ChangeCity

Advice 22 4 0Resources 23 3 0

Country Advice 26 9 1

Resources 23 11 2

A review of the Country participants who decreased in centrality was also carried out. Of the 11 decreasing in direct ties one was a system trainer that returned to their original role, three left the organization, four belonged to one group and the other three to separate groups. The group with the largest proportion of users decreasing ties (hereafter referred to as Group A) prompted further investigation. Further analysis revealed that Group A and another (B) had different network change profiles to the remaining two Country groups (C and D). Groups C and D demonstrated similar change patterns to their City counterparts (e.g. sizable increases density and direct ties for the majority of users). In contrast, groups A and B reported no significant change in density and rather than an increase in ties, the data indicated a shift in the pattern of existing ties. This is explored further below.

161

A modified graph of each network illustrates the changes over time for groups A and B (See Figures 7.09 - 7.12 below). The modifications were performed to improve interpretation and to emphasize the nature of the change. Advice and Resource network ties were merged based on the high correlation of the networks within each time period (Ibarra, 1993). No external ties to others within the network are shown. It should be noted that response rates for Group B were poor (38%) at Time 1 and (63%) at Time 2 and may account for the pattern of results. Network graphs for all City and Country graphs accompanied by a range of summary statistics can be located in Appendix A.06.

Figure 7.09 Country Group A networks (advice and resources merged) Time 1 and Time 2

Note: Red ties indicate reciprocal ties; Blue ties indicate non-reciprocal ties; Arrow indicates direction; ties dichotomized at >=1 (1=weak tie; 2=moderate tie; 3=strong tie)

Group A reported the highest amount of actors (4) as decreasing in ties in Country. Overall the mean density of tie value reported no change over time. The graph above shows that while some non-reciprocal ties have decreased other reciprocal ties between actors have increased. Actors such as #4, #5 and #9 have either maintained or increased their centrality levels while others have decreased. All three of these actors were line supervisors.

Time 1 Time 2

162

Figure 7.10 Country Group B networks (advice and resources merged) at Time 1 and Time 2

Note: Red ties indicate reciprocal ties; Blue ties indicate non-reciprocal ties; Arrow indicates direction; ties dichotomized at >=1 (1=weak tie; 2=moderate tie; 3=strong tie): User #6 left Erudio prior to Time 2.

Group B was the most geographically isolated group and was responsible for a small rural campus. As with group A, this group saw little change in the mean value of tie density over time (T1 1.08; T2 1.07). However, highly central actor #6 did leave prior to Time 2. The stable density score indicates that interactions intensity between the remaining members increased, compensating for #6’s departure or in response to the system’s introduction.

Figure 7.11 Country Group C networks (advice and resources merged) at Time 1 and Time 2

Time 1 Time 2

Note: Red ties indicate reciprocal ties; Blue ties indicate non-reciprocal ties; Arrow indicates direction; ties dichotomized at >=1 (1=weak tie; 2=moderate tie; 3=strong tie)

Time 1 Time 2

163

Group C was another small campus but close in relative terms to group A (2 hours drive) and interaction between the two groups was common. Interaction between actors clearly increases over time (Density T1 .85 : T2 1.04, p<.05). This is despite actors #4 and#5 becoming increasingly isolated. Interactions patterns have shifted noticeably as well with users #1, #2, #3 and #6 all developing non-reciprocal ties. In this instance it appears that the flow of resources and advice throughout the group has been altered, and in most instances ties between actors have increased.

Figure 7.12 Country Group D networks (advice and resources merged) at Time 1 and Time 2

Time 1 Time 2

Note: Red ties indicate reciprocal ties; Blue ties indicate non-reciprocal ties; Arrow indicates direction; ties dichotomized at >=1 (1=weak tie; 2=moderate tie; 3=strong tie)

Group D was located within the primary administration campus and was the location of Country’s senior management team. Group D’s density can be seen to increase over time (T1 .67: T2 .78, p<.10). The increase in ties appears to consist of reciprocal ties with few actors maintaining exclusive control over the distribution of advice or resources.

In summary, three observations can be made in comparing the groups A and B to groups C and D. One, groups A and B overall do not experience the increase in ties as compared to C and D (or the City groups). Chapter Eight will explore possible explanations for this difference in adoption patterns between both City and Country sites, as well as the difference between the Country groups. Two, all groups appear to experience an increase in reciprocity in addition to a reconfiguration in non-reciprocal ties. As discussed in Chapter Two, the introduction of ES technology has appeared to result in a reduction in the centralisation of resources and a more even distribution

164

of resources throughout the network. Three, the mean scores for acceptance indicate a difference in each group’s acceptance levels. Groups A and B, those exhibiting an alternate adoption pattern reported lower Perceived Usefulness and Perceived Ease of Use scores see Table 7.07 below. These points will be discussed further in Chapter Eight.

Table 7.07 User Acceptance scores of Country Groups

Country Perceived Usefulness

Perceived Ease of Use

M(SD) M(SD)Group A 1.90(.97) 2.19(1.30)

Group B 2.58(1.58) 2.81(1.29)

Group C 2.79(1.21) 3.50(1.39)Group D 2.67(1.29) 3.08(1.30)

7.2.4 E-I Index

An examination of Krackhardt’s (1988) E-I index as it relates to Resource and Advice structures provides an alternative approach to examining the change in tie distribution. Rather than examining the network for its overall patterns (i.e. density) or dyads (i.e. in-degree ties) the E-I index allows the researcher to examine the degree of interaction between and within groups in the network. In simple terms Krackhardt’s (1988) E-I index provides a ratio of ties reported outside the immediate workgroup to all those observed. Membership of groups is identified a-priori by the researcher. In this case group membership was based on formal workgroup arrangements.

The E-I index routine offered by UCINET6 provides a raw E-I score (External – Internal / External + Internal ties), expected score (ratio of maximum and minimum scores given the size of sub-units and density of the network) and a re-scaled index (raw score rescaled to fit within the expected score). E-I scores are provided for the overall network, each nominated group and for each actor. A random permutation significance test of the overall E-I score is also provided. E-I index scores range from -1 to 1 with a positive ratio indicating a tendency to interact with others outside of their nominated groups. A negative score indicates a tendency for internally focused actor interaction and group closure. In short, participants in groups with negative scores tend to prefer to interact with their own group members, rather than with others outside the group. E-I ratios were calculated for both networks at Time 1 and Time 2, with the results presented below in Table 7.08 and 7.09.

165

Table 7.08. E-I index of Country and City Advice networks at Time 1 and Time 2

City CountryT1 T2 T1 T2

E-I index .188 .257 -.659 -.638Expected .483 .483 .509 .509

Rescaled E-I -.286* -.619* -.659* -.638*

Note: * Observed E-I index is different to what might be expected by chance p<.05

Table 7.09. E-I index of Country and City Resource Exchange Networks at Time 1 and Time 2

City CountryT1 T2 T1 T2

E-I index -.061 .280 -.818 -.767Expected .483 .483 .509 .509

Rescaled E-I -.238* -.500* -.818* -.767*

Note: * Observed E-I index is different to what might be expected by chance p<.05

Both networks reflect work contexts performing similar functions within the same organization. However their scaled E-I indices highlight subtle differences in post-implementation network change. As is typical (McGrath & Krackhardt, 2004) both groups report negative Advice and Resource exchange ratio’s at Time 1 and Time 2. However, overall the each reported different trends in terms of whether groups increased or decreased their inward focus.

City: At Time 1 both City’s networks report a low level of within group dependence (Advice = -.286; Resources = -.238). At Time 2 however both networks report a moderate to high “within group” dependence (Advice = -.619; Resources = -.638). The moderate to strong negative orientation of the E-I index for the City networks shows a higher proportion of strong ties within groups than across groups at Time 2. Following the ES implementation City users appear to have increased their reliance on group members for advice, rather than seek sources of advice outside their work groups. A similar pattern is observed for Resource exchange between members.

Country: At Time 1 both Country’s networks report a strong degree of “within group” dependence (Advice = -.659; Resources = -.818). The indices for Resource exchange is demonstrably higher than that of Advice, indicating that users predominantly rely

166

heavily on team members for resources (and to a lesser extent advice) rather than others in the wider network. At Time 2 Country’s groups reported a modest change in E-I scores of a similar magnitude for both network types. While still indicating a strong internal focus groups reported a slight increase in their proportion of ties outside the group for both Advice (-.638) and Resources (-.767). It should be noted that while Country did increase between-group communication, in relative terms Country groups remained far more insular than their City counterparts. It is possible that their geographical separation may have something to do with this finding .

Comparison: A number of observations can be made when comparing the two sites at each data collection point, as well as across time. The E-I indices of Country’s networks differ in at least three aspects to their City counterparts. One, Country’s groups are far more insular than City, demonstrating a strong negative score at Time 1 for both Advice (-.659) and Resource exchange (-.818). Two, whereas City’s networks at Time 1 were comparatively similar, Country’s Resource exchange network index is demonstrably lower than its advice network. Three, Country reported a different pattern of change to City. For example, resource flows altered at each site but not in the same manner. A useful comparison is to examine the Advice networks of both sites. While noticeably different at Time 1 (City -.29; Country -.62) both sites have similar E-I scores at Time 2 (City -.62; Country -.64). While City’s rate of change was far higher, both sites reported similar rates of internal focus at Time 2 when considering advice exchange. This degree of similarity when comparing Advice networks is not as evident when comparing both Resource networks. City Resource flows became increasingly internally focused whereas Country demonstrated a small but significant shift towards a higher proportion of external transactions.

It appears that network configurations at Time 1 as well as other contextual elements such as group membership and geographical dispersion may have impacted on the manner in which the system was adopted. This is especially interesting given the aims of the current system in terms of process standardization and consistency (see as discussed in Section 1.5.1). This will be discussed further in the next chapter, Chapter Eight.

7.2.5 Summary - Network Change

This section of the chapter reported on four separate analyses aimed at progressively documenting and examining the nature of network change following an ES implementation.

167

While unable to conclusively determine whether network change had occurred, QAP analysis was able to confirm a reduction in the similarity of all matrices across time. The QAP results were consistent for both City and Country. While Advice and Resource exchange networks were highly correlated within each time period they reported a reduced correlation across time. While the QAP analyses were useful in disconfirming the stability of network structures over time, they were unable to highlight the specifics of the change.

A comparison of network density indicated a mean increase in ties for both sites. Only City however, reported an increase that was significant at p<.05. An examination of IN-degree ties for users at both sites identified a different pattern of network change at Time 2. While the majority of City users increased their ties by a sizable amount (10-15 ties) Country users reported a mean increase of only one tie. A closer examination of tie distribution at Country indicated that up to 44% of users actually decreased their number of ties at Time 2. A review of the Country demographic data indicated that group ownership determined the nature of the change. Two of the four Country groups exhibited a similar adoption profile to City, while the remaining two demonstrated little change in network structure.

The final analysis further examined the tie distribution throughout the organization to determine the influence of group levels ties on network change. Krackhardt and colleague’s E-I index was used as a measure of group and network interaction. An examination of E-I indices over time again indicated that the degree and nature of network change was determined by network site. City work groups demonstrated a tendency to become increasingly insular in their sharing of resources and advice. In contrast Country users reported a slight increase in their advice seeking and resource exchange outside their immediate workgroups.

Overall the analyses supported the notion that ES technology can act as a catalyst for network shock. The results only provided partial support for hypotheses 6a and 6b. While one network (City) reported increases in Advice density over time, one did not (Country). A similar result was found for each site’s Resource exchange networks. Further investigation at a group level of analysis revealed that two of the four Country groups exhibited change patterns consistent with City. Chapter Eight will investigate some of the possible contextual elements that may have led to the difference in network change across groups.

168

7.3 Network Centrality and User Acceptance

A fundamental aspect of the theory presented in Chapter Two was the argument that changes in a user’s network centrality or brokerage status would demonstrate a positive relationship with user acceptance levels. This section presents a series of multiple regressions aimed at testing the hypotheses concerned with this argument.

H7: Users will report a positive relationship between change in network centrality and user acceptance.

H8: Users will report a positive relationship between change in structural autonomy and user acceptance.

Given the high degree of correlation between Advice and Resource exchange networks at each given time point (.66 – .76) the decision was made to merge the networks prior to the regression analysis (Ibarra, 1993). Four network measures were chosen to represent the notions of centrality or brokerage as presented in Chapter Two. Freeman’s (1979) In-degree and (1991) Flow-betweenness were used as measures of centrality, while Burt’s (1992) structural autonomy measures of Efficiency and Constraint were used as measures of brokerage. While Bonacich’s (1987) Power centrality was discussed in Chapter Two it was removed from the analysis schedule due to the less than 100% response rate. Measures such as Bonacich’s (1987) Power and (1972) Eigenvector centrality have been shown to be susceptible to missing data and therefore considered potentially unstable in this case (Costenbader & Valente, 2003). Davis’ (1989) Technology Acceptance Measures of Perceived Usefulness and Perceived Ease of Use were used as the dependant variables. Where required and appropriate normalized scores were used as the measure of each users centrality.

In order to test the hypotheses concerning the link between network change and user acceptance a hierarchical regression technique was used, with the user acceptance variables at Time 2 (Perceived Usefulness; Perceived Ease of Use) as the dependant variables. The technique employed was similar to that reported in Chapter Five when investigating the relationship between job characteristics and user acceptance following the introduction of ES technology.

In the first step centrality or brokerage scores at Time 1 were entered to control for existing network position and to test for lagged effects (Parker et al., 2003). In the second step centrality or brokerage scores at Time 2 were entered. Any variance accounted for in the final model is the contribution of network centrality or brokerage

169

on user acceptance at Time 2 (Axtell & Parker, 2003). While causality remains untested, controlling for Time 1 network position allows a stronger test of association between user acceptance and centrality or brokerage at Time 2 (Parker et al., 2002).

The vector regression routine offered by UCINET 6 was used to run the regressions described above. Conventional OLS beta weights and intercept scores are provided, however the standard error terms were calculated using the random permutation method (outlined in section 4.3.3 of Chapter Four). In brief, UCINET6 uses multiple random permutations of the dependant variable values to derive estimates of standard errors (Borgatti, et al., 2002). In situations demonstrating the potential for autocorrelation the permutation approach has been shown to have significant advantages (Hannemann et al., 2005). The result is a far more stringent test of significance than would otherwise be applied when the independence of observations was assumed (Krackhardt, 1988). Unfortunately while the use of UCINET6’s vector regression routine is more appropriate for this type of data it does not report the value of ∆R2 or its significance level. Given that this is the statistic of interest when attempting to measure the effect of change on the dependent variable, an alternative solution was adopted. Where appropriate significance levels reported by UCINET6’s pseudo probability measure and conventional OLS regression standard error calculations were both provided. In short, the reported values and significance levels of ∆R2 are calculated by SPSS, the remainder by UCINET6’s vector regression routine. While the results will have to be interpreted with some degree of caution this approach allows the researcher to determine whether the change in effect size is significantly different to zero.

The first series of regressions examined the relationship between centrality and brokerage scores at Time 2 on Usefulness at Time 2, while controlling for the DV and IV measured at Time 1. The results for the four reported regression equations failed to support the hypothesis that there would be a positive relationship between centrality or brokerage position and Usefulness at Time 2. The results are presented in Table 7.10 below.

170

Table 7.10 Perceived Usefulness at Time 2 as Predicted by Network Change at Time 2 (n=49)

Predictors β R2 ∆R2 F change

Sig.

In-DegreeAt Time 1 -.07 .00At Time 2 .15 .01 .01 .654 ns

Flow BetweennessAt Time 1 -.17At Time 2 -.13 .08 .01 .507 ns

EfficiencyAt Time 1 -.05 .03At Time 2 -.25 .08 .05 2.42 ns

ConstraintAt Time 1 -.10 .00At Time 2 .07 .01 .00 .120 ns

Note: All regressions were non-significant at p<.05

The second set of regressions were carried out with Perceived Ease of Use as the dependant variable. Four regressions were run each measuring the relationship between centrality or brokerage at Time 2 and Ease of Use at Time 2 by controlling for the centrality or brokerage at Time 1 (see Table 7.11 below). Only one measure, the brokerage measure of Efficiency, reported a significant relationship to the DV. Efficiency at Time 2, negatively predicted Ease of Use (r2= .188, ∆R2 = .14, F change = 4.989, p<.05) after controlling for Efficiency and Ease of Use at Time 1. The bivariate coefficient for the effect of Efficiency at Time 2 showed a moderate negative relationship (β=-.43; p<.018).

Efficiency ratio scores range from 0 to 1, with a maximum positive value indicating contacts to non-redundant ties (Hanneman, 2005). The degree of redundancy present within an actors (ego) network is captured by the effective size and efficiency of their ego-network. An efficiency ratio of 1 indicates every contact within the network is non-redundant. Therefore the results indicate that the more redundant ties people develop following the system’s introduction (reduced brokerage opportunities and control over rare resources) the more likely they are to find the system easy to use. Consequently the results would indicate support for a hypothesis contrary to the one put forward in Chapter Two.

8 p<.05 calculated by UCINET6 random permutation test 9 p<.01 calculated by SPSS using conventional OLS.

171

An examination of the descriptive statistics for Efficiency scores record a decrease over Time (T1 - .63; T2 - .56). The mean decrease in scores over time does indicate a tendency for a reduction in Efficiency and therefore increased links with redundant providers of information over time. It is possible that consistency and reliability of information is more valued at this point in an implementation. Anecdotally users commented on the “complexity” of the system tasks and the “HAL way” at Time 2. It is possible that at this stage of the implementation users valued consistency of information over variety. Varied sources (and possibly contradictory) sources of information may have been perceived as confusing and less valuable than those tapping into the same content domain.

Table 7.11 Perceived Ease of Use at Time 2 as Predicted by Network Change at Time 2 (n=49)

Predictors β R2 ∆R2F

change

In-DegreeAt Time 1 -.15 .01At Time 2 .10 .02 .01 .301 ns

Flow BetweennessAt Time 1 -.32 .10At Time 2 -.01 .10 .00 .001 ns

EfficiencyAt Time 1 .02 .04At Time 2 -.43**10 .18*10 .14 7.84 **11

ConstraintAt Time 1 .01 .00At Time 2 .02 .00 .00 .01 ns

Note: ** p<.01; * p<.05

10 p<.05 calculated by UCINET6 random permutation test 11 p<.01 calculated by SPSS using conventional OLS.

172

7.4 Chapter Summary

A fundamental element of the thesis concerned ES technology’s capacity to alter social networks within organizations. Previous research (Burkhardt, 1994; Burkhardt & Brass, 1990) had indicated the likelihood of changes in advice networks following IT implementations. However, the capacity of ES to fundamentally alter the exchange of resources between users was considered to have farther reaching implications. A number of techniques were used to examine whether advice and resource exchange networks altered following the introduction of ES technology.

The findings found support for the idea that both network types at two separate pilot sites were significantly dissimilar over time. However, further investigation revealed that the pattern of change and adoption within each pilot site was different, contrasting a key aim of the system’s implementation. The City site reported an overall increase in density and the majority of users increased their direct ties in both networks types. In contrast, Country failed to report a significant increase in density and a lower proportion of users increased their ties for the same network types. Deeper analysis indicated that network change (or lack of) was dependant on location and group protocol. The findings also suggested that groups tended to rely their own internal resources for advice, rather than explore alternate sources of information. In relation to resource exchange both sites again demonstrated different patterns of adoption. City appeared to significantly increase their within-group sharing of resources while Country marginally decreased their within-group exchange of resources.

The second aim of the analysis was to examine whether changes in network position over time determined user acceptance. Four previously used measures of network centrality and brokerage were regressed against the two user acceptance variables of Usefulness and Ease of Use. In summary, the findings failed to support the notion of a positive relationship between change in network centrality or brokerage and user acceptance. However, evidence was found to suggest that users located within redundant networks demonstrated higher levels of perceived Ease of Use. Contrary to existing theoretical approaches (McGrath &Krackhardt, 2003; Perry-Smith & Shalley, 2003) users appeared to favour redundant sources of advice over alternate or diverse sources of information.

Chapter Eight will consider the findings reported here. The key focus of the chapter will be to examine those findings that were unexpected and worthy of further investigation. The chapter will use supplementary sources of information in an attempt to contextualize the findings and allow a greater understanding of the results.

173

Chapter Eight — Network Change Post – Hoc

Analysis & Discussion

8.0 Introduction

Underpinned by structuration theory the overall intent of the thesis was to investigate the structural implications of introducing ES technology. It was suggested that the new information system represented a new set of structural properties that had the capacity to alter or replace existing properties. The impact of ES technology on existing properties such as job design and social networks were both considered. Chapter Seven was principally concerned with the extent to which ES technology provided the opportunity for network change and whether network change determined user acceptance levels.

The purpose of this chapter is to discuss the findings of this study reported in Chapter Seven. Where appropriate the findings will be considered in light of the research context, attempting to provide clearer understanding of what was found. In summary, the findings indicated support for the idea that both Advice and Resource exchange networks changed following the introduction of ES technology. However, in contrast to the stated hypotheses, and the aims of the system’s design the pattern of change was inconsistent across the two pilot sites. The findings relating to network change and user acceptance also provided some unexpected results. The findings failed to support any of the hypotheses relating to the predicted positive relationship between centrality and brokerage measures and user acceptance at Time 2. Instead, the findings suggested that users with increasingly redundant contacts reported higher acceptance levels. Consequently the aim of this chapter is to explore possible reasons for the results. Three key areas will be addressed. One; to investigate further the inconsistent adoption patterns between the pilot sites. Two; to consider why the acceptance hypotheses were not supported, and three; discuss further the relationship between reduced structural autonomy and user acceptance.

8.1 ES Technology and Patterns of Network Change

The third research question posed in Chapter One was “Will ES technology alter the structural elements based around existing social network configurations ?”. Expanding on the question, Chapter Two introduced the idea of network shock

174

– forced or non-emergent changes to established patterns of interaction. Sources of network shock included organizational re-structuring, downsizing, job re-design and the implementation of new technology. The chapter then went on to discuss the specific aspects of ES technology and its capacity to initiate network shock. Three key elements of ES design were offered as examples of this capacity. Characteristics such as ES’s tendency to decentralise large amounts of information to a greater number of users, its primary focus on automated and sequential transaction processing of simple and routine tasks, and the mandatory nature of the system, were all aspects potentially triggering network shock. As such a hypothesis was put forward arguing for an increase in resource exchange ties throughout a network following the introduction of ES technology. Citing previous research (e.g. Burkhardt & Brass, 1990) the chapter also identified the probability that advice networks would increase following the introduction of an ES. Given the likely introduction of new skills and the degree of uncertainty following the introduction of an ES it was argued that an increase in user’s advice network activity (new ties or increased interactions) would be observed.

Overall the results offered only partial support for either hypothesis. The network data collected from the City site indicated support for the hypotheses with Advice and Resource exchange networks both observed to have significantly increased over time. While the Country network data reported a similar trend the increase in either network was not significant at a p<.05 level. Further analysis was undertaken to determine possible reasons behind the Country results. The findings indicated that two of the four Country groups (labeled C and D) displayed similar adoption patterns to their City counterparts. Elements of this adoption pattern included an overall increase in density for both Advice and Resource exchange networks, a majority of users increasing their direct ties and an increasing dependence on group members for resources and advice. A different pattern of network change was observed for the other two Country groups (labeled A and B). Groups A and B maintained comparable levels of density and in the case of group A, a higher proportion of users experienced decreasing ties.

Discussions with Country users yielded some possible answers for the disparity in results. Users from all groups were contacted to gain a clear understanding of the context surrounding the results. Very few users could account for the result observed for group B (no change). It was suggested that perhaps the small size of the group (seven users) and their extreme geographical isolation (four hours drive from any other group) was a possible explanation. However, another user (user #19) expressed surprise at the result given that the group’s size was comparable to that of group C (increase in ties).

175

The general perception was that the adoption patterns of groups C and D were consistent with a successful implementation. A system trainer commented that group D had their teams structured in relation to the “system roles” prior to go-live. The implication being that minimizing conflict between existing processes and system processes resulted in a smoother adoption process. This observation resonates with findings reported by Mitchell & Zmud (1999). They found that the “tight-coupling” of work processes and IT strategies from an early stage reduces uncertainty and increases the chance of project success. Another senior manager indicated that “leadership and management style at [groups C and D] helped facilitate the process” (user #14). As observed, the adoption patterns of C and D were consistent with that of the City network, reporting an increase in Advice and Resource exchange networks.

A consistent theme emerged when comparing the managerial and contextual characteristics of group A (decrease in ties) with group C (increase in ties) to explain the findings. Users were able to provide a consistent set of possible explanations for group A’s different adoption pattern. Several users made the assertion that group A was a “problem group” or one that had difficulties with the implementation. Group A was described in negative tones as being hierarchically structured with a narrow span of management control and rigidly enforced communication channels. In contrast, group C was described as having cohesive, familial culture where interaction between management and users was fostered through a strong emphasis on training and trust. Users were encouraged and supported to solve problems without management involvement. A group C manager placed a strong emphasis on the need to multi-skill users across different system roles and functions. Training users in multiple roles was seen as a way in which to buffer the unit against lean staffing and to maintain productivity during “slow periods”. The poor availability of staff due to its rural location along with its small size appears to have driven a multi-skilled focus within the team. Representative comments from users included:

“If you want something from another area you would go to your supervisor and go from there”

(group A, user #17)

“[Group C] has always had a culture of sharing information and networking”

(group C, user #19)

“the communication channels down here [group C] are a lot more open…and I’ll leave it at that” [in comparing their group to group A].

(user #15)

176

An interesting and important tangent concerns the emphasis group C placed on training, given that it was the only site observed to address the issues surrounding the ES driven role segregation. This approach is in direct contrast to the scenario described at City, where users portrayed an emerging culture of task isolationism and a “needs only” training regime (see Section 6.1.2.2). Group C was also the group to record the highest group mean scores for both Perceived Usefulness and Perceived Ease of Use. While is it difficult to attribute any causality it is interesting to compare the acceptance levels of group A [Perceived Usefulness, M = 1.90 (.97); Perceived Ease of Use, M = 2.19(1.30)] with group C and its strong emphasis on training [Perceived Usefulness, M = 2.79 (1.21); Perceived Ease of Use, M = 3.50(1.39)]. Previous research has identified supportive training interventions pre-and post implementation as having strong direct effects on both Perceived Ease of Use and intrinsic motivation to use the system (Venkatesh et al., 2002). While not an immediate focus of this study it is cautiously suggested that these results indicate support for these previous findings.

Overall the results and user feedback indicated that the inconsistent patterns of network change may be due to contextual elements such as management style, existing group structures and communication protocols. The differences in adoption patterns between City and Country groups are offered up as examples of structuration in action - that previous structural constraints guided the manner in which the system was appropriated by users. By the same token the structural characteristics of the system appeared to guide future actions. For example, at City new constraints concerning access to the system drove a “culture of compartmentalization” and reduced access to training (user #4). As documented in the job design discussion, system driven task segregation appeared to be changing the career profile of Erudio employees, from one of a multi-skilled generalist (Administrative Officer 1, AO2 etc.) to specialist roles based around system requirements and password controls. In contrast, a pre-existing culture of training and open communication combined with a necessity for multi-skilling at Country group C resulted in users being proficient in an number of system areas and reporting higher acceptance levels. Alternatively at Country group A rigid communication and management controls coincided with those high in centrality at Time 1 maintaining that high degree of centrality at Time 2. It appears that in this group, highly central users were able to consolidate their position following the implementation.

The results also raise questions regarding the intent and outcomes of ES technology and the distribution of resources. A commonly cited capacity of ES is the standardisation processes and the improvement of information and process flows throughout an organisation (Lengnick-Hall et al., 2004). These aims were consistent with those

177

outlined by Erudio management. However there was anecdotal and document evidence to suggest that the introduction of the ES into Erudio was intended to achieve more than just improved technological capability. As reviewed in Chapter One, Erudio documents indicate the desire for HAL to facilitate social, structural and even cultural change throughout the organisation. A review of the network change results would indicate that at Time 2 this was yet to be achieved. The implications of this finding in light of the research questions will be addressed in the final chapter.

8.1.1 Summary

The aim of this section was to explore possible reasons for the inconsistent pattern of network change observed at Time 2. Particular focus was placed on the Country network, the primary source of change inconsistency. Despite all four Country groups being responsible for the same type of work, in a similar faculty environment and possessing similar user profiles, variations in network change were evident. Discussions with users indicated that aspects such as management policy concerning reporting structures and allocation of training may have impacted on the manner in which advice and resource exchange networks changed following the introduction of the ES. The network results also indicated that the aim of process consistency and ES driven organizational change had not yet been realized at Time 2. In fact there were indications that existing differences within networks were exacerbating differences between groups and education centres within the organization.

8.2 Network Change and Acceptance

Hypotheses 7 and 8 predicted a positive relationship between changes in either centrality or brokerage scores and user acceptance. The results reported in Chapter Six failed to support any of the hypotheses. Three explanations are put forward to explain the failure of the data to support the hypotheses. One, that at Time 2 a high degree of network disruption was still occurring; two, other more salient aspects of the system reduced the effect of network change on user attitude; and three, that a wider network boundary was required to capture those actors directly affected by changes in network interactions. These are discussed below.

First, there is evidence to suggest that the organization was still experiencing a high degree of change at Time 2. It is possible therefore that users were unclear at this point as to their standing within the network due to the current disruption and the prospect for future change. This may have prevented users from accurately gauging

178

their standing within the network. A user’s response to a query concerning the progress of the implementation at Time 2 was “mate, we haven’t even sorted out our roles yet12 (user #4). A senior manager (#6) later concurred, commenting that at Time 2 people still had “[the old system] in the head” and that “there was a lot of stress at Time 2”. It is possible therefore that for a sizable proportion of users old patterns of behaviour were still being enacted and that substantial change had yet to occur. In order for people to realize the value of their network connections they must first have an accurate picture of the network and their place within it (Casciaro, Carley & Krackhardt, 1999; Krackhardt, 1990). It may have taken a longer period of time for users to recognize the permanence of positive or negative changes to their networks and to adjust their attitude accordingly.

Second, it is possible that network change is too discrete a process to be acknowledged during a period of such turbulent change. It is possible that other more salient aspects such as learning system commands and screens, dealing with client complaints (due to slow processing times) and management pressure to make the system perform are more closely related to user attitudes at this point in an implementation. Brass et al. (2004) review a number of outcomes stemming from an actors network interactions such as job satisfaction, power, turnover and performance. Consequently it is also possible that alternate, broader measures relating to a user’s work context (e.g. morale, job satisfaction, stress) may have yielded a stronger relationship with network position at Time 2.

While discussions with users failed to provide any information to buttress the thoughts presented above, they did allude to an alternate explanation. A consistent observation was that the primary benefactors of the system were management due to its superior reporting capacity (#6; 8; 9; 17; 19). As outlined in Chapter One and reviewed in Chapter Two, a key element of ES technology is its capacity for increased data storage, retrieval and analysis for timely decision making (Klaus et al., 2000). Importantly, sophisticated access protocols can determine who, when and how this information can be accessed (Bhattacherjee & Hirschheim, 1997). Some support was also found to support Sia et al.’s (2002) findings that previously powerful actors were able to consolidate their power following the system’s introduction. One senior user (#4) alleged that this was made possible by their access to increased training, authority

12 Refers to the manner in which system protocols and security access levels determined the amount and type of job activities performed by users. Typically system “roles” did not conform to established job descriptions and caused established jobs performed by users to be split up into various components and redistributed.

179

and information resources. Consequently there are indications that this case parallels those highlighted by Koch (2001) and Orlikowski (1991) in which IS is used to further strengthen managerial prerogative and control over organizational processes.

Lower ranking users had a different perspective of the changes occurring at their level. Several lower ranking (but higher usage) users observed that “everyone is on the same playing field” and that any power or influence afforded to their colleagues by their previous system knowledge had been erased (#11). It is possible that user’s primary (and possibly only) source of power was expert power derived from previous system knowledge and the superseded business processes (French & Raven, 1960). Therefore it is also possible that the relationship between network position and acceptance is moderated by hierarchical position. This study primarily examined “direct” users of the technology rather than other interested parties such as management. The mandatory nature of this system (Brown et al., 2002) may have resulted in users already comparatively low in the organizational hierarchy having little opportunity to exercise any power afforded to them by their increased centrality. For example, Ibarra (1993) found centrality to be a predictor for involvement in activities involving changes in structure and administrative processes but not for activities directly connected to the primary work task. Up to 70% of the sample were AO3 or lower and hence less likely to be involved in activities such as the implementation and facilitation of any structural change resulting from HAL’s implementation. Therefore it is possible that centrality would have neither a positive or negative effect on this particular cohort of users. In contrast, those in management or involved in more strategic aspects of the organization are perhaps more likely to react to changes in network interactions.

In summary, it is possible that the hypotheses relating to network change and user acceptance were not supported due to a range of contextual and study design issues. It is possible that there was a high degree of flux at Time 2 and that users may have been pre-occupied with more salient aspects of the system’s introduction. These elements may have prevented users from recognizing the degree of structural change taking place or resulted in its relative lack of importance to users. It is possible however that the boundaries placed around the network were too narrow. It is possible that a network boundary encapsulating all members of the organizational context (not just the direct users of the system) may have provided more support for the theory.

180

8.3 Structural Autonomy and Perceived Ease of Use

Two separate and unexpected findings reported in Chapter Six displayed a degree of convergence when observing user’s reactions to the ES implementation. The first was observed when examining the relationship between acceptance and Burt’s (1992) Efficiency measure. The second was highlighted using Krackhardt and Stern’s (1988) ratio of external to internal ties (E-I index).

One measure of structural autonomy, Efficiency at Time 2 predicted Ease of Use (r2= .1813, ∆R2 = .14, F change = 4.9814, p<.05) after controlling for Efficiency at Time 1. The bivariate coefficient for the effect of Efficiency at Time 2 showed a moderate negative relationship (β=-.43; p<.0113). Efficiency ratio scores range from 0 to 1, with a maximum positive value indicating contacts to non-redundant ties (Hanneman, 2005). As such it appears that (in this case) users developing networks of increasing redundancy reported increased levels of user acceptance.

In a separate set of analyses E-I index scores for Advice networks at City indicated that users tended to increase their dependence on members of their own groups for advice at Time 2. At Time 1 users reported a low but significant tendency for within- group dependence (Advice = -.29). At Time 2 this had increased to a moderate - high degree of within-group dependence (Advice = -.62). Country did not report the same pattern, however only a slight increase in an external focus was recorded (T1 Advice = -.66; T2 Advice -.64) and remained moderately internally focused. Previous research has argued the value of high positive index scores in time of organisational crises (Krackhardt & Stern, 1988; McGrath & Krackhardt, 2004). It has been argued that those organisations with a higher E-I index are better able to cope with organisational change. Increased external ties are considered to be valuable sources of advice, support and information. However, there is little reported evidence of how actor,s E-I scores change over time and the manner in which E-I scores are affected by change.

Taken together the results indicate in this case at least, a tendency for users to remain or become increasingly dependant on their group members for advice during periods of significant change. Further, the more users developed a network of redundant contacts (offering similar types of advice and information) the more likely they were to report higher perceptions of Ease of Use. Perceived Ease of Use has strong support as having a direct effect on Perceived Usefulness (Venkatesh & Davis, 2000) and therefore it is

13 p<.05 calculated by UCINET6 random permutation test14 p<.01 calculated by SPSS using conventional OLS

181

suggested that future research investigate this issue in greater detail. It is tempting to consider the emergence of a “laager mentality”14 among group members following the introduction of the new system. In some cases users appear to have “closed ranks” and increased their reliance on fellow group members or users while adapting to the change. User interviews indicated a possible reason for this was that very few sources of help and advice were available at Time 2. System trainers had returned to their previous roles and supervisors were expected to provide any additional assistance. While a couple of users indicated the development of networks outside their immediate work group most talked about “depending on each other to get through” (#5). One manager likened it to bonding through adversity, “it forced everyone to communicate with others because they all had similar problems” (#6).

It is thought that Social Information Processing (SIP) theory may help to explain these results and dovetail with research carried out by the likes of Ellis (1992) and Chattopadhyay et al. (1999). Originated by Salanick and Pfeffer (1978) a substantial amount of work has adopted SIP as an alternate approach to the study of job attitudes (Torraco, 2005). SIP has also been used to explain employee attitudes to organizational and technological change (Burkhardt, 1994; Chattopadhyay et al., 1999; Ellis, 1992). It is suggested that the negative relationship reported between Efficiency and Ease of Use further supports the findings that in ambiguous situations, SIP facilitates the development of beliefs through a shared sense-making experience (Chattopadhyay et al., 1999). Further, that SIP is likely to be strongest in uncertain situations when referent sources possess high credibility (Ellis, 1992). In this instance due to the reduced availability of reliable information, users able to develop a network of contacts providing consistent information reported higher acceptance levels.

Perhaps the most challenging aspect of these results is the manner in which they challenge the understood theory relating to both areas. Structural autonomy theory suggests that those with increased access to non-redundant resources are in a better position to exercise power, maximizing their opportunities (Burt, 1992). A user with more diverse sources of information and system knowledge is considered more likely to report higher levels of Ease of Use and hence Usefulness (Davis, 1989). Similarly, the results also question Krackhardt and Stern’s (1988) findings concerning the presence of a higher E-I index as facilitating organizational change. The results here challenge this in that users tending to rely on their own work groups and those developing networks of redundant contacts reported higher user acceptance. However, Krackhardt and Stern’s (1988) research did focus on the presence of friendship ties rather than advice or resource exchange ties. It is possible that while expansive friendship ties are a

14 An Afrikans term to describe the temporary formation of a defensive structure by drawing wagons together in a circle to form a barricade. The term was later applied to the South African government in its response to international criticism of its policy of apartheid.

182

valuable source of emotional and social support during organizational crisis, they are less useful in relation to specific task or role based aspects such as system adaptation and learning.

8.4 Conclusion

Three aspects were addressed in relation to the network findings. One, why there were inconsistencies in the network change patterns observed at both sites. While no empirical data was obtained, anecdotal reports suggested that contextual elements such as management policy, existing group structures and group protocols were possible reasons for variations in the patterns of network change. Answers to why the data failed to support the presence of a positive relationship between change in centrality or brokerage and user acceptance were also sought. A restrictive network boundary and the possibility of continued network flux at Time 2 were identified as the most likely explanations for the failure of the hypotheses to be supported. Finally, plausible explanations were sought for the unexpected negative relationship between network efficiency and acceptance. This result was also examined in light of evidence to suggest that users tended to seek advice from within their groups rather than seek out new external sources of information.

The final chapter, Chapter Nine will examine the implications of the findings in light of the research questions and the research problem. The theoretical and practical outcomes of the research will also be considered. The chapter and thesis will conclude by acknowledging the research’s limitations and suggesting a number of areas for further research.

183

Chapter Nine — Conclusions

9.0 Introduction / Review

This final chapter concludes the thesis by reviewing the findings in light of the research questions laid down in Chapter One. Driven by the need to investigate further the structural implications of ES technology the aim of the thesis was to answer four broad research questions. Underpinned by structuration theory, the questions probed the relationship between three possible sources of structure and social action within an organization. The characteristics of ES technology, user’s job design and their social networks were identified as three relevant sources of structure and social action within an organization. In broad terms it was suggested that ES technology represented a set of structural properties that either replaced or were superimposed over existing structures (represented by job design or social networks). The tension arising from the clash of structural properties was thought to have significant user acceptance implications. The chapter will review the findings and consider their theoretical and practical implications. The thesis will conclude with an acknowledgment of its limitations, and a number of suggestions for future research.

9.1 Research Questions and the Overall Research Problem

Chapter One outlined the need to investigate further the structural implications of ES technology, in particular its impact on both job design and social networks. It was suggested that the degree of user acceptance demonstrated for these system types was closely related to the extent to which their introduction stimulated the change of both these structural elements. Research questions arising from this discussion centered around the potential for, and the nature of, changes that could occur. It was also queried whether users would report a relationship between structural change and their acceptance of ES technology. This section reviews the findings in relation to those questions, before situating them in the context of the broader research problem.

184

9.1.1 Job Design

The job design component of the study considered two questions.

Research Question 1: Will ES technology alter the structural elements based around existing job characteristics?

Research Question 2: Is there a relationship between perceived system driven changes in job design and user acceptance?

Chapter Two expanded on the potential for ES characteristics to affect job design and considered the user implications of those possible changes. Framed around Hackman and Oldham’s (1976) Job Characteristics Model it was argued that users of ES technology would both anticipate and experience a change in job characteristics. Further, it was argued that users would report a positive relationship between anticipated or perceived levels of job design at Time 2 and system acceptance. However a moderating variable was hypothesized, arguing the amount of system exposure reported by a user would determine the degree of change experienced, and therefore the strength of the relationship between job change and system acceptance. The findings and discussion are summarised below.

Job Design Change

At Time 1, prior to the ES implementation, users anticipated a modest but significant decrease in the characteristics of Skill Variety, Task Significance, Autonomy and Feedback. As such the results supported the hypotheses set down in Chapter Two concerning user anticipations of a decrease in three of these job characteristics, but not for those relating to Task Identity and Feedback.

However, the longitudinal component of the job design study failed to provide support for the hypothesis that over time users would perceive an actual increase in Feedback and an actual decrease in the remaining four job characteristics. The results indicated a bi-polar effect, whereby job change was dependant on whether users were high or low users of ES. Further, perceptions of job change at Time 2 appeared to be dependant on the nature of the position held by the user. In this instance users with higher formal authority were seen to increase in overall job enrichment, while those lower in formal authority were seen to decrease over time.

185

It was suggested that the predominantly strategic nature of ES technology in its capacity to facilitate a number of managerial aims was a direct driver of this result. The inherent characteristics of ES technology are regarded as a key source of competitive advantage within the marketplace (Lengnick-Hall et al, 2004). Key characteristics such as increased reporting capability, process efficiencies and data control are all considered elements allowing organizations significant advantages. It is suggested however that these aspects are primarily of value to management, principally end-users of the system’s output. The results in this case tend to support this assertion, that those senior in the hierarchy benefited from the system’s introduction. In contrast the research suggests that actual direct users of the system may in fact experience a reduction in characteristics related to job enrichment at the expense of the positive gains experienced by management as end-users of the system’s output. To date few studies have been conducted to examine the impact of such systems on direct users of ES technology (Nah, Tan & Teh, 2004). As such this finding is of some significance to both the ES and user acceptance literatures.

In sum, the results suggest the presence of a relationship between the inherent characteristics of ES technology and its capacity for job change. The nature of the job change however was not, as was suggested, a common experience for the majority of users. Rather, the results indicated that different categories of user were affected differently, both positively and negatively. In broader terms the results suggest that in addition to representing a new set of structural properties, ES technology has the capacity to alter the structural properties associated with existing job design characteristics.

Job Design Change and User Acceptance

The differential effects of ES technology on perceptions of job change is a continued theme when considering the relationship between job design change and user acceptance. “System Usage” was confirmed as a moderating variable at both pre- and post-implementation. However, in all but one instance the results reported a relationship that was contrary to the hypothesized relationship. It was hypothesized that high users would report a stronger relationship than low users between changes in all five job characteristics and user acceptance. The pre-implementation results indicated that this was only true for one characteristic, Skill Variety. Additionally, the relationship was negative, rather than positive as hypothesized. From these results, it was suggested that high users, who are likely to have a high level of skill variety and proficiency invested in the existing legacy system, may have anticipated a reduction in the variety of skills while having to learn the new system. Another unexpected finding was the positive relationship between Meaningfulness and perceived ease of use for

186

low users, while high users failed to report a significant relationship. One explanation for these results may lie in the relationship between usage and seniority within the organization. Anecdotal reports indicated that those lower in the organizational hierarchy (and typically higher users) received very little information about the system. Consequently they were less able to make an assessment of the system’s impact on their job design. Conversely, low users tended to be more senior in the organization and it is thought that they were exposed to more information concerning the system. Consequently these users were able to make a more informed judgment regarding the system’s impact on their job role and context.

The post-implementation results identified low users as reporting a positive relationship between levels of Task Identity at Time 2 and user acceptance, but not for high users. No other interaction effects were reported for the remaining four job characteristics or the composite variable of meaningfulness. The post-implementation result for Task Identity was explained in that low users still coming to grips with the technology would have difficulty identifying the new process flows and their place within them. Conversely, low users who were able to quickly adapt to the new system and able to exploit the new functionality were more likely to experience higher levels of Task Identity. In either case the close linkage between work function and system process was likely to result in the positive relationship reported.

In summary, this component of the study provided weak support for the idea that ES technology may stimulate a degree of structural change in relation to job design. Chapter Two argued for ES characteristics to universally affect user’s job design to a similar degree. In fact the findings reported here indicate that aspects such as usage and hierarchy may determine the job design outcomes of an ES implementation. The results concerning job change and acceptance were inconclusive, but did indicate that low users were more likely to report a positive relationship between job change and acceptance. Importantly the results validate the need to further investigate the nature of job change associated with ES technology, the degree to which it affects users, and the extent to which acceptance levels are determined by job change.

187

9.1.2 Social Networks

The second component of the thesis considered two particular research questions.

Research Question 3: Will ES technology alter the structural elements based around existing social network configurations?

Research Question 4: Is there a relationship between perceived system driven changes in social networks and user acceptance?

A key contribution of the thesis is its explicit acknowledgement of what was termed network shock and the attitudinal and behavioral outcomes of such an event. Chapter Two hypothesized the potential for ES-driven social network driven change, coined “network shock” - defined as forced or non-emergent changes to established patterns of interaction. The capability of ES technology to induce network shock was considered, and two hypotheses put forward. While it was considered likely that advice networks would increase, it was also argued that more permanent, substantial change would occur in relation to resource exchange networks. It was also argued that a positive relationship would be observed between changes in network centrality / brokerage and user acceptance. In keeping with the structuralist perspective, it was proposed that users would react to the perceived changes and use their network connections to facilitate their own agendas. While outside the empirical scope of the study, the proposition considered the potential for users to use their network position to influence others as to the system’s value.

Network Change

Partial support was found for the hypothesis that both Advice and Resource exchange networks would increase in density following the introduction of the ES. Of the two case study sites only City supported the hypotheses. In that case both network types exhibited a marked increase in density and a sizable increase in ties between users. A group level analysis of the other network (Country) revealed that two of the four groups present within the network also exhibited characteristics consistent with the hypothesized effect. Post-hoc investigations suggested that managerial policy within the other two Country groups may have been responsible for the lack of increase in Advice and Resource exchange ties.

The change in configuration of Resource exchange ties observed in six of the eight groups is particularly important. Whereas an increase in Advice ties between users may be expected following an implementation, it is also just as reasonable to expect

188

them to be transitory (Burkhardt, 1994). In contrast, changes to Resource exchange patterns represent a potentially more permanent, fundamental change in the interaction patterns of users. As argued in Chapter Two, Resource exchange networks represent sources of structural power based around embedded relationships of dependence existing between users (Walker et al., 2000). Perceived or actual changes in existing relationships have the potential for disaffected users to jeopardize the successful introduction of an ES (Schwarz & Brock, 1998). Previous research has indicated the potential for implementation failure, delay and even project abandonment due to the actions of users perceiving the system’s introduction as detrimental to their interests (Brown & Jones, 1998; Cavaye & Christiansen, 1996; Macri et al., 2001).

The findings of this study identified two areas worthy of future investigation. The first was that contrary to the organization’s expectation of process consistency and standardization, aspects such as geographical location and management policy affected the pattern of network change. The second was that a review of E-I indices over time indicated that City actors tended to increase their reliance on actors within their own groups. Country actors (already heavily insular) only showed a marginal increase in their external focus over time. Current thinking suggests that for successful change to occur networks should demonstrate high degrees of interaction between, rather than within groups (McGrath & Krackhardt, 2003). This single case example hints at the possibly “natural” tendency for groups to retreat internally during periods of significant change. As discussed on several occasions, the effect of change interventions on network configurations is an area lacking in empirical investigation. Section 9.6 below will discuss the potential for future research to better investigate these areas.

Network Change and Acceptance

In this particular instance no evidence was found to support the idea of a positive relationship between a change in network conditions and user attitude. As discussed in Chapter Eight, it is possible that the timing of the data collection may in part be responsible for this result. Anecdotal evidence suggested that a high degree of flux was still evident at Time 2, and that other, more salient issues may have eclipsed any attitudes concerning network change. Had the opportunity presented itself, it is considered likely that a further round of data may have yielded results better able to support the theory put forward in Chapter Two. Of interest was the negative relationship found between network efficacy and acceptance. Complementing to some degree the E-I indices results, this finding indicated that users developing a redundant set of contacts tended to report higher acceptance levels. It is possible that this result

189

reflects a process of social information processing, whereby users, particularly during periods of uncertainty and ambiguity, look to others around them to make sense of the changes (Chattopadhyay et al., 1999; Ellis, 1992).

In summary, while the study found evidence to suggest that ES technology was capable of producing network shock, no evidence was found to support the hypothesized positive relationship between network change and acceptance. The findings however did indicate a number of possible future areas of research, some of which will be dealt with below in Section 9.6.

9.1.3 Overall Research Problem - Summary

This study was primarily conceived around the notion that ES technology characteristics were fundamentally different to earlier forms of IS/IT. It was argued that the unique characteristics of ES technology presented the opportunity for what Barley (1990) referred to as an “occasion for structuring”. Underpinned by structuration theory it was suggested that changes in the three structural mechanisms of job design, social networks and ES technology would to some degree determine user acceptance. As such the broad research question asked “What are the structural implications of introducing ES technology into an organisation?”. While not providing any conclusive answers to the question posed above, the research indicated the potential for structural change following an ES technology implementation. It also provided weak support for the relationship between user acceptance and the degree of job change experienced by users following the introduction of an ES.

An important observation is that while the results demonstrated a degree of change for both job design and networks, the evidence suggests that the case organization neither anticipated nor showed concern about these structural implications. Structuration theory in the context of ES technology would suggest that a high degree of consideration should be placed on existing structural properties. Particular attention should be paid to the manner in which they may determine the manner in which ES’ are incorporated into an organization. For example, this study has argued that while ES technology represents a set of structural constraints, it itself must be overlayed or replace an existing set of structural conditions embedded within an organization (e.g. represented by job design or social networks).

190

While structuration theory and its variants gain increasing support within the IS/IT literature (Poole & DeSanctis, 2004) examples abound of organizations attempting to use IT as a principle driver of initiatives such as business re-engineering and widespread structural change (Bhattacherjee & Hirschheim, 1997; Paper, Tingley & Mok, 2003). The organization studied in this case appears to be no different. As outlined in Chapter One, the case example ES was viewed as a primary driver of change, designed to achieve processes standardisation and consistency. In one report the capacity of the system to facilitate culture change was even discussed. In contrast, a review of senior change management and implementation documents indicate a lack of awareness or understanding of the changes faced by the organisation and its users (See Chapter Six, Section 6.1.1). The network analysis component of this study in particular highlights the capacity for contextual variables to impact on the manner in which system characteristics are incorporated into everyday working arrangements. In the case reported here it appears that aspects such as existing job design, resource exchange networks and management policy may impact on the manner in which the system is appropriated by users. As such these results join a growing body of research highlighting the difficulties associated with using ES technology as the primary driver for widespread organisational change (Bhattacherjee & Hirschheim, 1997; Haines, 1999; Paper et al., 2003). As will be discussed in Section 9.6 below, the research also indicates the need for a greater appreciation of the structural dynamics following the introduction of an ES, a greater understanding of the nature of the change, and its potential outcomes.

9.2 Theoretical Contribution

The work contained within the thesis offers a number of valuable theoretical, methodological and empirical contributions to each of the user acceptance, social network and enterprise systems literatures. These are detailed below.

9.2.1 User Acceptance - Technology Acceptance Model

The study makes a worthy contribution to the user acceptance and Technology Acceptance Model (TAM) literature in its empirical examination of the social and contextual determinants of user acceptance. While TAM has received strong empirical support, some have called for an increased focus on the social and contextual antecedents of Perceived Usefulness and Perceived Ease of Use (Green, 1998; Karahanna & Straub, 1999; Mathieson, 1991; Mathieson, Peacock & Chin, 2001). This study represents an attempt to determine the relative effect of structural elements such as job design and

191

social networks on user acceptance. Further, although not conclusive, the results do provide some weak support for the idea that changes in structural elements (defined here as job design and social networks) following an ES implementation are related to perceptions of Ease of Use and Usefulness. As such the study offers new insight into the potential determinants of Percieved Usefulness and Perceived Ease of Use beyond the individual unit of analysis.

Importantly the findings also indicate an increasing need for user acceptance research to stretch beyond transitory, short term outcomes such as perceived ease of use, perceived usefulness, training and computer efficacy. For example, the results indicated that some users experienced a decrease in characteristics associated with job enrichment following the ES implementation. Job enrichment has strong empirical links to attitudinal outcomes such as job satisfaction and internal work motivation, and to a lesser extent, behavioural outcomes such as absenteeism (Parker et al., 2001). Consequently the findings from this thesis suggest that the introduction of an ES may herald longer-term, and more fundamental attitudinal and behavioural implications for organizations. Accordingly the study is an important step forward in acknowledging that while system characteristics may drive immediate attitudes, antecedents of user acceptance such as changes in structual elements may also have longer term attitudinal and behavioural consequences.

9.2.2 Network Shock – Attitudinal and Behavioral Implications

Another significant contribution of the thesis is its explicit acknowledgement and definition of network shock and the consideration of the attitudinal and behavioural outcomes that may result from such an event. Network shock is defined as forced or non-emergent changes to established patterns of interaction. Examples of shock producing events may include: job re-design activities such as multi-skilling, job sharing and job rotation (Krackhardt & Porter, 1986); downsizing and departmental restructuring (Shah, 2000); training and development activities; and the introduction of new technology (Burkhardt, 1994).

There has been an increased focus on the role played by networks in the area of change management in recent times (Stevenson, 2003). However, existing work tends to focus on the manner in which networks can help or hinder the change initiative (Stevenson, 2003). Others focus on the manner in which the change alters actors’ network interactions, for example in relation to power (e.g. Brass, 1984) or structural holes (e.g. Susskind et al., 1998). To date the consideration of social networks and

192

the effect of network disruption on “attitudes to change” has only received sporadic attention. What few examples there are tend to focus on “catastrophic” events such as downsizing, and place a reduced emphasis on resulting participant attitudes towards the change event. The work contained within the thesis highlights the potential for network change (either by design or unintended consequence) to affect attitudes towards change. Further, that network change may also be a catalyst for a range of behaviors aimed at either supporting or sabotaging the implementation. Therefore the theory put forward here helps to further our understanding of the significant role played by social networks in the context of change intervention success or failure.

9.2.3 Network Change

The study makes another contribution to the social network literature in its empirical investigation of network change over time. In a recent essay, Borgatti (2003) highlighted the changing nature of social network research. He noted that while a considerable amount of resources had been spent on identifying the consequences of networks, only recently was a greater focus being placed on the observation and study of network antecedents and change. Brass et al. (2004) concurs, acknowledging the critical need to engage in longitudinal network research in order to reduce the causal ambiguity created by cross-sectional research designs. The study reported here furthers the study of network change, following the lead of others such as Susskind et al. (1998). In particular the study offers a better understanding of user’s network interactions under conditions of duress. For example, contrary to theoretical prescription, groups appeared to withdraw within, or maintain a dependence on group members for advice and assistance. This and other findings have provided a number of avenues worth pursuing in terms of future research (addressed further in section 9.6 below).

9.2.4 Structural Implications of Enterprise Systems Technology

Given the ubiquitous and continued presence of ES technology the findings also represent a contribution to the growing body of literature dedicated to the study of enterprise systems. In this instance, a theory of user acceptance was developed specifically for the nature and characteristics of ES technology. As such, the study represents a response to what Orlikowski & Iacono (2001) referred to as the need to “theorize the IT artefact”. In doing so, this research acknowledges the need to consider the function and intent of the information technology under investigation, in this case enterprise systems (Orlikowski & Iacono, 2001). For example, Long (1993) reviews numerous studies investigating the effect of what is generically termed “information

193

technology” on the job quality of employees. Unsurprisingly the results are mixed, with some studies reporting positive effects and others reporting negative effects for high and low status employees. The ability of researchers to both predict and explain the effect of IT on user’s job design is hampered when the specific nature of the IT is not considered (Orlikowski & Iacono, 2001).

In the study reported here the attention paid to both the strategic nature of ES technology, as well as documenting the case organisation’s specific aims for their ES, enabled the researcher to better interpret the study’s findings. Two findings should be of especial interest to ES researchers. First, the inherent characteristics of the system may actually represent the potential for de-enrichment for jobs lower in the organisational hierarchy, particularly those responsible for organizational efficiency activities such as data capture and processing. Second, the findings call into question those that advocate ES technology as a driver of job enrichment and empowerment (e.g. Davenport, 1998; Hammer & Champy, 1993). Instead it adds to an emerging body of research suggesting that the enrichment capability of ES technology is limited to those responsible for the system’s introduction or a select senior group within organisations (Koch, 2001; Macri et al., 2001; Tansley et al., 2001; Sia et al., 2002).

9.2.5 Network Analysis and Structuration

Recently Poole and DeSanctis (2004, 213-14) outlined a set of requirements for researchers wishing to examine structuration in an IS/IT context. Among them were the need: to develop a description of how the social system works and a description of how the social context is reproduced; to identify the redefining and reproduction of structures; and to undertake a critical enquiry into the dynamics of power underlying the structuration process. It is suggested that this study successfully highlights the utility of network methodologies and analysis techniques in complying with these requirements. Arguably few approaches are better suited to capture the dynamic nature of structuration and allow researchers to capture tangible elements of the structuration process during an event such as the implementation of a new technology. In carrying out this study I have provided an example of the practical, methodological and analytical considerations that should be addressed by IS/IT researchers wishing to adopt network analysis in the context of structuration theory. It is hoped that future attempts at using network analysis in the study of structuration theory would further refine the approach. It is suggested that identifying those networks representing the most potent structural properties during IS/IT implementations, as well as investigating the manner in which actors redefine or reassert their network positions following an implementation, are both avenues worth pursuing.

194

9.3 Practical Implications

The findings have a number of implications for practitioners involved in the co-ordination and management of ES implementations, as well as those responsible for the longer term issues of change effectiveness and performance. While the findings indicated stronger support for the relationship between job design and ES technology, the network findings also offer guidance for practitioners involved in ES implementations. Overall, the findings support previous research documenting the difficulties experienced by organizations attempting to achieve widespread organizational change via the introduction of an enterprise system. The introduction of a system aimed at process standardization and consistency is unlikely to achieve its aims purely by being implemented within a mandatory use context. Management practice, organizational structure and existing job design all present as elements capable of impacting on the manner in which users adopt and use the system. While it is conceivable that ES technology may legitimately represent a key component of a wide scale change initiative, it should be complemented with a range of additional interventions to facilitate the desired changes represented by the ES technology. Some of these are outlined in detail below.

9.3.1 Job Design and User Acceptance

Despite vendor assurances, ES technology is notoriously inflexible and often requires organizations to decide between prohibitively expensive customisations or adopting “best practice” system protocols and processes (Hall, 2002). Consequently when adopting ES, organizations typically change work flows and alter user’s job characteristics (Lengnick-Hall et al., 2004). While some like Koch (2001) argue this is often a deliberate attempt by management to further enforce management control, it is also likely that system driven job design may also be unintended “collateral damage” of the system’s introduction. Either way, this study suggests that greater attention should be paid to the job design consequences of introducing an ES into an organization. Existing research concerned with the impact of social context of user attitude suggests that user attitude is not solely determined by a system’s technical attributes (Burkhardt, 1994; Venkatesh & Davis, 2000). In the context of this study, the very fact that a proportion of users anticipated negative job outcomes resulting from an ES implementation may have been enough of a catalyst to trigger resistance activities. Numerous case examples exist within the literature attributing implementation failure to the activities of militant users perceiving a threat from the system (e.g. Brown & Jones,1998; Cavaye & Christiansen, 1996). Addressing users

195

attitudes to the relationship between job design and ES introductions may at least in part help to reduce the short and long term effects of anticipated or perceived job design change following the introduction of ES technology. The possible short and long term effects are addressed briefly below.

In the short term it is suggested that in order to reduce the risk of concerns about system-driven job change and reduced user acceptance levels, attention must be paid to user’s concerns about job change after a system “goes live”. Where possible users should be given a clear picture of post-implementation job design in the context of their own job. It is important that trainers and facilitators not only educate users in system “mechanics” such as navigation and system protocols, but highlight how changes compare in terms of the significance of their role, the value of additional ES related skills and the continuation of aspects such as autonomy within the work unit, for example.

In the long term it is suggested that practitioners pay closer attention to the effect of process-oriented work design on user attitudes outside immediate areas of concern, such as user acceptance. A solid body of work over the last three decades supports the effect of job characteristics on satisfaction and motivation, and to a lesser degree performance, turnover and absenteeism (Parker et al., 2001). Therefore in addition to the obvious user acceptance concerns, the study’s findings suggest the introduction of ES may have longer term implications for organisations. Consequently system implementers are advised to adopt a longer term perspective, with measures taken to ensure that process changes are made in line with job enrichment principles. Where possible attempts should be made to minimise impacts on job characteristics, or additional interventions should be planned to balance any negative changes that occur in relation to job design. For example, Task Identity was identified as a salient element in “Low” user’s minds post-implementation. This result was complemented with anecdotal evidence to suggest that low users in particular lacked an understanding and knowledge of the new business processes. The study’s findings indicate that it is these users who are most affected by the structural changes associated with ES technology. Consequently implementation teams should pay particular attention to the training of management and supervisors in the process flow and sequential transaction aspects of the system. Given the vital role played by these roles and the position of influence they hold within an organization it is considered important that this issue is addressed before and during the implementation phase.

196

9.3.2 Social Networks and User Acceptance

While others have discussed the role of social network analysis (SNA) in an organizational context (e.g. Nelson, 1988) this study has highlighted the utility of SNA in an ES implementation context. Network analysis has the ability to help practitioners identify key opinion holders within organizations and identify the presence (or lack of) information sharing networks (Borgatti, 2003). Consequently SNA can help facilitate the development of support and advice networks for users during the implementation phase (Mohrman et al., 2003). Network analysis can also assist in mapping out the potential impact of ES technology process flows on existing resource exchange relationships. Even the use of simple network mapping using a software tool such as Netdraw (used in Chapter Seven) can help identify the impact of new process flows and access protocols on users. The use of more sophisticated network metrics such as those used in this study may also be used to identify possible sources of resistance or indeed endorsement of the new system. It should be stressed that the use of network analysis (like any research tool) should not be conducted without an awareness of the ethical and political consequences resulting from collecting and acting on network data (Borgatti & Molina, 2003). Misinformed or misguided interpretations of network data can have significant employment, promotion and career implications for employees even if they have chosen not to participate in the collection of the network data (Borgatti & Molina, 2003). Consequently it is important when using network analysis as a management tool to gain active support and participation, provide a degree of transparency when reporting results, and to clearly define the bounds of the data collection and its use (Borgatti & Molina, 2003).

9.4 Limitations

The study reported here has a number of acknowledged limitations to be considered when discussing the significance and applicability of the results discussed above.

The first and most obvious limitation concerns the source of the data. While the embedded single case design adopted for the study had a number of advantages, the potential difficulties of generalising to an alternate population are acknowledged (Yin, 2003). The results reflect one organisation’s experience when introducing an ES. The potential for poor generalisability is compounded when taking into account the specific nature of the organisation’s industry (education) and the peculiar nature of its organisational structure and geographical dispersion. Further, the network component of the study was only able to capture the experiences of two similar departments in

197

a diverse organisation. It is possible that different contextual factors may prevent a replication of the results reported here and therefore further weaken the study’s generalisability.

The second concerns the longitudinal nature of the research. While the longitudinal approach taken in this study is strongly advocated (Brass et al., 2004; Green, 1998; Jackson et al., 1997) there is evidence to suggest that a longer time frame needed to be taken. As Burkhardt (1994) suggests, the dynamic nature of network interactions following an IT implementation requires a long term perspective. It is possible that given the nature of the change the organization was still experiencing a high degree of flux when Time 2 data were collected. Had the opportunity presented itself it is likely that a third or even fourth round of data would have produced a better reflection of the change process. As discussed earlier, it is also possible that an extended longitudinal approach may have yielded results better able to support the hypotheses set down in Chapter Two.

The nature of the respondents for both the job design and network studies form the basis of the study’s third limitation. As outlined in Chapter Three an on-line survey was employed as the primary data collection tool for the job design component of the study. It is possible that the nature of the survey, along with the delivery method may have led to respondent bias. Those responding may have a higher degree of IT efficacy and familiarity with web browser technology. Post-hoc anecdotal comments received from system trainers indicated that a high degree of users had limited Graphical User Interface (GUI) or “Windows” navigation skills. It is possible that the use of a web-based survey instrument may have prevented responses from those unfamiliar with GUI based interfaces and therefore the participants of the study may represent those with a pro-technology bias. That acknowledged, a comparison of participant demographics failed to identify any observable differences to those participating in a paper-based job satisfaction survey carried out over the same period.

Similarly, the nature of the network data collection method may have also impacted on the results. The challenging nature of network data collection and the necessity of high response rates in part forced the decision to only look at two faculties. This consequently reduced the amount of subjects and responses able to be tested. The opportunity to examine a larger number of networks (and therefore participants) would have allowed a comparison of a greater number of networks and their patterns of network change.

198

9.5 Directions for Future Research

The results of the studies have identified a number of potentially fruitful avenues for future research. Three key areas are thought be worthy of consideration in order to further both the user acceptance and social network literatures.

First, future research should investigate further the differential effects that ES technology may have on various user types within organizations. The job design findings highlighted the existence of what was termed a bi-polar effect on users of ES technology. In addition to an observed difference between high and low users, differences between high users were observed based on hierarchy and pre-existing levels of job enrichment. However “usage” as a variable is fairly limited in its ability to clearly identify the relationship between users and the impact of ES technology on job design and vice versa. For example, “High usage” fails to distinguish between those using the system for analytic purposes or alternatively, data entry.

Instead it is suggested that future research should examine role orientation as a potential moderator of the relationship between ES technology and job change. If supported this construct offers a higher degree of specificity in relation to predicting the effects of ES technology on user’s job design. In this instance role orientation refers to whether a user’s primary role within the organization is concerned with organizational efficiency or organizational effectiveness. An accumulating body of evidence supports ES technology’s capability to facilitate both efficiency and effectiveness aims of an organization. The efficiency aims of an organization are fulfilled thanks to ES’s increased ability to standardize data capture and processing as well as its increased monitoring and control capabilities (Orlikowski, 1991; Sia et al., 2002). Effectiveness aims are facilitated due to increased and improved reporting and analytical capability, integrated business processes and real-time data processing (Klaus et al., 2000; Davenport, 2000). It is reasonable to expect that the primary benefits offered by ES technology are likely to result in increased job enrichment for those responsible for organizational effectiveness. In some cases however a conflict of interest may emerge between the aims of ES technology and the aims of job enrichment, particularly when considering users primarily responsible for organizational efficiency. Those responsible for efficiency (users responsible for the data capture, entry and processing aspects of the system) are more likely to be constrained by password controls, data-entry protocols and process flow requirements (Hall, 2002; Tansley et al., 2001). Therefore their comparative lack of autonomy prevents users realizing any benefits

199

that may result from an increase in aspects such as Skill Variety or Feedback (Dodd & Ganster, 1996). If supported the construct of role orientation offers a higher degree of specificity in predicting the effects of ES technology on user’s job design.

Second, a greater focus should be applied to the nature of network interactions during significant change (Brass et al., 2004; Stevenson, 2003). Researchers also need to broaden their field of enquiry beyond “catastrophic” events such as downsizing where significant numbers of actors are removed from the network (e.g. Shah, 2000; Susskind et al., 1998). The research reported here indicates the need to better understand what occurs to employee’s networks following change interventions where key players remain, but established relationships are artificially altered. Even less research has been conducted on the dynamics of E-I indices and their relationship to change success. For example, the original E-I theory discussed the presence of friendship ties, rather than advice or resource exchange ties as were measured here. Future research may want to examine the presence of friendship ties and their role in relation to change intervention success. An extension of this may be to also consider the presence or disruption of simillian ties in relation to attitudes towards change. Example questions include: What do groups do during and after periods of significant change? What are the performance and acceptance implications for groups exhibiting different reactions to the change intervention? Do friendship networks exhibit the same properties over time (for example, do friendships change?).

There also remains a need to better understand the relationship between network efficiency and user acceptance. The results reported here indicated a negative relationship between network efficiency and user acceptance. In simple terms users with increasingly redundant contacts reported higher levels of perceived ease of use. Future research could investigate in greater detail reasons behind the relationship as well as testing the finding’s generalisability to other ES implementations and even other organizational change interventions. For example, the assumption of causality of the relationship needs to be examined. It is possible that these results reflect a case of homophily, whereby people tend to interact with others sharing similar characteristics (Borgatti & Foster, 2003). It is entirely possible that users with similar acceptance levels gravitate to each other, producing an alternate explanation for the relationship between user acceptance and network efficiency.

Third, future research should revisit the theory concerning user attitude and network change, shifting the focus of the timing and population under examination. A key element of the theoretical component was the idea that perceived or actual changes in social network position would influence user attitudes towards the system. The

200

findings failed to support this theory. This study primarily examined “direct” users of the technology rather than other interested parties such as management. The mandatory nature of this system (Brown et al., 2002) may have resulted in users already comparatively low in the organizational hierarchy having little opportunity to exercise any power afforded to them by their increased centrality. In contrast, those in management or those involved in more strategic aspects of the organization are perhaps more likely to react to changes in network interactions. It is suggested that future research revisit this theoretical proposition earlier in the technology adoption process and establish a wider network boundary. My findings concur with previous research to suggest that senior levels of the organisation were the primary benefactors of the implementation (Koch, 2001, Sia et al., 2001). Another set of limited research indicates that powerful actors may exercise their power at the time of adoption, rather than later during implementation (Macri et al., 2001; Tansley et al., 2001). As such it is suggested that researchers broaden the network boundary to capture those involved in the adoption of the system while carrying out the process earlier, prior to the decision to adopt the system.

9.6 Conclusion

Despite the growth and saturation of enterprise system types, comparatively little research has been undertaken to examine the user and organizational issues surrounding their implementation. This research has demonstrated the capacity for the inherent design elements of ES technology to have significant structural implications for user’s job designs and social networks. Further, the job design results, and to a lesser extent the network efficiency results, demonstrate the effect of social context on user acceptance. As such they provide insight regarding the potential determinants of user acceptance beyond the individual unit of analysis. The findings indicate an increasing need for user acceptance research to stretch beyond the transitory, short term measures of user acceptance such as perceived ease of use and usefulness. Finally, the thesis contributes to a small, but growing literature examining the role of social networks in the process of organizational change. In particular, the thesis considered in detail the attitudinal and behavioral consequences of artificially altering established patterns of interaction. As such the study highlights the need to better understand the role of networks not only in the case of facilitating change, but also in terms of change intervention success. This and other findings contained within the thesis represent a step forward in understanding the structural and user acceptance implications of this technology, while identifying a number of promising future research avenues.

201

Appendix A.01

Survey Briefing Slides

202

203

Appendix A.02

Job Design and User Acceptance Survey Form

204

HAL Functionality and Work ContextThank you for agreeing to participate in this project. It is hoped that by contributing to this survey the researchers will gain a better understanding of the crucial nature of work processes, system design and organisational expectations when introducing new IT systems. This short questionnaire should take 5-10 minutes to complete.

The questionnaire is divided into two parts. Part A asks you to consider the new HAL system, what it is like and whether it has changed your working relationships. Part B asks you to consider your work environment. This includes aspects such as your job and work procedures. The final section asks you to provide some basic demographic data for statistical verification purposes.

We value highly your individual responses and participation, so please do not confer with your colleagues.

This research is being carried out as part of a larger QUT PhD research project, as such no reported data will identify you individually. The researcher conducting this research abides by the principles governing the ethical conduct of research and, at all times, avows to protect the interests and confidentiality of all participants. None of your responses will allow you to be identified individually.

Your Unique Code Identifier

This questionnaire is completely confidential and information reported cannot identify you as an individual. However, to match your responses across the survey period, while still maintaining your anonymity, a unique code identifier is required.

To create your unique code identifier, please fill in the following spaces:

a) What are your institutes initials? (e.gXX or YY)

b) What are the last two letters in your father’s first name?

c) What are the first two letters in your mother’s maiden name?

d) What is the date (day only) of your birthday?

205

Part A – The HAL System

This part of the questionnaire asks you to describe how and what you think about HAL and the effect you feel it has had on your working conditions. Try to be as accurate as possible, remembering that all of your responses are completely confidential.

HAL Performance

This final section of Part A asks you to consider how user friendly HAL is. Please check the box that best describes the extent to which you feel about HAL.

1 2 3 4 5 6 7Disagree Strongly

Disagree Disagree Slightly

Neutral Agree Slightly Agree Agree Strongly

1 2 3 4 5 6 7

Using HAL has improved my performance in my job � � � � � � �

My interaction with HAL is clear and understandable � � � � � � �

Using HAL in my job increases my productivity � � � � � � �

Interacting with HAL does not require a lot of mental effort

� � � � � � �

Using HAL enhances my effectiveness in my job � � � � � � �

I find HAL easy to use � � � � � � �

I find HAL to be useful in my job � � � � � � �

I find it easy to get HAL to do what I want it to do � � � � � � �

206

Part B – Your Work Environment

This part of the questionnaire asks you to consider what impact the new HAL system may have had on both your job tasks and work procedures.

The first section of Part 2 asks you to consider how the new system may have effected the way you go about your job. Try to consider the tasks and functions of your job – and consider if these have changed with the introduction of HAL.

1 2 3 4 5 6 7Disagree Strongly

Disagree Disagree Slightly

Neutral Agree Slightly Agree Agree Strongly

Now (Post-implementation) Before (Pre-implementation)

1 2 3 4 5 6 7 1 2 3 4 5 6 7

The job requires me to use a number of complex or high level skills

� � � � � � � � � � � � � �

The job is arranged so that I can do an entire piece of work from beginning to end

� � � � � � � � � � � � � �

Just doing the work required by the job provides many chances for me to figure out how well I am doing

� � � � � � � � � � � � � �

The job allows me to use a number of complex or high level skills

� � � � � � � � � � � � � �

The job is one where a lot of other people can be affected by how well the work gets done

� � � � � � � � � � � � � �

The job gives me a chance to use my personal initiative and judgement in carrying out the work

� � � � � � � � � � � � � �

The job provides me with the chance to finish the pieces of work that I begin

� � � � � � � � � � � � � �

After I finish a job I know whether I performed well

� � � � � � � � � � � � � �

207

The job gives me considerable opportunity for independence and freedom in how I do the work

� � � � � � � � � � � � � �

The job itself is very significant and important in the broader scheme of things

� � � � � � � � � � � � � �

208

Participant Demographics

To finish off, the questionnaire asks some simple details about yourself. Remember that none of this data can or will be used to identify you as an individual. These details are for statistical analysis only.

Gender Female Male

�

�

Number of years in current work group

Less than 1 1-2 3-4 5-10 11+

�����

Age 15 – 19 20 – 29 30 – 39 40 – 49 50 – 59 60+

������

Hierarchical level

Director / Board Member Senior Management Middle Management Supervisor Team Member

�����

Number of years employed by ERUDIO

Less than 11-2 3-4 5-10 11+

�����

Number of subordinates

0 1-2 3-4 5-10 11+

�����

Job Grade AO2 � AO3 � AO4 � AO5 � AO6+ �

Average HAL Usage Per Day (Hours)

0 � 1 – 2 � 3 – 4 � 5 - 8 �

Previous Usage Per Day (Hours)

0 � 1 – 2 � 3 – 4 � 5 - 8 �

Thank you for participating in this study, your involvement is greatly appreciated. If you require any additional information please contact the

principal researcher using the details below.

209

Appendix A.03Social Network Interview Form

210

HA

L Fu

nctio

nalit

y an

d W

ork

Con

text

Wor

k C

onte

xt In

tera

ctio

n Q

uest

ionn

aire

This

shor

t que

stion

naire

is d

esig

ned

to h

elp

you

iden

tify

who

you

inte

ract

with

on

a re

gula

r bas

is w

hile

car

ryin

g ou

t you

r job

. Th

ere

are

thre

e sc

enar

ios t

hat w

e w

ould

like

you

to c

onsid

er.

Toge

ther

thes

e th

ree

scen

ario

s will

pro

vide

an

accu

rate

pic

ture

of

the

wor

k co

ntex

t with

in w

hich

you

ope

rate

.

To h

elp

you

in y

our c

hoic

es w

e ha

ve tr

ied

to in

clud

e th

e na

mes

of t

hose

peo

ple

you

are

mos

t lik

ely

to e

ncou

nter

as p

art o

f you

r jo

b. H

owev

er, w

hile

eve

ry a

ttem

pt h

as b

een

mad

e to

incl

ude

the

mos

t lik

ely

peop

le in

the

lists

belo

w, fe

el fr

ee to

incl

ude

othe

rs

not l

isted

that

you

dea

l with

on

a re

gula

r or f

requ

ent b

asis.

Impo

rtan

t: A

ll in

divi

dual

res

pons

es a

re c

ompl

etel

y co

nfide

ntia

l, ar

e to

be

used

onl

y fo

r th

e co

mpl

etio

n of

the

PhD

res

earc

h an

d w

ill n

ot b

e re

leas

ed in

any

form

out

side

the

requ

irem

ents

of t

he th

esis.

For a

dditi

onal

info

rmat

ion

refe

r to

the

“HA

L Fu

nctio

nalit

y &

Wor

k C

onte

xt s

tudy

con

sent

form

” or

alte

rnat

ivel

y co

ntac

t Gle

n M

urph

y on

07

3314

806

1 or

via

gdm

urph

y@op

tush

ome.

com

.au

211

1. R

esou

rce

Dep

ende

nce

In m

ost w

ork

plac

es p

eopl

e de

pend

on

each

oth

er to

a c

erta

in d

egre

e to

get

thei

r job

s do

ne.

This

sec

tion

asks

two

ques

tions

. Pa

rt A

. ask

s yo

u to

thin

k ab

out w

ho d

epen

ds o

n yo

u fo

r req

uire

d re

sour

ces,

Part

B. a

sks

you

to c

onsi

der w

ho y

ou d

epen

d on

to m

ake

sure

you

can

fulfi

l you

r job

re

quire

men

ts.

In th

e bo

xes

prov

ided

we

ask

you

to c

onsi

der

wha

t peo

ple

in y

our

wor

k co

ntex

t dep

end

on –

them

fro

m y

ou a

nd y

ou f

rom

them

. Sp

aces

are

pr

ovid

ed fo

r you

to in

dica

te th

e ty

pe o

f thi

ng p

eopl

e de

pend

on,

such

as a

dvic

e, a

ppro

vals

, res

ourc

es, a

nd su

ppor

t.

“Res

ourc

es”

refe

rs to

ele

men

ts su

ch a

s fun

ds, h

uman

reso

urce

s, tim

e ,m

ater

ial s

uppl

ies o

r suc

h th

ings

as i

nfor

mat

ion

or re

ports

.

“Adv

ice”

refe

rs to

adv

ice

give

n ab

out w

ork

rela

ted

activ

ities

, sol

utio

ns to

pro

blem

s or g

uida

nce

on su

ch th

ings

as p

olic

y an

d pr

oced

ure.

“Sup

port

” re

fers

to v

erba

l app

rova

ls, p

ositi

ve re

info

rcem

ent o

f ide

as, o

r vis

ible

end

orse

men

ts fr

om im

porta

nt o

r hig

hly

rega

rded

peo

ple

in y

our

orga

nisa

tion.

Use

the

boxe

s to

indi

cate

a le

vel o

f dep

ende

nce

on o

r by

that

per

son

(hig

h =

3, m

oder

ate

= 2,

low

= 1

). A

HIG

H le

vel o

f dep

ende

nce

wou

ld

sugg

est t

hat t

he r

esou

rces

pro

vide

d by

that

per

son

are

esse

ntia

l for

you

to c

ompl

ete

your

job

resp

onsi

bilit

ies,

a L

OW

leve

l wou

ld su

gges

t th

e re

sour

ces p

rovi

ded

mak

e yo

ur jo

b ea

sier

, but

are

not

ess

entia

l.

212

A.

Who

dep

ends

on

YO

U

Thin

k ab

out t

he p

eopl

e de

pend

ing

on y

ou to

get

mat

eria

ls, a

dditi

onal

reso

urce

s, ad

vice

, or a

nyth

ing

that

hel

ps y

ou g

et y

our j

ob d

one.

As

befo

re,

incl

ude

any

rele

vant

nam

es n

ot li

sted

in th

e sp

aces

pro

vide

d. I

mpo

rtant

– re

mem

ber t

o in

dica

te th

e le

vel o

f dep

ende

nce

by p

uttin

g a

3 (H

igh)

2

(Mod

erat

e) o

r 1 (L

ow):

R

emem

ber:

Thi

nk a

bout

thos

e “i

nfor

mal

” co

ntac

ts (e

.g. p

eopl

e ou

tsid

e yo

ur fo

rmal

repo

rting

stru

ctur

e, s

uch

as y

our s

uper

viso

r or t

eam

lead

er)

that

dep

end

on y

ou to

ach

ieve

thei

r job

goa

ls.

“Who

inte

ract

s with

you

obt

ain

the

requ

ired

reso

urce

s (in

form

atio

n, d

ata,

adv

ice,

mat

eria

ls, a

ppro

val)

they

nee

d to

ca

rry

out t

heir

dut

ies/t

asks

/job

?”

Type

of R

esou

rce

3 =

Hig

h D

epen

denc

e

2

= M

oder

ate

Dep

ende

nce

1

= L

ow D

epen

denc

eN

ames

rem

oved

for

confi

dent

ialit

y re

ason

sA

dvic

e

Res

ourc

es

Adv

ice

R

esou

rces

Adv

ice

R

esou

rces

A

dvic

e

Res

ourc

es

Adv

ice

R

esou

rces

A

dvic

e

Res

ourc

es

Adv

ice

R

esou

rces

A

dvic

e

Res

ourc

es

Adv

ice

R

esou

rces

A

dvic

e

Res

ourc

es

Adv

ice

R

esou

rces

A

dvic

e

Res

ourc

es

Adv

ice

R

esou

rces

A

dvic

e

Res

ourc

es

Adv

ice

R

esou

rces

A

dvic

e

Res

ourc

es

Adv

ice

R

esou

rces

A

dvic

e

Res

ourc

es

Adv

ice

R

esou

rces

A

dvic

e

Res

ourc

es

213

B. Y

our J

ob

Thin

k ab

out t

he p

eopl

e yo

u de

pend

on

to g

et m

ater

ials

, add

ition

al r

esou

rces

, adv

ice,

or

anyt

hing

that

hel

ps y

ou g

et y

our

job

done

. Im

porta

nt

– re

mem

ber t

o in

dica

te th

e le

vel o

f dep

ende

nce

by p

uttin

g a

3 (H

igh)

2 (M

oder

ate)

or 1

(Low

). R

emem

ber:

Thi

nk a

bout

you

r “in

form

al”

cont

acts

(e

.g p

eopl

e ou

tsid

e yo

ur fo

rmal

repo

rting

stru

ctur

e, su

ch a

s you

r sup

ervi

sor o

r tea

m le

ader

) you

use

to a

chie

ve y

our j

ob g

oals

.

“Who

do

you

inte

ract

with

to o

btai

n th

e re

quir

ed re

sour

ces (

info

rmat

ion,

dat

a, a

dvic

e, m

ater

ials,

app

rova

l) in

ord

er

for y

ou to

car

ry o

ut y

our d

utie

s/tas

ks/jo

b ?”

Type

of R

esou

rce

3 =

Hig

h D

epen

denc

e

2

= M

oder

ate

Dep

ende

nce

1

= L

ow D

epen

denc

eN

ames

rem

oved

for

confi

dent

ialit

y re

ason

sA

dvic

e

Res

ourc

es

Adv

ice

R

esou

rces

Adv

ice

R

esou

rces

A

dvic

e

Res

ourc

es

Adv

ice

R

esou

rces

A

dvic

e

Res

ourc

es

Adv

ice

R

esou

rces

A

dvic

e

Res

ourc

es

Adv

ice

R

esou

rces

A

dvic

e

Res

ourc

es

Adv

ice

R

esou

rces

A

dvic

e

Res

ourc

es

Adv

ice

R

esou

rces

A

dvic

e

Res

ourc

es

Adv

ice

R

esou

rces

A

dvic

e

Res

ourc

es

Adv

ice

R

esou

rces

A

dvic

e

Res

ourc

es

Adv

ice

R

esou

rces

A

dvic

e

Res

ourc

es

Than

k yo

u fo

r par

ticip

atin

g in

this

stud

y, y

our i

nvol

vem

ent i

s gre

atly

app

reci

ated

.

Plea

se b

e ass

ured

that

all

the i

nfor

mat

ion

cont

aine

d w

ithin

this

docu

men

t will

rem

ain

confi

dent

ial a

nd a

ny d

ata

resu

lting

fr

om th

e st

udy

cann

ot id

entif

y yo

u in

divi

dual

ly.

If y

ou re

quir

e an

y ad

ditio

nal i

nfor

mat

ion

plea

se c

onta

ct th

e pr

inci

pal r

esea

rche

r usin

g th

e de

tails

bel

ow.

214

215

Appendix A.04Survey Consent Form

216

HAL Functionality & Work Context

Large scale, organisational wide information systems are becoming increasingly prevalent in the business environment. However, little is known about what impact these systems have on the operation and function of organisations. [Insert organisation name] has agreed to be involved in research aimed at better understanding how these systems effect the way individuals interact before and after they are implemented. This form accompanies the information pack that you will have received with the questionnaire regarding research being carried out in this area.

The researcher conducting this research abides by the principles governing the ethical conduct of research and, at all times, avows to protect the interests and confidentiality of all participants. By signing this form you will indicate that:

I have discussed with the researcher the aims of the research being undertaken;

I have read and understood the information package;

I have been informed as to how the data collected will be used, stored and distributed;

I understand any reports generated as a result of the data collection process will not identify myself as an individual;

I understand all data collected will remain anonymous to all parties involved except the researcher and will not be disclosed unless additional permission is received from myself;

I understand that I am free to withdraw from the study at any time, without comment or penalty;

I understand that I may contact the secretary of the University Human Research Ethics Committee on 3864 2902 if they have concerns about the ethical conduct of the project; and

That I agree to participate in the research conducted as part of the study entitled “Social network theory and enterprise wide information systems” undertaken as part of a PhD (Management) degree for Glen Desson Murphy.

You may direct any further queries and further questions to Glen Murphy at the QUT School of Management on 3314 8061, or via e-mail at gd.Murphy@qut.edu.au . Thank you for your participation in this project.

Tear off and retain for your own records---------------------------------------------------------------------------------------------------------------------------

Return this section to the researcher

Participant Name:

Signature: Date: / /

Researcher: Glen D. Murphy Supervisor(s): Dr Artemis Chang School of Management Dr Kerrie Unsworth Faculty of Business Queensland University of Technology

217

Appendix A.05

Post-Hoc Interview Notes

218

1. Senior Management (AO6 +); COUNTRY; scheduled interview (3 months post)

- Expectations of system performance “not there”

- Greater involvement of users in process design / change required

- Related an incident of data theft whereby CITY users adopted a COUNTRY template, overwrote the material and effectively deleted “stole” COUNTRY’s data

- Needed more people trained rather than more people as trainers [note: this point was echoed 6 months later at CITY in much the same phrase]

- Processing times seem slow, clients are waiting for outputs – indicated that a possible reason for this was due to the fact that the system in this case was running on a “mini server” until the remaining sites went live [further investigation of this point with the local implementation manager and the overall project manager failed to support this information, indicating that this may have been a “placating white lie” established by middle and senior managers at a local level]

- Indicated that there was a proposed restructure being considered that involved the establishment of “centres of excellence” where exiting locations would move from the current multi-tasked model (each centre completed all tasks to be done) to one where “Group D” would be a Client Service Centre, “Group A” would be a call centre and ‘Group C” a records processing centre. This restructure was justified in terms of job design (only briefly alluded to) but primarily to overcome training and user knowledge issues. Also indicated that faster processing times and a chance to be more involved in the new round of process re-design that would occur was also seen as a significant advantage.

- Noted that the greatest amount of resistance and concerns about the restructure were from senior administrators and supervisors – no explanation was given for this. [Note: A senior admin’er interviewed later that day framed the resistance in terms of “the job being boring – doing one type of job all the time – others in their section concurred with the sentiment of the statement.]

2. Middle management (AO4 / AO5); COUNTRY; scheduled interview but declined to participate in survey etc. (3 months post)

- Stated that the HAL “roles” were a mismatch with previous job descriptions (designs) proceeded to show me a list of HAL roles (specific tasks or small groupings of tasks performed by various classifications of users – appeared to be an extensive list of task statements that ordinarily in a JD would be grouped into larger task descriptions or not provided in such detail – if at all. Indicated that in several instances previously established jobs typically cut across HAL roles often with varying permission levels. The implication of this was that people [depending on their permission level set by head office] were unable to do the job they used to and relied on others to do certain steps.

- Previously jobs were designed around a logically, rational collection of activities. The design of HAL and the separation of tasks into “HAL roles” means that people may not be able to carry out the same set of activities due to aspects taken out of their job - for example done by others, the system or due to a lack of authority (password controlled access to system functionality).

- noted that the password controlled job role issue was compounded by small regionally isolated teams with only one or two people to carry out all that had to be done.

219

- noted that not one new job description had been sighted with the new role delineations as per HAL

- believed that system implementers hadn’t taken into account the importance (value?) placed on accountability and responsibility that people place on their job responsibilities. Stated that “people get de-motivated when they can’t see their job from point A to point z”.

- Gave examples of where HAL struggled to deal with the specific nature of TAFE such as “reverse accreditation” with CQU.

- Was aware of some staff may leave due to stressors placed on them resulting from HAL

3. Team members (AO2 & 3) COUNTRY (general conversation) 3 months post implementation

- Processing time was slow

- struggling to match the old system & businesses processes with the HAL processes

- noted that they were told that they had a “vanilla system” and perceived this as a root cause of the problem in that not enough money was spent on the program from the beginning

- noted that the CSC (Client Service Centre – customer contact) in particular was struggling to cope with the increased service demands

- noted that overtime had increased [verified by management] due to processing increases – gave several examples of increased key strokes per task as compared to the previous system.

4. Supervisor / super-user; CITY; scheduled interview (29/4/2005)

Q – not what but how?

- used the word “complex” and noted the belief that people had to undergo a “concept shift” and that many were still operating as if they were using the old system – for example still coming to terms with the notion of automated transactions in relation to the production of an award [note: this was touched on by user#7 - 7/5/2005] whereby a certificate is now produced automatically when several other elements are completed (e.g payment of fees and entry of marks / subject grade into the system).

- Noted a key aim of HAL was to standardize ERUDIO processes throughout all locations and that as a result of HAL certain elements of the business had become more centralized.

- Raised a belief that task orientated groups were better supported by HAL rather than multi-tasked / interdependent teams [Note: refer to COUNTRY comment regarding isolated teams]. Indicated that this was due to the tendency of HAL to better enforce permission protocols and its tendency to segment work tasks. Indicated that non-demarcation / multi-tasked teams suffered a conflict of security classifications [Note: Would this mean reduced levels of UA and a greater incidence of work arounds?]. Voiced a belief that HAL tended to compartmentalize job tasks – therefore “task-based’ teams were better able to focus on their “role” and their “tasks” [Note: need to identify what groups may be task vs team based].

- Raised the possibility of problems faced by task based environments when people leave and the difficulty of succession planning due to an increasingly narrow skill base centred around system design

220

- Raised the idea of a “culture of compartmentalisation” [did not indicate wether this culture was forced down by the system and management or whether it had emerged in response to the system changes]

- Felt that many staff didn’t know “what happens next” as a result of this culture

- Indicated that multi-disciplinary teams such as theirs were better able to understand the system and the process flow (newly formed, centralised unit)

- Felt that “work arounds” were beginning to exist – that people are developing contacts to people who have the training (knowledge/advice) or the security clearance (resources / power). [Note: this individual had already discussed with me soon after go-live that he had obtained a far higher security clearance from the implementation team due to their prior relationship and acknowledgement as a superuser – had stressed the need for this to resolve problems in several functional areas].

- noted that certain areas and procedures had been taken off teams and relocated to specialist teams (such as the DP unit) but access remained “open” for the teams, they just didn’t know it (security by ignorance)

- indicated that a possible reason for the increase in resource ties was due to the compartmentalisation of tasks [Note: concurs with user #2 - COUNTRY account of HAL roles]. Felt that this had no real impact on power distributions and in-fact had suggested that powerful people may have consolidated power – particularly if their previous standing & position had afforded them greater training, access and information.

- Discussed the lack of clarity about business processes, especially HOW people do their job, all you see is the output – goes to the need to understand the process throughout the organisation and that the lack of understanding, along with a more complicated process leads to an accumulation of errors that are amplified the further along the chain it goes.

- Noted that the previous system was far more “forgiving” of errors” and the tendency of HAL to enforce sign-off protocols and authorisations had highlighted a “culture of lazyness” within management in relation to signing off and checking work etc. [Note: this could also be an example of HAL forcing a reassertion of managerial prerogative and preventing legitimate attempts by management to devolve decision making / delegation etc. down the line – this element is discuss by user #7;CITY, below). However remarked that HAL has highlighted problems in the [OLD SYSTEM] (e.g. management control) and deficiencies in organisational planning.

- Also noted that the intended aim of HAL (facilitate standardisation across all locations) would be unlikely due to history, culture and having teachers as managers (as opposed to professional managers) wanting to do things “their own way” – “everyone is encouraged thru work arounds to maintain their individuality”

4b. Follow up interview (phone) 13/9/2005

- Efficiency: where are they getting the info from ? noted that the “pilot” status of the site offered them far more opportunity to have trained people available for consultation, however did note that they (the interviewee) were one of about 5 or 6 people that had significant amounts of system knowledge and were reliable sources of info

- Also noted the possibility that people early on were operating on “blind faith” assuming what people told them was true and acting on it [i.e. if more than one person says the same thing then I’ll go with that]

221

- Noted the trend towards centralisation throughout the institutes – acknowledged the fact that this was driven at least in some degree by HAL design

- AO issue: used the term “compartmentalised knowledge” and noted that it was reinforced by a mentality of “you don’t get training unless you do that job” no real formal process to apply for task training therefore preventing people shifting roles. – noted that the future of TAFE employees would tend to take on very specialised career paths [note: this may also have implications for the employee for their marketability in the employment marketplace – digital handcuffs??]

- Noted that even the process of informal knowledge sharing (OnTheJob training) between users to build up skill areas may not solve problem due to lack of hands on practice due access protocols preventing a log on [:>) enter the workaround, stage right !]

- “not so easy to write a position description any more as it contrasts the generalist approach adopted in Faculties

5. Team member; CITY; unscheduled interview (29/4/2005)

Discussed about “anticipation”

- commented that “if people are told about “big changes” then they’ll anticipate them and be unsure. If we had been told that ‘you’re doing the same role/task, just slightly differently’ it would have allayed our fears”

- talked about using alternate contact points to find out things they hadn’t been trained in. discussed the importance of being able to depend on each other to get through – “take lots of breaks and come back to it later”

- training about our areas – need to know what you are responsible for (felt that some stuff wasn’t trained or know that was really needed. [Note: could also indicate role / task conflict as reported by others and that roles have been re-defined by users ?]

- Had to establish links with people downstairs about how to do “this ‘n that”. Also indicated that the team had “followed the chain of communication provided” [Note: indicates the use of a “forced” and “emergent” network to get required resources/information]

Q. Would you say your area is tasked based or multi-tasked based ?

- responded that each (approx 4 people) have individual tasks but at least 2 people share similar roles so they are able to “work as a team”

- complained of no overview provided as to where “things fit in”

6. Senior manager; CITY; scheduled interview (7/5/2005)

- Indicated early on in the meeting that they (CITY senior management/ directorship) had advised [Government department] and the ministers office that the planned rollout to other sites should NOT occur – rollout went ahead anyway. [Note: Email correspondence from union representatives at that time to the researcher indicated that a number of activities driven by the union were ongoing – the e-mails also alleged management bullying, coercion and staff stress levels as a result of the political and management pressure to “make the system work”]

222

- Training cited as a key issue as not all users were trained in all the areas they needed – cited time as a key factor in this decision rather than other resources such as money (although acknowledged as a consideration) or available trainers

- Suggested that some people demonstrated reduced acceptance not because of “lack of control” but because they felt a “lack of comfort” about the new procedures and requirements driven by the system

- Indicated that the system in general required more steps to be undertaken to complete tasks (e.g. the way things are done now). Gave the example of instances where students had been enrolled twice. Whereas the old system required a fairly straight forward deletion of the duplicate record HAL requires a total of 84 key strokes to rectify and approx 20mins for a user to complete. This is further compounded by the ease by which duplication errors can occur – cited an example of 200 at CITY site in one semester (approx 60+ hours of rework). Also gave the example of enrolment forms requiring far more information than the previous system

- Far more data entry points, and the data entry / process has to be accurate, “no short-cuts exist”

- On the positive side far more reports are now available but expressed frustration in that the contents of those reports and “what they could give us” were unknown. Indicated that a staff member had been allocated to find out what reports were available and what they contained. Knew of 2000 various reports – “just trying to find out what they all do”. However also cited concerns over the accuracy of the reports they were using. Cited an example of “how many international students were enrolled in a course – 3 different reports, three different answers [Note: see #4’s point made about error accumulation].

- Didn’t feel that the increased security protocols / password access was such a big issue, and mentioned that increased control over security was a positive move in any case.

- Tended to focus on the inability of the system to perform processes required by the organisation. Discussed the loss of [Client name] as a corporate customer because the automated invoicing system didn’t allow additional information to be put on specific invoices. Also voiced concern about the centralisation of invoicing and consequently the reduced flexibility of the system to adapt to local requirements such as the [Client name] case (worth approx $100’000 per month/semester??). Indicated the system processes as they stood were struggling to address the organisations diverse client group and payment options (direct fee pay, subsidised fees; private provider, short course; trainees) “HAL can’t recognise some things”.

- Also gave an example of the increased interaction between sites that wouldn’t have happened under the old system whereby fees outstanding are addressed first regardless of where the fees are paid and what for. Under the old system each site was responsible for its own fee management and recovery.

- Felt that at least in the case of the issues discussed above it was management that bore the greatest burden due to increased scrutiny from above [Note: See #4’s comment about “lazy management” and HAL exposing the amount of work previously done by lower ranks to support this culture].

223

- Suggested that at T2 people still had “[OLD SYSTEM] in the head” and that there was a lot of stress at T2. Cited in particular the Implementation team manager that had to take stress leave following the projects “completion” [Note: Anecdotal evidence from anonymous sources suggested that this individual was a key source of management bullying in their previous role and possibly in this one]. Noted that the organisation had acknowledged the problem of stress to the point that they had agreed to appoint counsellors to each major team to help staff with the problems experienced [Note: might it be able to get data on how well/much the services were used for each group ?]

- Didn’t feel that relationships within the group had changed significantly – however began to discuss the “performance issues” surrounding the CITY group before and during the implementation. Indicated that a restructure within the group was planned regardless of the system due to “task based” productivity issues and a lack of communication between group members.

A cited aim of the change was to reduce task isolationism and to “improve job enrichment” [Note: this account completely contrasts user#7’s account and to a certain degree #4’s comments regarding compartmentalisation]. Felt that the implementation assisted this process as it “forced everyone to communicate with others because they all had similar problems” – bonding through adversity. The standardisation of processes also facilitated the aim of the restructure (this comment seemed to contradict the others).

The senior supervisor of the group was also provided with leadership and management training [Note: previous discussion with team members about the interviewee had indicated he was not well liked due to the increased scrutiny he placed on team members; it is also known that shortly after the implementation (possibly after the discussion about their leadership skills) the senior supervisor applied for a transfer out of the role- and subsequently got it 3 -4 months after this interview].

7. Team member; CITY; scheduled interview (7/5/2005)

- stated the belief that HAL creates jobs – for example #4’s newly created multi-disciplinary unit that was not required previously; and that the system as created more volume of work – indicated that they felt [OLD SYSTEM] was more “streamlined”.

- When asked about “team vs task” stated that this was team based but that HAL creates “specialists” but complained that some people got trained in all areas while others only got trained in some.

- A key theme that was stressed was that HAL was more complex in its requirement for multiple sub-tasks to be completed – sequential steps rather than the old system which was more of a “all-in-one” procedure. Gave an example of printing out a class role – couldn’t do this any more [Note: the reasons for this appeared to be more involved than just sequential processing and involved security controls, role changes, training and system knowledge. Interesting to note however that the perception appeared to be that this was a failing of the system rather than anything to do with management and their control over these other issues].

- The process driven / “steps to do’ nature of the system has increased the complexity of the system. Noted that it was more complicated but there was “less to do”. Indicated that more staff had been appointed since the system had gone live – everybody does a little bit or relies on other areas such as the DP unit (a centralised work unit responsible for work the interviewee and their group used to do)

224

- Noted that the work was less challenging and required less thinking (reduced cognitive element). Summed it up by saying “less work, less challenging, less responsibility and “no say” in the work that was being done.

- Began to discuss the characteristics of their role prior to HAL. Indicated that their role, position, knowledge and contacts allowed their a high degree of control and responsibility, particularly in relation to work done by junior members and work submitted to management [Note: appears to support the comments made by #4 regarding management signoffs etc.].

- Observed that some roles had been split up and others had been maintained [Note: again, no overt reference to the role of management in this decision, and discussed in context of the system]. Stressed that all group members should be multi-skilled and that currently some roles were blurred with some people trained and authorised and, others not. Pointed out that under the previous system there was clear task delineation and that [OLD SYSTEM] roles were clearly defined [indicates a task based group?] suggested that the team approach hadn’t been followed through.

- Complained of a lack of responsibility due to the need to sign-off the process with management [Note: See #4’s point about management]

- “with [the previous system] I knew where to go, I knew people to get help and “if you scratch my back I’ll scratch yours” old the old contacts were gone “cause it’s all different”

- [Note: while no other evidence supports this point as yet on the face of it would appear that this interviewee fits the profile of the network dependence argument – system protocols and design have reduced their power and their job enrichment and consequently is vocal about the system’s failings.]

8. Team member; CITY; scheduled interview (12/5/2005)

- duties are the same but took a long time to get used to the “HAL way”

- shared a perception that HAL was a database designed for a corporate environment and made to fit into a tertiary environment

- “HAL has more screens but I’ve got used to what you’ve got to do to get myself out of trouble”

- It’s now fun to train new members on the aspects of HAL they need to know

- Not afraid of it any more, but still maintain that the old system was easier

- “Not good for the traineeship area” – an area highly specialised to the organisation, involving the provision of training to apprentices and corporate trainees over a three-four year period

- Felt “out of my comfort zone” “very stressed” but have developed more of a “skill set”, noted that the “favourites” option helps the navigation of the system [could possibly indicate a “rote learning approach to the function rather than understanding the processes behind the function and role]

- In terms of the increase in advice ties #4 was always tied up with other duties so other sources of advice were needed

- Other areas were no help [unsure of this comment, it may have referred to the need to look for advice internally as other sections of the organizations were unable to assist]

- However went on to identify 2-3 people that previously had no ties or very weak ties as valuable sources of information and advice – such as an HAL trainer and another contact outside the faculty

225

- “In the downstairs team everyone was talking”

- The Geographic’s of the building hampered communication without #4 operating in a brokerage role

- Commented that the responsibilities are largely faculty driven, not system driven

- Indicated that tasks and responsibilities were diploma based

- The Head Of Departments / supervisors were changing on a regular basis, caused further pressure. Indicated that certain members of the faculty were put under lots of pressure to “just make it work”

- Resource dependence came with advice

- Stress levels were really high – performance levels were expected to remain the same

- Indicated that there was a significant amount of “water cooler” discussion about HAL

9. Team member; CITY; scheduled interview (12/5/2005)

- Q has the way you do your job changed – Ans. Yes definitely

- Commented that the nature of the system and the job means that you can’t finish a job in a complete process. The system now requires the task to be completed in stages that frequently requires another part of the system to verify each stage or to provide a required piece of information etc. Can take three or four days to complete a task rather than all in one hit like I used to be able to do.

- Indicated that this process was a source of frustration and highlighted and example of “delta details” that are regularly not available when required and have to ring someone to find them, requiring the file to be put aside

- So there are several steps you have to go through which are clear but you can’t do them all at once.

- Cited evidence of results being changed without knowing who or what has done it and suggests that results were changed to better fit the requirements of the system

- Perceived increase in paperwork and cited a frustration with the uncertainty of not knowing whether the paperwork is correct, indicating that the procedural requirements were changing regularly

- Only received an hour or so of training therefore the advice nw explosion was not unsurprising and indicated that the trainer was still being contacted.

- Indicated that they were being contacted on a frequent basis because she was now regarded as an “expert” in a particular system function

- Indicated that the advice sharing was “good” but difficult when the giving of advice was accompanied by doubt about the accuracy / correctness of the advice

- Big learning curve

- Indicated that the previous group leader and acknowledged super user was still contacted as some info was still unclear and information on different screen was unclear and confusing

- Increase in signing procedures since HAL

226

- HAL has changed their role slightly, it has increased partly as a result of the AO3 (group leader) being removed to a centralised unit (due to HAL proficiency) components of their role have been split up and passed on to the team members

- Has moved form a 4day a week job to a 5day week job with very little time to discuss issues outside of HAL such as team development, direction of the group etc. Indicated that “free time” had reduced as a consequence of the system

- Indicated that there were more internal audits and reports to be produced in order to monitor the increased error rate / HAL problems [Note: See BM’s comments below concerning this issue]

- Expressed an opinion that HAL had affected MNGT less than their peers (A02&3) and that management had little understanding of the complexities surrounding the system and the associated changes

- “you can just produce results on demand”….”but that’s what management wanted” for example an often repeated phrase from management was heard when asking for reports “there must me some way….”

- Management doesn’t have the knowledge of what HAL does and doesn’t do

- User #4 tended to cover both upstairs and downstairs, since they were removed interaction on a frequent basis between the two areas has decreased

- Paperwork / more work since HAL, reported that others feel the same way and didn’t have time to plan or communicate with each other like before

- People were put in at the deep end such as #10 (was moved upstairs into a new role after the 2nd round of data collection), there was no time for decent amount of on-the-job training ‘cause of pressures from the system and the fact that few people had the training to do the training (hehehe…Dilbert?)

- Suggested that more training (at this point) might not even help much (for additional areas not always used or needed )as transferability and knowledge retention is an issue

- Noted that had to rely more on other people now to “get things fixed” and this was compounded by the fact that you didn’t really know who to contact anymore to get things fixed – cited this as a major source of frustration.

10. Team member; CITY; scheduled interview (12/5/2005)

- Job didn’t change because didn’t really use [OLD SYSTEM] in previous role

- Felt that HAL was easy to use [Note: this subject was the youngest and newest member of the team, possibly indicating a greater acceptance of GUI technology and less entrenched attitudes towards the newly imposed system processes

- Advice sharing was evidence of people helping each other but commented that after a while the interruptions were “annoying”

Deeper Questions indicated…

- “that dependence is happening heaps more” …”have to wait for members of the team to finish their steps in the process”

- “In the past we didn’t rely on others as much” …referring to the interaction patterns in the Traineeships group prior to HAL

227

- Harder to get work done due to no HAL knowledge

- Developed a new set of contacts to find out info etc.

11. Team member; CITY; scheduled interview (12/5/2005)

- has changed role since HAL introduced [Note: this participant has operated in several roles in all locations of the network and hence has a unique perspective on the operations of the faculty]

- feels that HAL is actually easier to use due to the menu driven nature of [OLD SYSTEM] and the multiple screen GUI nature of HAL

- commented on the clearly defined areas that are defined by the HAL “roles” [Note: try and get a copy of HAL roles and an old JD for the appendix]

- When asked to clarify the difference between HAL roles and the “pre-HAL job duties” – commented that HAL has resulted in more clearly defined areas, and specific roles for people and tends to encourage “knowledge based areas” and people get to know “the main thing that they do”… “you don’t tend to venture out to other areas” – commented that she felt this was more effective for a number of reasons:

“More control over what’s happened” [from an audit trail perspective]

“can see what’s been done”

The system does provide more information / data

There are less people to deal with when identifying problems [as opposed to [OLD SYSTEM] where there were numerous people could be responsible or able to carry out a range of duties and therefore no clear audit trail]

- noted that with [OLD SYSTEM] that “people had free will to do what they wanted” but with HAL people are constrained

- Commented that HAL stops mistakes – because bad info affects a lot of things i.e in most instances if an error is made it fails to accumulate due to the process orientated nature of the system

- Made a comment that HAL is actually allowing the identification of errors that previously were not being picked up

- Believed that the institution “had more control now

- Observed that people don’t like the fact that they can’t get things done when they want them of if they need them done because with the old structure / system they were able to rely on established relationships – these are no longer relevant because people either are no longer responsible for that area, don’t have the authority or the training

- Also identified that “everyone is one the same playing field”, that power or influence afforded to them by knowledge of pre-HAL procedures (or contacts to get things done) was gone because overall people were now beginning from scratch in terms of knowledge of procedure and process

- Commented that the job classification system was changing to reflect an AO(role) culture [e.g. credit transfers] rather than the broad traditional numerical classification of AO2 which reflected the multi-skilled nature of the previous role design

228

- Commented about the lack of control resulting from the centralization of key areas such as the DP unit etc. “makes you wait…(less responsive to urgent requests)

- “Don’t know what your doing” because of the elements of the job that are centralised

- Commented on the development potential (or lack of) due to the insular nature of the skill set that the HAL roles tend to encourage; observed that it prevents movement sideways and up because you are not exposed to other areas of the business

- “not as busy as I was” due to: increased centralisation of key tasks; set in what you do; prevented from doing any more than your set role, if you want to do different duties you need a different job [Note: concurs with user #7’s account]

- Summarised the changes as:

A clearer defined system

You know exactly what specific person is doing in a specific role (“not 20 people doing the same role”)

More control over information / data

Clearer info for admin staff to find out about students etc, however admin have less control over the management of students and their enrolment etc.

12. Team member; CITY; scheduled interview (12/5/2005)

- considered that job change was more due to management

- tended to use [OLD SYSTEM] more – enquiries were easier (as opposed to HAL)

- [OLD SYSTEM] tended to be “all together”

Role vs Job Q

- described examples of where “jobs” had been split up into 3 or 4 roles performed by different people

- indicated that she would prefer to back to the original role structure given the choice. Asked why ? replied that “they don’t know my course like I do”

- indicated a problem about the AO2 / AO3 classifications and the work performed and “who does what”

- the change was initially difficult

13. Supervisor; COUNTRY; unscheduled interview (13/9/2005)

- DP’s are centralised as are a lot of other functions

- Every location has a designated task now (“GROUP D” –DP’s & TAS; “GROUP A” refunds & credit transfers; “GROUP D” Awards / results publishing; “GROUP B” archiving)

- Centralised approach may help smaller institutes

229

14. Implementation team member; COUNTRY; unscheduled interview (13/9/2005)

- locations such as “GROUP D” and “GROUP C” “responded to segregation better”

- segregation referred to the manner in which HAL forced a “narrower band of operations” [referred to elsewhere as functional and task isolation]

- noted that areas such as “GROUP A” had a wider “span of access” in relation to [OLD SYSTEM] – that people had access and the training to perform a wider range of functions (i.e. a multi skilled team – see #4’s notes about team type and ease of adoption) and that may have led to problems

- also indicated that leadership and management style at “GROUP D” and “GROUP C” helped facilitate the process [“GROUP A” mngt – damned by NO praise?]

- “GROUP A” maybe had more AO3 roles which are broader by definition and therefore why maybe different adoption pattern [correlation between user type and AO?]

- there’s the HAL roles and there’s what people “could do”.

- Indicated that similar users (similar segregations) had been put together at “GROUP D” (maybe one reason for the Efficiency result)

- Previous system allowed a wider span of access and fitted better with the AO structure

- Concurred with comments expressed in CITY regarding the conflict between the manner in which work was categorised under the AO system and the way that HAL required=s jobs to be designed and categorised

- Commented that the AO structure is too broad when considering HAL roles and designating people to those roles

- HAL will change titles and job classifications

- Task isolation “segregation” issue has caused “angst”

- People were used to having “carte blanch” with [OLD SYSTEM] – doesn’t happen any more, blocked by knowledge and access protocols

- Concurred with the assessment of power being lost to the new system that knowledge of the system was power and that the new system removed that power source completely

- However – when prompted about possible examples of resistance indicated that very little existed and that reaction was more about confusion and not understanding the system logic (or lack of?) about process flows and new business rules. “they couldn’t see the logic behind segregation”

- Lack of available sources offered as a possible reason for the eff result, indicated that the training strategy employed was poor and that training was insufficient for user’s needs.

- Also noted that there were a set number of contacts that people were encouraged to seek out (formal)

- highlighted a Feedback mechanism – people were “chased up’ for errors in the system.

230

15. Supervisor; Country; unscheduled interview 13/9/2005

- indicated that the difference between “GROUP C” and “GROUP A” was “a staff issue, not an HAL issue”

- “its more to do with the way they operate up there” [became guarded at this point]

- “the communication channels down here [“GROUP C”] are a lot more open…and I’ll leave it at that”

- Indicated that “GROUP C” users were in the same office and not as spread as “GROUP A” [my personal observation was that in terms of office layout that “GROUP A” wasn’t terribly different to “GROUP D” – my personal feeling (without any real evidence) is that this user was referring more to the cohesion of the group and the lack of an enforced chain of command and communication channel – organic? In contrast to perhaps the more hierarchical, rigid structure of “GROUP A” – especially given the management there]

16. Senior manager (COUNTRY); Scheduled interview 14/9/2005

- emphasised that the differences between the two different types of implementation adoption were due to the “cultural issues” at “GROUP A” and “GROUP B” [the “GROUP B” emphasis was interesting as I had previously written off the no-change at that location as being down to a small, isolated team that had reached density saturation]

- also stressed the lack of interaction between the various users at “GROUP A” – even between those with similar roles

- hinted at a culture at “GROUP A” regarding their inability to manage conflict resolution and had difficulty in “resolving issues” [didn’t pursue this further even though I wanted to…]

- When prompted the subject indicated that “GROUP A” & “GROUP B” had more hierarchical structures with more established/rigid communication channels [later supported by user #17 indicating that requests for help, data, resources had to be put through management, even if people were at the same location]

- Indicated that “GROUP D” and “GROUP C” in particular had a strong cohesive nature

- Talked about the “family environment” and the social nature of the “GROUP C” locale that had been established by “GROUP C” management – attributed this culture to the increase in communication and resource sharing [interesting to note that without any prompting it was indicated that “GROUP C”/”GROUP D” had a successful implen whereas “GROUP A” and “GROUP B” had problems – that the increase in both comm. And resdep was the ideal outcome, an indication that things had gone as they should]

- Noted that “GROUP C”/”GROUP D” “made an effort to get along” and “communicate” that all areas communicate and get along and indicated that these groups demonstrated a willingness to resolve issues

- When asked about possible differences in roles prior to implen (and hence a possible reason for the different culture etc.) rejected the idea – confirmed that prior to implen all groups worked on a decentralised model and carried out similar functions to other locations.

- [note it was a user from “GROUP A” that indicated possible resistance to the centralisation plan based around job enrichment concerns]

231

- Subject indicated that they felt that some of the problems experienced (in general) were caused by the need to fit exiting business processes around the new system. Suggested that the opportunity should have maybe been taken to re-examine existing processes (in light of a good understanding of the system’s [Old System]abilities) and redesign the processes to achieve maximum outcomes. Indicated that some existing processes were large and possibly not required, and a product of numerous changes over time

- Indicated that the centralisation process may be more efficient from a management (of staff) and administrative perspective but made service delivery more difficult.

- Indicated that difficulties centred around accessing of information and the need to “go to other areas” to get information

- Complained about less personal contact and that you had no knowledge of a person’s workload, of their peaks and troughs, when they were under pressure etc. when requesting a job

- Also acknowledged that they had less management control over when (and how) the work (requests) got done

- Commented that people tended to lay the blame for problems at each centralised point (e.g. “it’s ok for you guys, you’ve got the DP unit there, we don’t) even though the point was stressed that they had no control over the DP unit either

- Acknowledged that users were yet to grasp the flow on effects of each user action – and that a move had started faculty wide to reduce the amount of “re-work” associated with errors caused by users not completing procedure that has a flow on effect

17. Supervisor; Country; scheduled interview; 14/9/2005

- indicated that the possible diff between “GROUP A” and other sites was due to the design of their buildings [as noted previously I personally observed no significant differences, especially given the layout, for example, of the City location]

- suggested perhaps that “GROUP D” had closely located workstations and all the users were pooled in one area as a reason for the increase in communication

- did indicate however that in “GROUP A” if you wanted something from another area you would go to your supervisor and go from there [this point tends to concur with the comments made by #16, #15 regarding the way in which “GROUP A” users communicated and interacted]

- used the same language as a city colleague to describe the centralisation of the DP function – “it’s been taken off us” – indicated that centres had volunteered for various roles

- agreed that HAL encouraged the “segregated” way of doing work, that access is limited and that you have to ring up other people to see different areas

- didn’t feel that management had benefited from the system as they had brought it in for the reports but that wasn’t working – its not doing what they bought it for

232

18. HAL trainer; Country

- indicated that the business process was stressed in the user training, but felt that many users interpreted it as an overview – expressed difficulty in explaining the training process. Indicated that much of the training lacked “real world application”, that it tended to be a “recipe for entering data” without reference to problem solving strategies if things went wrong or users encountered a problem

- struggled to identify why “GROUP A” and “GROUP C” may be different

- noted that after go-live “everybody felt abandoned” as trainers were expected to back to their operational roles immediately following go-live, therefore reducing their exposure to users requiring their assistance

- did acknowledge that “GROUP D” already had their teams structured in relation to access protocols [readiness for the change ?]

- indicated that the response to the system was due to management at each location

- commented that the lack of participation by supervisors (didn’t specify any location) which was voluntary, was one reason for any problems that may have been encountered. Felt that their lack of participation led to “a reduced appreciation and understanding about what was going on” [comment related to the integrated data flow management] observed that the supervisors tended to “stand back and wait”

19. Manager; Country

- noted the fact that “GROUP C” “has always had a culture of sharing information and networking”

- however also indicated surprise that “GROUP B” was different to “GROUP C” given that they were similar size campuses and teams

- hesitantly commented that “there is a culture at “GROUP A”

- suggested that the structure of “GROUP C” may have assisted the process, in that “all of the administrative staff are together” and “are not broken up” in contrast to “GROUP A” where admin staff are located in functional (teaching) areas rather than an administrative “pool”

- when prompted agreed with the observation that “GROUP A” was hierarchical in nature, that staff were encouraged to go through management when communicating with other areas whereas “GROUP C” staff only went to management if they had a problem (tended to sort it out amongst themselves)

- strong focus on training, gave numerous examples of where “proactive” training had taken place to ensure multi-skilling during peak times and when staff levels were low (the need to backfill)

- [this approach to training appear to be novel as it has not been observed anywhere else, at either site, in any location – a cultural anomaly ?]

- Acknowledged the issue of segregation and the impact it has had on AO classifications but went on to point out that the training strategy employed was reducing the impact of this

- Suggested perhaps that “moderate users” were those that used the system primarily for enquires and hence needed to use it but not enough to get trained in the area properly or use it frequently enough to become fluent with its operation

233

Appendix A.06

Country & City teams as by location (Advice and Resource Exchange merged; No external ties shown;

Dichotomised at 1)

234

Country Group A

T1 - 3 Months prior to Go-live

Usefulness: 4.67(.85) Density: .47 E-I Index: -.580Ease of use: 3.98(1.03) Centralisation: 29%

Country Group A

T2 – 3 months after Go-live

Usefulness: 1.91(.97) Density: .47 E-I Index: -.373Ease of use: 2.18(1.3) Centralisation: 23%

t p < .10; * p<.01; ** p<.001

235

Country Group B

T1 - 3 Months prior to Go-live

Usefulness: 3.25(1.76) Density: 1.08 E-I Index: -.5Ease of use: 2.63(.884) Centralisation: 48%

Country Group B

T2 – 3 months after Go-live

Usefulness: 2.17(1.58) Density: 1.07 E-I Index: -.143Ease of use: 2.08(.803) Centralisation: 31%

t p < .10; * p<.01; ** p<.001

236

Country Group C

T1 - 3 Months prior to Go-live

Usefulness: 4.71(.88) Density: .85 E-I Index: -.419Ease of use: 4.78(.73) Centralisation: 48%

Country Group C

T2 – 3 months after Go-live

Usefulness: 2.79(1.22) Density: 1.04* E-I Index: -.548Ease of use: 3.5(1.39) Centralisation: 38%

t p < .10; * p<.01; ** p<.001

237

Country Group D

T1 - 3 Months prior to Go-live

Usefulness: 4.6(.78) Density: .67 E-I Index: -.714Ease of use: 4.07 (1.1) Centralisation: 22%

Country Group D

T2 – 3 months after Go-live

Usefulness: 2.67(1.13) Density: .78t E-I Index: -.667Ease of use: 3.07(1.29) Centralisation: 23%

t p < .10; * p<.01; ** p<.001

238

City Group 1

T1 - 3 Months prior to Go-live

Usefulness: 5.05(.91) Density: .42 E-I Index: .111Ease of use: 4.47 (.73) Centralisation: .46%

City Group 1

T2 – 3 months after Go-live

Usefulness: 3.31(1.8) Density: .84** E-I Index: .097Ease of use: 3.30(1.32) Centralisation: 39%

t p < .10; * p<.01; ** p<.001

239

City Group 2

T1 - 3 Months prior to Go-live

Usefulness: 4.06(.13) Density: 1.27 E-I Index: .250Ease of use: 4.00(.00) Centralisation: 63%

City Group 2

T2 – 3 months after Go-live

Usefulness: 1.5(.66) Density: 2.33** E-I Index: .619Ease of use: 1.75(1.08) Centralisation: 19%

t p < .10; * p<.01; ** p<.001

240

City Group 3

T1 - 3 Months prior to Go-live

Usefulness: 4.93(1.08) Density: .753 E-I Index: .086Ease of use: 4.06(.66) Centralisation: 36%

City Group 3

T2 – 3 months after Go-live

Usefulness: 2.25(.90) Density: .925 E-I Index: .182Ease of use: 2.00(.95) Centralisation: 33%

t p < .10; * p<.01; ** p<.001

241

City Group 4

T1 - 3 Months prior to Go-live

Usefulness: 4.93(.74) Density: 1.54 E-I Index: .538Ease of use: 4.0(.41) Centralisation: 12.5%

City Group 4

T2 – 3 months after Go-live

Usefulness: 3.25(.54) Density: 1.56 E-I Index: .556Ease of use: 3.5(1.39) Centralisation: 25%

t p < .10; * p<.01; ** p<.001

242

243

Reference List

Andreou, A. & Boone, L. 2002. The impact of information technology and cultural differences on organizational behavior in the financial services industry. Journal of Intellectual Capital, 3(3); 248-261.

Anderson, B. Butts, C. & Carley, K. 1999. The interaction of size and density with graph-level indices. Social Networks, 21: 239-267.

Axtell, C. & Parker, S. 2003. Promoting role breadth self-efficacy through involvement, work redesign and training. Human Relations, 56(1): 113-131.

Aydin, C. & Rice, R. 1991. Social worlds, individual differences, and implementation. Information and Management, 20: 119-136.

Baldwin, T. Bedell, M. & Johnson, 1997. The social fabric of a team based MBA program: network effects on student satisfaction and performance. Academy of Management Journal, 40(6): 1369-1397.

Barley, S. 1990. The alignment of technology and structure through roles and networks. Administrative Science Quarterly, 35: 61-103.

Battacherjee, A. & Hirchheim, R. 1997. IT and organizational change: Lessons from client / server technology implementation. Journal of General Management, 23(2): 31-46.

Benbasat, I. & Zmud, R. 1999. Empirical research in information systems: The practice of relevance. MIS Quarterly, 23(1): 7-15.

Berenson, M. & Levine, D. 1996. Basic business statistics: Concepts and applications (6th ed.). New Jersey: Prentice Hall.

Bernard, H. & Killworth, P. 1977. Informant accuracy in social network data. Human Communication Research, 4: 3-18.

Bernard, H. & Killworth, P. & Sailer, L. 1981. Summary of research on informant accuracy in network data, and on the reverse small world problem. Connections, 4(2): 11-25.

Besson, P. & Rowe, F. 2001. ERP project dynamics and enacted dialogue: Perceived understanding, perceived leeway, and the nature of task related conflicts. The DATA BASE for Advances in Information Systems, 32(4) Fall: 47-66.

Bingi, P. Sharma, M. & Godla, J. 1999. Critical issues affecting an ERP implementation. Information Systems Management, 16(3): 7-15.

Boonzaier, B. Ficker, B. & Rust, B. 2001. A review of research on the job characteristics model and the attendant job diagnostic survey. South African Journal of Business Management, 32(1): 11-34.

244

Bonacich, P. 1972. Factoring and weighting approaches to status scores and clique detection. Journal of Mathematical Sociology, 2: 113-120.

Bonacich, P. 1987. Power and centrality: A family of measures. American Journal of Sociology, 92: 1170-1182.

Borgatti, S. & Cross, R. 2003. A relational view of information seeking and learning in social networks. Management Science, 49(4): 432-445.

Borgatti, S. & Foster, P. 2003. The network paradigm in organizational research: A review and a typology. Journal of Management, 29(6): 991-1013.

Borgatti, S.P., Everett, M.G. and Freeman, L.C. 2002. Ucinet 6 for Windows. Harvard: Analytic Technologies.

Borgatti, S. & Molina, J. 2003. Ethical and strategic issues in organizational social network analysis. Journal of Applied Behavioural Science, 39; 337-349.

Borgatti, S.P. 2002. NetDraw: Graph Visualization Software. Harvard: Analytic Technologies

Borgattti, S. Jones, C & Everett, M. 1998. Network measures of social capital. Connections, 21(2): 1-36

Borgatti, S. 2003. The state of organisational social network research today. Retrieved 14th January, 2006, from http://www.analytictech.com/mb874.

Borgatti, S. (In press) Identifying sets of key players in a network. Computational, Mathematical and Organisational Theory.

Boudreau, M. Gefen, D. & Straub, D. 2001. Validation in information systems research: A state of the art assessment. MIS Quarterly, 25(1): 1-16.

Brandyberry, A. Rai, A. & White, G. 1999. Intermediate performance impacts of advanced manufacturing technology systems: An empirical investigation. Decision Sciences, 30: 993-1020.

Brass, D., Galaskiewicz, J., Greve, H. & Tsai, W. 2004. Taking stock of networks and organisations: A multilevel perspective. Academy of Management Journal, 47(6): 795-817.

Brass, D. & Burkhardt, M. 1993. Potential power and power use: An investigation of structure and behaviour. Academy of Management Journal, 36: 441-470.

Brass, D. 1995. A social network analysis on Human Resource Management. Research in Personnel and Human Resources Management, 13: 39-79.

Brass, D. 1984. Being in the right place: A structural analysis of individual influence in an organization. Administrative Science Quarterly, 29: 518-539.

245

Brown, A. & Jones, M. 1998. Doomed to failure: Narratives on inevitability and conspiracy in a failed IS project. Organization Studies, 19: 73-88.

Brown, A. 1995. Managing understandings: Politics, symbolism, niche marketing and the quest for legitimacy in an IT implementation. Organization Studies, 16: 951-969.

Brown, J. & Quarter, J. 1994. Resistance to change: The influence of social networks on the conversion of a privately-owned unionized business to a worker cooperative. Economic and Industrial Democracy, 15: 259-282.

Brown, S. Massey, A. Montoya-Weiss, M. & Burkman, J. 2002. Do I really have to ? User acceptance of mandated technology. European Journal of Information Systems, 11: 283-295.

Bryant, F. & Yarnold, P. 1995. Principal-Components analysis and exploratory and confirmatory factor analysis. In L.G. Grimm & P.R. Yarnold (Eds.) Reading and understanding multivariate statistics. (99-136), Washington, APA.

Burkhardt, M. & Brass, B. 1990. Changing patterns or patterns of change: The effects of a change in technology on social network structure and power. Administrative Science Quarterly, 35: 104-127.

Burkhardt, M. 1994. Social interaction effects following a technological change: A longitudinal investigation. Academy of Management Journal, 37(4): 869-898.

Burt, R. 1992. Structural Holes. Cambridge, MA. Harvard University Press.

Burt, R. 1997. The contingent value of social capital. Administrative Science Quarterly, 42: 339-365.

Burt, R. 2000. The network structure of social capital. In R.I. Sutton & B.M. Staw (Eds.), Research in Organizational Behavior (pp.345-423).

Burt, R. 1999. The social capital of opinion leaders. Annals of the American Academy of Political and Social Science. 566: 37-54.

Burt, R. 2002. The social capital of good ideas. In M. Guillen, R. Collins, P. England & M. Meyer (eds.) The new economic sociology: Developments in an emerging field. New York: Russell Sage Foundation.

Burt, R. 2004. Structural holes and good ideas. American Journal of Sociology. 110: 349-399.

Cavaye, A. & Christiansen, J. 1996. Understanding IS implementation by estimating the power of sub-units. European Journal of Information Systems, 5: 222-232.

Casciaro, T. Carley, K. & Krackhardt, D. 1999. Positive affectivity and accuracy in social network perception. Motivation and Emotion, 23(4): 285-306.

246

Chan, P. & Land, C. 1999. Implementing reengineering using information technology. Business Process Management, 5(4): 311-324.

Chattopadhyay, P. Glick, W. Miller C. Huber, G. 1999. Determinants of executive beliefs: Comparing functional conditioning and social influence. Strategic Management Journal, 20: 763-789.

Chung, S. & Snyder, C. 2000. ERP adoption: A technological evolution approach. International Journal of Agile Management Systems, 2(1): 24-32.

Cohen, J. Cohen, P. Aitken, S. & West, L. 2003. Applied multiple regression / correlation analysis for the behavioural sciences. Mahwah, NJ: L. Erlbaum & Associates.

Conway, J. & Huffcutt, A. 2003. A review and evaluation of exploratory factor analysis practices in organizational research. Organizational Research Methods, 6(2): 147-168.

Cook, K. & Emerson, R. 1984. Exchange Networks and the analysis of complex organisations. Research in the Sociology of Organisations, 3: 1-30.

Cooper, D. & Emory, C. 1995. Business research methods (5th ed.). Boston, Irwin.

Couper, M. Traugott, M. Lamias, M. 2001. Web survey design and administration. Public Opinion Quarterly, 65; 230-253.

Couper, M. 2000. Web surveys: A review of issues and approaches. Public Opinion Quarterly, 64; 464-494.

Corston, R. & Colman, A. 2000. A crash course in SPSS for Windows. Oxford; Balckwell Publishers.

Cortina, J. 1993. What is coefficient alpha ? An examination of theory and applications. Journal of Applied Psychology, 78(1); 98-104.

Costenbader, E. & Valentae, T. 2003. The stability of centrality measures when networks are sampled. Social Networks, 25: 283-307.

Cross, R. Borgatti, S. & Parker, A. 2001. Beyond answers: dimensions of the advice network. Social Networks, 23: 215-235.

Cummings, J. & Cross, R. 2003. Structural properties of work groups and their consequences for performance. Social Networks, 25(3): 197-210.

Davenport, T. 2000. The future of enterprise system-enabled organizations. Information Systems Frontiers, 2(2): 163-180.

Davenport, T. 1998. Putting the enterprise into the enterprise system. Harvard Business Review, 76(4): 121-131.

247

Davis, A. Cox, T. & Beale, D. 1991. Exposure to Repetitive Computer-Based Work. Personnel Review, 20(1): 3-13.

Davis, F. 1989. Perceived usefulness, perceived ease of use and user acceptance of information technology. MIS Quarterly, 13: 319-339.

DeSanctis, G. & Poole, M. 1994. Capturing the complexity in advanced technology use: Adaptive structuration theory. Organization Science, 5(2): 121-147.

DeVellis, R. 1991. Scale development: Theory and applications. California, Sage Publications.

Dillman, D. 2000. Mail and internet surveys: The tailored design method, New York; Wiley.

Dillman, D. 1991. The design and administration of mail surveys. Annual Review of Sociology, 17: 225-49.

Dodd, N. & Ganster, D. 1996. The interactive effects of variety, autonomy and feedback on attitudes and performance. Journal of Organizational Behavior, 17: 329-347.

Dommeyer, C. & Moriarty, E. 2000. Comparing two forms of an e-mail survey: Embedded vs. attached. International Journal of Market Research, 42; 39-50.

Dube, L.& Pare, G. 2003. Rigor in information systems positivist case research: Current practices, trends and recommendations. MIS Quarterly, 27(4): 597-635.

Edwards, J. 1994. Regression analysis as an alternative to difference scores. Journal of Management, 20(3): 683-689.

Edwards, J. 2001. Ten difference score myths. Organization Research Methods, 4(3): 265-288.

Ellis, B. 1992. The effects of uncertainty and source credibility on attitudes about organizational change. Management Communication Quarterly, 6(1): 34-57.

Emerson, R. 1962. Power-dependence relations. American Sociological Review, 27: 31-40.

Fife-Schaw, C. 1995. Questionnaire design. In G. Breakwell, S. Hammond and C. Fife-Schaw (eds.) Research methods in psychology. London, Sage Publications.

Fink, A. 1995. The survey handbook. California, Sage Publications.

Fombrun, M. 1982. Strategies for network research in organizations. Academy of Management Review, 7: 280-291.

Fowler, F. 1995. Survey research methods (2nd ed.). California, Sage Publications.

Frazer, L. & Lawley, M. 2000. Questionnaire design and administration: A practical

248

guide. South Australia; John Wiley & Sons.

Freeman, L. C., Borgatti, S. P., & White, D. R. 1991. Centrality in valued graphs: A measure of betweenness based on network flow. Social Networks, 13: 141-154.

Freeman, L. 1979. Centrality in social networks: Conceptual clarification. Social Networks, 1: 215-239.

French, J. & Raven, B. 1960. The bases of social power in group dynamics. In D. Cartwright & P. Zander (Eds.) Organizational Dynamics: Research and theory (pp 607-23). New York: Harper Row.

Fried, Y. & Ferris, G. 1989. The validity of the job characteristics model: A review and meta-analysis. Personnel Psychology, 40: 287-322.

Fulk, J. Schmitz, J. & Steinfield, C. 1990. A social influence model of technology use. In J. Fulk & C. Steinfield (Eds.), Organizations and communication technology: 117-142. Newbury Park, CA: SAGE.

Fulk, J. 1993. Social construction of communication technology. Academy of Management Journal, 36: 921-950.

Ghani, J. & Al-Meer, A. 1989. Effect of end user computing on job satisfaction: An exploratory study. Information & Management, 17(4): 191-195.

Giddens, A. 1999. Elements of the theory of structuration. In Contemporary Social Theory, A. Elliott (Ed.), 119-130. Cornwall, Blackwell.

Giddens, A. 1993. Social theory and modern sociology. Cambridge, England: Polity.

Giddens, A. 1984. The constitution of society: Outline of the theory of structuration. Cambridge, Polity Press.

Gill, T. G. 1996. Expert systems usage: Task change and intrinsic motivation. MIS Quarterly, 20(3): 301-329.

Gould, J. and Fernandez, J. 1989. Structures of mediation: A formal approach to brokerage in transaction networks. Sociological Methodology: 89-126.

Grant, G. & Uruthirapathy, A. 2002. The effects of an enterprise resource planning system (ERP) implementation on job characteristics: A study using the Hackman and Oldham job characteristics model. 2002 Information Resources Management Association International Conference - Issues and trends of information technology management in contemporary organizations, Seattle, Washington, USA, May 19-22, 2002.

Green, C. 1998. Normative influence on the acceptance of information technology: Measurement and effects. Small Group Research, 29(1): 85-123.

Grey, C. & Mitev, N. 1995. Re-engineering organizations: A critical appraisal. Personnel Review, 24(1): 6-18.

249

Griffith, T. & Northcraft, G. 1996. Cognitive elements in the implementation of new technology: Can less information provide more benefits? MIS Quarterly, 20(1); 99-110.

Griffith, T. 1999. Technology features as triggers for sensemaking. Academy of Management Review, 24: 472-488.

Gunter, B. Nicholas, D. Huntington, P. Williams, P. 2002. Online vs. Offline research: Implications for evaluating digital media. Aslib Proceedings,54; 229-239.

Hackman, J. & Oldham, G. 1976. Motivation through the design of work: Test of a theory. Organizational Behaviour and Human Performance, 16: 250-279.

Hackman, J. & Oldham, G. 1980. Work Redesign. Reading, MA: Addison-Wesley.

Haines, D. 1999. Letting “the system” do the work: The promise and perils of computerization. The Journal of Applied Behavioural Science, 35: 306-324.

Hall, R. 2002. Enterprise resource planning systems and organisational change: Transforming work organisation? Strategic Change, 11: 263-270.

Hammer, M. 1984. Explorations into the meaning of social network interview data. Social Networks, 7: 189-201.

Hammer, M. & Champy, J. 1993. Reengineering the corporation: A manifesto for business revolution. Harper Business: New York.

Hanneman, Robert A. and Mark Riddle. 2005. Introduction to social network methods. Riverside, CA: University of California, Riverside (published in digital form at http://faculty.ucr.edu/~henneman/.).

Hinds, P., Carley, K., Krackhardt, D. & Wholey, D. 2000. Choosing Work Group Members: Balancing Similarity, Competence, and Familiarity. Organizational Behaviour and Human Decision Processes, 81(2): 226-251.

Hinkin, T. 1989. A brief tutorial on the development of measures for the use in survey questionnaires. Organizational Research Methods, 1; 104-119.

Hinkin, T. 1995. A review of scale development practices in the study of organisations. Journal of Management, 21; 967-988.

Hislop, D. Newell, S. Scarbrough, H. & Swan, J. 2000. Networks, knowledge and power: Decision making, politics and the process of innovation. Technology Analysis & Strategic Management, 12(3): 399-411.

250

HLOAR - HAL Project: High Level Organisational Analysis Report, ERUDIO, Document Version 1.0, December 2002.

Holton, E. & Russell, C. 1997. The relationship of anticipation to newcomer socialization processes and outcomes: A pilot study. Journal of Occupational and Organizational Psychology, 70: 163-172.

Hong, W., Thong, J. Wong, W., & Tam, K. 2001. Determinants of user acceptance of digital libraries: An empirical examination of individual differences and systems characteristics. Journal of Management Information Systems, 18(3); 97-125.

Howell, J. & Higgins, C. 1990. Champions of technological innovation. Administrative Science Quarterly, 35: 317-341.

Huber, G. 1990. A theory of the effects of advanced information technologies on organisational design, intelligence, and decision making. Academy of Management Review, 15(1): 47-71.

Hubert, L. 1985. Combinatorial data analysis: association and partial association. Psychometrika, 50: 449-467.

Ibarra, H. 1993. Network centrality, power and innovation involvement: Determinants of technical and administrative roles. Academy of Management Journal, 36: 471-501.

Ibarra, H. & Andrews, S. 1993. Power, social influence and sensemaking: Effects of network centrality and proximity on employee perceptions. Administrative Science Quarterly, 38: 277-303.

Igbaria, M. Guimaraes, T. & Davis, G. 1995. Testing the determinants of microcomputer usage via a structured equation model. Journal of Management Information Systems, 11(4): 87-114.

Jaccard, J. & Turrisi, R. 2003. Interaction effects in multiple regression (2nd Ed.). New York, Harper Collins.

Jaing, J. Muhanna, W. & Klien, G. 2000. User resistance and strategies for promoting acceptance across system types. Information & Management, 37: 25-36.

Jackson, C. Chow, S. & Leitch, R. 1997. Toward an understanding of the behavioural intention to use an information system. Decision Sciences, 28(2): 357-389.

Karahanna, E. & Straub, D. 1999. The psychological origins of perceived usefulness and ease of use. Information & Management, 35: 237-250.

Khan, W. Cross, R. & Parker, A. 2003. Layers of diagnosis for planned relational

251

change in organizations. Journal of Applied Behavioural Science, 39(3): 259-280.

Killworth, P., & Bernard, H. 1976. Informant accuracy in social network data. Human Organisation, 35: 269-86.

Klaus, H. Rosemann, M. & Gable, G. 2000. What is ERP? Information Systems Frontiers, 2(2); 141-162.

Koch, C. 2001. Enterprise resource planning: Information technology as a steamroller for management politics. Journal of Organizational Change Management, 14(1): 64-78.

Knoke, D. & Kuklinski, J. 1982. Network analysis. Beverly Hills, SAGE.

Krackhardt, D. & Porter, L. 1986. When friends leave: A structural analysis of the relationship between turnover and stayer’s attitudes. Administrative Science Quarterly, 30: 242-261.

Krackhardt, D. 1987. Cognitive social structures. Social Networks, 9: 109-134.

Krackhardt, D. 1988. Predicting with networks: Nonparametric multiple regression analysis with dyadic data. Social Networks, 10: 359-381.

Krackhardt, D. & Stern, R. 1988. Informal networks and organizational crises: An experimental simulation. Social Psychology Quarterly, 51(2): 123-140.

Krackhardt, D. 1990. Assessing the political landscape: Structure, cognition and power in organizations. Administrative Science Quarterly, 35: 342-369.

Krackhardt, D., Blyth, & McGrath, C. 1994. Krackplot 3.0: An improved network drawing program. Connections, 17(2); 53-55.

Krebs, V., 2004. Personal communication. Response to SOCNET posting.

Krosnick, J. 1999. Survey research. Annual Review of Psychology, 50; 537-67.

Lau, T. Wong, Y. & Law, M. 2001. Information technology and work environment – does it change the way people interact at work ? Human Systems Management, 20: 267-279.

Lengnick-Hall, C., Lengnick-Hall, M. & Abdinnour-Helm, S. 2004. The role of social and intellectual capital in achieving competitive advantage through enterprise resource planning (ERP) systems. Journal of Engineering Technology Management, 21: 307-330.

Liker, J. Haddad, C. Karlin, J. 1999. Perspectives on technology and work organisation. Annual Review of Sociology, 25: 575-596.

252

Long, R. 1993. The impact of new office information technology on job quality of female and male employees. Human Relations, 46(8); 23-35.

Luce, R. & Perry, A. 1949. A method of matrix analysis of group structure. Psychometrica, 14: 95-116.

Macri, D. Tagliaventi, M. & Bertolotti, F. 2001. Sociometric location and innovation: How the social network intervenes between the structural position of early adopters and changes in the power map. Technovation, 21: 1-13.

Manfreda, K. Batagelj, Z. Vehovar, V. 2002. Design of web survey questionnaires: Three basic experiments. Journal of Computer Mediated Communication, 7; 1-31.

Marsden, P. 1990. Network data and measurement. Annual Review of Sociology, 16: 435-63.

Martin, J. 1999. A general permutation-based QAP analysis approach for dyadic data from multiple groups. Connections, 22(2): 1-11.

Martin, M. 1998. Enterprise Resource Planning software. Fortune, February 2: 149-151.

Mathieson, K. 1991. Predicting user intentions: Comparing the technology acceptance model with the theory of planned behaviour. Information Systems Research, 2; 173-189.

Mathieson, K. & Keil, M. 1998. Beyond the interface: Ease of use and task/technology fit. Information and Management, 34: 221-230.

Mathieson, K. Peacock, E. Chin, W. 2001. Extending the technology acceptance model: The influence of perceived user resources. The DATABASE for advances in Information Systems, 32; 86-112.

Maxwell, S. & Delaney, H. 1990. Designing experiments and analysing data: A model comparison perspective. Pacific Grove, CA: Brook/Cole Publishing.

McGrath, C. & Krackhardt, D. 2003. Network conditions for organizational change. Journal of Applied Behavioural Science, 39(3): 281-300.

Meyer, G. 1994. Social information processing and social networks: A test of social influence mechanisms. Human Relations, 47:1013-1037.

Miller, V. Johnson, J. & Grau, J. 1994. Antecedents to willingness to participate in a planned organizational change. Journal of Applied Communication Research, 22: 59-80.

Mitchell, V. & Zmud, R. 1999. The effects of coupling IT and work process strategies in redesign projects. Organization Science, 10(4): 424-438.

253

Mitev, N. 1996. Empowerment, change and information technology: Socio-technical design and business process re-engineering. Personnel Review, 25(4): 56-66.

Mohrman, S. Ramkrishnan, T. & Mohrman, A. 2003. The role of networks in fundamental organizational change: A grounded analysis. Journal of Applied Behavioural Science, 39(3): 301-323.

Mouzelis, N. 1989. Restructuring structuration theory. The Sociological Review, 37(4): 613-635.

Nelson, R. 1989. The strength of strong ties: Social networks and intergroup conflict in organizations. Academy of Management Journal, 32(2): 377-401.

Nelson, R. & Mathews, K.M. 1991. Network characteristics of high performing organizations. The Journal of Business Communication, 28(4): 367-386.

Nohria, N. 1992. Is a network perspective a useful way of studying organizations ? In Nohria, N & Eccles, R. (Eds.) Networks and Organizations, (pp. 1-22), Boston: Harvard Business Press.

O’Mahoney, S. & Barley, S. 1999. Do digital telecommunications affect work and organisation ? The state of our knowledge. Research in Organisational Behaviour, 21: 125-161.

Oliver, D. & Oliver, L. 2002. ERP adoption: Selling the system. IFIP WG 8.2 Working Conference Global and Organizational Discourse about IT. Barcelona, December 12-14.

Oliver, D. & Romm, C. 2002. Justifying enterprise resource planning adoption. Journal of Information Technology, 17: 199-213.

Oppenheim, A. 1992. Questionnaire design, interviewing and attitude measurement. New York, Pinter.

Orlikowski, W. 1991. Integrated information environment or matrix of control ? The contradictory implications of information technology. Accounting, Management and Information Technologies, 1(1): 9-42.

Orlikowski, W. 1992. The duality of technology: Rethinking the concept of technology in organisations. Organization Science, 3(3): 398-427.

Orlikowski, W. 2000. Using technology and constituting structures: A practice lens for studying technology in organisations. Organization Science, 11(4): 404-428.

Orlikowski, W. & Barley, S. 2001. Technology and institutions: What can research of information technology and research on organisations learn from each other ? MIS Quarterly, 25: 145-165.

254

Orlikowski, W. & Iacono, C. 2001. Desperately seeking the “IT” in IT research – A call to theorizing the IT artefact. Information Systems Research, 12(2): 121-134.

Orlikowski, W. & Robey, D. 2001. Information technology and the structuring of organizations. Information Systems Research, 2(2): 143-169.

Oshagbemi, T. 1997. Academics and their managers: A comparative study in job satisfaction. Personnel Review, 28(1/2): 108-123.

Papa, J. 1990. Communication network patterns and employee performance with new technology. Communication Research, 17; 344-368.

Paper, D. Tingley, K. & Mok, W. 2003. The relationship between BPR and ERP systems: A failed project. Annals of Cases on Information Technology, 5: 45-55.

Parker, S. Griffin, M. Sprigg, C. & Wall, T. 2002. Effect of temporary contracts on perceived work characteristics and job strain: A longitudinal study. Personnel Psychology, 55(3): 689-719.

Parker, S. Wall, T. & Cordery, J. 2001. Future work design research and practice: Towards an elaborated model of work design. Journal of Occupational and Organizational Psychology, 74: 413-440.

Pavett, C. & Lau, A. 1983. Managerial work: The influence of hierarchical level and functional speciality. Academy of Management Journal, 26(1): 170-177.

Perry-Smith, J. & Shalley, C. 2003. The social side of creativity: A static and dynamic social network perspective. Academy of Management Review, 28(1): 89-106.

Pfeffer, J. 1992. Managing with power: Politics and influence in organisations. Boston, Harvard Business School Press.

Podolny, J. & Page, K. 1998. Network forms of organisation. Annual Review of Sociology, 24: 57-76.

Podsakoff, P. & Organ, D. 1986. Self reports in organizational research: problems and prospects. Journal of Management, 12(4): 531-544.

Pollock, N. & Cornford, J. 2004. ERP systems and the university as a unique organisation. Information Technology and People, 17(1): 31-52.

Poole, M. & DeSanctis, G. 2004. Structuration theory in information systems research: Methods and controversies. In M.E. Whitman & A.B. Woszczynski (Eds.) The handbook of information systems research (206-249). Idea Group.

Poole, M. Seibold D. & McPhee, R. 1986. Group decision-making as a structurational process. Quarterly Journal of Speech, 71; 74-102.

255

Postmes, T. Spears, R. & Lea, M. 2000. The formation of group norms in computer-mediated communication. Human Communication Research, 26(3): 341-371.

Preece, D. 1987. New technology and job design – Lessons from the print industry ? Employee Relations, 9(2): 3-8.

Pruijt, H. 1998. Multiple personalities: The case of business process engineering. Journal of Organizational Change Management, 11(3): 260-72.

Quinn, R. 1988. Beyond rational management. San Francisco; Jossey-Bass.

Rafferty, A., & Griffin, M. 2001. Expanding organizational diagnosis by assessing the intensity of change activities. Organizational Development Journal, 19(3): 3-14.

Rice, R. & Aydin, C. 1991. Attitudes toward new organizational technology: Network proximity as a mechanism for social information processing. Administrative Science Quarterly, 6: 219-244.

Rice, R.E. & Richards, W. Jr. 1985. An overview of communication network analysis methods and programs. In B. Dervin & M. Voigt (Eds.), Progress in communication sciences, 6, 105-165. Norwood, NJ: Ablex.

Richardson, D. 2002. MAGE. Little River Institute, Biochemistry Department, Duke University. North Carolina.

Robertson, D. 1989. Social determinants of information systems use. Journal of Management Information Systems, 5(4): 55-71.

Robey, D. 1979. User attitudes and management information system use. Academy of Management Journal, 22: 527-538.

Robey, D. & Zmud, R. 1992. Research on the organization of end-user computing: Theoretical perspectives from organisational science. Information Technology and People, 6(1): 11-27.

Robey, D., Ross, J. & Boudreau, M. 2002. Learning to implement enterprise systems: An exploratory study of the dialectics of change. Journal of Management Information Systems, 19(1): 17-32.

Roth, P. & BeVier, C. 1998. Research rates in HRM/OB survey research: Norms and correlates, 1990-1994. Journal of Management, 24(1): 97-117.

Rusbult, C. Farrell, D. Rogers, G. & Mainous, A. 1988. Impact of exchange variables on exit, voice, loyalty and neglect: An integrative model of responses to declining job satisfaction. Academy of Management Journal, 31: 599-627.

Ryker, R. Nath, R. & Henson, J. 1997. Determinants of computer user expectations and their relationships with user satisfaction: An empirical study. Information Processing & Management, 33(4): 529-537.

256

Sackett, P. and Larson, R. 1990. Research strategies and tactics in industrial and organisational psychology. In M.D. Dunnette, L.M. Hough, (Eds.). Handbook of Organisational and Industrial Psychology, Paolo Alto, Consulting Psychologists Press; 421-489.

Salancik, G. & Pfeffer, J. 1978. A social information processing approach to job attitudes and task design. Administrative Science Quarterly, 23: 224-253.

Saxon, D. Garrat, D. Gilroy, P. Cairns, C. 2003. Collecting data in the information age: Exploring web-based research survey methods in educational research. Research in Education, 69; 51-66.

Seabright, M. Levinthal, D. & Fichman, M. 1992. Role of individual attachments in the dissolution of interorganizational relationships. Academy of Management Journal, 35(1): 122-161.

Schmitt, N. & Klimowski, R. 1991. Assessing employee attitudes and opinions. In N.W. Zikmund, W. 1991. Business research methods (5th ed.). Orlando, Florida; Dryde.

Schmitt N. & R. Klimowski (Eds.) Research Methods in Human Resource Management, Cincinnati, South Western Publishing Company.

Schmidt, K. & Daume, B. 1993. Job characteristics and voluntary employee turnover: Direct and moderated relationships. European Work and Organizational Psychologist, 3(1): 29-42.

Schmitz, J. & Fulk, J. 1991. Organizational colleagues, media richness and electronic mail. Communication Research, 18(4): 487 –524.

Schwarz, G. & Brock, D. 1998. Waving hello or waving good-bye ? Organizational change in the information age. The International Journal of Organizational Analysis, 6(1): 65-90.

Scott, J. 2000. Social network analysis (2nd ed.). London, Sage publications.

Shah, P. 2000. Network destruction: The structural implications of downsizing. Academy of Management Journal, 43(1): 101-112.

Sia, S. Tang, M. Soh, C. & Boh, W. 2002. Enterprise Resource Planning (ERP) systems as a technology of power: Empowerment or panoptic control? The DATA BASE for Advances in Information Systems, 33(1) Winter: 23-38.

Snijders, T. & Borgatti, S. 1999. Non-parametric standard errors and tests for network statistics. Connections, 22(2): 61-70.

Songini, M. 2004. University bedevilled by portal rollout problems. Computerworld, 38: 39.

257

Sparrow, R. Liden, R. Wayne, S. & Kraimer, M. 2001. Social networks and the performance of individuals and groups. Academy of Management Journal, 44(2): 316-325.

Spector, P. 1992. A consideration of the validity and meaning of self report measures of job conditions. In C.L.Cooper & I.T. Robertson (Eds.) International Review of Industrial and Organizational Psychology, 7: 123-151.

Spector, P. 1994. Using self report questionnaires in OB research: A comment on a controversial method. Journal of Organizational Behaviour, 15: 385-392.

SPSS marketing Department. 2001. SPSS Base 10.0 applications guide. Chicago; SPSS.

Stephenson, K. & Zelen, M. 1991. Rethinking Centrality. Social Networks, 13.

Stevenson, W. 2003. Planned organizational change and organizational networks. Journal of Applied Behavioral Science, 39(3): 238-242.

Stevenson, W. Bartunek, J. & Borgatti, S. 2003. Front and backstage processes of an organizational restructuring effort. Journal of Applied Behavioral Science, 39; 243-258.

Susskind, A. Miller, V. & Johnson, J. 1998. Downsizing and structural holes: Their impact on layoff survivors’ perceptions of organisational chaos and openness to change. Communication Research, 25(1): 30-64.

Swanson, E. 1982. Measuring user attitudes in MIS research: A review. OMEGA, 10: 157-165.

Symon, G. 2000. Information and communication technologies and the network organisation: A critical analysis. Journal of Occupational and Organisational Psychology, 73: 389-414.

Szajna, B. 1996. Empirical evaluation of the revised technology acceptance model. Management Science, 42(1): 85-92.

Szajna, B. 1993. The effects if information system user expectations on their performance and perceptions. MIS Quarterly, 17(4): 493-516.

Tabachnick, B & Fidell, L. 1989. Using multivariate statistics (2nd Ed.). California, SAGE.

Tansley, C. Newell, S. & Williams, H. 2001. Effecting HRM – style practices through an integrated human resource information system: An e-greenfield site ? Personnel Review, 30(3): 351-370.

258

Teigs, R. Tetrick, L. Fried, Y. 1992. Growth need strength and context satisfaction as moderators of the relations of the job characteristics model. Journal of Management, 18; 575-593.

Tenkasi, R. & Chesmore, M. 2003. Social networks and planned organizational change: The impact of strong network ties on effective change implementation and use. Journal of Applied Behavioural Science, 39(3): 281-300.

Tichy, N. Tushman, M. & Fombrun, C. 1979. Social network analysis for organizations. Academy of Management Review, 4: 507-519.

Thomson, L. Surface, E. Martin, D. & Sanders, M. 2003. From paper to pixels: Moving personnel surveys to the web. Personnel Psychology, 56; 197-227.

Tolsby, J. 2000. Taylorism given a helping hand: How an IT system changed employees’ flexibility and personal involvement in their work. Journal of Organizational Change Management, 13(5): 482 – 501.

Torraco, R. 2005. Work design theory: A review and critique with implications for human resource development. Human Resource Development Quarterly, 16(1): 85-109.

Totterdell, P., Wall, T., Holman, D. & Diamond, H. 2004. Affect networks: A structural analysis of the relationship between work ties and job related affect. Journal of Applied Psychology, 89(5): 854-867.

Venkatesh, V. & Davis, F. 1996. A model of the antecedents of perceived ease of use: Development and test. Decision Sciences, 27: 451-481.

Venkatesh, V. 1999. Creation of favourable user perceptions: Exploring the role of intrinsic motivation. MIS Quarterly, 23: 239-260.

Venkatesh, V. 2000. Determinants of perceived ease of use: Integrating control, intrinsic motivation and emotion into the technology acceptance model. Information System Research, 11(4); 342-365.

Venkatesh, V. & Davis, F. 2000. A theoretical extension of the technology acceptance model: Four longitudinal field studies. Management Science, 46(2): 186-204.

Venkatesh, V. Speier, C. & Morris, M. 2002. User acceptance enablers in individual decision making about technology: Toward an integrated model. Decision Sciences, 33(2): 297-316.

Walker, H. Thye, S. Simpson, S. Lovaglia, M. Willer, D. & Markovsky, B. 2000. Network exchange theory: Recent developments and new directions. Social Psychology Quarterly, 63(4): 324-337.

259

Wall, T., Corbett, M., Martin, R., Clegg, C. & Jackson, P. 1990. Advanced manufacturing technology, work design and performance: A change study. Journal of Applied Psychology, 75(6): 691-697.

Wasserman, S. & Faust, K. 1994. Social network analysis: Methods and applications. Melbourne, Cambridge University Press.

Weisberg, H. Krosnick, J. & Bowen, B. 1996. An introduction to survey research and data analysis (3rd ed.). California, SAGE.

WERP. 2005. Work Effectiveness Profile for Erudio. Unpublished Report. Work Effectiveness Research Program, Australian Centre for Business Research, Faculty of Business, Queensland University of Technology.

Williams, L. Cote, J. & Buckley, M. 1989. Lack of method variance in self-report affect and perceptions at work: reality or artefact ? Journal of Applied Psychology, 74(3): 462-468.

Whetton, D. 1989. What constitutes a theoretical contribution? Academy of Management Review, 14(4); 490-495.

Wright, S. 2001. Toward understanding user acceptance of enterprise resource planning systems. American Accounting Association, 2001 Annual Meeting, Atlanta, GA.

Yin, R. 1994 Case study research: design and methods 2nd ed. California, SAGE.

Yin, R. 2003 Case study research: design and methods 3rd ed. California, SAGE.

Zeffane, R. 1994. Correlates of job satisfaction and their implications for work redesign: A focus on the Australian telecommunications industry. Public Personnel Management, 23(1): 61-88.

Zikmund, W. 1997. Business research methods (5th ed.). Orlando, Florida: Dryde.

Zikmund, W. 2000. Business research methods (6th ed.). Orlando, Florida: Dryde.

This page has been left intentionally blank.