Final Cam PHD Error Free Thesis

i
CERTIFICATION
The undersigned certify that they have read and hereby recommend for acceptance
by the University of Dar es Salaam a thesis entitled: Modelling ICT Adoption and
Use for SMEs Competitiveness and Growth, in fulfilment of the requirements for
the degree of Doctor of Philosophy (Engineering Management) of the University of
Dar es Salaam.
…………..……………………………….
Prof. E. Mjema
(Supervisor 1)
Date: …………...……………………
…………..……………………………….
Prof. B. A. T. Kundi
(Supervisor 2)
Date: …………...……………………
ii
DECLARATION
AND
COPYRIGHT
I, Cuthbert Ed’Benny Msuya Amos, declare that this thesis is my own original
work and that it has not been presented and will not be presented to any other
University for a similar or any other degree award.
Signature ------------------------------------------
This thesis is copyright material protected under the Berne Convention, the
Copyright Act 1999 and other international and national enactments, in that behalf,
on intellectual property. It may not be reproduced by any means, in full or in part,
except for short extracts in fair dealings, for research or private study, critical
scholarly review or discourse with an acknowledgement, without the written
permission of the Director of Postgraduate Studies, on behalf of both the author and
the University of Dar es Salaam.

iii
ACKNOWLEDGEMENT
First and foremost, I would like to thank my GOD my creator for giving me the
strength, good health, knowledge and understanding that enabled me to achieve this
important goal in my life.
Second, I am especially indebted to Professor E. Mjema and Prof. B. A. T Kundi of
the University of Dar es Salaam for supervising this research. Their assistance on
academic issues, positive criticism and direction in this work have been invaluable.
Their guidance, comments and the massive editorial work made this work a success
and appear in a coherent manner.
My appreciation also goes to University of Dar es Salaam Management for the
financial support which made this work possible. I am particularly very thankful to
Prof. M. Maboko the then Deputy Vice Chancellor on Academic affairs for enabling
me to secure a UDSM SIDA SAREC sponsorship for this research. Without this
sponsorship, I would not have been able to pursue my studies.
Finally, but definitely not least, I would like to thank Dr. S. Kihwele for giving me
access to an E-Library with vast collection of reference books. I am also indebted to
all staff members of the MIE Department and fellow postgraduate students for all the
assistance they rendered to me during the course of this study. My wife Melania and
our children Amos, Alex and Anneth deserve my sincere gratitude for giving me the
reason to work harder. To all of you I say thank you very much and GOD bless you.
iv
DEDICATION
This work is dedicated to my father Amos W. Msuya, my mother Naojwa Herini, my
late brother Marcelino A. W. Msuya, my late cousin Grace Y. Mngara and my whole
family. Thank you very much for being there for me.
v
LIST OF ABBREVIATIONS
AGFI Adjusted Goodness-Of-Fit Index
AM Alternative Model
AMOS Analysis of Moment Structures
CFI Comparative Fit Index
DBERR Department of Business, Enterprise and Regulatory Reform
DF Degree of Freedom
DOI Diffusion of Innovation
DTPB Decomposed Theory of Planned Behaviour
EDI Electronic Data Interchange
EM Expectation Maximisation
GDP Gross Domestic Product
GFI Goodness of Fit Index
ICT Information and Communication Technology
IDT Innovations Diffusion Theory
IS Information Systems
ISPs Internet Service Providers
IT Information Technology
LANs Local Area Networks
LDC Least Developed Countries
MAR Missing At Random
MCAR Missing Completely At Random

vi
MG Model generating
ML Maximum likelihood
N Population
n Sample Size
NFI Normed Fit Index
NTP National Telecommunication Policy
OECD Organisation for Economic co-operation and Development
PC Personal Computer
RMSEA Root Mean Square Error of Approximation
ROI Return on Investment
SC Strictly Confirmatory
SCT Social Cognitive Theory
SEM Structural Equation Modelling
SMC Squared Multiple Correlations
SME Small and Medium Scale Enterprise
TAM Technology Acceptance Model
TAM2 Technology Acceptance Model 2
TCP/IP Transmission Control Protocol/Internet Protocol
TCRA Tanzania Communications Regulatory Authority
TLI Tucker-Lewis coefficient Index
TPB Theory of Planned Behaviour
TPC Tanzania Posts Corporation
TPTC Tanzania Posts and Telecommunications Corporation
TRA Tanzania Revenue Authority

vii
TRA Theory of Reasoned Action
TTCL Tanzania Telecommunication Company Limited
UDSM University of Dar es Salaam
ULS Unweighted Least Squares
UN United Nations
URT United Republic of Tanzania
UTAUT Unified Theory of Acceptance and Use of Technology
WWW World Wide Web

viii
ABSTRACT
The Small and Medium Sized Enterprises (SMEs) sector in Tanzania is characterised
by stagnant micro enterprises that have little incidence of growth. In addition, there
is enough evidence to the effect that SMEs in Tanzania are not utilising ICTs to their
advantage despite that fact that ICT adoption and use has become critical for any
business survival and growth. Thus, this research focused at developing an ICT
adoption model to aid Tanzania SMEs in the adoption process.
Using multiple research design, methodological triangulation and survey data from a
sample of 341 respondents, a parsimonious explanatory Structural Equation Model
that identifies critical adoption factors, test for effectiveness of government policies
as well as test the post adoption SMEs performance was developed and validated.
Research resulted as follows: (i) Perceived benefits, customer pressures and
marketing strategy are the critical determinant factors of ICT adoption in Tanzania
SMEs; (ii) Government policies in the form of taxes and tariffs, support programs
and support infrastructure do not help SMEs in ICT adoption situations; (iii) By
2014, adoption and use of ICTs in SMEs had grown from nonexistence in 2004 to
80.1% computer, 68.9% Internet access, 56.6% website, and 72.2% email; (iv) 47%
of the SMEs were competitive due to ICT adoption and use.
The developed model was able to account for 47.8% of ICT adoption and 16.5% of
SMEs performance using 5 predictor variables only proving to be reliable and easy to
ix
use tool. Therefore, the model is recommended for use in Tanzania and other
countries with similar social and economic conditions to Tanzania.
LIST OF PAPERS PUBLISHED IN ASSOCIATION WITH THIS THESIS
Msuya, C. A., Mjema, E and Kundi, B (2016), “Assessment of ICT Adoption and
Use in Tanzania SME”, Proceedings of the 4th International Conference on
Mechanical and Industrial Engineering, September, 2016, AICC, Arusha
Tanzania.
Msuya, C. A., Mjema, E and Kundi, B (2017), “ICT Adoption and Use in Tanzania
SMEs”, Accepted by Tanzania Journal of Engineering and Technology, May
2017.
Msuya, C. A., Mjema, E and Kundi, B (2017), “Towards the Development of the
SMEs Sector in Tanzania: The Impact of Government Interventions for
aiding ICT Adoption and Use by SMEs” Submitted to Tanzania Journal of
Engineering and Technology, May 2017.

x
TABLE OF CONTENTS
Page
Certification...................................................................................................................i
Declaration and Copyright...........................................................................................ii
Acknowledgement.......................................................................................................iii
Dedication....................................................................................................................iv
List of Abbreviations....................................................................................................v
Abstract.....................................................................................................................viii
List of Papers Published in Association with this Thesis............................................ix
Table of Contents..........................................................................................................x
List of Tables............................................................................................................xxii
List of Figures..........................................................................................................xxvi
CHAPTER ONE: INTRODUCTION......................................................................1
1.1 General Introduction.........................................................................................1
1.2 Definition of Terms..........................................................................................4
1.2.1 Definition of Technology.................................................................................4
1.2.2 Definition of ICT..............................................................................................4
1.2.3 Definition of Computer.....................................................................................5
1.2.4 Definition of the Internet..................................................................................5

xi
1.2.5 SMEs Definition...............................................................................................6
1.3 Background to the Problem..............................................................................7
1.4 Statement of the Problem................................................................................10
1.5 Research Objectives........................................................................................12
1.5.1 The Main Objective........................................................................................12
1.5.2 Specific Objectives:........................................................................................12
1.6 Significance of the Research..........................................................................12
1.7 Scope of the Research.....................................................................................13
1.7.1 Study Area......................................................................................................14
1.7.2 Sectors Studied...............................................................................................14
1.7.3 Annual Turnover.............................................................................................15
1.7.4 Types of ICTs.................................................................................................16
1.7.5 Respondents....................................................................................................16
1.8 Thesis Organization........................................................................................16
1.9 Chapter Summary...........................................................................................21
CHAPTER TWO: LITERATURE REVIEW I: THE SME AND ICT
SECTORS IN TANZANIA..........................................................................22
2.1 Introduction.....................................................................................................22
2.2 The SME Sector in Tanzania..........................................................................22
2.2.1 Historical Background of the SME Sector in Tanzania..................................22

xii
2.2.2 Current Status of the SME Sector in Tanzania...............................................25
2.3 ICT Development in Tanzania........................................................................28
2.3.1 Computer Technology....................................................................................28
2.3.2 The Telecommunications Sector....................................................................29
2.3.3 The Current Status of the ICT Sector in Tanzania.........................................31
2.4 ICT Adoption by Tanzania SMEs..................................................................39
2.5 Chapter Summary...........................................................................................41
CHAPTER THREE: LITERATURE REVIEW II: ICT ADOPTION IN SMEs
........................................................................................................................42
3.1 Introduction.....................................................................................................42
3.2 The Role of ICTs to Business.........................................................................42
3.2.1 Benefits of ICT Adoption in SMEs................................................................45
3.3 Review of ICT Adoption Theories and Models.............................................47
3.3.1 Theoretical Foundation of Adoption Models.................................................47
3.3.2 The Diffusion of Innovation (DOI) Theory....................................................49
3.3.3 The Theory of Reasoned Action-(TRA).........................................................50
3.3.4 The Theory of Planned Behaviour-(TPB)......................................................52
3.3.5 The Technology Acceptance Model-(TAM)..................................................53
3.4. Models of ICT Adoption in SMEs..................................................................54
3.5 The Main Factors Influencing ICT Adoption in SMEs..................................63

xiii
3.6 Emerging Issues from Literature Review-Knowledge Gap Analysis.............65
3.7 Chapter Summary...........................................................................................66
CHAPTER FOUR: THEORETICAL FRAMEWORK AND HYPOTHESES.67
4.1 Introduction.....................................................................................................67
4.2 Exploration of Research Variables.................................................................72
4.2.1 Personal Factors..............................................................................................72
4.2.1.1 Owner-Manager Education.............................................................................72
4.2.1.2 Knowledge of ICT..........................................................................................73
4.2.1.3 Perceived Benefits..........................................................................................74
4.2.2 Organisation Factors.......................................................................................77
4.2.2.1 The Firm........................................................................................................78
4.2.2.2 Organisational Readiness...............................................................................82
4.2.2.3 Business Strategy...........................................................................................85
4.2.2.4 External Pressures..........................................................................................90
4.2.3 Economic Factors...........................................................................................92
4.2.3.1 Access to Financial Support..........................................................................93
4.2.3.2 Costs of ICT Systems....................................................................................95
4.2.4 Government Policies.......................................................................................97
4.2.4.1 Support Programs........................................................................................102
4.2.4.2 Taxes and Tariffs.........................................................................................102

xiv
4.2.4.3 Support Infrastructure..................................................................................105
4.2.5 ICT Adoption................................................................................................105
4.2.5.1 Computers.....................................................................................................106
4.2.5.2 The Internet..................................................................................................108
4.2.6 SME Performance........................................................................................113
4.2.6.1 Return on Investment...................................................................................116
4.2.6.2 Innovations..................................................................................................118
4.3 Summary of Model Factors..........................................................................120
4.3.1 The Conceptual Model.................................................................................122
4.4 Hypotheses....................................................................................................123
4.5 Chapter Summary.........................................................................................125
CHAPTER FIVE: RESEARCH DESIGN AND METHODOLOGY...............126
5.1 Introduction...................................................................................................126
5.2 Research Designs..........................................................................................129
5.2.1 Exploratory Research....................................................................................129
5.2.2 Descriptive Research....................................................................................130
5.2.3 Explanatory (Causal) Research.....................................................................131
5.2.4 Relationships among Research Designs.......................................................132
5.2.5 Research Design Used in the Current Study.................................................133
5.3 Research Philosophy.....................................................................................133

xv
5.3.1 Philosophical Paradigm Adopted in the Current Research...........................137
5.4 Research Approaches....................................................................................137
5.4.1 Deductive Approach.....................................................................................137
5.4.2 Inductive Approach......................................................................................138
5.4.3 Research Approach Employed in the Current Study....................................139
5.5 Research Strategies.......................................................................................139
5.5.1 Case Study....................................................................................................140
5.5.2 Survey...........................................................................................................142
5.5.3 The Research Strategy Used in the Current Study.......................................143
5.6 Research Choices..........................................................................................144
5.6.1 Quantitative Research Method.....................................................................145
5.6.2 Qualitative Research Method.......................................................................146
5.6.3 Mixed Methods and Multi Methods.............................................................148
5.6.4 The Research Methods Used in the Current Study.......................................149
5.7 Research Time Dimension............................................................................150
5.8 Data Collection.............................................................................................151
5.8.1 Development of Research Instrument..........................................................151
5.8.2 Pilot Case Studies.........................................................................................151
5.8.3 Questionnaire Design....................................................................................152
5.8.4 Questionnaire Pre-test...................................................................................157

xvi
5.8.5 Pilot Survey..................................................................................................159
5.8.6 Reliability Analysis of the Research Instrument..........................................160
5.8.7 The Validity of Research Instrument............................................................163
5.8.7.1 Content Validity...........................................................................................164
5.8.7.2 Criterion-related Validity.............................................................................164
5.8.7.3 Construct Validity........................................................................................165
5.8.8 The Need for Sampling.................................................................................166
5.8.8.1 Sampling Procedures...................................................................................168
5.8.8.2 Study Population..........................................................................................169
5.8.8.3 Sampling Frame...........................................................................................169
5.8.8.4 Sample Size.................................................................................................170
5.8.8.5 Sampling Methods.......................................................................................174
5.9 Data Collection Procedures..........................................................................177
5.9.1 Unique Problems during Data Collection.....................................................178
5.9.2 Response Rate Analysis................................................................................180
5.10 Data Analysis Methods.................................................................................182
5.10.1 Preliminary Data Analysis Methods............................................................183
5.10.2 Structural Equation Modelling.....................................................................183
5.10.2.1 Structural Equation Modelling Goodness-of-Fit Statistics....................187
5.10.2.2 Multiple Group Analysis........................................................................192

xvii
5.10.3 Sample Size Requirements...........................................................................193
5.11 Considerations for Multivariate Data Analysis............................................193
5.11.1 Missing Data.................................................................................................194
5.11.2 Multivariate Outliers.....................................................................................197
5.11.3 Multivariate Normality.................................................................................200
5.11.4 Multicolinearity and Singularity...................................................................202
5.12 Chapter Summary.........................................................................................204
CHAPTER SIX: PRELIMINARY DATA ANALYSIS AND FINDINGS.......205
6.1 Introduction...................................................................................................205
6.2 Missing Data Analysis..................................................................................206
6.3 Testing Data for Multivariate Outliers.........................................................207
6.4 Reliability Analysis......................................................................................209
6.5 Convergent Validity Analysis.......................................................................211
6.6 Testing Data for Multivariate Normality......................................................212
6.7 Descriptive Analyses....................................................................................212
6.7.1 Demographic Statistics.................................................................................212
6.7.2 ICT Adoption................................................................................................216
6.7.2.1 Computer Adoption.....................................................................................216
6.7.2.2 Internet Adoption.........................................................................................218
6.7.2.3 Website Adoption........................................................................................219

xviii
6.7.2.4 Email Adoption............................................................................................220
6.7.3 ICT use by Tanzania SMEs..........................................................................221
6.7.3.1 Use of Computers........................................................................................223
6.7.3.2 Use of the Internet.......................................................................................225
6.7.4 Post Adoption Performance Evaluation........................................................227
6.8 Discussion of Descriptive Analyses Findings..............................................229
6.9 Chapter Summary.........................................................................................232
CHAPTER SEVEN: THE ICT ADOPTION MODEL DEVELOPMENT......233
7.1 Introduction...................................................................................................233
7.2 Constructs of the Research Model................................................................234
7.3 Construct Reliability.....................................................................................237
7.5 Discriminant Validity of Constructs.............................................................243
7.5.1 Measurement Model One.............................................................................245
7.5.2 Measurement Model Two.............................................................................250
7.5.3 Measurement Model Three...........................................................................253
7.6 Model Estimation..........................................................................................258
7.6.1 Unstandardised and Standardised Estimates................................................258
7.6.2. Squared Multiple Correlations (SMC)..........................................................259
7.7 The ICT Adoption Model for Tanzania SMEs.............................................260
7.7.1 Variables of the Developed ICT Adoption Model.......................................266

xix
7.8 Hypotheses Testing-Direct Path Hypotheses................................................268
7.9 Hypotheses Testing-Moderating Hypotheses...............................................272
7.9.1 Effect of Support Programs on ICT Adoption..............................................275
7.9.2 Effect of Taxes and Tariffs on ICT Adoption..............................................283
7.9.3 Effect of Infrastructure on ICT Adoption.....................................................290
7.10 Interpretations of the Developed ICT Adoption Model...............................297
7.11 Chapter Summary.........................................................................................300
CHAPTER EIGHT: CONCLUSIONS AND RECOMENDATIONS..............306
8.1 Introduction...................................................................................................306
8.2 Conclusions...................................................................................................307
8.2.1 The ICT Adoption Model.............................................................................307
8.2.2 Factors Influencing ICT Adoption by Tanzania SMEs................................309
8.2.3 The Relationship between the Relevant Factors and their Effect on ICT
Adoption by Tanzania SMEs........................................................................310
8.2.4 The Status of ICT Adoption and Use by Tanzania SMEs............................313
8.2.5 The Influence of ICT Adoption on Tanzania SMEs Performance...............314
8.3 Research Contributions.................................................................................315
8.3.1 Contribution to Knowledge..........................................................................315
8.3.2 Methodological Contributions......................................................................316
8.3.3 Practical Contributions.................................................................................317

xx
8.4 Recommendations for Successful ICT Adoption by SMEs.........................318
8.4.1 General Recommendations for Practice/use.................................................319
8.4.2 Recommendations for Owner-Managers......................................................319
8.4.3 Recommendations to the Government..........................................................321
8.5 Research Limitations....................................................................................322
8.6 Direction for Future Research......................................................................324
REFERENCES.......................................................................................................326
APENDICES...........................................................................................................368
Apendix I: Survey Research....................................................................................368
A: Cover Letter for Survey................................................................................368
B: Research Instrument.....................................................................................369
Appendix II: Descriptive Statistics..........................................................................376
A: Missing Data Analysis..................................................................................376
B: Multivariate Normality Test.........................................................................379
C: ICT Adoption and Frequency of Use............................................................380
Appendix III: Multicolinearity and Discriminant Validity Tests...........................382
A: Sample Correlation Standardized Residual Covariance’s and Implied
Correlations for Exogenous Latent Constructs.............................................382

xxi
B: Sample Correlation Standardized Residual Covariance’s and Implied
Correlations for Endogenous Latent Constructs...........................................388
C: Sample Correlation Standardized Residual Covariance’s and Implied
Correlations for all Latent Constructs...........................................................393
D: Sample Correlation and Standardized Residual Covariance’s for the ICT
Adoption Model............................................................................................395
Appendix IV: Multiple Group Analysis-Nested Model Comparison.....................397
A: Support Programs.........................................................................................397
B: Taxes and Tariffs..........................................................................................398
C: Support Infrastructure...................................................................................399
xxii
LIST OF TABLES
Table 1.1: Tanzania SMEs Classification................................................................7
Table 2.1: Small Business Distribution by Area....................................................27
Table 2.2: Computer Population in Tanzania up to 1993......................................29
Table 2.3: Licensed Operators between 2003~2009..............................................33
Table 2.4: Licensed Telecom Operators................................................................33
Table 2.5: Weighted Average Prepaid Tariff for Voice Telecom..........................34
Table 2.6: Subscriptions and Teledensity..............................................................35
Table 2.7: Internet Subscribers by Subscription Type...........................................36
Table 2.8: Estimates of Internet Users by Subscription Type...............................36
Table 3.1: Theories of 20 Most Cited Articles and Books on ICT Adoption........48
Table 3.2: Inhibitors for Internet Adoption in SMEs.............................................56
Table 3.3: Drivers for Internet Adoption in SMEs................................................57
Table 3.4: Summary of Relevant Studies of Technology Adoption in SMEs.......60
Table 3.5: Summary of Main Adoption Factors in SMEs.....................................64
Table 4.1: Summary of ICT Adoption Factors in the Current Study.....................69
Table 4.2: Individual Factors for the ICT Adoption Model.................................120
Table 5.1: The Relationship between Paradigms and Strategies.........................136
Table 5.2: Major Differences between Deductive and Inductive Approaches to
Research..............................................................................................138
Table 5.3. The Differences in Emphasis in Qualitative vs. Quantitative Methods....
............................................................................................................147
xxiii
Table 5.4: Pilot Survey Reliability and Internal Consistence Analysis Results. .163
Table 5.5a: Sample Distribution in Participating Regions....................................176
Table 5.5b: Sample Distribution in Participating Districts....................................177
Table 5.6a: Summary of Survey Data from Secondary Sampling Units...............181
Table 5.6b: Summary of Survey Data from Primary Sampling Units...................182
Table 5.7: Summary of SEM Fit Measures Used in this Research......................191
Table 5.8: Sample Size Requirements for Different Statistical Analysis............193
Table 6.1: Mahalanobis Distance Test Results....................................................208
Table 6.2: Main Survey’s Reliability and Internal Consistence Analysis Results....
............................................................................................................210
Table 6.3: Demographic Information..................................................................213
Table 6.4: Summary of Computer Adoption.......................................................217
Table 6.5: Networked Computers (LAN)............................................................217
Table 6.6: Summary of Internet Adoption...........................................................219
Table 6.7: Summary of Website Adoption..........................................................219
Table 6.8: Summary of Email Adoption..............................................................221
Table 6.9a: Use of Computers for all Accounting Activities................................224
Table6.9b: Use of Computers for Market Research.............................................224
Table 6.9c: Use of Computers for Business Transactions Using Software Packages.
............................................................................................................224
Table 6.9d: Use of Computers for all Business Transactions................................224
Table 6.10a: Use of Email for external Communication.....................................225
Table 6.10b: Use of Internet for obtaining Information from Suppliers..............225
Table 6.10c: Use of Internet for Sending Purchase Orders to Suppliers.............226

xxiv
Table 6.10d: Use of Internet for Product and Market Research...........................226
Table 6.10e: Use of Websites to Offer Information to Customers......................226
Table 6.10f: Use of Websites to Receive Orders from Customers......................226
Table 6.10g: Use of Internet for Offering Online Payments................................227
Table 6.11a: Increase of Profit Due to Computer Adoption................................227
Table 6.11b: Increase of Profit Due to Internet Adoption....................................227
Table 6.11c: Increase of Market Size Due to Internet Adoption.........................228
Table 6.11d: Increase of Market Size Due to Computer Adoption......................228
Table 6.11e: Process Innovations Due to Computer Adoption............................228
Table 6.11f: Process Innovations Due to the Internet.........................................229
Table 7.1: Constructs of the Model......................................................................236
Table 7.2: The SMC for the 41 Indicators Measuring Exogenous Constructs........239
Table 7.3: The SMC for the 22 Indicators Measuring Endogenous Constructs......240
Table 7.4: Summary of Highly Correlated Variables for the Three Exogenous Latent
Constructs..............................................................................................242
Table 7.5: Summary of Highly Correlated Variables for the Two Endogenous Latent
Constructs..............................................................................................243
Table 7.6: Correlation between Latent Constructs for Investigating Discriminant
Validity..................................................................................................246
Table 7.7: Implied (for all remaining variables) Correlations for Three Exogenous
Latent Constructs for Investigating Discriminant Validity....................247
Table 7.8: Standardized Residual Covariance’s for Measurement Model One...248
Table 7.9: Implied (for all variables) Correlations for two Endogenous Latent
Constructs for Investigating Discriminant Validity............................251

xxv
Table 7.10: Standardized Residual Covariance’s for Measurement Model Two. .251
Table 7.11: Correlation between Five Latent Constructs for Investigating
Discriminant Validity.........................................................................255
Table 7.12: Correlation between Five Latent Constructs (after recording variables)
for Investigating Discriminant Validity..............................................258
Table 7.13: Parameter Summary for the Developed ICT Adoption Model...........263
Table 7.14: Summary of developed ICT adoption Model Variables.....................266
Table 7.15: Regression Weights for ICT Adoption Model....................................269
Table 7.16: Squared Multiple Correlations for the ICT Adoption Model.............270
Table 7.17: Standardized Regression Weights......................................................271
Table 7.18: Covariances of the ICT Adoption Model...........................................271
Table 7.19: Parameter Summary for the Final ICT Adoption Model....................273
Table 7.20: Frequency Distribution for Government Support Programs..............276
Table 7.21a: Regression Weights (Structural weights)-Inadequate Support.......283
Table 7.21b: Regression Weights (Structural weights)-Adequate Support.........283
Table 7.22: Frequency Distribution regarding Taxes and Tariffs..........................284
Table 7.23a: Regression Weights (Structural weights)-Taxes not an Obstacle...289
Table 7.23b: Regression Weights (Structural weights)-Taxes an Obstacle.........290
Table 7.24: Frequency Distribution on Support Infrastructure..............................291
Table 7.25a: Regression Weights (Structural weights)-Adequate Infrastructure 296
Table 7.25b: Regression Weights (Structural weights)-Inadequate Infrastructure....
.........................................................................................................296
Table 7.26: Summary of Direct Hypotheses Results.............................................303
Table 7.27: Summary of Moderating Hypotheses Results....................................304

xxvi
Table 7.28: Summary of the Implication of the Tested Hypotheses......................304
LIST OF FIGURES
Figure 1.1: Thesis Outline.......................................................................................20
Figure 2.1: SMEs broad sector allocation by area..................................................27
Figure 2.2: Tanzania National ICT Broadband Backbone Network Map...............38
Figure 3.1: The Innovation Adoption Process........................................................49
Figure 3.2: Theory of Reasoned Action..................................................................51
Figure 3.3: The Theory of Planned Behaviour-TPB...............................................53
Figure 3.4: The Original Technology Acceptance Model.......................................54
Figure 3.5: Factors Affecting ICT Adoption by SMEs...........................................58
Figure 3.6: Model of ICT Adoption in Micro SMEs..............................................59
Figure 4.1: The Conceptual Model for Accessing ICT Adoption in SMEs..........122
Figure 5.1: Research Onion...................................................................................128
Figure 5.2: Relationships among Research Designs.............................................132
Figure 5.3: Sampling Process................................................................................168
Figure 6.1: Company Size-Number of Employees................................................214
Figure 6.2: Company Size-Capital Investment......................................................215
Figure 6.3: Adoption and Frequency of Use of Computers..................................217
Figure 6.4: Adoption and Frequency of Use of the Internet.................................218
Figure 6.5: Adoption and Frequency of Use of Website.......................................220
Figure 6.6: Adoption and Frequency of Use of Email..........................................221

xxvii
Figure 7.1: Measurement Model One with Standardised Estimates for Three
Exogenous Latent Constructs.............................................................249
Figure 7.2: Measurement Model Two with Standardised Estimates for Two
Endogenous Latent Constructs...........................................................252
Figure 7.3: Measurement Model Three with Standardised Estimates for all Five
Latent Constructs................................................................................254
Figure 7.4: Measurement Model Three with Standardised Estimates for all Latent
Constructs using recorded Variables..................................................257
Figure 7.5: The Initial ICT Adoption Model with Unstandardised Estimates......261
Figure 7.6: The Initial ICT Adoption Model with Standardised Estimates..........262
Figure 7.7: The ICT Adoption Model with Unstandardised Estimates.................264
Figure 7.8: The ICT Adoption Model with Standardised Estimates.....................265
Figure 7.9: The Final ICT Adoption Model with Unstandardised Estimates.......274
Figure 7.10: The Baseline Model with Unstandardised Estimates for Multiple
Group Analysis-Adequate Support Programs..................................277
Group Analysis-Inadequate Support Programs................................278
Figure 7.12: The Structural Weights Model with Unstandardised Estimates for
Multiple Group Analysis-Adequate Support Programs....................281
Multiple Group Analysis-Inadequate Support Programs.................282
Group Analysis-Taxes not an Obstacle............................................285

xxviii
Group Analysis-Taxes an Obstacle..................................................286
Multiple Group Analysis-Taxes not an Obstacle..............................287
Multiple Group Analysis-Taxes an Obstacle....................................288
Group Analysis-Adequate Infrastructure..........................................292
Group Analysis-Inadequate Infrastructure.......................................293
Multiple Group Analysis-Adequate Infrastructure...........................294
Multiple Group Analysis-Inadequate Infrastructure.........................295

1
CHAPTER ONE
INTRODUCTION
1.1 General Introduction
The importance of the Small and Medium Enterprise (SME) sector as the cornerstone
of most economies is widely recognised (Abdullah and Bakar, 2000; Akhtar, 1997;
Costello, 2009; DBIS, 2013a; Hall, 2002; Mbamba, 2004; Olomi, 2005). The
literatures show that, almost 90% of the total numbers of businesses across the world
are SMEs (Mira and Robert, 2007). SMEs play a significant role in creating
employment opportunities (Hall, 2002), and they contribute substantially to the
economy. For example, in the Netherlands, in year 1999, SMEs accounted for
98.8% of all private sector companies and contributed 31.6% to Gross Domestic
Product (GDP). The sector also employed 55% of the total workforce (EIM, 1999).
In United Kingdom, in year 2007, about 99.9% of the nation’s 4.3 million business
enterprises were SMEs and they accounted for well over half of employment (58.9%)
and had a turnover of (51.9%) (DBERR, 2007).
The Observatory of European SMEs (2003) note that, in year 2003, about 92% of all
European enterprises were SMEs. More recent figures show that there are more than
20, 000,000 SMEs in EU representing about 99% of all businesses (European
Commission, 2014). These SMEs are considered to be the true back-bone of the
European economy (European Commission, 2014). They are primarily responsible
for innovation, employment, wealth generation, economic growth and social

2
integration (European Commission, 2014). At present, it is estimated that SMEs
provide two out of three of the private sector jobs and contribute to more than half of
the total value added created by businesses in the European Union (European
Commission, 2014).
Previous studies by Abdullah and Bakar (2000) and Urata (2000) suggested that,
SMEs were vital to the emergence of healthy private sectors, especially in
developing countries. This has been corroborated by newer studies. According to the
IFC (2010), formal SMEs contribute up to 45% of employment and up to 33% of
GDP in developing economies; these numbers are significantly higher when taking
into account the estimated contributions of SMEs operating in the informal sector
(IFC, 2010). The literature (DBIS, 2013a; IFC, 2010) has also shown that in
developing economies, SMEs could contribute more to economic development than
they currently do. Some of the wider issues that have been mentioned as
bottlenecks to the development of the sector in many countries include: Limited
access to finance; Too expensive labour force; Lack of skilled labour;
Implementation issues related to new technology; Implementation issues related to
new forms of organization; Lack of quality management; Problems with
administrative regulations; Problems with infrastructure and Problems with
purchasing power of customers (Observatory of European SMEs, 2007).
On the same note, in recent years, adoption and effective use of ICTs in business has
been identified as being critical to business survival and growth (Apulu; 2012;
Costello, 2009; Mbamba, 2004; Naul, 2012; Olomi, 2005; Van Akkeren and Cavaye,
3
1999b). In general, the use of ICTs in SMEs has been considered vital for making
SMEs competitive (Ahmed, et al., 2010; Apulu, 2012; Consoli, 2012; Costello,
2009; European Commission, 2014; Modimongale, 2009; Naul, 2012; Olise et al.,
2014; Skoko and Cerić, 2010; Tarutė and Gatautis, 2014).
According to the United Nations (2007), the use of ICTs in SMEs can result in
several significant benefits including increased productivity, increased efficiency of
internal business operations, improved internal and external communications,
reduction of communication costs and geographical barriers for both internal and
global clients and suppliers, expanding the market size (client base) through the use
of e-marketing.
The European Commission (2008) has also noted that, the use of ICTs in SMEs can
help SMEs to grow and become more innovative. Thus, the commission
recommends that ICTs adoption and use in SMEs should be encouraged (European
Commission, 2008). According to the European Commission (2008), the use of
ICTs in SMEs can help to improve the technical and managerial skills of the owner-
managers and thus enabling SMEs to take advantage of the available e-business
solutions. This in turn helps the SMEs to address the issue of high cost of
ownership of ICT equipment (European Commission, 2008).
In addition, Love et al. (2005) claim that, the use of ICTs affords many benefits to
the SMEs at both operational level, tactical level and strategic level. Moreover,
Ongori (2009) observed that, in the present era of globalisation, the adoption of ICT
4
would change the way businesses operate. In particular, Ongori (2009) envisaged
changing of business structures and business operations in order to create
competitive advantage for businesses. Consequently, Ongori (2009) concludes that,
“for SMEs to grow and become successful, they must have the ability to compete and
dynamically respond to rapidly changing markets. This means that SMEs need to
adopt ICT and connect to the digital marketplace”.
This study looks at the SMEs sector in the context of Tanzania. In particular, the
research investigates at how Tanzania SMEs can use ICTs in order to become
competitive and achieve sustainable growth.
1.2 Definition of Terms
1.2.1 Definition of Technology
According to the Open Learn (2014), depending on their viewpoint and context,
different people attach different meaning to the word technology. This research uses
the definition by Dictionary.com (2014) which define technology as: “The branch of
knowledge that deals with the creation and use of technical means and their
interrelation with life, society, and the environment, drawing upon such subjects as
industrial arts, engineering, applied science, and pure science”. Different
technologies involved in this research a defined in subsequent sections.

5
1.2.2 Definition of ICT
Information and Communication Technologies (ICT) is a broad concept that covers
Information Systems (IS) Information Technology (IT) and digitalisation
(Sophonthummapharn, 2008). This research adopts the definition by Beckinsale and
Ram (2006). Beckinsale and Ram (2006) define ICT as “any technology used to
support information gathering, processing, distribution and use” According to
Beckinsale and Ram (2006), this covers all forms of technologies such as computers,
Internet, websites, fixed line telephones, mobile phones and other wireless
communications devices, computer networks, broadband and various specialised
devices (Manueli et al., 2007). The two technologies are defined in the subsequent
sections.
1.2.3 Definition of Computer
According to the American Heritage Science Dictionary (2005) computer is defined
as “a programmable machine that performs high speed processing of numbers, as
well as of text, graphics, symbols, and sound. All computers contain a central
processing unit that interprets and executes instructions; input devices, such as a
keyboard and a mouse, through which data and commands enter the computer;
memory that enables the computer to store programs and data; and output devices,
such as printers and display screens, that show the results after the computer has
processed data”. In this research, the use of computers for automation (digitisation)
of the business process as well as a strategic management tool in SME business
settings is explored.
6
1.2.4 Definition of the Internet
The Internet is an important aspect of ICT (Apulu, 2012; Beckinsale and Ram, 2006;
Manueli et al., 2007; Sophonthummapharn, 2008). It is defined as “a publicly
available computer network consisting of a worldwide network of computer
networks that use the TCP/IP network protocols to facilitate data transmission and
exchange” (Kripanont, 2007; Peterson and Davie, 2007). Kripanont (2007) report
that, in popular idiom, the Internet often refers to the World Wide Web (WWW),
electronic mail (email) and online chat services that operates on the Internet.
Sometimes, the Internet is simply referred to as “the Net”.
According to the literature, the Internet physically uses a portion of the total
resources of the currently existing public telecommunication networks (Kripanont,
2007). However, the Internet is distinguished from other types of communication
networks by its use of the use of TCP/IP network protocols to facilitate data
transmission and exchange (Kripanont, 2007; Peterson and Davie, 2007). It should
be noted that, while computers facilitate for data processing, storage and
interpretation, the Internet allows for easy ways of transmission and sharing the
information. In this research, the use of Internet as strategic marketing and
communication tool is explored.
1.2.5 SMEs Definition
The SMEs nomenclature is used to mean micro, small and medium enterprises
(URT, 2002). Due to the wide diversity of businesses, there is no universally
accepted definition of SMEs (Harindranath et al., 2010; Massawe, 2003; Pass et al.,
7
2000; Wangwe, 1999). Different countries use various measures of size depending
on their level of development. The commonly used yardsticks are the total number of
employees, total investment and sales turnover (URT, 2002). This study adopts the
definition of SMEs as postulated by the SMEs Development Policy of Tanzania. See
URT (2002). In the context of Tanzania, SMEs are defined according to sector,
employment size and capital investments in machinery (URT, 2002). Accordingly,
SMEs are defined as micro, small and medium size enterprises in nonfarm activities,
including manufacturing, mining, commerce and services (URT, 2002). Micro
enterprises are those engaging up to 4 people, or employing capital amounting up to
TShs. 5 million. Small enterprises are those engaging between 5 and 49 employees or
with capital investment from TShs. 5 million to TShs. 200 million. Medium
enterprises employ between 50 and 99 people or use capital investment from TShs.
200 million to TShs. 800 million (URT, 2002). Table 6.1 gives Tanzania SMEs
classification system.
Table 1.1: Tanzania SMEs Classification
Category Number of Employees Capital Investment in Machinery

in TShs. (“000 000”)
Micro Enterprise 1-4 Up to 5
Small Enterprise 5-49 Above 5 to 200
Medium Enterprise 50-99 Above 200 to 800
Large Enterprise 100+ Above 800
Source: (URT, 2002)
According to URT (2002) in the event of an enterprise falling under more than one
category, then the level of investment becomes the deciding factor (URT, 2002).
8
1.3 Background to the Problem
In Tanzania, the SME sector has also been recognized as a significant sector in
employment creation, income generation, and poverty alleviation (Bagachwa, 1983;
Mbamba, 2004; NBS, 2014; Olomi, 2005; Olomi, 2006; URT, 2002; URT, 2012).
The sector has also been considered a base for industrial development of the country
(Bagachwa, 1983; Mbamba, 2004; URT, 2002). At present, about 27% of the
Tanzania ‘s GDP originates from the SME sector (URT, 2012). Furthermore, it is
observed that, while employment has become a very significant issue in Tanzania,
over the past two decades, the SME sector has been the main source of new
employments (Mlingi, 2000; Olomi, 2005; Olomi, 2006; URT, 2002). According
the literature (Olomi, 2005; Olomi, 2006; URT, 2002) it is estimated that over 90%
of the new entrants into the Tanzania labour force every year end up in the SME
sector (URT, 2002). Olomi (2005; 2006) and URT (2002) conclude that, SMEs
development remains to be the main source of jobs in the future (Olomi, 2005; URT,
2002).
However, despite the importance of this sector as shown by the statistics, the SME
sector in Tanzania is still underdeveloped (Mbamba, 2004; Olomi, 2005; Olomi,
2006; URT, 2002; URT, 2012). Available literature shows the existence of a number
of constraints hampering the development of the sector (Calcopietro and Massawe,
1999; Mbamba, 2004, Olomi, 2005; Olomi, 2006; URT, 2002). For example,
Calcopietro and Massawe (1999) identified and broadly classified the factors
hindering SMEs development in Tanzania into five categories namely; macro-
economic and policy environment, physical and technological infrastructure, the

9
banking and finance structure, legal and regulatory framework, and market
conditions. A closer analysis of the issues raised by Calcopietro and Massawe
(1999) shows that they are more or less the same issues raised by the Observatory of
European SMEs (2007) for European SMEs. In a different study, Olomi (2006)
specifically identified that, the limited capacity of people who start and operate the
businesses and the poor state of support institutions to the sector as being some of the
most serious problems hampering the development of the SME sector in Tanzania.
This is the same as lack of quality management and poor support infrastructure as
highlighted by the Observatory of European SMEs (2007). Other research work has
shown that the use of ICTs in Tanzania SMEs is still very low (Babiker, 2001;
Baynit, 2002; Issack, 2006; Isaya, 2009; Jonathan, 2004; Kavenuke, 2009; Mbamba,
2004; Mlay, 2009; Muttayabalwa, 2007; Naul, 2012).
In response to the challenges faced by the sector, there have been numerous efforts
by the Government of Tanzania to address some of the basic issues identified by the
literature (Mbamba, 2004; Olomi, 2005; URT, 2002; URT, 2012). The enacting of
the SME policy of Tanzania, the ICT policy of Tanzania and consequently the setting
up and coordination of support institutions for the SMEs and tax incentives are some
of government efforts to develop the sector (URT,2002; URT, 2003a). However,
despite these efforts, most recent studies (URT, 2012) indicate that, most of the
identified constraints have persisted to date. For example, according to the URT
(2012), ICT use among Tanzania SMEs is still very low despite the fact that among
other things, the policies are aimed at encouraging ICT adoption in businesses and
business growth in general (URT,2002; URT, 2003a). Overall, it can be noted that
10
there is lack of significant growth in many individual SMEs and the sector as a whole
(URT, 2012). Available data points out that, for more than three decades now, the
sector has been dominated by informal micro enterprises. In addition, these
businesses have an extremely low incidence of graduation to formal small and
medium size enterprise (Mlingi, 2000; NBS, 2014; NISS, 1991; Olomi, 2005, URT,
2002; URT, 2012).
Moreover, despite the fact that the number of SMEs has increased from 1.7 million
businesses in 1991 to about 3.2 Million businesses in 2010 and that the number of
people employed in the sector has increased from 3.0 million to 5.2 million jobs in
the same period, the sector contribution to the Tanzania economy has remained at
about 30% of the GDP (NBS, 2014; NISS, 1990; URT, 2002; URT, 2012). This
suggests that most of the individual SMEs remain stagnant for a big part of their life
cycle, a phenomenon also observed by Olomi (2006). Olomi (2006) noted that there
are variations in performance among SMEs; while some are growing, a majority of
small enterprises have permanently remained micro or informal without any marked
growth. As a result, these enterprises have limited access to markets, limited access
to some support services as well as limited capacity to pay taxes and create high-
quality jobs. Olomi (2006) further foresee that, with the ongoing globalization this
situation is likely to worsen as competition intensifies.
1.4 Statement of the Problem.
Tanzania SMEs are faced with a serious problem: after start-up, most SMEs remain
stagnant and do not grow up into mature business as are expected (Mbamba, 2004;
Mlingi, 2000; NISS, 1991; Olomi, 2005; Olomi, 2006; URT, 2002). Moreover,
11
despite the fact that there is enough evidence to the effect that adoption and use of
ICTs is critical for business survival and growth (Ahmed, et al., 2010; Apulu, 2012;
Consoli, 2012; Costello, 2009; European Commission, 2014; Harindranath et al.,
2010; Hazbo and Arnela, 2010; Naul, 2012; Olise et al., 2014; Tarutė and Gatautis,
2014), ICT adoption and use in Tanzania SMEs is still very low (Baynit, 2002;
Issack, 2006; Isaya, 2009; Jonathan, 2004; Kavenuke, 2009; Mbamba, 2004; Mlay,
2009; Muttayabalwa, 2007; Naul, 2012; URT, 2012). Previous studies on the
problem in other countries show that, economic growth of a firm or that of a country
depends on the competitiveness of that firm or that nation (Rainer and Kazem 1994;
Syed and Mohammad, 2009; World Economic Forum, 2013). The literature further
shows that, in today’s knowledge economy, adoption and use of Information and
Communication Technologies (ICTs) represents the fundamental source of
competitiveness and the basis for SMEs survival and growth (Ahmed, et al., 2010;
Apulu, 2012; Consoli, 2012; Costello, 2009; European Commission, 2014; Hazbo
and Arnela, 2010; Modimongale, 2009; Naul, 2012; Olise et al., 2014; Tarutė and
Gatautis, 2014).
Based on the observed facts, the literature suggests various models for ICT adoption
by SMEs (Al-Qirim, 2005; Beckinsale and Ram, 2006; Costello, 2009; Grandon and
Pearson, 2004a; Manueli et al., 2007; Mehrtens et al., 2001; Pedersen, 2005; Van
Akkeren and Cavaye, 1999b; Windrum and de Berranger, 2002; Zappala and Gray,
2006). However, most of these models have been developed on the level of
economic development and culture of a particular country, and none has been tested
to suit the Tanzania situation. According to Kilele (2011) and Mbamba (2004), the
12
most effective model has to relate to the local environment, economic and social
priorities of a country. Thus, because of the social context, the available models
cannot be used in Tanzania since they lack constructs and or elements that captures
the uniqueness of Tanzania situation. Review of the literature does not show any
work related to ICT adoption model that has been designed and tested specifically for
guiding SMEs ICT adoption process in the context of Tanzania. This is a knowledge
gap in the literature and a shortcoming in Tanzania SME sector. This work is an
attempt to bridge this gap by developing and testing a model that will aid ICT
adoption process in Tanzania SMEs.
1.5 Research Objectives
1.5.1 The Main Objective.
The overall objective of this research is to develop a model for guiding Tanzania
SMEs in the ICT adoption process. To achieve the main objective, the following
specific objectives will be undertaken.
1.5.2 Specific Objectives:
i. To analyse the factors influencing ICT adoption by Tanzania SMEs.
ii. To establish the relationship between the relevant factors and their effect on
ICT adoption by Tanzania SMEs.

13
iii. To establish the status of ICT adoption and use by Tanzania SMEs.
iv. To establish the influence of ICT adoption on Tanzania SMEs performance.
1.6 Significance of the Research
The results of this research have both academic and practical importance.
Academically the research contributes to the general literature that fills in the
knowledge gaps that has been highlighted by the literature review. In practical terms,
the developed model will enable SMEs in Tanzania to adopt ICTs successfully by
aiding the ICT adoption process. Also, the outcome of this research will draw
attention to issues that hold back efficient use of ICTs in SMEs after adoption and
explain how SMEs can use ICTs to gain competitive advantage. This will result in
SMEs growth and economic development of the country. Moreover, the results of
this research will serve as a yardstick on the relevance and effectiveness of existing
ICT policies. This will enable for timely policy reviews, implementation and support
strategies for SMEs by policy makers such as the government, regulators and
professional institutions. The developed model can also be used in other countries
that have similar social and economic conditions like Tanzania, such as the
developing countries south of the Sahara Desert.

14
1.7 Scope of the Research
This research has concentrated on the development and validation of an ICT
adoption model for aiding SMEs in Tanzania to adopt ICTs successfully. The model
can be replicated in other countries with similar social and economic conditions to
Tanzania. The Internet and Computer Technologies were empirically investigated in
the context of Tanzania SMEs in order to examine what potential determinant factors
are involved and how much they influence the adoption decisions. Post adoption
evaluation was also carried out in order to establish how adoption of ICTs influences
SMEs competitiveness and growth. In order to make the study manageable in the
given time frame and resource limitations, the research adopted the following
strategies:
1.7.1 Study Area
Data was collected in four regions only in mainland Tanzania namely Dar es Salaam,
Arusha, Mwanza and Mbeya. The selection of these regions was justified by the fact
that, the four regions and in particular the four cities in these regions have the highest
business density than any other region or city in the country (TRA, 2013; URT,
2012). For example, according to the URT (2012) in year 2010, business
concentration in Mbeya city was 46% of all businesses in the Mbeya region while for
the Dar es Salaam city, business concentration was 45% of all businesses in Dar es
Salaam region. The high concentration of businesses in relatively small area made
data collection easier and reduced travel costs to reach the respondents. In addition,
the four regions offered the best opportunity to collect data about SMEs in rural
settings, as well as the data about SMEs in urban contexts, which was a prerequisite
15
for the study. It should be noted that, the four regions apart from being the home to
the four largest cities in Tanzania, they also have vast rural areas (NBS, 2014).
1.7.2 Sectors Studied
The literature has classified different business sectors as service sector,
manufacturing sector and trading sector (Costello; 2009; Peou; 2009;
Sophonthummapharn, 2008). According to Sophonthummapharn (2008), most of the
previous studies on SMEs have treated SMEs as a single homogeneous group of
businesses and included SMEs from all sectors in the studies (Sophonthummapharn,
2008). Other researchers however have argued that, this has limited the
generalisation of results since SMEs in different sectors are radically different in the
nature of their business, and thus each sector could contribute to the research
uniquely (Sophonthummapharn, 2008). The author notes that, the choice of what
approach is taken will much depend on the nature of the study.
In this study, SMEs from two sectors, the service and trading sectors were
researched. This choice was motivated by the fact that, the two sectors account for
over 85% of all SMEs in Tanzania. According to the URT (2012), the general
trading sector account for 55.4% of all SMEs while the service sector account for
30.5% of all SMEs. Therefore, the choice of the two sectors was expected to give
results that are relevant to the majority of SMEs in Tanzania. The choice also made
the data collection exercise easier. In addition, studying more than one sector created
the possibility of studying the two sectors separately as recommended by the
literature (Sophonthummapharn, 2008). However, due to sample size requirements,

16
this thesis report finding made on analyses that treat the SMEs from the two sectors
as a single homogeneous group.
1.7.3 Annual Turnover
This research concentrated on SMEs with an annual turnover of TShs 12,000,000
and above and excluded those below this threshold. This choice was influenced by
the fact that it was much easier to get detailed information about SMEs in this
category than in the other categories. According to data available from Tanzania
Revenue Authority (TRA), and the Districts/Municipal Directors, during the time
frame of this research, Dar es Salaam had about 25,386 businesses in this category,
Arusha had about 11,612 businesses, Mwanza had about 4855 businesses and Mbeya
had about 1146 businesses.
1.7.4 Types of ICTs
It should be noted that, ICT is a broad term covering a number of different
technologies (Beckinsale and Ram, 2006; Costello, 2009; Manueli et al., 2007;
Mpofu et al., 2009; Sophonthummapharn, 2008). However, this study focused on
the Computer and Internet technologies only, and this can also be attributed to
research constraints, that is, the time frame and the research budget. The choice of
these technologies was influenced by the fact that they have been shown to be the
fundamental source of competitiveness and growth for businesses in the present day
digital economy (Apulu, 2012; Costello, 2009; European Commission, 2014; Hazbo
and Arnela, 2010; Olise et al., 2014; Tarutė and Gatautis, 2014).
17
1.7.5 Respondents
Respondents to this research were the owners-managers of SMEs in the service
sector and general trading sector. The choice of respondents was influenced by the
fact that, in SMEs, owners-managers make all key decisions including the decisions
on what type of technology is appropriate for their business (Apulu, 2012; Costello,
2009; Levy and Powell, 2005; Rashid and Al Qirim, 2001; Van Akkeren and
Cavaye, 1999b). Thus, only the data obtained from owners and or managers was
included in the analysis and data obtained from company employees was rejected.
1.8 Thesis Organization
The organization of this thesis is summarized in Figure 1.1 and is explained as
follows:
Chapter One: Introduction
This chapter presents a brief general introduction on the background of the research
along with definitions of the key concepts of this research, that is, ICTs, Computers,
the Internet and SMEs before spelling out the research problem. The chapter also
outlines the research main objective, together with specific objectives. This is
followed by the significance of the study as well as the scope of the research. The
chapter ends with an outline of the thesis.

18
Chapter Two: Literature Review I: The SME Sector and ICT Development in
Tanzania
This chapter explores the SME sector and ICT development in the context of
Tanzania. The chapter is arranged as follows: First, a historical account of the SME
sector in Tanzania is made. Second, an overview of ICT development in Tanzania is
made. Finally, the use of ICTs in Tanzania SMEs is explained before a summary is
provided.
Chapter Three: Literature Review II: ICT Adoption in SMEs
In this chapter, literature reviews regarding many aspects of ICT adoption in SMEs
are explored. A thorough review of past research and current theories and models
related to ICT adoption and use in SMEs is made. Finally, the shortcomings of the
existing models and knowledge gaps that justify the need for this study are identified.
Chapter Four: The Conceptual Model and Hypotheses
This chapter proposes and discusses a conceptual model for this study. The
conceptual model is developed using constructs and elements that have been firmly
established by the available literature that they are critical to adoption issues in
SMEs. The conceptual model has three groups of key determinants that are
expected to influence ICT adoption and usage behaviour by SMEs together with
o n e g r o u p o f moderators that are expected to moderate the influence of these
key determinants. Post adoption evaluation is also captured in the conceptual model
in a group of key performance indicators. From the developed conceptual model,
operational hypotheses for the research are formulated.

19
Chapter Five: Research Design and Methodology
This chapter presents the research design and methodology used in this study. Using
the metaphor of the “Research Onion”, philosophical paradigms related to IS
research are briefly discussed, and the philosophical standpoint of the author is
acknowledged. After that, the diverse alternatives to research approaches, designs,
strategies, and time dimension are discussed, and justification of specific choices and
uses made in this study are then clarified. Moreover, this chapter elaborates the
details of research methodologies in regard to sampling procedure and questionnaire
development, as well as data collection in both pilot study and the main survey.
Finally, the chapter provide a summary of data analysis methods and tools used in
the study.
Chapter Six: Preliminary Data Analysis and Findings
This chapter give account of some of analyses carried out on the data, and the
findings observed. The testing commences with analyses that are (almost always/
mandatory) carried out first before other types of data analysis can take place. These
include: Internal consistency reliability test, which is used to demonstrate the
reliability of the measurement instrument (Kripanont, 2007; Peou, 2009;
Sophonthummapharn, 2008; SPSS Survival Manual, 2007). This is followed by
missing data analysis which show and mitigate the impact of missing data, constructs
convergent validity tests, which is used to show the validity of the constructs, testing
for the univariate and multivariate normality of data using the Skewness and Kurtosis
statistic measures and testing for multivariate outliers using Mahalanobis distance
20
statistic. Thereafter, descriptive analyses are performed and inferences are made.
Results of the analyses are given in tables, figures and plots in order to enhance the
understanding of presented statistical results.
Chapter Seven: The ICT Adoption Model Development
This chapter presents development and testing the ICT adoption model. Construct
reliability and discriminant validity, multicolinearity and singularity tests are
performed first before the ICT adoption model is developed and analysed using the
Structural Equation Modelling techniques (SEM). The IBM AMOS version 20
software is used in the implementation of SEM. Research hypotheses are tested and
finally the model implications are discussed and explained.
Chapter Eight: Conclusion and Recommendations
This chapter gives the overall conclusions of the study. The key findings of the
study are highlighted and then conclusions, recommendations, contributions of the
study, as well as the implications of the study, are made. The chapter also
acknowledges limitations of the study and ends by giving suggestions for further
research.
PROBLEM IDENTIFICATION,
RESEARCH OBJECTIVES,
CHAPTER 1 SIGNIFICANCE OF THE STUDY,
SCOPE AND LIMITATIONS
LITERATURE REVIEW I: THE SME

CHAPTER 2 SECTOR AND ICT DEVELOPMENT IN
TANZANIA
CHAPTER 3 LITERATURE REVIEW II: ICT ADOPTION

IN SMES
21
CHAPTER 7 ICT ADOPTION MODEL DEVELOPMENT
CHAPTER 8 CONCLUSIONS AND RECOMMENDATIONS
Figure 1.1: Thesis Outline

1.9 Chapter Summary
This chapter has introduced the background information on the importance of the
SME sector in the economy of any country. The chapter has also highlighted the state
of the SME sector in Tanzania with great emphasis on the need for development of
the sector as a means of addressing critical social economic challenges faced by the
22
country such as employment creation, poverty reduction and the overall economic
growth of the country. The chapter further outlined the problem statement and set out
the research objectives. Moreover, a brief explanation of the justification for the
study as well as the scope of the research was made. The chapter ends by giving an
outline of the thesis.
CHAPTER TWO
LITERATURE REVIEW I: THE SME AND ICT SECTORS IN TANZANIA

23
2.1 Introduction
This chapter explores the SME sector and ICT development in the context of
Tanzania. The chapter is arranged as follows: First the SME sector is thoroughly
discussed, and the historical account and the current status of the SME sector in
Tanzania are made. Second, an overview of ICT development and the
telecommunication infrastructures in Tanzania are discussed whereby the National
ICT Policy and the Regulatory Framework are highlighted. Furthermore, important
statistics related to E-readiness are given. In conclusion, the chapter describes the
level of ICTs use in Tanzania SMEs.
2.2 The SME Sector in Tanzania
2.2.1 Historical Background of the SME Sector in Tanzania
The history of the SMEs sector in Tanzania is defined by three different times. The
colonial era which cover the timeframe from year 1891 to year 1960; the time soon
after independence in 1961 to the mid 1980s during which the country embraced
socialism political ideology and centrally planned economy; and the post
independence time from the mid 1980s to date, when the country switched from the
centrally planned economy to market oriented economy. According to the literature,
prior and during the early days of colonial era, the elements and skills of
entrepreneurship among the indigenous African people were represented by the
presence of the black-smiths and other crafts men (Naul, 2012).

24
During the colonial era, there were deliberate efforts in restricting the participation of
indigenous Africans in business activities (Naul, 2012; Rugumamu and Mutagwaba,
1999). Available literature shows that, during this time, the participation of
indigenous Africans in business was limited to ownership of very small enterprises
such as tiny shops whereas Arabs were allowed to operate as retailers only, and
Asians as medium wholesalers and retailers (Naul, 2012; Rugumamu and
Mutagwaba, 1999).
Soon after independence in year 1961, and in particular following the Arusha
Declaration in year 1967, Tanzania embraced socialism political ideology and
centrally planned economy (Naul, 2012; Rugumamu and Mutagwaba, 1999).
According to Naul (2012), the Arusha declaration focused on promoting public
ownership of the economy. This made the public sector the most important source of
investment and employment in the formal sector throughout this period. During this
time, marked achievement in social development especially in primary education,
health services delivery as well as water supply and sanitation were recorded in the
1970s and early 1980’s as a result of the socialism approach (Hamisi, 2011; Naul,
2012; Olomi, 2001a; Olomi, 2001b; Temu and Due, 2000). However, it was also
during this time, that the private sector was seriously marginalized (Naul, 2012;
Olomi, 2001a; 2001b). Within this socialism political framework, private business
sector was actively discouraged in favour of public enterprises that were government
owned, community based, or cooperative owned ventures (Naul, 2012; Olomi,
2001a; 2001b). Furthermore, all major undertakings in the private sector were
nationalized, and regulations restricting civil servants and leaders of the ruling
25
political party from engaging in business activities were instituted. This left
whatever was remaining of the business activities in the private sector to people who
had little or no education at all, since at that time, almost all indigenous educated
people were members of the civil service. Moreover, within this socialism policy,
decision making was based on a top down approach with the government being the
only organ that made all key decisions. According to Olomi (2001a; 2001b), the
reliance on Government discretion in decision making, contributed to the stifling of
development of entrepreneurial values such as the need for achievement, personal
initiatives and or creativity, willingness to take risks and related behaviours (Olomi
2001a; 2001b).
In the mid-1980s, Tanzania experienced a massive economic crisis that lasted for
over a decade and left the economy in shambles (Naul, 2012). During this time, the
government of Tanzania was forced to move from centrally planned public sector led
economy towards a more market oriented one, thus permitting the private sector
development. It was in this background that the SME sector in Tanzania was
established (Naul, 2012; Olomi, 2001a; 2001b). Since this time, the government
has taken a number of positive steps to ensure that the private sector and, in
particular, the SME sector is developed. These include for example, the formulation
of regulatory frameworks and policies such as the SME Policy of Tanzania (URT
2002) and the Tanzania ICT Policy (URT, 2003a), as well as the formation of
support institutions for the sector such as Small Industries Development Organisation
(SIDO) and the Commission for Science and Technology (COSTECH).
Nevertheless, looking at the historical background, it can be argued that, the SME
26
sector in Tanzania is a product of structural adjustment policies rather than of a
design and thus there are still a lot to be done to develop the sector further.
2.2.2 Current Status of the SME Sector in Tanzania
Olomi (2006) observed that, there is a shortage of comprehensive, accurate and up to
date data on the state of the sector (Olomi, 2006). According to Olomi (2006), for
almost two decades, most reports on the sector have relied on the data from 1991
National Informal Sector Survey (NISS) which became outdated and unreliable as
time went by. Olomi (2006) further noted that, although, in the same period, other
studies such as the 1992 Rural Informal Sector Survey (RISS) and the 1995 Dar es
Salaam Informal Sector Survey (DISS) were carried out and were the basis of some
of the reports on the sector, by far, in this period, the NISS (1991) remained to be the
only nationwide study of the informal sector. The author of this research observed
the same phenomenon. The most recent comprehensive study in the sector is the
URT (2012) which became available in 2013. It is thus natural that most of the
statistics given in this section are based on the URT (2012) and the NISS (1991)
reports.
According to the URT (2012), by September of year 2010, about 2,754,697
individuals in Tanzania owned and ran small businesses. The report further shows
that, while 86.2% of the small businesspersons owned and ran only one business, the
rest owned and ran two (12.8%) or more (1%) businesses. In total, all owners owned
and ran about 3,162,886 SMEs. The URT (2012) further reveals that, about 96.4%
27
of small business owners were sole proprietors, and about 3.5% were in partnerships
in which 77% of all partners were related (URT, 2012).
The URT (2012) report also show that, only 3.9% of all small businesses were
formally registered with the Business Registrations and Licensing Agency (BRELA);
of those unregistered, 96.4% were micro businesses, and 84.4% were classified as
small businesses. It is also reported that, only 44.7% of small business owners kept
financial records. Based on the survey findings, it was estimated that small
businesses contributed about 27% to Tanzania's GDP in 2010 and also made a
valuable contribution to poverty reduction (URT, 2012).
The major broad sectors identified by the URT (2012) include, the trade sector which
had about 1,758,896 SMEs and accounted for 55.4% of all SMEs, the service sector
which hand about 964,235 SMEs and accounted for 30.5%, of all SMEs and the
manufacturing sector which had about 430,381 SMEs and accounted for 13.6% of all
SMEs in Tanzania. The remaining 0.5% of SMEs were categorized as other services
(URT, 2012). Most of the SMEs in the other services category were related to
agriculture, forestry, and fishing (0.4%) and the remaining 0.1% were related to
water supply, sewerage, and waste management (URT, 2012). Out of the 3,162,886
SMEs in Tanzania over 90% were located in mainland Tanzania and the rest were
located in Zanzibar (URT, 2012). Table 2.1 show the general distribution of the
SMEs across the country while Figure 2.1 show the sectoral distribution of SMEs in
the country.
28
Table 2.1: Small Business Distribution by Area
Total Dar es Salaam Other urban Rural areas

3,162,886 466,049 935, 256 1,675,385
Source: URT (2012)
Source: URT (2012)
Figure 2.1: SMEs broad sector allocation by area
On employment creation, the URT (2012) report that, about 2.9% of all small
businesses had five or more employees; 66.1% were one person businesses, and
31.0% had two to four employees. Overall, the SMEs employed about 5,206,168
people; of those, 3,447,469 were owners/spouses; 964,246 were paid workers;
718,663 were relatives and friends; and 79,390 were apprentices. According to the
URT (2012) report, these figures indicate that employment in small business is
29
mainly a family affair. The low number of apprentices also indicates insufficient
skills development within the MSME segment. About one fifth of employed people
in the MSME segment were underemployed (URT, 2012).
NISS (1991) and Olomi (2005) observed that, the SME sector in Tanzania is
characterised by informal micro enterprises that had an extremely low incidence of
graduation to formal small and medium size. Olomi (2005) further argued that, these
enterprises remain marginal players, creating and sustaining low quality jobs. The
URT (2012) observations seem to confirm what has been observed by previous work.
2.3 ICT Development in Tanzania
2.3.1 Computer Technology
Development of the ICT sector with respect to use of the computer technology can
be traced back to year 1956 when the Public Works Department under the British
Administration introduced the hand punch input device for computer cards (Sawe,
1999). According to Sawe (1999) the device was used to capture data for processing
elsewhere. In year 1965, the government of Tanzania through the Ministry of
Finance introduced the first computer in the country (Mbamba, 2004; Naul, 2012).
This was immediately followed up by acquisitions of computers for other
government departments and parastatal organisations (Mbamba, 2004; Naul, 2012).
However, this investment failed to meet the expectations anticipated earlier and the
government of Tanzania banned the importation of computers for almost two
decades in the time from year 1974 to year 1993. Despite the ban on the importation
30
of computers, when trade liberalization started in year 1986, some organizations
started to import computers (Naul, 2012). A study conducted by Baker (1993) on
behalf of Tanzania Commission for Science and Technology (COSTECH) show that
by 1993 computer use in Tanzania was still very low as shown in Table 2.2. In
recognition of the importance of computer use in the social and development of the
country, the government of Tanzania lifted the ban on the importation of computers
as part of economic liberalization (URT, 2003a). Moreover, the government of
Tanzania abolished all taxes and duties on computers and peripherals since the
financial year 2001/2002 as a stimulus to the development of the ICT sector (URT,
2003a).
Table 2.2: Computer Population in Tanzania up to 1993
Year Number of Computers

Mainframes Mini Computers Microcomputers
1965 1 - -
1978 5 7 -
1984 13 15 79
1986 16 37 470
1993 - - 570
Source: Baker (1993)
2.3.2 The Telecommunications Sector
According to Behitsa et al. (2010), prior to economic liberalization of 1986 and up to
1993, the telecommunication sector was the monopoly of a state-owned company
Tanzania Telecommunication Company Limited (TTCL) operating under the then
Tanzania Posts and Telecommunications Corporation (TPTC). The TTCL held a
monopoly in the provision of communication, in particular, the telephone

31
communication service. In addition, TTCL was also responsible for the regulation of
the telecommunication sector (Behitsa et al., 2010).
As a part of a wider economic liberalization policy in the country, in year 1993, the
government of Tanzania paved the way for the liberalization of the
telecommunication sector through the Communication Act of 1993 (Behitsa et al.,
2010). This led to the split of the TPTC into the Tanzania Posts Corporation (TPC),
the Tanzania Telecommunications Company Limited (TTCL) and the Tanzania
Communication Commission (Behitsa et al., 2010). The Communication Act of
1993 was followed up by the National Telecommunication Policy (NTP) in year
1997 (URT, 1997). The National Telecommunication Policy provided the framework
for further reforms and private-sector engagement in the sector (Behitsa et al., 2010;
URT, 1997).
In year 2003, the Tanzania Communications Regulatory Authority (TCRA) was
established through the Act of the Parliament (Behitsa et al., 2010). TCRA is an
independent agency for the regulating and licensing of postal, broadcast and
communications industries. In particular, the TCRA is mandated to promote
competition and economic efficiency, protect consumer interests, grant licenses and
enforce license conditions, regulate tariffs, and monitor performance (Behitsa et al.,
2010). In the same year 2003, three other policies meant to provide frameworks and
guidance for further development and management of the sector was passed. These
include the National Information and Broadcasting Policy (URT, 2003b), the
National Postal Policy (URT, 2003c), and the Tanzania National ICT Policy (URT,
2003a). According to Behitsa et al. (2010), of the four mentioned policies, the
32
National Telecommunications Policy of 1997 is the basic document that provides
vision to 2020 (Behitsa et al., 2010). The policy aims at ensuring accelerated
development of telecommunications infrastructure and services in order to accelerate
access to telecommunication services by all sectors of the national economy as part
of the national development strategy (Behitsa et al., 2010). Tanzania’s ICT Policy
was approved by parliament in 2005 to provide Government guidance on ICT issues
(Behitsa et al., 2010). The policy aims at providing a national framework that will
enable ICT to contribute towards achieving national development goals; and to
transform Tanzania into a knowledge-based society through the application of ICT
(Behitsa et al., 2010; URT, 2003a).
The sector became fully liberalized in year 2005 and by year 2009; Tanzania had a
fully competitive telecommunications sector (Behitsa et al., 2010). As a result of
these efforts, in year 2009, Tanzania’s telecommunications sector was the fastest
growing sector of the economy (Behitsa et al., 2010; URT, 2010). In that year, the
sector recorded a 21.9% growth, up from 20.5% in 2008 and contributed 2.5% to the
GDP up from 2.1 % in 2008 (URT, 2010). The full impact of these efforts can be
explained better by looking at the trend of the sector development as explained in
section 2.3.3. Naul (2012) notes that, to date, ICT has grown in importance and
usage, and it is applied in almost all aspects of life from domestic, business and the
academic arena.
33
2.3.3 The Current Status of the ICT Sector in Tanzania
Available data show that since liberalization of the telecommunication sector began,
there has been remarkable growth. However, in order to fully comprehend the
figures and statistics that show the current status of the ICT sector in Tanzania, it is
important to take note on how the data in the sector has been archived over the years.
Under the TTCL, telecommunications sector licenses were based on the so called
vertical regime, whereby the license granted the operator “the right to operate a
telecom or a broadcasting network and right to provide services on that network”.
According to TCRA (2014a) and Behitsa et al. (2010), the convergence of
technology in the telecommunications sector necessitated the adoption of a
technology and service neutral licensing regime, technically known as the Converged
Licensing Framework (CLF). TCRA introduced the CLF in February 2006 with four
license clusters namely: The Network Facilities License (NFL), the Network
Services License (NSL), the Applications Service License (ASL) and the Contents
Services License (CSL). These licenses cover international, national, regional and
district market segments. At the same time, in a phasing out arrangement, TCRA
continued to issue licenses issued under the old regime in the areas of Public Data
Operators, Internet Service Providers (ISPs), Private Dedicated Data
Communications, and Postal and Courier Operators up until 2009.
Table 2.3 shows the trend of Voice Mobile Operators and ISP/Data operators in the
time frame from year 2003 to year 2009. In this time frame, the number of Voice
Mobile Operators increased from 5 in year 2003 to 7 in year 2009 while the number
of Internet Service Providers (ISP) grew from 22 to 62 in the same time frame.
34
Table 2.4 shows a summary of Telecommunications Operators licensed under the
Converged Licensing Framework.
Table 2.3: Licensed Operators between 2003~2009
Number/Year 2003 2004 2005 2006 2007 2008 2009

Voice Mobile Operators 5 5 5 6 6 6 7
ISP/Data Operators 22 23 23 25 34 60 62
Source: Behitsa et al. (2010).
Table 2.4: Licensed Telecom Operators
Type of License Market segment Number of Licenses as at

September 2014
Network Facilities International and National 8
National 12
Regional 2
Network Services International and National 8
National 7
Regional 3
Application Services International and National 15
International 1
National 55
Regional 11
District 1
Content Services National TV Stations 5
Regional TV Stations 1
District TV Stations 20
National Radio Stations 6
Regional Radio Stations 17
District Radio Stations 61
Source: TCRA (2014a)

35
As at September 2014, under the CLF, the TCRA has issued 22 Network Facility
Licenses, of which 8 are both international and national, 12 are national and 2 are
regional. Within the same time frame, 18 Network Services Licenses were issued, of
which 8 are both international and national, 7 are national, and 3 are regional.
Moreover, a total of 83 Application Services Licenses were issued, of which 15 are
both international and national, 1 is international, 55 are national, 11 are regional and
1 district license. Also, a total of 110 Content Services Licenses have been issued, or
which 5 are national TV stations, 1 regional TV station, 20 district TV stations, 6
national Radio stations, 17 regional radio stations and 61 district radio stations.
Behitsa et al. (2010) noted that, it is not easy to determine the actual number of
operational licensees, due to the nature of the licensing process. They recommended
a study to determine the different levels of operation for the different licensees.
Table 2.5 document the trend of the tariffs for prepaid voice services in the time from
year 2005 to year 2013. It can be observed that there has been a significant
reduction in the cost in this time frame. According to Behitsa et al. (2010), the
prepaid voice (telecommunication) per minute tariffs decreased by about 40% over
the period between year 2003 and year 2009.
Table 2.5: Weighted Average Prepaid Tariff for Voice Telecom
Year On-Net Off-Net East Africa International

Tariffs Tariffs Tariffs Tariffs
2005 219 292 499 1,580
2006 223 284 362 608
2007 199 286 331 465
2008 212 284 332 485
2009 147 292 332 453
36
2010 93 263 330 468

2011 62 187 331 476
2012 89 182 326 485
2013 165 240 399 752
Source: TCRA (2014b)
Table 2.6 show the trend on the subscription for both fixed and mobile networks and
the overall Teledensity for the time from year 2005 to year 2014. It can be observed
that, during this time, voice subscribers (both fixed and mobile) increased from
3,118,157 to 30,530, 609. On the same note, the Teledensity increased from 10% in
year 2005 to 68% in year 2015. This by all means is a major growth of the sector.
However, Behitsa et al. (2010) cautions on the meaning and use of this data since it
is known that most mobile phone users can have up to four registered lines with
different operators. When this is taken into account, the actual Teledensity may be
much lower than recorded here. Moreover, it is also known that, most of mobile
phone users are located in urban areas, and thus this figure does not tell anything
about the actual digital divide between rural and urban dwellers. Nevertheless, the
observed growth is remarkable.
Table 2.6: Subscriptions and Teledensity
Year Fixed Network Mobile Network Total Teledensity

Subscriptions Subscriptions Subscriptions (Penetrations)
2005 154,420 2,963,737 3,118,157 10%
2006 151,644 5,614,922 5,766,566 15%
2007 163,269 8,322,857 8,486,126 21%
2008 123,809 13,006,793 13,130,602 32%
2009 172,922 17,469,486 17,642,408 43%
2010 174,511 20,983,853 21,158,364 50%
2011 161,063 25,666,455 25,827,518 59%
2012 176,367 27,450,789 27,627,156 61%
2013 164,999 27,442,823 27,607,822 61%
37
2014 150,149 30,430,460 30,580,609 68%

Note: 2014 Statistics collected up to September 2014
Table 2.7 and Table 2.8 show the data for Internet subscribers by subscription type
and the estimates of Internet users by subscription type respectively. Compared to
voice subscription, it can be observed that internet penetration in Tanzania is still
very low.
Table 2.7: Internet Subscribers by Subscription Type
Year Internet Cafes Organizations/ Household Total

Institutions /Individuals
2008 350 3,055 248,433 251,838
2009 599 3,235 393,688 397,522
2010 736 3,329 533,090 537,155
2011 602 3,323 567,487 571,412
2012 636 3,245 1,133,417 1,137,298
2013 652 3,952 1,429,588 1,434,192
Table 2.8: Estimates of Internet Users by Subscription Type
Year Internet Cafes Organizations Household Total

/Institutions /Individuals
2008 126,000 2,444,000 993,732 3,563,732
2009 215,640 2,588,000 1,574,752 4,378,392
2010 264,960 2,663,200 2,132,360 5,060,520
2011 216,720 2,824,550 2,269,948 5,311,218
2012 228,960 2,758,250 4,533,668 7,520,878
2013 234,720 3,359,200 5,718,352 9,312,272

38
In a study conducted in year 2005, Souter et al. (2007) concluded that Internet
penetration in Tanzania was negligible. Behitsa et al. (2010) argue that, until 2009
Tanzania's Internet sector had remained underdeveloped as confirmed by data from
Table 2.8. Available literature (URT, 2003a) show that, one of the major constraints
on ICT development in Tanzania has been very high costs of service provision, a
problem attributed to the lack of adequate communications infrastructure.
According to Behitsa et al. (2010) this was true during the time when internal
communications infrastructure depended heavily on the traditional fixed-line
network that have limited reach and the international communications infrastructure
depended much on satellite and microwave links both of which are significantly
more expensive to operate and offer much less capacity than optical fibre cables.
In line with the ICT policy (URT, 2003a) in year 2009, Tanzania embarked on
building a high capacity state-of-art National ICT Broadband Backbone (NICTBB)
with a view to achieving its ICT vision (Pazi and Chatwin, 2013; Pazi et al., 2013;
Yonazi, 2012). The NICTBB project has four phases. Phase one is intended to
connect all regional administrative headquarters, phase two to connect some district
headquarters, phase three to connect all remaining district headquarters and phase
four to connect all villages in mainland Tanzania. The NICTBB phase one and two
were completed and became operational in year 2010 and 2012 respectively. The
two phases covers 7560 km out of the intended 10,000km for the whole project (Pazi
and Chatwin, 2013) and has connected 21 regional and some district headquarters.
39
At present NICTBB has Point of Presence (PoPs) in the administrative headquarters
of all districts of Tanzania Mainland.
Moreover, by design, the NICTBB addresses national needs and those of the
landlocked neighbouring countries by offering both cross-border connectivity and
access to International submarine cables (Pazi and Chatwin, 2013). On the same
context, the first international submarine fibre cable arrived in Tanzania in July 2009,
and became operational in year 2010 (Pazi and Chatwin, 2013; Yonazi, 2012). At
present Tanzania is connected to two operational international submarine cables by
SEACOM and EASSY cables in Dar es Salaam. Figure 2.2 show the NICTBB map
with regional and cross-border PoPs as well as the two international submarine
cables highlighted.
40
Source: URT (2014)
Figure 2.2: Tanzania National ICT Broadband Backbone Network Map
The advent of the NICTBB and the landing of international submarine optical fibre
cables have had a great impact on the development of the ICT sector in Tanzania.
For example, according to Yonazi (2012), by year 2012, the backhaul tariffs in
Tanzania were reduced by 99%, compared to year 2009 tariffs as a result of the
NICTBB. Yonazi (2012) further observes that, the abundance of national and
international bandwidth is fast enabling e-markets in Tanzania to emerge and are

41
growing, leveraging the growing popularity of e-transactions already backed by
Mobile electronic money transfer services. In conclusion, while a number of
constraints still remain to bring up ICT usage in Tanzania, this section has shown a
number of very positive initiatives that has been taken to push forward the sector.
2.4 ICT Adoption by Tanzania SMEs
As pointed out in section 2.2.2, there is very little literature on the SME sector in
Tanzania. Academic studies carried out about ICT adoption in SMEs in Tanzania
include the work of Isaya (2009), Jonathan (2004), Mbamba (2004), Muttayabalwa
(2007) and Naul (2012). Other studies such as that of Babiker (2001), Baynit
(2002), Issack (2006), Kavenuke (2009) and Mlay (2009) explores ICT adoption and
use in big business or government institutions. These studies together paint a
picture that ICT adoption and use in the country, in general, is still very low despite
government efforts such as enacting of ICT and SMEs policies and the tax incentives
for the acquisition of ICTs in particular computers. This opinion was also echoed by
three of the respondents during the pilot study of this research. The three
respondents on different occasions and locations were of the opinion that the country
as a whole was not ready to engage in serious business models involving massive use
of ICTs. One respondent cited the lack of a national payment system, local credit
cards, and a legislative framework appropriate for e-business. The other two cited as
examples, lack of adequate infrastructure to support their business ideas, high cost of
ICT business solutions, lack of trust in online business transactions and unplanned
frequent power cuts. The low usage of ICTs among Tanzania SMEs is also
42
highlighted by the URT (2012) which is one of the most recent comprehensive
studies on the topic.
According to the URT (2012), there are very few small businesses that own a laptop
or a computer. Those who own a computer use it mainly for business correspondence
(0.5%), business records (0.5%), accessing the internet (0.3%), providing computer
services to customers (0.1%), selling products and services online (0.0%), or private
use (0.1%). Also, most of the small businesses surveyed did not use the internet for
doing business. Only a few (0.5%) used the internet for getting information about
products and markets, and almost none used it for orders with suppliers and
customers. A very few (0.6%) used it for sending and receiving emails.
Note, however, that, the URT (2012) baseline survey choice of respondents included:
SMEs that conducted business by the road side (16.6%), SMEs that conducted
business at home (39.8%), and mobile businesses (8.4%). Only 18.4% of the
respondents of the URT (2012) were located in formal commercial area. In this
research, only SMEs located in formal commercial area were considered. In
addition, only SMEs with annual turnover of from 12,000,000/= and above were
considered in this research. The choices made were based on the fact that, it was
easier to reach the SMEs in these categories and get their detailed information than
the SMEs in the other categories. Given, the differences on the selection of the unit
of analysis for the two studies, that is the current study and the URT (2012) baseline
survey, this research is expected to give slightly different statistics on the state of
ICT adoption and use among Tanzania SMEs.

43
2.5 Chapter Summary
This chapter has highlighted the developments in the SME sector and ICT sector
respectively in the context of Tanzania. The chapter has shown that while there are
significant developments in the two sectors since structural adjustment policies were
implemented, a number of challenges still remain. One notable problem is the low
usage of ICTs among Tanzania SMEs. This problem is one of the primary concerns
of this research since the importance of ICT usage in businesses is well
acknowledged in the literature. Preliminary inquiries conducted by the author
suggest that, lack of regulatory framework, inadequate infrastructure, high costs of
ICT solutions, mistrust of online business transactions and frequent power cuts to be
some of the wider issues affecting ICT adoption and use in SMEs.
44
CHAPTER THREE
LITERATURE REVIEW II: ICT ADOPTION IN SMEs
3.1 Introduction
This study covers a topic that has been the focus of much research in the past decade,
and, for this reason, a lot of literature exists in the area with a variety of
terminologies used. An excellent summary of the research in this respect is found in
the work of Costello (2009), Korpelainen (2011), Mpofu et al. (2009) and
Sophonthummapharn (2008). Terminologies commonly found in the literature
include Information System (IS), Information Technology (IT), and Information and
Communication Technologies (ICT). According to Sophonthummapharn (2008),
these terms are not distinct from each other. They are connected to technology and
sometimes substituted and used interchangeably (Sophonthummapharn, 2008). For
this reason, all studies related to IS, IT, and ICT adoption are applicable to this
research.
The rest of this chapter is organised as follows: First a review of the literature is
made on the impact and importance of ICTs use in SMEs. Second, a review of past
research and current theories and models related to ICT adoption and use in SMEs is
made. Finally, the shortcomings of the existing models and knowledge gaps are
identified.
45
3.2 The Role of ICTs to Business
The literature has shown that, in recent years, ICT implementations in business have
become indispensable (Apulu, 2012; Costello, 2009; Galloway and Mochrie, 2005;
Hazbo and Arnela, 2010; Mbamba, 2004; Modimongale, 2009; Mpofu et al., 2009;
Naul, 2012; Sophonthummapharn, 2008). According to Apulu (2012), the use of
ICTs in business has revolutionised business practices, and this has made adoption
and use of ICT critical to all business. Previous studies show that, benefits of ICT
cut across all sectors of the economy and all fields of human activities (Apulu, 2012;
Ion and Andreea, 2008). A number of benefits attributed to ICT implementations in
business have been identified (Apulu, 2012). Generally speaking, ICTs are said to
enhance business operations, as well as organisational efficiency (Apulu, 2012; Ion
and Andreea, 2008; Udo and Edoho, 2000).
Naul (2012) note that, the benefits of ICTs to any business can be categorised into
two groups namely primary activities and supportive or secondary activities.
According to Naul (2012) primary activities involve such activities as inbound and
outbound logistics, operation and manufacturing programs, marketing, sales and
services. Inbound logistics consist of relationships with suppliers that include all
activities required to receive, store and disseminate outputs (Naul, 2012). Outbound
logistics involve activities required to collect, store and distribute the output such as
e-commerce and e-marketing through website (Naul, 2012). Operation and
manufacturing programs include all activities required to transform inputs into
outputs of products and services (Naul, 2012).

46
Secondary activities involve Procurement, Human Resource Management,
Technological Development and infrastructure that serves the company needs and
ties its various parts together (Naul, 2012). Infrastructure consists of functions or
departments such as accounting, legal, finance, planning, public affairs, government
relations, quality assurance and general management (Naul, 2012). Procurement is
the acquisition of inputs for the firm, while Human Resource Management consists
of all activities involved in recruiting, hiring, training, developing, compensating and
if necessary dismissing or laying off personnel (Naul, 2012). Technological
Development is concerned with the equipment, hardware, software procedures and
technical knowledge that is brought to bear in the firm’s transformation of inputs into
outputs (Naul, 2012).
According to Moreton and Chester (1997), there are some organisational initiatives
that are impossible to realise without the use of ICT. They further argue that “ICT
provides an unparalleled method of processing; analysing and communicating the
information from both inside and outside business that is needed to detect and
understand the patterns and pace of change” (Moreton and Chester, 1997). Apulu
(2012) claim that, ICTs offers many potential benefits to organisations. They make
the organisations more efficient, effective and competitive (Apulu, 2012). Fulantelli
and Allegra (2003) and Jonathan (2004) observe from the literature that, ICTs give
enterprises many possibilities for improving their competitiveness. Fulantelli and
Allegra (2003) claim that, “ICT provides mechanisms for gaining access to new
market opportunities and specialized information services such as distance
consulting, continuous training and new advisory modes”. The implication of this
47
observation is that, organisations that are capable of exploiting the benefits offered
by ICT can handle innovative processes such as knowledge management, customer
relationship management and supply chain management more effectively
(Fulantelli and Allegra, 2003, Jonathan, 2004).
3.2.1 Benefits of ICT Adoption in SMEs
Harindranath et al. (2010) noted that, regardless of the differences that exist between
big businesses and small businesses, SMEs are actually not different from larger
firms in one key aspect. That is, in order to survive and prosper in competitive
business arenas, both SMEs and larger firms need to successfully maintain a
competitive advantage. According to Dyerson et al. (2009), “the adoption and use
of ICT is widely seen as critical for the competitiveness of SMEs in the emerging
global market and has resulted in more effective use of time”. Skoko et al. (2008)
claim that, the role of ICT in SMEs must be considered very important since ICTs
have become “a major catalyst and enabler of organisational change”.
The literature (Evans and Wurster, 1997; Jannex et al., 2004) observes that, ICTs can
increase the richness and reach of the SMEs. Using ICTs, SMEs get the opportunity
to conduct business anywhere in the world. Moreover, companies can communicate,
collaborate and conduct transactions with their customers, distributors and suppliers
via the internet (Evans and Wurster, 1997; Jannex et al., 2004). In other words,
ICTs give SMEs the ability to participate in the global digital economy (Golding et
al., 2008). According to Ashrafi and Murtaza (2010), ICTs give SMEs the ability to
gain access to new markets, supply new products and services, add value to products,
48
changes the company business processes, increase performance and productivity of
the business, the ability to employ new business outlets and the ability to maintain
competitiveness edge.
According to Pavic et al. (2007) the advances in ICTs have given SMEs the
opportunity to attain a competitive advantage through process and product
innovations, marketing, and better customer responsiveness. Frempong (2007) and
Apulu (2012) observe that, the contributions of ICT to business development have
been all-encompassing and therefore businesses of all size including SMEs find that
without adequate ICT infrastructures, it is becoming increasingly difficult to
compete effectively in the world markets (Apulu, 2012; Costello, 2009; Frempong,
2007). The United Nations (2007) conclude that, the use of ICT enables SMEs to
participate in the knowledge economy and offers vast opportunities to narrow down
the social and economic inequalities that will assist SMEs in achieving broader
development goals (United Nations, 2007).
However, despite the recognition of the importance of ICTs in businesses, there is
evidence to the effect that, SMEs have not been responding quickly enough to
changes in ICT (Harindranath et al., 2010). That is, the take-up of such technologies
by SMEs has not been widespread (Harindranath et al., 2010). This shortcoming has
been a subject of intense research (including this work) in the past two decades. In
the next section, ICT adoption theories and models are thoroughly explored.
49
3.3 Review of ICT Adoption Theories and Models
3.3.1 Theoretical Foundation of Adoption Models
Gallivan (2001) as cited in Jeyaraj et al. (2006) and Korpelainen (2011) distinguishes
four theories namely the Diffusion of Innovation Theory (DOI), the Theory of
Reasoned Action (TRA), the Theory of Planned Behaviour (TPB), and the
Technology Acceptance Model (TAM) as the core theoretical frameworks upon
which most of the technology adoption studies are based. According to the literature
(Manueli et al., 2007; Venkatesh and Davis 2000), several hundred studies may be
found applying one or a combination of these theories to explain end-users' adoption
and acceptance of different kinds of ICT-systems and applications (Manueli et al.,
2007; Venkatesh and Davis, 2000). Apart from the four theories, the Social
Cognitive Theory (SCT) has also been used in some adoption studies
(Sophonthummapharn, 2008). In Korpelainen (2011) study, a ranking of most cited
theories of ICT implementation and adoption is made as depicted in Table 3.1. In
this ranking, TAM was found to be the most cited theory of the literature. Another
previous literature has also distinguished that in the past two decades, the TAM has
been the most influential model (Chuttur, 2009; Jeyaraj et al., 2006; Korpelainen,
2011). A summary for each of the four core theoretical constructs are presented
below while full details are presented in other studies such as Costello (2009),
Kripanont (2007), Manueli et al. (2007), Sophonthummapharn (2008) and Venkatesh
and Davis (2000).

50
Table 3.1: Theories of 20 Most Cited Articles and Books on ICT Adoption
THEORY AUTHORS, YEAR
1. Diffusion of Innovations (DOI)

DOI Rogers, (1965) (different
editions); Moore and Benbasat,
1991
TRA and DOI Combination Karahanna et al. (1999)
2. Theory of Reasoned Actions (TRA)

TRA Fishbein and Ajzen (1975;
1980)
TRA and TAM (comparison) Davis et al. (1989)
3. Technology Acceptance Model (TAM)

TAM Davis (1989)
TAM and TPB (comparison) Mathieson (1991)
TAM (replication) Adams et al. (1992)
TAM, TPB, and DTPB (comparison) Taylor and Todd (1995)
TAM2 (Extension of TAM) Venkatesh and Davis (2000)
4. Theory of Planned Behaviour (TPB)

TBP Ajzen (1991)
5. Unified Theory of Acceptance and Use

of Technology (UTAUT)
UTAUT combines eight models: TRA, Venkatesh et al. (2003)
TAM, and TPB
6. Model of the ICT Implementation Cooper and Zmud (1990)

Process
7. Information Systems Success Model Delone and McLean (1992)
Source: Korpelainen (2011)

51
3.3.2 The Diffusion of Innovation (DOI) Theory
The framework of innovation adoption process was introduced in 1962 and has been
extensively cited in the literature since then (Rogers, 2003). Rogers (1965; 2003),
comprehensively developed the Diffusion of Innovation (DOI) also known as the
Innovation Diffusion Theory (IDT) to explain how, why, and at what rate new ideas
and technology spread through any given group of individuals or organizations. DOI
is defined as “the process by which an innovation is communicated through certain
channels over time among the members of a social system” (Rogers, 2003). DOI has
been widely used in several disciplines including sociology, anthropology, education,
marketing, management, and others. Rogers (2003) categorized adopters of new
innovations or ideas as innovators (2.5%), early adopters (13.5%), early majority
(34%), late majority (34%), and laggards (16%) based on a bezel-shaped curve as
shown in Figure 3.1.
Source: Rogers (2003)

52
Figure 3.1: The Innovation Adoption Process
Rodgers (2003) explained that the innovation stretch through society in an S curve.
He was able to demonstrate through plotting variables how an adoption grew
exponentially (Rodgers, 2003). Rodgers (2003) also demonstrated that, change can
be promoted and will create a domino effect within a specified population. Rodgers
(2003) study examined the belief that an innovation decision was based on a cost-
benefit analysis and that people would adopt if they were convinced that the
innovation could produce some relative advantage to the idea it replaced. Diffusion
theory also states that "opinion leaders and their attitudes directly affect the
subsequent adoption” (Rodgers, 2003). According to Costello (2009) critics of the
DOI theory have remarked that, the theory attempts to explain a very complex reality
in a very simplified a fashion. However, Costello (2009) notes that, “this theory has
remained as a starting point upon which much IS research in this area is based”.
Costello (2009) further notes that, since this early work, research has consistently
revealed that perceived need (relative advantage) is key issue in the adoption
decision (Costello, 2009).
3.3.3 The Theory of Reasoned Action-(TRA)
TRA was formulated in 1967 in an attempt to provide consistency in studies of the
relationship between behaviour and attitudes (Ajzen, 1988; Fishbein and Ajzen,
1975; Werner, 2004). The main assumption of TRA is that individuals are rational in
considering their actions and the implications of their actions (decision-making). The
TRA model is shown in Figure 3.2. The TRA has two main concepts: “principles
53
of compatibility” and the concept of “behavioural intention” (Ajzen, 1988; Fishbein
and Ajzen, 1975). Principles of compatibility specify that in order to predict a
specific behaviour directed to a specific target in a given context and time, specific
attitudes that correspond to the specific target, time and context should be assessed
(Ajzen, 1988; Fishbein and Ajzen, 1975).
Belief’s and Attitude toward

Evaluation Behaviour
Behaviour Actual
Intention Behaviour
Normative Belief’s Subjective

and Norm
Motivation to copy
Source: Fishbein and Ajzen (1975)
Figure 3.2: Theory of Reasoned Action
The concept of behaviour intention states that, “an individual’s motivation to engage
in behaviour is defined by the attitudes that influence the behaviour” (Fishbein and
Ajzen, 1975). Behaviour intention indicates how much effort an individual would
like to commit to perform such behaviour. Higher commitment is more likely to
mean that behaviour would be performed. Behaviours intention is determined by
attitudes and subjective norms (Ajzen, 1991; Fishbein and Ajzen, 1975). An attitude
refers to an individual’s perception (either favourable or unfavourable) toward
specific behaviour (Werner, 2004). “Subjective norm” refers to the individual’s

54
subjective judgment regarding others’ preference and support for a behaviour
(Werner, 2004).
According to Werner (2004), TRA have some limitations in predicting behaviour
(Werner, 2004). The first limitation is that intention determinants are not limited to
attitudes, subjective norms, and perceived behavioural control (Ajzen, 1991). There
may be other factors that influence behaviour. Empirical studies showed that only
40% of the variance of behaviour could be explained using TRA or TPB (Ajzen,
1991; Werner, 2004). The second limitation is that there may be a substantial gap of
time between assessment of behaviour intention and the actual behaviour being
assessed (Werner, 2004). In that time gap, the intention of an individual might
change. The third limitation is that, TRA is a predictive model that predicts an
individual’s action based on certain criteria. However, individuals do not always
behave as predicted by those criteria (Werner 2004). In terms of ICT adoption, TRA
(as part of TAM) have been used to explain the adoption process from individual
perspectives (See Table 3.4).
3.3.4 The Theory of Planned Behaviour-(TPB)
The Theory of Planned Behaviour is an extension of TRA (Werner 2004) and thus
has the same main assumptions as TRA. TPB try to overcome TRA’s weakness by
adding Perceived Behavioural Control as an additional factor in determining
individual behaviour (Ajzen, 1991). Perceived behavioural control is an individual
perception on how easily a specific behaviour will be performed (Ajzen, 1991).
Ajzen (1991) argued that, perceived behavioural control might indirectly influence
55
behaviour. TPB have the same limitations as TRA in predicting behaviour, since
they are both based on the same underlying assumptions (Werner, 2004). TPB has
also been used to explain the adoption of IT (See Table 3.4). The TPB model is
shown in Figure 3.3.
Behavioural Attitude
Belief’s Toward the
Behavior
Normative Subjective Behaviour

Belief’s Norm Intention Behaviour
Control Perceived Actual

Beliefs Behaviour Behaviour
Control Control
Source: Ajzen (1991)
Figure 3.3: The Theory of Planned Behaviour-TPB
3.3.5 The Technology Acceptance Model-(TAM)
The Technology Acceptance Model (TAM) developed by Davis (1986), is an
extension of the Theory of Reasoned Action developed by Fishbein and Ajzen
(1975). Based on the TRA constructs, Davis (1986; 1989) proposes TAM to explain
and predict the behavioural intentions of individuals toward IT usage (Davis, 1986;
Davis, 1989; Davis et al., 1989). The TAM model hypothesizes that actual
behaviour is directly determined by behavioural intention to use, which is in turn
influenced by a user’s attitude toward perceived usefulness. Furthermore, an
individual’s attitude and perceived usefulness are influenced by perceived ease of
use. According to Korpelainen (2011), since TAM has been introduced, it has been
56
the most influential and widely used model in predicting and explaining the usage
behaviour of technology related products and services (Costello, 2009; Gallivan,
2001; Jeyaraj et al., 2006; Korpelainen, 2011). Figure 3.4 shows the TAM.
Perceived
Usefulness
Attitude Behavioural Actual use

External Towards intention to use
Variables
Perceived
Ease of use
Source: Davis (1989)
Figure 3.4: The Original Technology Acceptance Model
On the other hand, Van Akkeren and Cavaye (1999b) point out that, TAM has been
criticized for possible limitations since it emphasizes only the effect of technology
(Van Akkeren and Cavaye, 1999b). According to Van Akkeren and Cavaye (1999b),
critics of TAM conclude that, a limitation of the model is the fact that, outside
influence, economic factors, and influences on personal behaviour are excluded (Van
Akkeren and Cavaye, 1999b). In response to this, a number of extension models of
TAM such as TAM2 (Venkatesh and Davis, 2000) have been developed in an
attempt to address the issues raised by critics.

57
3.4. Models of ICT Adoption in SMEs
Having established the theoretical foundation of adoption models in general, this
section looks at previous studies in the context of ICT adoption in SMEs as an
important part of the literature review. Several models for ICT adoption by SMEs
has been suggested in the literature (Al-Qirim, 2005; Beckinsale and Ram, 2006;
Costello, 2009; Grandon and Pearson, 2004a; Levy and Powell, 2005; Manueli et al.,
2007; Mehrtens et al., 2001; Rashid and Al-Qirim, 2001; Van Akkeren and Cavaye,
1999b; Zappala and Gray, 2006). In these studies, inhibitors, drivers and factors that
influence the adoption of specific technology related products and or services by
SMEs were determined as summarised by Levy et al. (2005).
In a previous study, Levy et al. (2005) developed a set of eBusiness inhibitors and a
set of eBusiness drivers shown in Table 3.2 and Table 3.3 respectively. The
inhibitors and drivers were developed from data collected from a survey of 354
SMEs in the UK and the literature (Costello, 2009; Levy et al., 2005). According to
Costello (2009) and Levy et al. (2005), many of the identified drivers can be
explained as perceived benefits while as, many of the inhibitors are understood as
being due to management limitations. Costello (2009) claim that, a review of these
factors shows that, almost all of the factors could be seen as being drivers/inhibitors
of any technology. Furthermore, Costello (2009) argues that, the issues highlighted
by Levy et al. (2005) appear so often in the literature as to be considered very
significant in research.
58
Table 3.2: Inhibitors for Internet Adoption in SMEs
Inhibitor Source
Cost Implementation Costs Santarelli and D’Altri (2003);
Kendall et al. (2001); Grandon and
Pearson (2004a); Van Akkeren and
Cavaye (2000); Lawson et al.
(2003)
Limited financial resources Sharma et al. (2004); Chapman et
al. (2000);
Riemenschneider et al. (2003)
Need for immediate return Van Akkeren and Cavaye (1999b)
on investment
Security Concerns about Santarelli and D’Altri (2003);
confidentiality Kendall et al. (2001); Lawson et al.
(2003)
Fear of fraud Van Akkeren and Cavaye (2000)
Management Insufficient time spent on Bianchi and Bivona (2002);
planning Grandon and Pearson (2004a)
Insufficient knowledge or Klein and Quelch (1997);
experience of IS Premkumar and Roberts (1999);
Zhu et al. (2003); Sharma et al.
(2004); Kowtha and Choon (2001)
Inexperienced owner Van Akkeren and Cavaye (2000);
Klein and Quelch (1997);
Premkumar and Roberts (1999)
Technology Complexity requiring new Kowtha and Choon (2001); Van
skills Akkeren and Cavaye (2000);
Existing IS limiting future Van Akkeren and Cavaye (2000);
development Zhu et al. (2003)
Lack of trust in external IS Chapman et al. (2000)
suppliers
59
Limited in-house IS skills Santarelli and D’Altri (2003);

Kendall et al. (2001); Poon and
Swatman (1999); Sharma et al.
(2004); Chapman et al. (2000);
Lawson et al. (2003);
Source: Levy et al. (2005)
Table 3.3: Drivers for Internet Adoption in SMEs
Driver Source
Reduced operating costs Standing et al. (2003); Quayle and Christiansen
(2004); Quayle (2002); Kendall et al. (2001);
Sales and purchasing Quayle and Christiansen (2004); Jeffcoate et al.
cost reduction (2004); Tse and Soufani (2003); Riemenschneider et
al., (2003)
Improved range and Quayle and Christiansen (2004); Jeffcoate et al.
quality of services to (2004); Tse and Soufani (2003); Mehrtens et al.
customers (2001); Teo and Pian (2003); Sadowski et al. (2002);
Santarelli and D’Altri (2003); Quayle (2002); Daniel
and Grimshaw (2002); Riemenschneider et al., (2003)
Increased speed in Tse and Soufani (2003)
dispatch of goods
Finding suppliers Dandridge and Levenburg (2000); Teo and Pian
(2003);
Santarelli and D’Altri (2003)
Avoiding loss of market Santarelli and D’Altri (2003); Kendall et al. (2001);
share Riemenschneider et al. (2003)
Increase market share Standing et al. (2003); Quayle and Christiansen
(2004);
Daniel and Grimshaw (2002); Kendall et al. (2001)
Market intelligence Quayle and Christiansen (2004); Jeffcoate et al.
(2004); Mehrtens et al. (2001)
Improved trading Quayle and Christiansen (2004); Mehrtens et al.
relationships (2001)
Source: Levy et al. (2005)

60
In the work of Van Akkeren and Cavaye (1999a), lack of business benefit, mistrust
of the ICT industry and lack of time were identified as the main inhibitors to internet
adoption in SMEs. In a subsequent study, Van Akkeren and Cavaye (1999b), cited
by Costello (2009), Manueli et al. (2007), and Mpofu et al. (2009) identified and
classified key ICT adoption attributes in SMEs into two main categories: owner-
manager characteristics and small firm characteristics. Figure 3.5 below depicts the
model with three components: owner-manager characteristics; return on investment
(ROI); and firm/business characteristics for factors that contribute to ICT adoption in
small businesses.
OWNER/ MANAGER CHARACTERISTICS
 Perceived Benefit
 Computer Literacy
 Assertiveness
 Perceived Control
 Subjective norm
 Mistrust of IT industry
 Lack of time
RETURN ON INVESTIMENT IT ADOPTION IN SMEs
SMALL FIRM CHARACTERISTICS
 Organization readiness
 External pressure to adopt
 Customer/Supplier dependency
 Structural sophistication of the firm
 Size sector status
 Information Intensity
Source: Van Akkeren and Cavaye (1999b)
Figure 3.5: Factors Affecting ICT Adoption by SMEs

61
Van Akkeren and Cavaye (1999b), suggested that, more factors should be discovered
by research. In particular, Van Akkeren and Cavaye (1999b), called for inclusion of
return on investments in future research on adoption models.
Building up from the work of Van Akkeren and Cavaye (1999b), Costello (2009)
proposed a conceptual model of ICT adoption as presented in Figure 3.6. Costello
(2009) extended the work of Van Akkeren and Cavaye (1999b) by incorporating
return on investment in the model, as suggested by Van Akkeren and Cavaye
(1999b). Costello (2009) called for other factors such as the government
interventions (the moderating factors) and constraints variables to be studied in
future work as they were not tested in her model.
Owner/Manager Issues
 Education Experience Moderator
 Knowledge of IT Government Interventions
 Perceived Value
Measure of
ICT ADOPTION Investment
(Post Evaluation)
Firm Issues
 Organizational Readiness Constraints Variables
 External Pressure Size, Sector, Age, Status
 Strategy
Source: Costello (2009)

62
Figure 3.6: Model of ICT Adoption in Micro SMEs
Many other models in the literature (Levy et al., 2005; Peou, 2009) follow a similar
pattern. That is, they build up from one or more previous models, to explain a
particular situation or a new phenomenon. Table 3.4 give a summary of some of the
relevant studies of ICT adoption in SMEs as they appear in the literature. The
summary provides the main theoretical construct underlying each study, additional
factors, the technology examined in the study as well as the numbers of SMEs
examined.
Table 3.4: Summary of Relevant Studies of Technology Adoption in SMEs
Prior studies Theoretical Additional Factors Technology in No of

Basis the study SMEs
Al-Qirim (2005) DOI -Cost e-commerce 129
-Size adoption SMEs
-Information
intensity
-Vendor support
-External pressure
-CEO’s
Innovativeness and
Involvement
Celuch et al. TPB, SCT n. a Using Internet 303
(2007) for purchase SMEs
and
information
management
Chwelos et al. DOI -External pressure EDI adoption 286
(2001) -Organizational SMEs
readiness
Dawkins and TPB n. a Employee 87

Frass (2005) Involvement SMEs
(EI) program
Dholakia and SCT -Firm Size Internet 45
Kshetri (2004) -Prior technology adoption SMEs
use
-Prior media use
63
-Perceived
Competitive
Pressure
Grandon and TPB n. a e-commerce n. a
Mykytyn (2004) adoption
Grandon and TAM -Organizational e-commerce 100
Pearson (2004a) readiness adoption in SMEs
-External pressure USA
-Perception of
strategic value
Source: Sophonthummapharn (2008)
Table 3.4 (Continued)

Grandon and TAM -Organizational e-commerce 83
Pearson (2004b) readiness adoption in SMEs
-External pressure Chile
-Compatibility
-Organizational
support
-Managerial
productivity
-Strategic decision
aids
Harrison et al. TPB n. a IT adoption 162
(1997) SMEs
Iacovou et al. DOI -External pressure EDI adoption 7
(1995) -Organizational SMEs
readiness
Igbaria et al. TAM -Intra-organizational Personal 203
(1997) -Extra- computer SMEs
organizational
Jeon et al. -Management e-business 204

(2006) characteristics adoption SMEs
-Organizational
characteristics
-Environmental
Characteristics
Kaynak et al. TAM* -Perceived e-commerce 237
64
(2005) limitations adoption SMEs

Kuan and Chau DOI -Organizational EDI adoption 575
(2001) factors SMEs
-Environmental
factors
Kohn and Husig TAM, DOI n. a Innovation 33
(2006) software SMEs
Kula and Tatoglu TPB* -Attitude toward Internet 237
(2003) Internet use adoption SMEs
Lee (2004) SCT, DOI -Financial slacks Internet 71
-Innovativeness technologies SMEs
-Image
-Competitive
Pressure
Prior studies Theoretical Additional Factors Technology No of

Basis in the study SMEs
Lertwongsatien DOI -Organizational e-commerce 386
and factors adoption SMEs
Wongpinunwatana -Environmental
(2003) factors
Lin and Wu TAM n. a End-user 195
(2004) computing SMEs
acceptance
Mehrtens et al., DOI -Organizational Internet 7
(2001) readiness adoption SMEs
-External pressure
Mirchandani and DOI -Management’s e-commerce 62
Motwani (2001) time and adoption SMEs
Enthusiasm
-Computer
knowledge
-Information
intensity
-competition
-cost
Mole et al. (2004) DOI -Firm Size Soft process 218
-Firm technologies SMEs
characteristics
-Firm capabilities
-Competition
Olson and Boyer TAM -Comfort Internet 416
65
(2003) -Attitude purchasing SMEs

-Contextual factors
Poon and TAM n. a Internet 23
Swatman (1999) commerce SMEs
Premkumar and DOI -Top management EDI, Online 78
Roberts (1999) support access, e-mail, SMEs
-Firm Size Internet
-External pressure

Riemenschneider TAM, TPB n. a IT adoption 156
et al. (2003) SMEs
Riemenschneider TPB n. a Web-based 232
and McKinney e-commerce SMEs
(2001-2002)
Seyal et al. (2004) DOI, TAM
-Organizational e-commerce 54
factors SMEs
-Environmental
factors
Thong (1999) DOI -Management Information 166
characteristics System (IS) SMEs
-Organizational adoption
characteristics
-Environmental
Characteristic
Wymer and Regan TAM, TPB, n. a e-commerce 102
(2005) SCT, DOI adoption SMEs
Note: * = not explicitly stated in the study, n. a = not available

66
3.5 The Main Factors Influencing ICT Adoption in SMEs
It can be observed from a review of various models developed to aid ICT adoption
by SMEs (Table 3.4 above) that a number of factors that can influence ICT adoption
exist. In an attempt to determine the main ICT adoption factors in SMEs, Grandon
and Pearson (2004a) cited in Costello (2009) analysed ten ICT adoption models from
the literature. Grandon and Pearson (2004a) found that many of the same issues
recurred throughout the literature. This can also be deduced from Table 3.4 above.
In another study, Costello (2009) thoroughly analysed the literature and summarized
the main adoption factors as they appear in the literature. The summary of these
factors is given in Table 3.5. According to Costello (2009), the fact that many
factors are similar in so many models may indicate that the technology to be adopted
is irrelevant. The factors indicated in Table 3.4 together with other factors found in
the literature form the basis for this study.
Table 3.5: Summary of Main Adoption Factors in SMEs
FACTORS MODEL DEVELOPMENT SOURCE

The entrepreneur, The Firm, Longitudinal survey on SME Storey (1994)
The Strategy. issues
Owner manager 6 Literature sources – no Van Akkeren
characteristics, survey and
Firm Characteristics. Cavaye (1999b)
Organisational Readiness, 7 SMEs Case Studies Mehrtens et al.,
External Pressures, (Australia) (2001)
Strategy, Perceived Benefits.
Individual factors, 3 Literature sources – no Rashid and Al-
Organisational factors, survey (New Zealand) Qirim (2001)
Environmental factors,
Technological factors.
Organisational Readiness, 9 Literature sources, 100 Grandon and
External Pressure, email surveys Pearson (2004a)
67
Perceived Usefulness
Perceived Ease of Use
Entrepreneur, Underpinned by Storey and Levy and Powell
Firm, Strategy. surveys with WM SMEs (2005)
Personal factors, Underpinned by Rashid and National B2B
Economic Factors (cost and Al-Qirim and survey of 60 (2005)
resource), WM SMEs.
Environmental factors,
Relevant solution.
Source: Costello (2009)
3.6 Emerging Issues from Literature Review-Knowledge Gap Analysis
A systematic review of the existing adoption theories and models shows that
currently there are a number of ICT adoption models being offered to aid ICT
adoption by SMEs. However, looking at these models as they are summarized in
Table 3.4, it can be clearly seen that while some studies employ a single model,
many combine more than one model to better explain the adoption situation of a
particular technology. Additionally, it is observed that most studies include a variety
of additional factors in their studies. This reflects the insufficiency and inconsistency
of existing adoption models (Sophonthummapharn, 2008). The literature concludes
that, to date, the most effective model that support and aid SMEs in their adoption of
new technology has not been determined (Costello, 2009; Mpofu et al., 2009;
Sophonthummapharn, 2008). This is a research gap in the literature.
The literature also suggests the need for the development of mid-range theories, that
is, theories that provides for better understanding of the factors that drive or inhibit
the adoption and use of ICT within the specific context of SMEs (Al-Qirim, 2004;
68
Apulu, 2012; Beckinsale and Ram, 2006; Bharati and Chaudhury, 2006; Caldeira and
Ward, 2002; Chibelushi and Costello, 2009; Finchman, 2000; Mpofu et al., 2009;
Zappala and Gray, 2006). Available literature presents many empirical observations
from developed nations that have social economic settings that are quite different
from those of the so-called Least Developed Countries (LCD’s) like Tanzania while
comprehensive and systematic studies on ICT adoption and use in SMEs in LDC’s
countries south of Sahara including Tanzania are visibly missing (Apulu, 2012;
Mbamba, 2004; Mpofu et al., 2009). This is another research gap in the literature.
Costello (2009) and Sarosa and Zowghi (2003) observe that, despite the number of
technology adoption models available, most of research work to date has given most
attention to identification of drivers and barriers to adoption of ICT and there is still a
lack of strategy, models or frameworks that actually guide SMEs in the adoption
process. Costello (2009) further points out that, the most frequently asked question
by any SMEs is how to adopt successfully. In Tanzania, the studies carried out with
regard to ICT adoption by SMEs include the work of Isaya (2009), Jonathan (2004),
Mbamba (2004) Muttayabalwa (2007) and Naul (2012). The contribution of these
studies is mainly based on the identification of the factors that impede ICT adoption
or SMEs growth. None of these works offer a framework to help with strategic
intent of SMEs in their adoption situation as suggested by Costello (2009).
Moreover, these studies do not address post adoption issues such as how ICT can be
used to strengthen the competitiveness of the SMEs in order to ensure their survival
and growth. This is yet another research gap. In this study, an attempt is made to
bridge the second and third knowledge gaps identified from the literature. In
69
particular, the research concentrates on developing a model for aiding ICT adoption
by SMEs in the context of Tanzania.
3.7 Chapter Summary
In this chapter, the importance of adoption and use of ICTs by SMEs has been
explained. In addition, a thorough literature review of the available models of ICT
adoption by SMEs, including the theoretical foundation of the models has been
made. The chapter wind up by highlighting the knowledge gaps of the existing
models and set up the basis of this study.
CHAPTER FOUR
THEORETICAL FRAMEWORK AND HYPOTHESES
4.1 Introduction
In the two previous chapters the literature review that forms the basis of this study
has been covered. Many well-known theories and models which are useful for the
theoretical background for this research have been discussed. In view of the
literature review and gap analysis made, a summary of ICT adoption factors in this
research is presented in Table 4.1. These factors are in reality a summary of critical
issues raised by other models found in the literature (Costello, 2009). On a similar
note, a conceptual model for aiding SMEs ICT adoption process within the context
of Tanzania is proposed as shown in Figure 4.1. In the proposed model, the factors
of Table 4.1 are categorised into six major groups namely: Personal Factors,
70
Organisational Factors: Economic Factors; Government policy; ICT adoption and
SMEs Performance. Within each group, individual factors explored in this research
are shown. In Table 4.2, individual factors explored in this study within each
group are indicated together with the academic reference of each item.
The categorisation of model constructs is influenced by previous research by Van
Akkeren and Cavaye (199b) and Costello (2009). According to Costello (2009),
although the categorisation of factors into firm characteristics and owner-manager
characteristics is found in much research, it is often not as easy to define lines
between these areas Costello (2009). Costello (2009) argues that the small
company sophistication may be due to the owner-manager or the level of technology
within the firm already. In addition, the organisational readiness of a firm could be as
much linked to strategy or lack of strategy in the company as to the owner-manager’s
attitude to adoption (Costello, 2009). In order to ensure that this research is
thorough and that it captures all important elements, like in Costello (2009), evidence
from existing models both in practice and in academic literature and evidence from
SMEs literature surrounding the subject were used (Costello, 2009).
The proposed model builds up on Van Akkeren and Cavaye (1999b) and Costello
(2009) models and adds other factors found in the literature that capture the
uniqueness of the Tanzania situation. For example, the literature (Nkonoki, 2010;
Mbamba, 2004) have shown that, economic factors and in particular access to
financial support (Nkonoki, 2010), is one of the critical issues affecting the
development of the SME sector in Tanzania. Thus, economic factors are included
71
and tested in this study as an independent latent construct. Moreover, Costello
(2009) called for inclusion and testing for the moderating effect of Government
policies on the main adoption factors. In this work, the moderating effects of
Government policies on SMEs are tested, in the form of support programmes,
support infrastructure and taxes and tariffs.
It should be noted that, like in previous adoption studies in SMEs (Apulu, 2012;
Costello, 2009; Mpofu et al., 2009; Van Akkeren and Cavaye, 1999b) the
justification for inclusion of factors and or element in this study is based on the fact
that:
i. The element appears in credible academic literature as of concern to both
small and medium scale companies.
ii. The element appears in adoption and or growth models as of concern to micro
companies or SMEs.
iii. The element is proven by previous research to be of significance in the
adoption situation
iv. The element is proven by previous research to be of significance to SMEs.
The rest of this chapter is arranged as follows: First the operational definition of the
research variables and constructs are reviewed in order to acknowledge their validity
72
and importance in this study. Secondly the conceptual model is presented in order to
show the existing relationships between the model variables and constructs. Thirdly
the hypotheses are developed and finally a summary is given.
Table 4.1: Summary of ICT Adoption Factors in the Current Study
Factor in Name of the Factor in Source

current study Different Studies
Personal Individual factors Rashid and Al-Qirim (2001)
Factors Entrepreneur Levy and Powell (2005)
Starting resources of the Storey (1994)
Entrepreneur
Owner/manager Van Akkeren and Cavaye (1999b)
Characteristics
Owner/Managerial Mpofu et al. (2009)
Attributes
Owner Manager Issues Costello (2009)

Perceived Technological (innovation) Rashid and Al-Qirim (2001)
Benefit factors
Perceived Benefits Alarm and Noor (2009)
Mehrtens et al. (2001)
Perceived Usefulness + Grandon and Pearson
Perceived Ease of Use (2004a)
Owner manager characteristics Van Akkeren and Cavaye
(1999b)
Perceived Value Costello (2009)
Owner/Managerial Attributes Mpofu et al. (2009)
Perceived Benefits Chung (2006)
External Environmental Factors Rashid and Al-Qirim (2001)
Pressures Costello (2009); Grandon
External Pressures and Pearson (2004a);
Firm Characteristics Van Akkeren and Cavaye
(1999b)
Environmental Attributes Mpofu et al. (2009)
Costello (2009); Levy and
Strategy Powell (2005); Levy et al.
(2005); Storey (1994)
73

(1999b)
Strategic Value Grandon and Pearson
(2004a)
Chung (2006); Grandon and
Organisational readiness Pearson (2004a); Lee and
Cheung (2004)
Organization Context Sophonthummapharn
(2008)
Economic Context Skoko et al. (2008)
Costs Alarm and Noor (2009);
Love et al. (2005)
Resources Ghobakhloo et al. (2011b)
Ghobakhloo et al. (2012)

Organisational The Firm Levy and Powell (2005);
Factors Storey (1994)
(1999b)
Organization Context Sophonthummapharn (2008)
Firm Issues, Costello (2009)
Organisational Attributes Mpofu et al. (2009)
Organisational Factors Rashid and Al-Qirim (2001)
Organisational Readiness Grandon and Pearson
(2004a); Costello (2009)
(1999b)
Organisational Attributes Mpofu et al. (2009)
Government Environmental Factors Sophonthummapharn (2008);
Policies Teo et al. (1999)
Government Role Mpofu et al. (2009)
Government Support Alam and Noor (2009);
Apulu (2012)
Environmental Context Gono et al. (n.d)
74
Government Ghobakhloo et al. (2012)

ICT Adoption ICT Adoption Alarm and Noor (2009);
Apulu (2012); Costello
(2009); Mpofu et al. (2009)
Internet Adoption Chung (2006); Mehrtens et
al. (2001)
IT Adoption Van Akkeren and Cavaye
(1999b); Ghobakhloo et al.
(2012)
Electronic Commerce Adoption Grandon and Pearson (2004a)
SME Return on Investment Van Akkeren and Cavaye
Performance (1999b)
Measure of Success Costello (2009)
Performance Wu (2009)
4.2 Exploration of Research Variables
4.2.1 Personal Factors
From the SME and IS literature, there are a number of factors that directly affect ICT
adoption and business performance in general that can be attributed to the owner-
manager of the business (Apulu, 2012; Costello, 2009; Ghobakhloo et al., 2011a;
Kripanont, 2007; Sophonthummapharn, 2008). According to the literature, the
characteristics of the owner-manager can include: Perceived benefits, perceived
control, computer literacy, subjective norm and assertiveness (Apulu, 2012; Chung,
2006; Costello, 2009; Ghobakhloo et al., 2011a; Grandon and Pearson, 2004b; Levy
and Powell, 2005; Mehrtens et al., 2001; Rashid and Al-Qirim, 2001;
Sophonthummapharn, 2008; Van Akkeren and Cavaye, 1999b). Costello (2009)
noted that the subject area is very wide, meaning at times it is not possible to include
all variables and factors in a single study. Thus, in this research, the education
experiences of the owner-manager, the perceptions and attitudes o f the owner-

75
manager toward ICT and the ICT knowledge of the owner-manager, were explored
as Education experience, Perceived benefits, and Knowledge of IT, while perceived
control subjective norm, and assertiveness were not explored. The three variables
are further discussed in the next sections in order to establish the legitimacy of their
inclusion in this research.
4.2.1.1 Owner-Manager Education
According to Costello (2009), of all the factors that impact business growth the
education of the owner-manager tops the list. This may be attributed to the fact that,
education imparts people with knowledge and intellectual capabilities that enable
them to successful manage their daily undertakings. Costello (2009) suggest that,
there is a dichotomy as to whether the greater knowledge facilitates the development
of greater knowledge or simply more confidence in dealing with customers and
suppliers. According to Apulu (2012), to manage any business successfully, the
owner-manager must have both technical as well as general business knowledge.
Costello (2009) claim that, all previous research related to SME growth and
successful adoption of technology, has shown that the education of the owner-
manager is the most important factor as a proxy for business growth (Costello, 2009).
This justifies the inclusion of this factor in this model so that relationship between
this factor and other issues can be established.
4.2.1.2 Knowledge of ICT
In house knowledge of ICT is an important factor for a business to benefit from the
technology (Costello, 2009; Rahab and Hartono 2012; Sophonthummapharn, 2008).

76
According to the literature, small businesses typically lack in specialized ICT
knowledge and technical skills when compared to large businesses (Alarm and Noor,
2009; Apulu, 2012; Costello, 2009; Ghobakhloo et al., 2012; Mpofu et al., 2009;
Rahab and Hartono, 2012; Sophonthummapharn, 2008). Previous studies have also
shown that, small businesses failed to utilize ICT to due to lack of ICT knowledge
(Rahab and Hartono 2012). In Rahab and Hartono (2012), it is further shown that,
owners-managers that lack knowledge and awareness of ICT tend to reject the notion
that ICT could be of any use to their business as they had no idea of the benefits that
ICT could potentially offer (Costello, 2009; Rahab and Hartono 2012). This
would seem to imply that if these leaders could be educated on the benefits of ICT,
they may be more willing to adopt such technology and reap the benefits.
Van Akkeren and Cavaye (1999b) also cited in Costello (2009) established that, if
the CEO is innovative and knowledgeable about IT then adoption was more likely to
be successful. This was later corroborated by Mehrtens et al. (2001). Costello (2009)
claims that, although knowledge of ICT is a critical issue for all companies, “if that
knowledge has been gained through the educational experience of the owner-
manager that has an impact on the company turnover, perceived value as well as
adoption”. Costello (2009) argue that, since the perceived benefits to the company
are most often, perceived by the most influential individual in the company, the
owner-manager, then it is very important that they have enough knowledge of ICT to
either, make the decision themselves or recognise their lack of knowledge and seek
appropriate advice. Thus, this element is included in this study to complement the
owner-manager education.
77
4.2.1.3 Perceived Benefits
Perceived benefit (or perceived usefulness + perceived ease of use) is one of the core
constructs of behaviour intention for the different versions of the TAM model (Davis
1989; Venkatesh and Davis, 2000), the SCT model and the DOI models
(Sophonthummapharn, 2008). It is also a direct determinant of usage behaviour
(Kripanont, 2007). This construct has been used in many technology adoption
studies including computer acceptance (Igbaria et al., 1997; Lin and Wu, 2004), use
of Internet purchasing (Olson and Boyer, 2003), Internet commerce (Poon and
Swatman, 1999; Walczuch et al., 2000), Internet adoption (Chung, 2006; Kripanont
2007; Mehrtens et al., 2001), e-commerce adoption (Al-Qirim, 2005; Grandon and
Pearson, 2004a, Grandon and Pearson, 2004b; Kaynak et al., 2005; Mirchandani and
Motwani, 2001), IT adoption (Riemenschneider et al., 2003), EDI adoption (Iacovou
et al., 1995; Premkumar and Roberts, 1999), PDA acceptance (Yi et al., 2006),
eCRM adoption (Sophonthummapharn, 2008) and ICT adoption (Alam and Noor,
2009; Costello, 2009), and has been proven to be a major determinant of the adoption
decision.
According to Van Akkeren and Cavaye (1999b), perceived benefits affect technology
adoption in terms of the perceived ease of use and/or usefulness of the technology.
They argue that if an owner-manager does not perceive the technology in a positive
way, he or she will be reluctant to adopt (Van Akkeren and Cavaye, 1999a; Van
Akkeren and Cavaye, 1999b). Costello (2009) state that “SMEs adopt technology
because they perceive the advantage it represents over and above traditional
78
methods previously employed by the company”. Alam and Noor (2009) claim that,
existing literature has proved that the greater the benefits perceived by the SMEs the
higher the possibility of ICT adoption.
Costello (2009) identified efficiency, marketing, reliability and business
development as some of perceived benefits of ICT adoption in SMEs. According to
Chung (2006) perceived benefits of the Internet by SMEs include both direct and
indirect benefits. Cost reduction on operations and improved internal efficiencies are
some of the direct benefits while improved business processes and customer services
are some of the indirect benefits (Chung, 2006). OECD (2005) report that, “ICT is
able to improve information and knowledge management inside the firm and
increase the speed and reliability of transactions for both business-to-business (B2B)
and business-to-consumer (B2C) transactions”. Alam and Noor (2009) claim that
“ICT is able to offer enterprise a wide range of possibilities for improving their
competitiveness such as provide mechanisms for getting access to new market
opportunities and specialized information services” (Alam and Noor, 2009).
However, the literature shows that the outcome of adoption is not a simple and
straightforward issue (Alam and Noor, 2009; Apulu, 2012; Costello, 2009; Mpofu et
al., 2009; Sophonthummapharn, 2008) and this may discourage some SMEs from
adoption (Alam and Noor, 2009). Costello (2009) argue that, while there is evidence
that adoption of new technology can lead to improvements, on the other hand there is
not always a direct correlation straight forward as would be identified by those
involved in adoption initiatives (Costello, 2009). According to Alam and Noor

79
(2009), there are still many organisations that are not taking advantage of ICT
adoption despite the many perceive benefits that have been made available through
adoption with ICT. They thus recommend for perceived benefits to be taken into
consideration as one of the factors that affects ICT adoption in SMEs (Alam and
Noor, 2009). Costello (2009) stated that, “perceived benefits appear in so many
technology adoption models and research studies that it would have a natural place
in any new model”. Thus, in this research, perceived benefits are used to explore the
adoption and use of computer and Internet technologies in SMEs in the Tanzania
context. Reduced operation costs, competitive advantage, improved revenue and
profits, improved communications with customers, suppliers and employees and
advantage over traditional methods will be explored in this research under the
context of perceived benefits.
4.2.2 Organisation Factors
Organizational characteristics were also some of the most frequently studied factors
concerning ICT adoption in SMEs and SMEs growth (See Table 3.5 and Table 4.1)
According to Scupola (2003) organizational context is a source of structures,
processes and attributes that influence the uptake of new technological innovations
(Scupola, 2003). Previous research show that organisational characteristics can
include, organisational readiness, external pressure to adopt, customer and or
supplier dependency, business strategy, structural sophistication of the firm, sector
and status of the firm (Costello, 2009; Grandon and Pearson, 2004a; Levy and
Powell, 2005; Mehrtens et al., 2001; Rashid and Al-Qirim, 2001; Storey, 1994;
Thong and Yap, 1995; Van Akkeren and Cavaye, 1999b). In the work of Costello
80
(2009), extensive literature review on how organisation factors play out in small
business technology adoption situation is made. Other research such as Mpofu et al.
(2009) and Apulu (2012) also contribute in this area. In the next sections,
organisational readiness, external pressure to adopt, business strategy, age, sector and
status of the firm are explored in order to justify their inclusion in this research. In
this study, the structural sophistication of the firm was not measured, but was
captured in the strategy of the firm, as justified by Costello (2009). According to
Costello (2009) the evidence from the literature review shows that companies
become more structurally complex as they grow and that the larger the organisation
the more likely they are to consider strategy. This implies that structural
sophistication of the firm can therefore be captured via strategy (Costello, 2009). On
a similar note, external pressures were examined and the pressure of customer and/or
supplier is noted within this element. Thus, all the factors which represent the firm
characteristics are encapsulated into the firm, organisational readiness, strategy and
external pressures. Further discussion on the individual factors under this category is
given in the next sections.
4.2.2.1 The Firm
This element refers to status of the business, and the sector in which the business
operates (Costello, 2009; Mpofu et al., 2009). Business status can include the age of
the firm and size of the firm (Costello, 2009; Mpofu et al., 2009). Costello (2009)
noted that, factors relating to the firm itself have been shown to impact on adoption.
In fact, the literature has enough evidence about the importance of this factor in
81
adoption situations (Costello, 2009; Dholakia and Kshetri, 2004; Mpofu et al., 2009;
Thong and Yap, 1995) and thus the factor was measured and evaluated in this study.
The elements of this factor are further evaluated:
Business Age: Costello (2009) argue that, there are many reasons for incorporating
this element in a new study. One major reason is the fact that, it has been shown that
many SMEs experience rapid growth upon start up. In order to keep up with this
growth, SMEs adopt technology at a faster rate than at any other time during their
lifecycle (Costello, 2009; Levy and Powell, 2005; Storey, 1994). In view of this
observation, it would be anticipated that companies that have been in operation for a
number of years would adopt technology in a different pattern compared to those at
start up (Costello, 2009). That is, the technology infrastructure would be less
established and would perhaps have more problems regarding implementation.
According to Costello (2009), many researchers use age of company as a control
variable. Previous research found out that there was a negative relationship between
growth and age of the firm since many firms grow quickly initially and then slow
down (Storey, 1994). Costello (2009) suggests that newer studies should anticipate
the same results as Storey (1994) since as stated earlier, it seems that fast growing
young companies adopt ICTs more quickly than older, slower growing companies
(Costello, 2009).
Firm Size: Thong and Yap (1995) noted that, small businesses suffer from a special
condition commonly referred to as resource poverty. According to Thong and Yap
(1995), resource poverty results from various conditions unique to small businesses,
82
such as operating in a highly competitive environment, financial constraints, lack of
professional expertise, and susceptibility to external forces. In addition, Thong and
Yap (1995) found out that because of these unique conditions, small businesses are
characterized by severe constraints on financial resources, a lack of in-house IT
expertise, and a short-range management perspective. The observation by Thong
and Yap (1995) were later confirmed in other research. Dholakia and Kshetri (2004)
reported that firm size is an important determinant of a firm’s involvement and
decision process in acquiring ICTs. They argued that, larger organizations tend to
have the resources and tools that are necessary in initiating and deployment of new
information technologies (Dholakia and Kshetri, 2004). In addition, Dholakia and
Kshetri (2004) found out that, size as measured by the number of employees, is
particularly important for e-mail benefits and mobile communications.
In agreement with Thong and Yap (1995), Sophonthummapharn (2008) suggested
that, larger firms tend to have more of the necessary resources for adopting techno
relationship innovation than smaller firms. Sophonthummapharn (2008) identified
the common constraints seen in SMEs as limited financial resources, lack of in-house
information technology expertise, and a short-range management perspective. On a
similar note, Costello (2009) found out that, both size of company and age can be a
major indicator of success. When both elements are present, they can be a
compelling argument (Costello, 2009). Costello (2009) claim that, since small
companies are born and die with increasing frequency, a business that have been
around for a number of years with a number of employees can be an indicator of
future success too (Costello, 2009). However, Costello (2009) caution that, many
83
companies are sole traders and intend to stay that way choosing to run the business
for ‘life-style’ purposes despite the fact that they also have adoption needs. Costello
(2009) notes that, in the UK, the trend to run the business for ‘life-style’ purposes has
been on the increase.
In contrast with Dholakia and Kshetri (2004) observations, Locke (2004) did not find
any significant differences in ICT investment relating to firm size in terms of
turnover and number of employees. However, Costello (2009) claim that a number of
researchers have called for further work that includes size and age of firm. On a
similar note, Sophonthummapharn (2008) concludes that there enough reference to
support the inclusion of size in new studies such as Al-Qirim (2005), Dholakia and
Kshetri (2004), Lertwongsatien and Wongpinunwatana (2003), Premkumar and
Roberts (1999), and Thong (1999). Prior studies typically employ only one
dimension in measuring a firm’s Size (Sophonthummapharn, 2008) and this might
restrict the interpretation of the findings. In order to gain more insight about this
construct, this research measure business size in three dimensions, the number of
employees in the company, annual turnover, and capital investment.
Business Sector
Sheppard and Hooton (2007) suggest that business operations in some sectors
embrace ICT more readily than others. This view is supported by research such as
Dholakia and Kshetri (2004), Mpofu et al. (2009), Sophonthummapharn (2008) and
Thong and Yap (1995). The literature shows that, business in its particular industry
may experience the so-called competitiveness of environment (Thong and Yap,

84
1995) or the industry pressure to adopt (Sophonthummapharn, 2008). According to
Thong and Yap (1995), IT adoption can change a business environment in three
ways; First IT can change the industry structure and, in so doing, alter the rules of
competition. Second, IT adoption can also create competitive advantage by giving
businesses new ways to outperform their rivals. Finally, IT spawn’s new businesses,
often from within existing operations of the business. That being the case, a
business in an environment that is more competitive would feel a greater need to turn
to IT to gain a competitive advantage (Thong and Yap, 1995). On the other hand,
Thong and Yap (1995) note that, a business in a less competitive environment would
not be faced with a push to be innovative. It should be noted that, competitiveness of
environment or the industry pressure, is a kind of external pressure and thus the issue
is explored under the context of external pressure.
4.2.2.2 Organisational Readiness
Organisational Readiness in this context refers to the firm’s ability and or readiness
to deal with technology adoption issues, in this case computer and Internet
technologies. The literature has shown that there is a relationship between
organizational readiness and the level of technology adoption in a firm (Chwelos et
al., 2001; Costello, 2009; Grandon and Pearson, 2004a; Kuan and Chau, 2001;
Mehrtens et al., 2001; Sophonthummapharn, 2008; Thong, 1999; Venkatesh et al.,
2003). According to Sophonthummapharn (2008), the construct has a positive
impact on the decision to adopt. Chwelos et al. (2001) noted that, a firm’s IT
sophistication is positive and significantly contributes to the intent to adopt EDI.
Kuan and Chau (2001) observed the same trend as Chwelos et al. (2001). Mehrtens
85
et al. (2001), and Grandon and Pearson (2004a) found that, the availability of
technological resources in a firm is significantly related to e-commerce adoption.
The literature identifies availability of in-house knowledge of ICT, as one of the key
factors that contributes to organisation readiness (Costello, 2009; Jeon et al., 2006;
Mehrtens et al., 2001; Sophonthummapharn, 2008; Thong, 1999). In-house
knowledge of ICT can be represented by the availability of IT professionals in the
firm, and or the availability of IT knowledgeable employees who are non-IT
professionals (Costello, 2009; Mehrtens et al., 2001). According to Thong (1999),
SMEs with more IS knowledgeable employees were more likely to adopt IS. This
view was corroborated by Jeon et al. (2006) who confirmed that, the knowledge of
IT and e-business of employees has a positive relationship to the decision to adopt e-
business in SMEs. Mehrtens et al. (2001) identified the level of Internet use among
non-IT professionals rather than IT professionals as one of key determinants of
Organisation readiness. In addition, they argued that the availability of adequate
computer systems to access and use the Internet without major problems as being
equally important (Mehrtens et al., 2001). Sophonthummapharn (2008) observed
that, firms which do not have Technological Expertise may be unaware of new
technology. On the other hand, such firms may not want to take the risk of adopting
new technology and thus decide to postpone adoption of such technology until they
have sufficient internal expertise (Sophonthummapharn, 2008).
Costello (2009) claim that, the literature has identified the advice sought by a firm,
internal pressure, specialist skills in the company and implementation issues linked
86
to skills as indicators of organisation readiness towards ICT adoption. According to
Costello (2009), if companies had to seek advice regarding ICT adoption outside of
their own organisation, then it can be assumed that the necessary knowledge to make
the decision was not available in house. The literature shows that small business
owners-manager more often rely on advice from the social capital they have access
to (Cooke, 2007; Costello, 2009; Mpofu et al., 2009). Costello (2009) note that in
such situations, the majority of advice is given in informal way and it may not be the
right advice.
Furthermore, Costello (2009) argues that, internal pressures may be an influence on
the decision to adopt as well as an influence on the success or failure of an adoption.
According to Costello and Thompson (2004 and Costello (2009), internal pressure to
adopt is directly linked with the availability of ICT knowledgeable employees in a
company. Costello and Thompson (2004) and Costello (2009) note that, a company
with a higher level of ICT knowledgeable employees can experience internal
pressure to adopt (Costello and Thompson, 2004; Costello, 2009). Costello (2009)
argue that if employees are knowledgeable about ICT then they are likely to be
aware of the availability of, and the need for systems that can increase capacity,
capability or aid a marketing strategy (Costello, 2009). That can result in internal
pressure to adopt.
In addition, Costello (2009) note the fact that, there is enough well documented
evidence to prove that lack of right technological skills in a company can severely
impact an adoption. However, Costello (2009) argues that, there are technology
87
skills and knowledge needed to make the decision to adopt in the first place. Costello
(2009 claim that if the skills to align technology with the business are not available
then the decision to adopt will be flawed (Costello, 2009). The literature (Costello,
2009; Mehrtens et al., 2001; Mpofu et al., 2009) note that, most of post adoption
problems can be avoided if the adoption process is guided by well informed
decisions (Costello, 2009; Mehrtens et al., 2001; Mpofu et al., 2009).
The literature has also established that in adoption situations small companies mostly
concentrate on the implementation and the cost of the technology (Costello et al.,
2006). According to Levy and Powell (2005), this observation is attributed to the
fact that very few SMEs focus on strategy. Thus, as technology trends escalate and
data volumes increase the issue simply get worse (Costello, 2009). Costello (2009)
observe that, most small companies do not invest in training and employees learn
through experimentation. In such a scenario, implementation problems are likely to
be at the centre of adoption issues within small companies (Caldeira and Ward, 2002;
Costello, 2009). Costello (2009) further notes that, the use of outside help may also
aggravate the situation as they do not understand the business. Therefore, it can be
argued that many of the implementation problems are linked to skills as not having
the right skills can impact on the problems surrounding the post-adoption period
(Costello, 2009).
In view of this discussion, there is enough evidence from the literature to the effect
that organisation readiness is a key factor in ICT adoption (Chwelos et al., 2001;
Costello, 2009; Grandon and Pearson, 2004a; Kuan and Chau, 2001; Mehrtens et al.,
88
2001; Sophonthummapharn, 2008; Thong, 1999; Venkatesh et al., 2003) and thus is
included in this study. In this research, the advice sought by the firms in adoption
situation, the internal pressure to adopt, the availability of skilled IT personnel in a
firm, the level of ICT knowledge among non-ICT professionals, availability of
resources in the firm and implementation issues reflecting these concerns were
explored to establish the readiness of small firms in ICT adoption.
4.2.2.3 Business Strategy
The word strategy can be defined in different ways (Jonas, 2000; Mintzberg, 2001).
According to Mintzberg (2001), a strategy can be defined as a plan or some sort of
consciously intended course of action, a guideline (or set of guidelines) t o deal
with a situation (Mintzberg, 2001). In the context of business management, a
business strategy can simply be defined as “a plan that allows a company to
accomplish its mission (Jonas, 2000). Levy and Powel (2005) label strategy as “the
actions taken by the firm once in business”. Levy and Powel (2005) site market
positioning, new product introduction and technological sophistication as the key
drivers. According to Storey (1994) it is the relationship between the owner-
manager, the business strategy and its context that is important for business growth
(Levy and Powel, 2005; Storey, 1994). Costello (2009) argue that, there is a clear
connection between businesses performance and the strategic planning process of a
company. According to Costello (2009), IS theorists believe that ICT adoption and
investment should be made in the context of IS/ICT strategy and should support
business strategy.
89
However, small businesses are generally known to lack in strategic planning as a
management tool (Costello, 2009; Nkonoki, 2010) and this has a big impact on their
decisions (Costello, 2009). Thus, this construct has been used in many technology
adoption studies in order to explain how SMEs could be assisted in the adoption
process. It is for this same reason that the construct is included in this research.
Nkonoki (2010), cited lack of business strategy as one of the major factors affecting
SMEs development in Tanzania. Some of notable studies in the literature that have
strategy as one of studied factors include: e-commerce adoption (Grandon and
Pearson, 2004b), ICT adoption (Costello, 2009; Van Akkeren and Cavaye, 1999b),
Internet adoption (Levy and Powel, 2005) and e-business (Levy and Powel, 2005).
In the work of Costello (2009) strategy is interpreted as “the elements that are most
likely to influence the owner-manager in their decision-making when considering
adoption of new technology” This interpretation reflects on a previous work in the
area of strategic intent by Levy et al. (2001). This research adopts the same
interpretation of strategy. According to Costello (2009) other elements including
technological sophistication may be an important part of strategy. However, Costello
(2009) notes that, this information can be captured via the infrastructure. On a
similar tone, Costello (2009) notes that rolling programs of technology updates and
frequent use of the latest technologies to improve the business can be considered to
be strategic decisions. On the other hand, if a change that take place in a company
are not planned and the company reacts to customer demands in a reactive manner
that can be taken as lack of strategy (Costello, 2009).

90
Costello (2009) identified the level of decision making undertaken, reason for
purchase and Significance of purchase, as key elements for measuring strategy
(Costello, 2009). These elements are also used in this study to explore issues related
to strategy and they are explored in this section.
The level of decision making undertaken: According to Costello (2009), the
literature has shown that in spite of the simplified structure in micro-companies,
decision making can still be very complex within small companies. Costello (2009)
argues that, simplified structure in micro-companies does not necessarily lead to less
complexity in decision making as personal beliefs and knowledge can have a major
impact. The work of Levy and Powell (2005) established that small companies adopt
different management styles as they grow. Thus, who makes the decision to adopt
can be an indication of their strategic maturity (Costello, 2009). According to
Costello (2009), delegation of decision making may show that a company is starting
to grow and adopt a more formal management style. On the other hand, lack of
delegation of decision making demonstrates the autocratic style of management
which is typical for small companies (Bianchi, 2002; Costello, 2009). The literature
has shown that management capacity is limited in micro-company (Costello, 2009;
Matlay 1999; Mpofu et al., 2009; Sophonthummapharn, 2008; Thong and Yap,
1995). Therefore, just like in previous research (Costello, 2009), it is anticipated that
most of SMEs studied will have an owner-manager who makes the decision
themselves (Costello, 2009). After all, ICT adoption decisions are not fundamentally
different to other decisions and thus are prone to the same lack of understanding as
other areas (Costello, 2009; Remenyi et al., 1991).

91
Reason for Adoption: Costello (2009) claim that, when the reasons for adoption is
included within strategy, it helps to define the objectives that a company may have
within an adoption situation. Available literature shows that the reasons given by
companies to justify their investments in ICTs are many and varied (Costello, 2009;
Costello and Sloane, 2003; Levy et al., 1998; Storey, 1994). Costello (2009) note
that, while some of the given reasons are deliberate strategic investments, others
reflects a decision to adopt brought about by peer pressure. Poon et al. (1996)
discovered that, even the excitement surrounding the technology can drive small
companies to adopt. Therefore, Costello (2009) suggests that, when the reason to
adopt is explored, it helps to shed light on the thinking within the company (Costello,
2009). Costello (2009) argue that, many SMEs are not aware of strategy, and do not
implement strategy using a formal method. Thus, Costello (2009) recommends that
including the reason for the adoption as an element in Adoption models will provide
evidence of strategic intent as described by Levy et al. (1998).
Significance of Adoption: Costello (2009) allege that the significance of the
purchase (adoption) has been defined for exploration from the literature regarding
strategy. This area according to Costello (2009) provides insights into the attitudes
the company or the owner-manager may have towards technology being adopted.
Costello (2009) note that the significance’ for a particular company may vary from
being a financially significant purchase which may indicate that the company is
focused on short term, cost reducing, profit driven strategies, or merely that they
consider ICT as a yet another infrastructure cost. In addition, a company may decide
92
that an adoption is of technological significance thus suggesting that technological
competence could be an issue (Costello, 2009).
Previous research has shown that, exploring the significance of the purchase as
defined by the owner-manager within a micro-company will give useful insights to
the adoption situation and the strategy that they have adopted (Costello, 2009; Levy
and Powell, 2005; Mehrtens et al., 1999; Van Akkeren and Cavaye, 1999a). Levy et
al. (2005) also cited by Costello (2009) identified drivers and inhibitors of internet
adoption and attributed them to issues as diverse as an inexperienced owner,
complexity requiring new skills to security concerns. Thus, this element will
highlight the emphasis that the owner-manager and/or decision maker places on their
investments (Costello, 2009).
It should be noted that, the reason of the purchase differs from the significance of the
purchase in that the reason of the purchase may be shown to be due to pressure from
external as well as internal sources (Costello, 2009). However, the significance of
the purchase may indicate attitudes, bias and culture towards the adoption (Costello,
2009).
4.2.2.4 External Pressures
External Pressures refer to pressures emanating from outside the firm and they can
affect the firm in a positive or negative way (Sophonthummapharn, 2008).
According to Sophonthummapharn (2008), external pressure can be categorised as
customer pressure, competitive pressure and industry pressure. Ghobakhloo et al.

93
(2011b) and Ghobakhloo et al. (2012) cite the literature to support the notion that, for
many companies, pressures to keep up with the competition, providing a means to
enhance survival and/or growth, managing changes, promoting services to
customers, and staying competitive and/or enhancing innovation abilities have forced
SMEs to adopt IT (Ghobakhloo et al., 2011b; Ghobakhloo et al., 2012). In
addition, they note that, as small businesses are susceptible to customer pressure.
That is, these firms adopt IT as a result of demand from customers to develop the
efficiency of their inter-organizational dealings (Ghobakhloo et al., 2011b;
Ghobakhloo et al., 2012).
Customer pressure can be defined as the behaviour and or demand of customers that
force a firm to adopt a technology in order to keep and satisfy customers
(Sophonthummapharn, 2008). Review of the literature has shown that small
businesses are increasingly under pressure from customers and suppliers to include
technology into the way they run their business (Costello, 2009; Ghobakhloo et al.,
2011b; Grandon and Pearson, 2004a; Grandon and Pearson, 2004b; Mehrtens et al.,
2001; Mpofu et al., 2009; Van Akkeren and Cavaye, 1999b). Based on this fact,
Costello (2009) conclude that, it is thus very important to include external pressures
within any new technology adoption model (Costello, 2009).
According to Costello (2009), when many of the SME customers are adopting new
technology at a fast rate, the SME may find it very difficult to keep pace, and that
exerts pressure on the SME to adopt (Costello, 2009). However, Costello (2009)
notes that, most often the SMEs fail to see that external issues may be exerting
94
pressure on them to adopt (Costello, 2009). She argues that, SMEs should view the
external pressure as strategic method or indeed recognise that this is a driver
(Costello (2009). Thus, Costello (2009) explored external pressures by looking at
the influences on the purchase and by identifying external pressure in the decision-
making process. Sophonthummapharn (2008) note that, customer satisfaction leads
to customer loyalty which eventually leads to a firm’s profitability
(Sophonthummapharn, 2008). The implication of this is that, SMEs may feel
pressured to adopt a technology in order to please their customers.
Competitive Pressure can be defined as “the intensity level of competition that causes
a firm to feel it is underperforming.” (Sophonthummapharn, 2008). According to
Sophonthummapharn (2008), in tough and competitive situations, firms are pressured
to seek and retain competitive advantages. When a firm recognise that a particular
technology has made its rivals more profitable than it is, that firm suffers from
Competitive Pressure and will feel forced to adopt the same or similar technology.
Additionally, the sector in which a firm operates may also influence the decision to
adopt a particular technology (Costello, 2009; Ghobakhloo et al., 2011b;
Sophonthummapharn, 2008; Thong and Yap, 1995). According to the literature
(Sophonthummapharn, 2008), when a firm see more and more companies in its
sector adopting a particular technology, (it could be either competitors or business
partners), that firm may feel the pressure to adopt the same technology, in order to
retain its competitive advantage. This is the Industry pressure. This review shows
that external pressure can play a key role in adoption situations. This justifies the
95
inclusion of external pressure in this study. In this research, external pressures are
explored by looking at customer pressure, supplier’s pressure, competitive pressure
and industry pressure. In addition, influences on adoption are also explored in an
attempt to obtain other significant insights into the social and economic circles that
SMEs move in and are influenced by including the social capital that they have
access to (Costello, 2009).
4.2.3 Economic Factors
Economic Factors is another construct that appear widely in technology adoption
studies. This factor refers to the availability of financial resources for adoption and
maintenance of technology for a company (Sophonthummapharn, 2008; Ghobakhloo
et al., 2011a; Ghobakhloo et al., 2012; Costello, 2009). In general, adoption costs
may include: acquisition costs, implementation costs, operational costs, and
maintenance costs (Love et al., 2005; Sophonthummapharn, 2008). In this
research, Economic factors are studied. In particular, this research looks at access to
financial support and costs of ICT systems. The two elements are further explained
in the next sections.
4.2.3.1 Access to Financial Support
According to Nkonoki (2010), capital constraints and lack of access to finances are
obviously the most discussed by many scholars as major factors limiting business
growth. Ghobakhloo et al. (2012) argue that, “financial resources are one of the
most crucial resources which are known as the key SMEs performance requirements
96
and are the critical success factors”. Review of other previous research has shown
that, firms with adequate financial resources will be better prepared in an adoption
situation (Chwelos et al., 2001; Kuan and Chau, 2001; Lee, 2004;
Sophonthummapharn, 2008). Chwelos et al. (2001) found that, the intent to adopt
EDI in SMEs is also influenced by the availability of financial resources. While
examining EDI adoption in SMEs, Kuan and Chau (2001) noted that, firms that
adopt perceive lower financial costs than non-adopters do. Sophonthummapharn
(2008) argue that, “firms that have adopted technology, are more likely to have more
financial resources to implement the technology than non-adopting firms”. Costello
(2009) point out that, financial factors have strong influence on ICTs diffusion to
SMEs. In light of these observations, it can be argued that, access to financial
support can play a critical role in technology adoption situation and thus its inclusion
in this study justified.
Generally, SMEs are known to lack in financial resources when compared to large
businesses (Costello, 2009; Ghobakhloo et al., 2012; Ghobakhloo et al., 2011a).
Ghobakhloo et al. (2012) claim that, most SMEs suffer from the lack of sufficient
financial resources and most owner-managers invest their own personal assets.
Ghobakhloo et al. (2012) further observe that, for obvious reasons, limited financial
resources make small businesses to be very careful about their investment and capital
spending, since an inaccurate IT investment decision can cause severe financial
consequences for SMEs (Ghobakhloo et al., 2012). In extreme situation that may
lead to insolvency and economical failure (Ghobakhloo et al., 2012). Peou (2009)
argued that small firm financing is a major concern in Cambodia. According to Gono
97
et al. (2014), SMEs have been found to be financially constrained due to lack of
access to external financing, in both developed and developing countries. Nkonoki
(2010) noted that, capital constraints and in particular in-access to finance is one of
the major factors hindering SMEs growth in Tanzania. In-access to finance is a type
of capital constraint, and it refers to inaccessibility of funds for expanding, running
or maintaining a business undertaking (IFC, 2010; Nkonoki, 2010). Other studies
have also shown that one of the major problems affecting the SME sector in
Tanzania is limited access to finance (Kuzilwa, 2005; Olomi, 2006; URT, 2002).
According to DBIS (2013b), overdrafts and bank loans are the most common sources
of additional finance for SMEs. In addition, DBIS (2013b) argue that, the most
significant advantage overdrafts and bank loans have over raising equity is that,
neither involves relinquishing any share of ownership or control of the business. In
this research, the interactions between Entrepreneurs and financial institutions are
explored.
4.2.3.2 Costs of ICT Systems
According to Ghobakhloo et al. (2012), past studies has shown that cost of ICT
systems affects adoption of the technology in SMEs (Alam et al., 2009; Costello,
2009; Ghobakhloo et al., 2012; Ghobakhloo et al., 2011a; Sophonthummapharn,
2008). Ghobakhloo et al. (2012) state that, owner-managers considers the elements
of IT costs (hardware and software) closely during IT adoption process within SMEs.
In a study conducted in Holland, Walczuch et al. (2000) found out that, high costs
were the major reason for Dutch SMEs not having internet access and their own
98
websites. Ghobakhloo et al. (2012) claim that, most businesses in United States of
America (USA), have significant difficulty affording the costs of ICT tools while
90% of these businesses consider lack of financial resources and skills as the main
barriers to ICT adoption. Ghobakhloo et al. (2011b) further observes that, in
general SMEs cannot afford to adopt ICT or reap its benefits through the effective
use of ICT, in short or medium period of time because of the financial constraints
experienced by the majority of SMEs, as well as the high start-up costs of ICT or
very expensive software or ready-to-use online packages.
On the other hand, Premkumar (2003) argues that adoption cost is not a significant
factor in determining IT adoption within SMEs. According to Ghobakhloo et al.
(2012), this observation is empirically supported through a study conducted in
Malaysia in 2009, and found out that “despite IT costs being one of the major risks
perceived by Malaysian SMEs, there are no significant associations between high
ICT infrastructure costs and ICT adoption in these businesses” (Ghobakhloo et al.,
2012). Alam and Noor (2009) argues that, the literature has shown that there is a
direct and significant relationship between cost and adoption of technology. The
lower the cost of adoption the higher the new innovation such as the ICT will be
adopted by the company and vice versa.
However, Ghobakhloo et al. (2012) maintain that, in spite of decreases in the initial
and direct costs of IT adoption in recent years, cost is still regarded as an essential
issue when it comes to adoption and implementation of ITs in SMEs. Love et al.
(2005) also note that, although the prices of hardware and software have decreased
99
noticeably and have become more affordable, the estimation of IT adoption cost is
difficult (Love et al., 2005). This difficulty leads to uncertainty about anticipated IT
benefits, and thus cost is still a significant barrier to IT investment in SMEs (Love et
al., 2005). In addition, Love et al. (2005) observe that, direct costs are usually
underestimated and viewed as the cost of hardware, software and installation while in
actual fact, IT’s direct costs result from the implementation of new technology (Love
et al., 2005).
The literature suggests that, beside initial costs of software and hardware and
installation costs, IT adoption expenses also go beyond direct costs and include costs
of staff training and motivation, transformation from old to new systems in terms of
procedures and organizational structure, as well as post IT implementation expenses,
cost of management time and effort, productivity losses and finally expenses
encompassing the costs of maintenance and development (Love et al., 2005;
Ghobakhloo et al., 2012).
In view of this discussion, it is seen that, cost of ICT systems affects the adoption of
the technology, especially when we take into account that, most SMEs are
characterised by restricted financial resources (Costello, 2009; Ghobakhloo et al.,
2012; Ghobakhloo et al., 2011a). Thus, the inclusion of this element in this study is
justified.
100
4.2.4 Government Policies
Government policies have featured in previous technology adoption studies mainly
appearing under the context of environmental factor (Sophonthummapharn, 2008).
Due to the peculiar nature of SMEs, Costello (2009) suggested the inclusion of this
construct in any new model of ICT adoption. After all, available literature shows
that, there is a significant positive relationship between IT adoption in businesses and
government support (Alam and Noor, 2009; Costello, 2009; Ghobakhloo et al., 2012;
Sophonthummapharn, 2008).
According to Ghobakhloo et al. (2012), SMEs generally depend more on external
resources and support than other companies, because of their size and lack of
resources. Sophonthummapharn (2008) note that, SMEs are typically limited in
financial resources and IT knowledge as compared to larger firms and therefore,
SMEs need more government support than larger firms. Costello (2009) argue that,
due to lack of resources such as expertise in the planning process and in the
technology, SMEs have a strong need for impartial advice independent from vendors.
In addition, many SMEs need practical help in the form of trusted ICT Consultancy
as well as grants or loans (Costello, 2009). This is the type of support that can be
provided by the government. Ghobakhloo et al. (2012) claim that, government
initiatives and policies could directly and or indirectly stimulate the development of
IT infrastructure and information provision to energize faster technology diffusion.
This notion is shared by other authors such as, Alam and Noor (2009), Costello
(2009), Sophonthummapharn (2008) and Mpofu et al. (2009). Costello (2009)
suggests that, ever since it was established that eBusiness is a major source of
101
competitive advantage, governments around the world are creating funding schemes
and initiatives to facilitate ecommerce adoption in small businesses (Costello, 2009).
In addition, there have also been many endeavours by authors around the globe to
familiarise and group barriers and enablers in this major area of concern (Costello,
2009). In view of this, the literature shows that, government support for facilitating
information transfers to SMEs is incrementally increasing (Alam and Noor, 2009;
Costello, 2009; Ghobakhloo et al., 2012; Sophonthummapharn, 2008).
However, the literature suggests that, generally, governmental assistance have not
been advantageous to the SMEs (Costello, 2009; Ghobakhloo et al., 2012; Nkonoki,
2010). According to Costello (2009) the literature has suggested that, many
intervention projects have addressed the needs of the policy makers’ career rather
than the intervention itself. As an example, Costello (2009) claim that, in spite of
several years of intervention packages by the UK government, the small business
sector still has problems with technology adoption and in fact SMEs are often left
suspicious of those that work to help them such as Business Link and Chambers of
Commerce (Costello, 2009).
In another study on SMEs in six different European countries, Ghobakhloo et al.
(2012) observed that, although governments have tried to assist SMEs in adopting IT
through increasing public spending on technology projects, nevertheless, there was
indication that, there are adoption barriers built into governmental agency
mechanisms designed to help these businesses (Ghobakhloo et al., 2012). These

102
adoption barriers are attributable to the gap between what is really required by SMEs
and what is provided by the government (Ghobakhloo et al., 2012).
Nkonoki (2010) suggests that, even though the government of Tanzania recognise
the importance of SMEs sector, and thus has set up policies and support institutions
to support the sector, most small business owners are not made aware of the services
(Nkonoki, 2010). According to Nkonoki (2010), it is a surprise to note that, the
minority ones benefiting from these schemes or services are the large-scale
businessmen. In the same context, Fink (1998) found out that, government grants do
not appear to be a significant factor supporting IT adoption within Australian SMEs.
On the other hand, in spite of the above-mentioned results indicating that government
intervention has not generally been found to be helpful, latest studies, especially in
developing countries, have shown that IT adoption in SMEs has considerably
improved as a result of government policies and initiatives. For example, available
literature explains that, IT adoption and e-readiness within Iranian SMEs has
improved significantly as a result of Iranian government plan for ICT development
(Ghobakhloo et al., 2012).
On a similar note, recent studies from Malaysia have shown that, Malaysian SMEs
generally do not agree with the observation that, cost is a significant determinant of
ICT adoption (Alam and Noor, 2009; Ghobakhloo et al., 2012; Tan et al., 2009).
According to these studies, the rationale behind this observation is that, all types of
financial support have been provided by government to these businesses for ICT
103
adoption (Alam and Noor, 2009; Ghobakhloo et al., 2012; Tan et al., 2009).
Through government agencies such as the Malaysian SME bank, Malaysian
Technology Development Corporation, Multimedia Super Corridor (MSC) and Small
and Medium Scale Industries Development Corporation (SMIDEC), SMEs in
Malaysia have received adequate financial support and a number of necessary
training programs (Alam and Noor, 2009; Ghobakhloo et al., 2012; Tan et al., 2009).
Thus, the literature concludes that, IT adoption process seems to be considerably
simplified for Malaysian SMEs and this is attributed to the supportive policies and
comprehensive IT support provided by the Malaysian government (Alam and Noor,
2009; Ghobakhloo et al., 2012; Tan et al., 2009).
The lesson learned from this review is that, given the right approach by governments
in their effort to support the SME sector, positive results will be achieved. Costello
(2009) argues that, “Policy makers must address the issue for the benefits of SMEs,
rather than strive to hit government targets. Their major concern should be the SMEs
needs and not personal career development (Costello, 2009). From a similar
perspective, Alam and Noor (2009) suggests that, given the globalisation of the ICT
industry, there is a need to understand government role and responsibility in
contributing to the success of ICT development in businesses. Nkonoki (2010) notes
that, in spite of the lack of enough research on how government policies boost
growth of small firms’ the fact remains that, if the schemes and policies are put into
use for the benefit of the majority, positive results will be achieved.
104
The literature shows that, governmental interventions are multidimensional (Alam
and Noor, 2009; Costello, 2009; Ghobakhloo et al., 2012; Ghobakhloo et al., 2011a;
Mpofu et al., 2009; Sophonthummapharn, 2008; Tan et al., 2009). For example, if
the government encourage SMEs to adopt the Internet and e-commerce, it must also
come up with related laws to support transactions over the Internet. If that is not
done, then it is likely that, the adoption rate will be slow (Sophonthummapharn,
2008). It can thus be argued that, the role of the government includes: Setting up of
Policies, Regulatory Frameworks and Laws, development of key infrastructure such
as telecommunication infrastructure, roads, energy and water, establish support
institution and assign clear roles to key actors, create good environment for business
to operate (Alam and Noor, 2009; Costello, 2009; Ghobakhloo et al., 2012;
Ghobakhloo et al., 2011a; Mpofu et al., 2009; Nkonoki, 2012; Sophonthummapharn,
2008; Tan et al., 2009; URT, 2002). In this research, Support programs, Taxes and
Tariffs and Support Infrastructure are studied under the government policies
context.
4.2.4.1 Support Programs
In section 4.1.4, the importance of government interventions and support in SMEs is
thoroughly discussed. The literature shows that, one of the key areas that the
government can help SMEs is the creation and coordination of Support institutions
(Alam and Noor, 2009; Apulu, 2012; Costello, 2009). Alam and Noor (2009)
conclude that, the success of ICT adoption by Malaysian SMEs is attributed to
supportive government policies. Through government agencies, SMEs in Malaysia
get comprehensive IT support such as access to finances and necessary training
programs (Alam and Noor, 2009; Ghobakhloo et al., 2012). In this research,
105
government programs aimed at providing SMEs with support on financial assistance,
skill training programs and access to market information are looked at.
4.2.4.2 Taxes and Tariffs
The literature shows that, taxation is a very important topic for policy makers
(Atawodi and Ojeka, 2012; CEBR, 2009; Freedman, 2003; Mungaya et al., 2012;
Mwangi and Nganga, 2012; OECD, 2009). This is attributed to the fact that, taxation
is the major way of funding of governments. Atawodi and Ojeka (2012) cite the
literature and define tax as “a statutory compulsory contribution imposed by
government exacted from a person’s or entity’s income, property or transaction for
the purpose of funding governance”. A tax can either be of three basic structures;
proportional, regressive or progressive (Atawodi and Ojeka, 2012). Tax is said to be
proportional when the taxpayer is levied an amount that is an indirect proportion of
his income (Atawodi and Ojeka, 2012). A regressive tax is one that charges a higher
rate to persons receiving lower income, and finally a progressive tax levies a higher
rate to higher income earners (Atawodi and Ojeka, 2012). According to Mungaya et
al. (2012), a poorly conceived taxation system can have catastrophic consequences
for the concerned economy. They thus argue that, a tax structure must be adequately
designed to the specific environmental conditions; otherwise it may create a greater
burden to the tax-paying organizations (Mungaya et al., 2012).
In the context of this research, previous studies on ICT adoption by SMEs have
shown that, tax incentives for SMEs can stimulate ICT adoption in the SMEs (Alam
and Noor, 2009). Alam and Noor (2009), cite a study carried out in Israel to identify
106
three most important factors that had contributed to the creation ICT industry in
Israel. The study, found government’s tax incentives to be one of the success factors
(Alam and Noor, 2009). However, the literature shows differing arguments for and
against the targeting of tax incentives at SMEs (Atawodi and Ojeka, 2012; CEBR,
2009; Freedman, 2003; Mungaya et al., 2012; Mwangi and Nganga, 2012; OECD,
2009).
One of the major issues raised in the literature against the targeting of tax incentives
at SMEs is the fact that, the provision of tax reliefs and exemptions for small
businesses is against the key principle of neutrality (Freedman, 2003; OECD, 2009).
Freedman (2003) also note that, since it is extremely difficult to target growth
oriented SMEs only, reliefs aimed at all small businesses are available, inevitably, to
non-growth firms. This is viewed as wastage of resources. Furthermore, businesses
can be created, or reorganised partly in order to obtain the tax advantages and or for
the purpose of tax avoidance (Freedman, 2003). In addition, it is also argued that,
tax incentives and reliefs to small business may result in economic inefficiency
(Freedman, 2003). Resources should be allocated to larger, more efficient firms
rather than to small, less efficient firms. Alternatively, they may not be effective at
all.
On the other hand, the literature shows a great support for tax incentives to SMEs
(Atawodi and Ojeka, 2012; CEBR, 2009; Freedman, 2003; Mungaya et al., 2012;
Mwangi and Nganga, 2012; OECD, 2009). According to CEBR (2009), there is
empirical evidence showing that increasing business taxation provides disincentives

107
for small businesses to engage in activities that they have particular strengths in:
entrepreneurial activity, investment and innovation, and employment (CEBR, 2009).
Mwangi and Nganga (2012) observe that, there are various types of taxes which the
government use in raising required revenue but those taxes may have a negative
effect on the SME’s sector growth. Mwangi and Nganga (2012) recommend that, the
level of taxation set must be friendly and not stifle the running of business. Atawodi
and Ojeka (2012) found out that, there is a significant negative relationship between
taxes and the business’ ability to sustain itself and to expand. They thus
recommended for a tax policy that will stimulate the growth of the SME sector
(Atawodi and Ojeka, 2012).
URT (2002) state that, “SMEs are generally, confronted with unique problems
including heavy costs of compliance resulting from their size”. On a similar context,
Nkonoki (2010) came to the conclusion that, reducing the corporate taxes charged to
small firms, as well as reducing fees for registering small firms, can result in growth
of the SME sector in Tanzania. This study looks at how tax and tariffs affect
adoption and use of ICTs in small business.
4.2.4.3 Support Infrastructure
Previous research has shown that, infrastructure provided by the government and or
its institutions, is vital to help both economic growth and technological progress of
SMEs (Apulu, 2012; Costello, 2009; Ghobakhloo et al., 2012; Ghobakhloo et al.,
2011b; Mpofu et al., 2009). In addition, the literature shows how telecommunication
infrastructure impacts on ICT adoption by small business (Apulu, 2012; Behitsa et
al., 2010; Costello, 2009; Ghobakhloo et al., 2012; Ghobakhloo et al., 2011b; Mpofu
108
et al., 2009; Nchunge et al., 2013). The literature has shown that, poor or
underdeveloped communications infrastructure, results in poor, unreliable and
expensive service provision (Behitsa et al., 2010). This is known to have a negative
impact on ICT adoption (Apulu, 2012; Behitsa et al., 2010; Mpofu et al., 2009;
Nchunge et al., 2013).
In view of this, section 2.3 of this research discussed how Telecommunication
infrastructure plays into the overall picture of ICT development in the context of
Tanzania. On a similar tone, Nchunge et al. (2013) argue that, availability of other
support infrastructure such as roads and reliable energy are also important factors in
ICT adoption situations. Thus, support infrastructures are considered important in
this research and they refer to all infrastructures that support technology transfer,
such as Telecommunication infrastructure, roads, water, and reliable electricity.
4.2.5 ICT Adoption
ICT adoption is one of the dependent variables and is a core issue in this research,
since adoption of ICTs is believed and thus hypothesised to be a solution to the
observed problems. In section 3.1 the benefits of ICT adoption by SMEs in general,
are presented. This section complements section 3.1 by specifically targeting on
Computer and the Internet technologies, the two elements studied under the context
of ICT adoption.
109
4.2.5.1 Computers
There is evidence to the fact that, the benefits of ICTs, cut across all fields of human
activities (Apulu, 2012; Ion and Andreea, 2008; Onasanya, 2002a; Onasanya,
2002b). Osaat and Oyet (2012) claim that, the discovery of computer so far remains
the most important development in the ICT revolution”. It can thus be argued that,
computer technology cut across all aspects of human life. According to Onasanya
(2002a, 2002b), computer today is involved in reshaping the entire world and had
been more deeply involved in the texture of human life than any other dormant
technology ever.
Computer technology is credited with the ability to compute, store, retrieve,
organize, transmit, and algorithmically transform any type of information that can be
digitized (Brynjolfsson and Hitt, 2000). Numbers, text, video, music, speech,
programs, and engineering drawings are some of information type that can be
processed by a computer (Brynjolfsson and Hitt, 2000). Typically, a computer can
solve highly complicated problems quickly and accurately. Massive amount of data
can fast be manipulated and results can be viewed, stored or shared in the most
convenient way. Using different applications software’s, a computer can be used in
almost any discipline including such areas of specialisation as communications,
business, arts, education, medicine, governance, space exploration, engineering and
science (Anyaogu, 1991; Onasanya, 2002a; Onasanya, 2002b; Brynjolfsson and Hitt,
2000).
110
In the context of business, Feigenbaum (2015) state that, “In an age of booming
technology, running a business without computers is like trying to breathe without
lungs”. Feigenbaum (2015) further argue that, whether we like it or not, computers
have become an integral part of the way business is done. Sison (2012) claim that,
computers are used for all aspects of running a business. According to Sison (2012)
businesses use computers to keep records, develop budgets and forecasts, prepare
marketing documents, research and stay in contact with other employees and
customers. In general, computers are used in business, for word-processing,
spreadsheets, accounting activities, communication, marketing, education and
research (Anyaogu, 1991; Crowder, 2015; Feigenbaum, 2015; Onasanya, 2002b;
Sison, 2012; Suttle, 2015). In addition, computers are used for maintaining
computerised databases and information systems of a company (Anyaogu, 1991;
Onasanya, 2002b).
Available literature has shown that, adoption and use of computer system can be of
great benefit to a company (Anyaogu, 1991; Brynjolfsson and Hitt, 2000; Crowder,
2015; Feigenbaum, 2015; Onasanya, 2002b; Sison, 2012). Sison (2012) suggests
that, computer use in business enhances productivity and provides results that might
not otherwise be possible. On a similar note, the literature have enough evidence
that, when used effectively, computers save time, money, and increase business
efficiency and productivity (Apulu, 2012; Costello, 2009; Crowder, 2015
Feigenbaum, 2015; Sison, 2012; Suttle, 2015). In view of this observation, Sison
(2012) recommends that businesses should look at their operation and ensure that
they are using their computers to the fullest extent. Brynjolfsson and Hitt (2000)
111
noted that, as the computers become cheaper and more powerful, the business value
of computers is most limited by the ability of managers to invent new processes,
procedures and organizational structures that leverage this capability and less by the
computational capability of computers. It could thus be argued that, in today’s
business environment, a computer is a strategic management tool that when used to
its full potential can boost business efficiency, productivity and growth.
Previous research on SMEs concludes that, computers can benefit small business the
same way they can benefit big organisations (Apulu, 2012; Costello, 2009;
Ghobakhloo et al., 2012). In particular, the literature show that effective use of
computers results in increased efficiency, competitiveness and performance of a firm
(Apulu, 2012; Costello, 2009; Galloway and Mochrie, 2005; Harindranath et al.,
2010; Hazbo and Arnela, 2010; Lerman et al., 2004; Modimongale, 2009; Naul,
2012; European Commission, 2014; Sophonthummapharn, 2008). In this research,
the adoption and the level of computers use in SMEs and its contribution to the
overall performance of the firm are explored.
4.2.5.2 The Internet
From the definition of the Internet, section 1.1.3, it can be summarised that, the
Internet technology brings together computer technologies, computer networking
technologies and telecommunication technologies to form a global information
system. The biggest strength of the Internet is its ability to offer abundant
information resources using faster communication links at a low cost. Some authors
have argued that, the Internet is a virtual treasure trove of information (Ernst, 2001).
112
In the past decade, there has been a dramatic growth of the Internet and the World
Wide Web (Chung, 2006). Kripanont (2007) claim that, the Internet today is a
public, cooperative, and self-sustaining facility accessible to hundreds of millions of
people worldwide.
Since its advent, the Internet has transformed the world like no other technology
before (Apulu, 2012). Apulu (2012) report that, the Internet is considered to be one
of the “defining symbols” of the 21st century innovation. According to Apulu (2012),
the Internet has transformed the conceptual notions of how people value knowledge
to create a new economy. Sellitto and Martin (2003) cited by Apulu (2012) claim
that, the internet has become an extremely important modern-day technology for
businesses. Manyika and Roxburgh (2011) note that the Internet have changed the
way we work, socialise, create and share information and organise the flow of
people, idea and things around the globe. In view of this, it can be concluded that,
the Internet today affects every aspect of human life. Some of the areas that has
benefited from the Internet technology include: communication, education and
research, business and economy and leisure and entertainment (Kripanont, 2007).
The literature provides the proof to the effect that adoption and use of the Internet
has advantages to the business (Apulu, 2012; Ifinedo, 2011; Ion and Andreea, 2008;
Kula and Tatoglu, 2003; Udo and Edoho, 2000). Ernst (2001) argue that, the Internet
has become the means for conducting growing numbers of transactions between
suppliers and large international corporations. This is attributed to the speed,
flexibility, and efficiency that the Internet offers (Ernst, 2001). Ernst (2001)
113
concludes that, the Internet has opened new markets to developing Asia and has
accelerated the diffusion of knowledge throughout the region (Ernst, 2001).
In the context of the SMEs, there is enough literature to support the fact that, Internet
adoption is a key success factor to small business (Alam, 2009; Ana et al., 2006;
Chung, 2006; Dholakia and Kshetri, 2004; Gilmore et al., 2007; Hsu et al., 2008;
Ifinedo, 2011; Kaynak et al., 2005; Kula and Tatoglu, 2003; Martin and Matlay,
2003; Mehrtens et al., 2001; Nathan, 2013; Tan et al., 2010). Ifinedo (2011) claim
that, the emergence of the internet has presented businesses (small and large) with an
opportunity to improve performance and enhance revenue generation. Similarly,
Ghobakhloo et al. (2011b) report that, in the global business environments, SMEs are
incrementally using information and communications technologies (ICT)-based
electronic commerce (EC) to gain competitive advantages and to have access to
global markets.
Martin and Matlay (2003) argue that, business could gain competitive advantage
from Internet usage if they can achieve the right mix of managerial capacity and
marketing focus in terms of image, brand and customer needs. Dholakia and Kshetri
(2004) noted that, the Internet can be a critical factor in enhancing a firm’s market
reach and operational efficiency. They further argued that, Internet based
technologies provide small firms the opportunity to overcome the limitations of size
and compete more effectively and/or in larger markets with bigger sized
establishments (Dholakia and Kshetri, 2004). In addition, Dholakia and Kshetri
(2004) report that, there is some evidence to suggest that the Internet has increased
114
international opportunities for SMEs (Dholakia and Kshetri, 2004). On a similar
context, Gilmore et al. (2007) suggest that, the use of the internet for e-commerce
impacted upon the barriers to export entry and has provided hope to millions of small
and medium sized enterprises (SMEs) attempting to enter foreign markets by
allowing them to communicate globally as efficiently as any large business.
Kaynak et al. (2005) identified the most prominent benefits of conducting business
on the Internet as:
i. The ability of business to offer direct links with customers, suppliers and
distributors and to facilitate transactions
ii. The ability of business to facilitate information transfer
iii. The ability of companies to develop new products and services for existing
and new customers and
iv. The opportunities for companies to market their products around the world
without physically contacting customers or advertising in other parts of the
world.
In addition, Kaynak et al. (2005) report that, among the principle benefits of internet-
based EC that are more directly relevant for small business include: direct savings
115
such as product promotion, new sales channels, quick product delivery, more
satisfaction of customers, inexpensive advertising medium, enhanced company
image, new business opportunities, efficiency in information gathering and better
support from suppliers.
On the other hand, a systematic literature review has conclusively shown that, in
spite of the obvious benefits, the rate of adoption of Internet and its related
technologies or services by SMEs is low (Alam, 2009; Ana et al., 2006; Apulu,
2012; Chung, 2006; Dholakia and Kshetri, 2004; Ghobakhloo et al., 2011b; Gilmore
et al., 2007; Hsu et al., 2008; Ifinedo, 2011; Kaynak et al., 2005; Kula and Tatoglu,
2003; Martin and Matlay, 2003; Mehrtens et al., 2001; Nathan, 2013; Tan et al.,
2010). Furthermore, the literature claims that, much research that has been
conducted in order to improve the situation has been conducted in developed
countries (Alam, 2009; Ana et al., 2006; Apulu, 2012; Chung, 2006; Dholakia and
Kshetri, 2004; Ghobakhloo et al., 2011b; Gilmore et al., 2007; Hsu et al., 2008;
Ifinedo, 2011; Kaynak et al., 2005; Kula and Tatoglu, 2003; Martin and Matlay,
2003; Mehrtens et al., 2001; Nathan, 2013; Tan et al., 2010).
Ghobakhloo et al. (2011b) and Tan et al. (2007) notice that, SMEs in developing
countries face challenges different from those in developed countries and differs
greatly in adopting and benefiting from e-commerce (Ghobakhloo et al., 2011b; Tan
et al., 2007). Ghobakhloo et al. (2011b) further observes that, e-commerce adoption
in these businesses, has only recently gained attention in the academic press.
According to Ghobakhloo et al. (2011b) this scenario makes research related to e-

116
commerce implementation in developing countries context scarce. Thus, the
literature recommends for researches that are robust enough to capture most, if not
all, of the idiosyncrasies observed (Apulu, 2012; Ghobakhloo et al., 2011b; Mpofu et
al., 2009). In view of the discussion above, the inclusion of Internet technology in
this research is justified.
4.2.6 SME Performance
The final construct in the proposed model is concerned with the post adoption
evaluation of surveyed SMEs here termed as SME Performance. The rationale of
including this construct in this research is based on recommendations from previous
work (Costello, 2009; Lefebvre and Lefebvre, 1996; Van Akkeren and Cavaye,
1999b). In their assessment of current models in the literature, Van Akkeren and
Cavaye (1999b) stated that, small companies needed an immediate return on
investments, and thus called for a model which included return on investment (ROI).
Van Akkeren and Cavaye (1999b) noted that “SME owners are only concerned with
a return on investments; hence they are reluctant to make substantial investments
when short-term returns are not guaranteed”. According to Costello (2009), a recent
review of the literature revealed that the inclusion of ROI in a model as still not
being fulfilled. In addition, Costello (2009) noted that, there is concern from the
literature review that “with no pre-evaluation and no post-evaluation measures,
micro-companies will continue to make decisions based on a poor assessment of
their last adoption” (Costello, 2009).
Costello (2009) further discovered that, the vast majority of small businesses rely on
“gut instinct” to make a decision as to whether the adoption is successful or not. She
117
noted that, 97% of the micro-companies studied did not evaluate purchases or
measure the success of those purchases in any tangible way (Costello, 2009).
Costello (2009) argue that, “this, may indicate a gap in knowledge of micro-
companies regarding some factors that are important in measuring adoption success
as well as lack of understanding of business success”. Costello (2009) identified a
number of factors that could have been measured but were not. According to
Costello (2009) this demonstrates micro-companies’ inability to use historic data to
inform future investment decisions.
St-Pierre and Sylvain (2006) observe that, “the importance of evaluating
performance is obvious for business enterprises in general, and for SMEs in
particular”. However, Haber and Reichel (2005) note that, performance
measurement is not a simple task. They argue that, the difficult lies in defining
effectiveness and performance (Haber and Reichel, 2005). Wu (2009) acknowledge
that, performance is a widely-used concept in many areas. He claims that usually,
performance is a measure of how well a mechanism/process achieves its purpose
(Wu, 2009). According to Wu (2009), in enterprise management, organization’s
performance is defined as “how well the organization is managed” and “the value
the organization delivers for customers and other stakeholders.” In the context of
this research, “performance” is related to achieving the small business owner-
manager interests after ICT adoption. The literature has shown that, in the context of
SMEs, perceptions of success are directly linked to the perceived benefits of
adoption by the owner-manager (Costello, 2009; Lefebvre and Lefebvre, 1996;
Mehrtens et al., 2001; Thong and Yap, 1995; Van Akkeren and Cavaye 1999b).
118
Haber and Reichel (2005) claim that, studies in different industries have primarily
relied on financial variables to evaluate business performance. According to Haber
and Reichel (2005) this reliance, reflect the shareholder interests of the SMEs
owner–manager. Haber and Reichel (2005) report that frequently used financial
measures include return on investment (ROI), profit and revenues. On the other
hand, Haber and Reichel (2005) note that, assessing performance solely on the basis
of financial measures neglects other relevant performance dimensions of small
ventures that relate to managerial aspects such as market share, number of employees
and revenue per employee (Haber and Reichel, 2005). On a similar note, Wu (2009)
claim that, measuring performance is a multi-dimensional concept. He argues that
“effectiveness “and “efficiency” are the two fundamental dimensions of
performance. According to Wu (2009), to illustrate efficiency, effectiveness, and
the value delivered, multi-measures should be used.
The literature notes that, Innovation can also be used as firm’s performance indicator
(DBIS, 2014; Gunday et al., 2009; OECD, 2005). According to Gunday et al.
(2009), innovativeness is one of the fundamental instruments of growth strategies to
enter new markets, to increase the existing market share and to provide the company
with a competitive edge. Gunday et al., 2009, further notes that innovation is
embedded in the organizational structures, processes, products, and services within a
firm and is an essential component of competitiveness (Gunday et al., 2009). DBIS
(2014) claim that, there is a large body of evidence to support the notion that,
innovation is at the core of business productivity growth, and therefore at the core of
119
economic growth overall. In view of this review, in this research, return on
investment and innovation are explored as indicators of performance after ICT
adoption in small business. The factors are discussed in the next sections.
4.2.6.1 Return on Investment
Botchkarev and Andru (2011) report that Return on Investment (ROI) is one of the
most popular performance measurement and evaluation metrics used in business
analysis. According to Botchkarev and Andru (2011) initially ROI was identified as a
financial term and defined as a concept based on a rigorous and quantifiable analysis
of financial returns and costs. Investopedia (2011) define ROI as “a performance
measure used to evaluate the efficiency of an investment or to compare the efficiency
of a number of different investments”. According to Investopedia (2011), to
calculate ROI, the benefit (return) of an investment is divided by the cost of the
investment; the result is expressed as a percentage or a ratio (Investopedia, 2011;
Botchkarev and Andru, 2011). Thus, ROI formula is:
However, Cresswell (2004) note that “ROI analysis in general is a rather diverse
collection of methods, skills, tools, activities, and ideas. They all may be useful for
assessing the relative value over time of some investment” (Cresswell, 2004).
Investopedia (2011) explain that, the calculation for ROI and, therefore the
definition, can be modified to suit the situation. According to Investopedia (2011), it
all depends on what is included as returns and costs. In addition, Investopedia

120
(2011) and Botchkarev and Andru (2011) note that the definition of ROI in the
broadest sense just attempts to measure the profitability of an investment and, as
such, there is no one "right" calculation (Investopedia, 2011; Botchkarev and Andru,
2011).
Botchkarev and Andru (2011) argue that the large number of ROI methods today,
has led to the situation where ROI is often experienced as a non-rigorous, amorphous
bundle of mixed approaches, prone to the risks of inaccuracy and biased judgement.
This can further be explained by the existence of two extreme views regarding the
use of ROI. While one view claim that ROI is “the most popular” metric to use
when comparing the attractiveness of one information technology (IT) investment to
another, the other view is to “forget ROI” altogether (Botchkarev and Andru, 2011).
Nevertheless, Botchkarev and Andru (2011) claim that, when applied correctly, ROI
analysis is a powerful tool for evaluating existing information systems and making
informed decisions on software acquisitions and other projects (Botchkarev and
Andru, 2011). At present, ROI is widely recognized and accepted in business and
financial management in the private and public sectors (Botchkarev and Andru,
2011; Investopedia, 2011) although a meaningful analysis of returns on investment in
information technology is far easier said than done (Cresswell, 2004).
Haber and Reichel (2005) argue that, using objective financial measures would seem
to be the simplest way to assess performance. However, they point out that, these
data are sometimes confidential and difficult to obtain from the respondents (Haber
and Reichel, 2005). They argue that small firms are known for their inability and
121
unwillingness to provide desired information. In some cases, objective financial data
on small ventures are not publicly available (Haber and Reichel, 2005). According
to Haber and Reichel (2005), in such scenario, previous research has shown that,
subjective measures are more flexible and thus useful. They argue that, given the
need for valid performance measures and the difficulty in collecting valid data, there
is merit in the use of both objective and subjective performance measures
simultaneously (Haber and Reichel, 2005). Costello (2009) discovered that the
majority of SMEs do not evaluate performance after adoption and those who do, tend
to use a single or simple metrics. In view of the observations made by Haber and
Reichel (2005) and Costello (2009), in this research, increase in market size, increase
in capital, increase in sales, increase in profit, and the number of employees are
explored as indicators of return on investment.
4.2.6.2 Innovations
According to DBIS (2014), “Innovation is the application of new knowledge to the
production of goods and services” The literature show that innovativeness is one of
the fundamental instruments of growth strategies (DBIS, 2014; Gunday et al., 2009;
OECD, 2005). It enables a business to enter into new markets, to increase the
existing market share and to provide the company with a competitive edge (Gunday
et al., 2009; OECD, 2005). DBIS (2014) claim that, there is enough evidence from
the literature to the effect that innovation is at the core of business productivity
growth, and therefore at the core of economic growth overall.
According to the literature (Gunday et al., 2009; OECD, 2005) innovation as a term
is not only related to products and processes, but is also related to marketing and
122
organization (Gunday et al., 2009; OECD, 2005). Gunday et al. (2009) claim that
different types of innovation are described in the literature. OECD (2005) identify
four different types of innovation as: product innovation, process innovation,
marketing innovation and organizational innovation. The OECD (2005) notes that,
product and process innovations are closely related to the concept of technological
developments.
Product innovation is defined as “the introduction of a good or service that is new or
significantly improved regarding its characteristics or intended uses; including
significant improvements in technical specifications, components and materials,
incorporated software, user friendliness or other functional characteristics
(OECD ,2005). The OECD (2005) further defines process innovation as “the
implementation of a new or significantly improved production or delivery method.
This includes significant changes in techniques, equipment and/or software. Process
innovations can be intended to decrease unit costs of production or delivery, to
increase quality, or to produce or deliver new or significantly improved products”
(OECD, 2005).
In addition, marketing innovation is defined as “the implementation of a new
marketing method involving significant changes in product design or packaging,
product placement, product promotion or pricing” (OECD, 2005). Finally, the
OECD (2005) define organisational innovation as: “a new organizational method in
the firm’s business practices, workplace organization or external relations.
Organizational innovations have a tendency to increase firm performance by

123
reducing administrative and transaction costs, improving workplace satisfaction
(and thus labour productivity), gaining access to non-tradable assets (such as non-
codified external knowledge) or reducing costs of supplies (OECD, 2005). From the
operational definitions of the four types of innovations, it can be seen that computer
and Internet adoption by a firm can affect both types of innovation. In this research,
under the Innovation construct, the emergence of new processes and methods that
enhance company efficiency and effectiveness, are explored in evaluating firm
performance after Internet and computer adoption,
4.3 Summary of Model Factors
Having established the validity of different constructs and factors, this section give a
summary of factors used in this study. The summary is given a tabular form as
depicted by Table 4.2.
Table 4.2: Individual Factors for the ICT Adoption Model
Category Individual Factor Reference

Personal Factors Educational experience Chung (2006); Costello (2009);
Knowledge of IT Grandon and Pearson (2004a);
Perceived benefit Rashid and Al-Qirim (2001);
Van Akkeren and Cavaye
(1999b)
The Firm Age, Size, Sector Costello (2009); Van Akkeren
and Cavaye (1999b)

124
Category Individual Factor Reference

Organisational Advice sought Chung (2006); Costello (2009);
Readiness Internal pressures Grandon and Pearson (2004a);
In house ICT skills Mehrtens et al. (2001); Rashid
Adequate resources and Al-Qirim (2001); Van
Akkeren and Cavaye (1999b)
Strategy Decision making Costello (2009);
Reason for adoption Levy et al. (2005)
Significance of adoption
External Pressure Customer pressure Chung (2006); Costello (2009);
Competitive pressure Ghobakhloo et al. (2012);
Suppliers pressure Grandon and Pearson (2004a);
Industry pressure Mehrtens et al. (2001); Rashid
and Al-Qirim (2001);
Sophonthummapharn (2008);
Van Akkeren and Cavaye
(1999b)
Economic Factors Access to financial Alam (2009); Ghobakhloo et al.
support, (2012); Mpofu et al. (2009);
Sophonthummapharn (2008);
Cost of ICT systems Walczuch et al. (2000);
Government Support programmes Ghobakhloo et al. (2012);
Policies Taxes and tariffs Mpofu et al. (2009);
Support infrastructure Sophonthummapharn (2008)
ICT Adoption Computer systems Apulu (2012); Brynjolfsson and
The Internet Hitt (2000); Chung (2006);
Dholakia and Kshetri (2004);
Galloway and Mochrie (2005);
Gilmore et al. (2007); Mehrtens
et al. (2001); Modimongale
(2009); Onasanya (2002b)
SMEs Performance Return on Investment Botchkarev and Andru (2011);
Costello (2009); Cresswell A. M
Innovations (2004); DBIS 2014; Gunday et
al. (2009); Haber and Reichel
(2005); Investopedia (2011);
OECD (2005)
4.3.1 The Conceptual Model
Figure 4.1 depict the conceptual model used in this study.
Explanatory Variables
PERSONAL FACTORS-PF
 Education Experience SME PERFOMANCE-SP
 Knowledge of IT  Return on Investment
 Perceived Benefit  Process Innovation
125
Figure 4.1: The Conceptual Model for Accessing ICT Adoption in SMEs
In this study, descriptive analysis was used in preliminary data analysis. Thereafter
Structural Equation Modelling (SEM) was used in model development and
validation.
4.4 Hypotheses
From the conceptual model developed the hypotheses for this research were
formulated and they were arranged into five groups. Group one had three hypotheses
126
(H1a, H1b and H1c) that were meant to test the individual effect of explanatory
variables (Personal Factors, Organisation Factors and Economic Factors) on ICT
adoption (the dependent variable). Group two had three hypotheses (H2a, H2b and
H2c). These hypotheses tested the combined effect of two explanatory variables on
ICT adoption. Group three had one hypothesis H3 which tested the combined effects
of all explanatory variables on ICT adoption. Group four had one hypothesis H4
which tested the effect of ICT adoption on the performance of SMEs (SMEP).
Finally, the fifth group had three hypotheses (H5a, H5b and H5c). The hypotheses in
group five tested the moderating effect of government policies (support programs,
taxes and tariffs and support infrastructure) has on the influence of predictors
(explanatory variables) of ICT adoption. The hypotheses are summarised as:
H1a: Personal factors have a direct positive significant relationship with ICT
adoption in SMEs.
H1b: Organizational factors have a direct positive significant relationship with ICT
adoption in SMEs.
H1c: Economic factors have a direct positive significant relationship with ICT
adoption in SMEs.
H2a: Personal factors and Organizational factors have a direct positive significant
relationship with ICT adoption in SMEs.

127
H2b: Personal factors and Economic factors have a direct positive significant
H2c: Organizational factors and Economic factors have a direct positive significant
H3: Personal factors, Organizational factors and Economic factors have a direct
positive significant relationship with ICT adoption in SMEs.
H4: ICT adoptions have a direct positive significant relationship with SMEs
performance.
H5a: The influence of personal factors, organisation factors and economic factors
(predictors) on ICT adoption in SMEs is moderated by the taxes and tariffs
charged by the government on the SMEs.
H5b: The influence of personal factors, organisation factors and economic factors
(predictors) on ICT adoption in SMEs is moderated by support programs
provided by the government to SMEs.
H5c: The influence of personal factors, organisation factors and economic factors
(predictors) on ICT adoption in SMEs is moderated by the support
infrastructure provided by the government to SMEs.

128
4.5 Chapter Summary
In this chapter, the conceptual model in this research have been developed and
presented. The model constructs and variables have been explored in order to
establish their validity in this study. A total of eleven hypotheses have been
formulated from the model for the aim of testing the effect of explanatory variables
on dependent variables. The methodology used in carrying out this research is given
in the next chapter.
CHAPTER FIVE
RESEARCH DESIGN AND METHODOLOGY
5.1 Introduction
One of the formidable problems that a researcher faces after defining the research
problem is the preparation of the design of the research project, popularly known as
the “research design” (Apulu, 2012; Kothari, 2004; Saunders et al., 2009;
Sophonthummapharn, 2008). However, according to Apulu (2012), proper
understanding of research design is one of the most challenging issues in research
due to the fact that, scholars use various terms to describe research approaches and
strategies (Apulu, 2012) as seen in the given definitions for research design and
methodology.
129
Kothari (2004) defined research design as “the arrangement of conditions for
collection and analysis of data in a manner that aims to combine relevance to the
research purpose with economy in procedure.” This means that, both data and
methods, and the way in which these will be configured in the research project, need
to be the most effective in producing the answers to the research questions, taking
into account practical and other constraints of the study (Apulu, 2012; Kothari, 2004;
Saunders et al., 2009; Sophonthummapharn, 2008). Apulu (2012) note that,
research design is also known as a research plan or research strategy. Saunders et al.
(2009) defined research strategy (or research design) as an overall approach that a
researcher adopts in order to answer the research question (Saunders et al., 2009). On
a similar note, Crotty (1998) claimed that, a research strategy or plan of action is the
design that shapes a researcher’s choice and use of particular methods and links them
to the desired outcomes (Crotty, 1998).
Research methodology has been defined as “a strategy, process, plan of action or
design lying behind the choice and use of particular methods and linking the choice
and use of those methods to the desired outcomes” (Apulu, 2012). Remenyi et al.
(2003) described methodology as “the overall approach to a problem which could be
put into practice in a research process, from the theoretical underpinning to the
collection and analysis of data”. Collis and Hussey (2003) identified methodology as
the “overall approach to the entire process of the research study”. According to
Sophonthummapharn (2008), research methodology is focused around the problems
to be investigated in a study and thus its definition can vary according to the
problems to be investigated.
130
This study adopts the description of the different research terms as advocated by
Saunders et al. (2009). According to Saunders et al. (2009), the selection of
methods and technique or techniques used to obtain data, along with procedures to
analyse these data, represents only the final decision about the overall research
design. Thus, using the metaphor of the “Research Onion”, Saunders et al. (2009)
presented the overall research methodology in which the thoughts with regard to the
research problem lie in the centre. That way, several layers (of the onion) have to be
“peeled away” before reaching this central position. According to Saunders et al.
(2009), these layers are the important aspects to be considered in determining the
research methodology for a particular research study. Figure 5.1 shows the
“research Onion” as suggested by Saunders et al. (2009). The onion identifies
philosophy, approaches, strategies, choices, time horizon, and techniques layers.
These layers illustrate the steps of research design. It should be noted that, while
there are different definitions and classifications of these terms, classification set
forward by Saunders et al. (2009) is preferred here, since in my opinion, it provides a
clear overall framework for the complete research process.

131
Source: Saunders et al. (2009)
Figure 5.1: Research Onion
This chapter presents the research design and methodology used in this study.
However, prior to that, an overview of different types of research designs are given
in order to identify the type to be used in the current study. This helps to understand
the decisions made on the choice of methods and techniques in the study. After
research type identification, the rest of the chapter is arranged as follows: First the
philosophical paradigms related to IS research are briefly discussed and the
philosophical standpoint of the author is acknowledged. Second, different
alternatives for research approaches, designs, strategies, and time dimension are
discussed, and the specific choices made in this study are clarified. Third, this
chapter elaborates the details of research methodologies in regard to sampling
procedure, questionnaire development and the pilot survey. Finally, the chapter
presents an overview of the methods, techniques and procedures used in data
collection and data analysis.

132
5.2 Research Designs
From the literature, the author notes that there are different ways of classifying
research. Some authors classify research based on the method used such as
qualitative research or quantitative research. Other authors classify research
according to the observational method used such as survey research or case study
research. However, Yin (2009) and Costello (2009) point out that, the main reasons
for conducting any research are either to explore, to describe or to explain a
phenomenon (Costello, 2009, Yin 2009). Thus, research designs can be classified as
exploratory, descriptive, and explanatory (causal) research (Saunders et al., 2009;
Sophonthummapharn, 2008). The three different types of research are further
explained in the next section.
5.2.1 Exploratory Research
Exploratory research is a type of research undertaken in order to provide the
researcher with a better understanding of a situation or a research problem
(Sophonthummapharn, 2008; Costello, 2009). This is common when there is little
theory available to guide predictions like when a researcher examines a new interest
or when the subject of study itself is relatively new (Sophonthummapharn, 2008;
Costello, 2009). According to Sophonthummapharn (2008), exploratory research put
a major emphasis on the discovery of ideas and insights and it is used when seeking
insight into the general nature of a problem, the possible decision alternatives, and
relevant variables that need to be considered (Sophonthummapharn, 2008).
Exploratory research can be conducted through a number of techniques including

133
literature review, interviews, Delphi technique, focus group, case study, project test,
experience survey, and ethnography (Sophonthummapharn, 2008).
5.2.2 Descriptive Research
Descriptive research is a type of research that seeks to provide an accurate
description of observations of phenomenon being studied (Sophonthummapharn,
2008). The researcher observes and then describes what was observed
(Sophonthummapharn, 2008). Usually, descriptive research designs are structured
and t h e y a r e specifically designed to measure the characteristics d e s c r i b e d in
a research question (Sophonthummapharn, 2008). Hypotheses, derived from the
theory, usually serve to guide the process and provide a list of what needs to be
measured. As a rule, descriptive research does not answer questions about
how/when/why the characteristics occurred. Rather it addresses the "what" question.
For example, “What are the characteristics of the population or situation being
studied?” (Sophonthummapharn, 2008). In general, things are described by
providing measures of an event or activity and descriptive research often
accomplishes this by using descriptive statistics (Sophonthummapharn, 2008). These
include frequency counts (how many), measures of central tendency (mean or mode),
and a measure of variation (standard deviation). Descriptive research can be
conducted through sample surveys, an omnibus panel, a true panel, and longitudinal
study (Sophonthummapharn, 2008).

134
5.2.3 Explanatory (Causal) Research
The primary purpose of explanatory research design is on studying a situation or a
problem in order to explain the causal relationships among variables (Saunders et al.,
2009; Sophonthummapharn, 2008). As such they are designed to test whether one
event causes another. Explanatory studies are characterized by research hypotheses
that specify the nature and direction of the relationships between or among variables
being studied (Saunders et al., 2009; Sophonthummapharn, 2008). It is concerned
with determining cause-and-effect relationships, which are studied via experiments
(Sophonthummapharn, 2008). Explanatory research aims to develop precise theory
that can be used to definitively explain the phenomena, which leads to the
generalization from the research and thus probability sampling is normally a
requirement. In more detail, it tests whether or not some event causes another. This
type of research is the most complicated (Sophonthummapharn, 2008). It often takes
a long time from planning to execution and it can be very expensive
(Sophonthummapharn, 2008). The data are quantitative and almost always require
the use of a statistical test to establish the validity of the relationships. For example,
explanatory survey research may investigate the factors that contribute to customer
satisfaction and determine the relative weight of each factor, or seek to model the
variables that lead to shopping cart abandonment. Explanatory research is typically
conducted through laboratory and field experiments.
5.2.4 Relationships among Research Designs
From the discussion above it can be seen that, the three types of research differ in
several aspects including research purpose, the way research questions or hypotheses
135
are formulated, and the way data are collected. However, despite the distinctions
among the three types of research, they nonetheless have complementary roles in
many research projects (Sophonthummapharn, 2008). The distinctions are not
absolute because in reality any research project is likely to serve several purposes
and therefore require more than one research design (Sophonthummapharn, 2008).
According to Sophonthummapharn (2008), the three research designs are interrelated
as stages in a continuous process as depicted in figure 5.2
Source: Sophonthummapharn (2008).
Figure 5.2: Relationships among Research Designs
In a typical case where a research project employs more than one research design,
exploratory research is more often used as a starting point in order to get a clear
understanding of the problem (Sophonthummapharn, 2008). After a successful
exploratory research, the possible explanations or hypotheses can then be developed

136
and serve as guides for the subsequent descriptive or causal research projects.
Descriptive research design can also be a starting point. However, the choice of
initial step will depend on whether researchers can be sufficiently specific in
formulating the problem (Sophonthummapharn, 2008).
5.2.5 Research Design Used in the Current Study
This study employs multiple research designs as suggested by Costello (2009) and as
depicted in figure 5.2. The study begins with an exploratory phase involving
extensive literature review about the topic of interest, that is, adoption and use of
ICTs in SMEs. Research questions and hypotheses are then developed for
subsequent descriptive study. Finally, an attempt is made to explain the cause of
some of the observed characteristics.
5.3 Research Philosophy
Research philosophy or sometimes called the research paradigms can simply be
defined as “a set of shared assumptions or ways of thinking about some aspect of
this world” (Oates, 2006). This means that different paradigms will result in
different views of the world. Saunders et al. (2009) define philosophy as the belief
and thinking that an individual has about knowledge and how it is created and
developed. According to the literature (Apulu, 2012; Saunders et al., 2009;
Sophonthummapharn, 2008), the researcher beliefs and experiences somehow
influences the way a particular research project is conducted. Sophonthummapharn
(2008) further argues that, “the way a researcher views environments or the world
137
surrounding him/her will underlie the choice of which research practices should be
used including research approaches and research strategies”.
Apulu (2012) observed that, there are three factors that influence the need to
understand philosophical paradigms in a research study. Firstly, philosophical
paradigms help to define the research design, the type of evidence that is required,
how it will be gathered and interpreted, and how this will give answers to the
research questions. Secondly, philosophical paradigms assist the researcher to
identify which research design will work for a particular study. Thirdly,
philosophical paradigms can help the researcher to create research designs that may
not be related to the researcher’s experience.
Saunders et al. (2009) identifies ten different philosophies that guide research (see
Figure 5.1 in page 131) including Positivism, Realism, Interpretivism, Objectivism,
Subjectivism, and Pragmatism. According to Saunders et al. (2009) the different
philosophies are a result of how we think about research philosophy and they identify
three major ways of thinking as epistemology, ontology and axiology (Saunders et
al. (2009). The literature (Apulu, 2012; Costello, 2009; Saunders et al., 2009;
Sophonthummapharn, 2008), show that research paradigms are often shared within a
particular research field and as such researchers from the same field will normally
tend to have the same philosophy (Costello, 2009). Costello (2009) noted that, in
the field of IS research, the positivist, interpretivism and critical research
philosophical paradigms find extensive use with the positivist paradigm being the
most widely used paradigm (Costello, 2009).

138
According to Hinkelmann (2010) positivist paradigm is the oldest approach and has
taken centuries to develop. The paradigm has its grounding within the physical
sciences world (Hinkelmann, 2010). Often, the positivist paradigm is referred to as
“the scientific method” since this reflects the underlying assumptions within it.
That is, it believes in the possibility to observe and describe reality from an objective
viewpoint (Hinkelmann, 2010). Using observations that are repeatable, Positivists
observe the world in some neutral and objective way, discover “general”
relationships and “universal” laws, derive theories and test them (Costello, 2009;
Hinkelmann, 2010).
On the other hand, Interpretivism paradigm is not as popular as Positivist paradigm
in IS research (Costello, 2009). H o w e v e r , the paradigm is growing in acceptance
among IS researchers (Costello, 2009). Interpretivists believe that, it is necessary to
understand differences between humans in our roles as social actors (Costello, 2009;
Oates, 2006). This approach allows researchers to be concerned with the social
context of systems and how the system is influenced and can influence the
setting (Costello, 2009; Oates, 2006). Interpretivism research recognizes multiple
subjective realities and can therefore make the quality of research more difficult
to ascertain (Costello, 2009; Oates, 2006).
The third paradigm, that is, critical research is relatively less well accepted in IS
research than interpretivism (Costello, 2009). According to Apulu (2012) critical
research is often portrayed as a third alternative to the first set of paradigms, that is,
positivism and interpretivism. The critical research philosophy accepts social realty
139
as interpretivism does, but claims that social reality has objective properties that
can dominate the way we perceive the world (Apulu, 2012; Costello, 2009; Oates,
2006). Apulu (2012), noted that, the central issue of critical research is that it aims at
changing social reality and promoting emancipation.
Apulu (2012) and Costello (2009) claim that, research paradigms are often
associated with particular research strategies. In addition, it is also possible to
have a strategy that can be interpreted via multiple paradigms (Apulu, 2012;
Costello, 2009; Oates, 2006). Table 5.1 shows research paradigms with associated
strategies.
Table 5.1: The Relationship between Paradigms and Strategies
Philosophical Paradigm Research Strategy

Positivism Experiments
Positivism Surveys
Interpretive
Interpretive Ethnography
Critical Research
Interpretive Case studies
Critical Research
Interpretive Action Research
Positivism
Critical Research
Interpretive Design and Creation
Positivism
Critical Research
Source: Oates (2006)

140
5.3.1 Philosophical Paradigm Adopted in the Current Research
The author of this work has a strong background in Engineering and thus naturally
most of his views are Positivists in nature. This research test hypotheses based
on ICT adoption theories in order to explain ICT adoption situation in the
context of Tanzania SMEs. Therefore, this research subscribes primarily to the
Positivists paradigm. However, it should also be noted that the research also
has some elements of interpretism as evidenced by the use of multiple research
design.
5.4 Research Approaches
In research, there are two broad methods of logical reasoning namely the deductive
and inductive approaches (Apulu, 2012; Saunders et al., 2009; Sophonthummapharn,
2008). The two approaches are further explained in the next section.
5.4.1 Deductive Approach
Deductive reasoning works from the more general to the more specific, and is often
referred to as a top down approach (Sophonthummapharn, 2008). In this school of
thought, theories or hypothesis are developed and then tested through empirical
observation (Apulu, 2012). According to Sophonthummapharn (2008), researchers
might begin by examining theories related to their topic of interest. They then narrow
those theories down to more specific research questions or hypotheses that can be
tested. Then, the researchers answer questions or confirm hypotheses through a

141
number of research methods, mainly in quantitative ways in order to be able to
generalize the findings (Saunders et al., 2009; Sophonthummapharn, 2008).
5.4.2 Inductive Approach
Inductive reasoning works the other way, moving from more specific observations to
broader generalizations and theories (Apulu, 2012; Sophonthummapharn, 2008).
This is often called a bottom up approach. In inductive reasoning, the researchers
begin with specific observations and measures, begin to detect patterns and
regularities, and then formulate some tentative hypotheses that they can explore and
finally end up developing some general conclusions or theories (Apulu, 2012;
Sophonthummapharn, 2008). According to Apulu (2012) the hypothesis and or
theories developed in inductive research try to explain the empirical observations of
the real world. Table 5.2 provides a summary of comparison of the two approaches
to research.
Table 5.2: Major Differences between Deductive and Inductive Approaches

to Research
Deduction emphasises Inductive emphasises

Scientific principles Gaining an understanding of the
meanings humans attach to events
Moving from theory to data A close understanding of the research
context
The need to explain causal relationships The collection of qualitative data
between variables
The collection of quantitative data A more flexible structure to permit
changes of research emphasis as the
research progresses
The application of controls to ensure A realization that the researcher is part
validity of data of the research process
The operationalisation of concepts to Less concern with need to generalise
ensure clarity of definitions
A highly-structured approach
142
Researcher independent of what is

being researched
The necessity to select samples of
sufficient size in order to generalise
conclusions
5.4.3 Research Approach Employed in the Current Study
This study is a deductive research. Technology adoption models and theories in
SMEs, together with similar studies were reviewed and discussed in order to develop
a conceptual model to guide this research. Then, a number of research hypotheses
were formulated and tested in order to understand the adoption of ICTs among SMEs
in Tanzania. This being a deductive research, the findings were expected to be
generalised.
5.5 Research Strategies
Saunders et al. (2009) defined research strategy as “the general plan of how the
researcher will go about answering the research questions”. On a similar note,
Sophonthummapharn (2008) claim that, research strategy refers to the research
procedure used to answer research question(s) and fulfil the purposes of the research.
Research strategy is also called modes of observation (Sophonthummapharn, 2008).
The literatures show that there are a number of research strategies that can be
employed for exploratory, descriptive, and explanatory research (Saunders et al.,
2009; Sophonthummapharn, 2008). Moreover, Saunders et al. (2009) point out that,
appropriate research strategy has to be chosen based on research questions and
objectives, the extent of existing knowledge on the subject area to be researched, the
143
amount of time and resources available, and the philosophical underpinnings of the
researcher. According to Sophonthummapharn (2008), no particular research
strategy is inherently superior or inferior to another. Each strategy has its own
strengths and weaknesses for any particular research situation (Saunders et al., 2009;
Sophonthummapharn, 2008; Wedawatta et al., 2010). In addition, research strategies
can be used either jointly or separately in any research project (Saunders et al., 2009;
Sophonthummapharn, 2008; Wedawatta et al., 2010).
The literature shows several major research procedures that can be labelled as
research strategies. These includes experiment, survey, case study, action research,
grounded theory, ethnography, archival research, cross sectional studies, longitudinal
studies and participative enquiry (Saunders et al., 2009; Sophonthummapharn, 2008;
Wedawatta et al., 2010). This research adopts the case study and survey research
strategies. In the following sections, the case study and survey strategies are briefly
described and the reasons for their choice are explained.
5.5.1 Case Study
The literature (Gerring, 2004; Yin, 2009; Zucker, 2009) has multiple definitions and
understandings of what a case study is. Yin (2009) for example defines a case study
as “An empirical inquiry about a contemporary phenomenon (e.g., a “case”), set
within its real-world context especially when the boundaries between phenomenon
and context are not clearly evident”.

144
On the other hand, Sophonthummapharn (2008) cites other literatures and further
explain that a case study refers to “A strategy for doing research which involves an
empirical investigation of a particular contemporary phenomenon within its real life
context using multiple sources of evidence”. According to the literature (Costello,
2009; Saunders et al., 2009; Sophonthummapharn, 2008; Yin, 2009) case studies can
be used in both exploratory, descriptive and explanatory research and they have the
ability to answer the ‘why’ ‘what’ and ‘how’ questions. Yin (2009) noted that, a
case study strategy should be considered when:
i. The focus of the study is to answer “how” and “why” questions
ii. You cannot manipulate the behaviour of those involved in the study
iii. You want to cover contextual conditions because you believe they are
relevant to the phenomenon under study or,
iv. The boundaries are not clear between the phenomenon and context.
According to the literature (Costello, 2009; Saunders et al., 2009;
Sophonthummapharn, 2008; Yin, 2009), case studies may either focus on a single
case or use a number of cases: A single case may form the basis of research on
typical, critical or deviant cases, while multiple cases may be used to achieve
replication of a single type of incident in different settings, or to compare and
contrast different cases (Sophonthummapharn, 2008). Case studies can also be

145
holistic or embedded (Sophonthummapharn, 2008; Yin, 2009). Typically, case
studies can be conducted using several data collection techniques including
interviews, observation, document gathering and analysis, focus groups, and
questionnaires. According to Costello (2009), case studies are the most common
qualitative method used in IS research.
5.5.2 Survey
A quick literature review reveals many different definitions of what constitutes a
survey (Leeuw et al., 2008). Leeuw et al. (2008), cites a number of books on survey
methodology that immediately describe the major components of surveys and of
survey error instead of giving a definition, and other books that give definitions
ranging from concise definitions to elaborate descriptions of criteria. According to
Leeuw et al. (2008), looking at the common things in these definitions, a survey can
be seen as “A research strategy in which quantitative information is systematically
collected from a relatively large sample taken from a population”.
The literature (Babbie, 2004; Saunders et al., 2009, Sophonthummapharn, 2008)
show that, survey research strategy is the most popular and common strategy for
social research, including business disciplines (Babbie, 2004; Saunders et al., 2009,
Sophonthummapharn, 2008). This strategy can be used to answer ‘who,’ ‘what,’
‘where,’ and ‘how’ questions and is mainly used in descriptive and exploratory
research (Sophonthummapharn, 2008). The strategy is generally associated with the
deductive research approach. One of the strengths of a survey strategy is the fact that
it allows researchers to collect a large amount of data from a substantial population at

146
a very low cost (Sophonthummapharn, 2008). As suggested by the definition, the
data are typically quantitative and are elicited from subjects by questioning. More
often than not, the questions are asked using questionnaires. The questions can also
be asked by face-to-face or telephone interviews or by using or a mixture of
questionnaires and interviews (Sophonthummapharn, 2008). It should be noted that,
the quality of respondents and the questions asked affect the research findings. Thus,
questionnaire construction and sampling procedures must be considered seriously
before data collection because the better they are, the more reliable and valid the data
is obtained and this leads to more accurate research generalizations. The collected
data can be easily compared and analysed using various statistical techniques. Survey
research strategy is usually the preferred by researchers who are interested in
collecting original data to explain a population that is too large to observe directly.
Careful probability sampling provides a group of respondents whose characteristics
may be taken to reflect those of the larger population, and carefully constructed
standardized questionnaires provide data in the same form from all respondents
(Babbie, 2004). In more detail, a survey strategy provides researchers more control
over the research process, and it is possible to generate findings that are
representative of the whole population at a lower cost than collecting the data for the
whole population (Saunders et al., 2009).
5.5.3 The Research Strategy Used in the Current Study
As mentioned in section 5.2.5 this research employs multiple research design as
depicted in Figure 5.2. This choice results in the use of multiple strategies for the
research since each research strategy has its strengths and weaknesses for a given
147
research design. The literature (Costello, 2009; Saunders et al., 2009;
Sophonthummapharn, 2008) show that no research strategy is inherently superior or
inferior to another and that research strategies can be used either jointly or separately
in any research project. A typical example is Costello (2009) who combines case
studies and surveys in one research project. According to Sophonthummapharn
(2008), a number of issues should be considered when choosing which research
strategies to use. These include among other things the researcher’s philosophical
belief, research approach, experience, time and budget available. Taking these
factors into account as explained in earlier sections of this thesis, the use of case
studies and survey strategies in this research are justified.
In this work, multiple case studies strategy was used in the exploratory phase of the
research in order to help the researcher get a better understanding of the problem
before research hypotheses to be tested were developed. Thereafter, the survey
strategy was used for the descriptive and explanatory phases of this research. Semi
structured questions and interviews were used as data collection techniques for the
case studies while for the survey, self-administered questionnaires were used.
5.6 Research Choices
Literature review (Costello, 2009; Saunders et al., 2009; Sophonthummapharn, 2008;
Yin, 2009) show that, there are two major research methods: quantitative and
qualitative. According to Sophonthummapharn (2008), each of the two methods is
significantly different in the way data are collected and analysed (Costello, 2009;
Saunders et al., 2009; Sophonthummapharn, 2008; Yin, 2009). In other words, the
148
two methods use data of different formats. While as quantitative research method
focus on numerical data, qualitative research method on the other hand focuses on
non-numerical data such as voice text and pictures (Apulu 2012; Babbie, 2004;
Saunders et al., 2009). Sophonthummapharn (2008) and Saunders et al. (2009) note
that, when applied to a particular phenomenon, both methods have advantages and
disadvantages. Furthermore, the literature (ACAPS, 2012; Costello, 2009; Babbie,
2004; Saunders et al., 2009; Sophonthummapharn, 2008) show that, the two methods
have complementary roles in research and that no one method is better or dominates
the other. They are both widely used in social research (ACAPS, 2012; Babbie,
2004; Costello, 2009; Saunders et al., 2009; Sophonthummapharn, 2008). Saunders
et al. (2009) refer to “research choice” as the way a researcher chooses to combine
quantitative and qualitative techniques and procedures in a research. The two
methods are further explained in the next two sections.
5.6.1 Quantitative Research Method
Saglam and Milanova (2013) note that, there are many descriptions to quantitative
research. Creswell (2003) define quantitative research as “a type of research in
which the investigator primarily uses post-positivist claims for developing knowledge
(i.e., cause and effect thinking, reduction to specific variables and hypotheses and
questions, use of measurement and observation, and the test of theories), employs
strategies of inquiry such as experiments and surveys, and collects data on
predetermined instruments that yield statistical data”.

149
Generally, the quantitative research deals with numeric data. The method poses the
who, what, when, where, how much, how many, and how questions (Saglam and
Milanova, 2013). According to Sophonthummapharn (2008), the aim of quantitative
methods is to establish whether the predictive generalizations of a theory hold true.
In other words, quantitative research, tests pre-determined hypotheses and produce
generalisable results (ACAPS, 2012; Sophonthummapharn, 2008). It can thus be
argued that, quantitative research method is best suited for testing theories and
generalising findings (ACAPS, 2012; Costello, 2009; Saglam and Milanova, 2013;
Saunders et al., 2009; Sophonthummapharn, 2008). The vital skills needed for
successful conduction of a quantitative study include the ability to develop proper
hypotheses, testing the hypotheses with proper statistical techniques, and the ability
to correctly interpret statistical information into descriptive information
(Sophonthummapharn, 2008).
5.6.2 Qualitative Research Method
Ritchie and Lewis (2003) noted that, most qualitative research literature begin with
some attempt to define what is meant by this term, either theoretically or practically,
or both. However, giving a precise definition of qualitative research is difficult since
the term is used as an overarching category (Ritchie and Lewis, 2003). That is, it
covers a wide range of approaches and methods found within different research
disciplines (Ritchie and Lewis, 2003).
For practical reasons, this research adopts the definition by Nkwi et al. (2001) who
explain qualitative research as “any research that uses data that do not indicate
150
ordinal values” In this definition, the focus is on the data generated and or used. In
reality, the data in qualitative research is non-numeric and can be in the form of text,
images, and sounds (Apulu, 2012; Hancock et al., 2007; Nkwi et al., 2001; Ritchie
and Lewis, 2003; Sophonthummapharn, 2008).
The literature (Apulu, 2012; Hancock et al., 2007; Ritchie and Lewis, 2003) also
show that, qualitative research mainly subscribes to the Interpretism philosophical
point of view (Apulu, 2012; Hancock et al., 2007; Ritchie and Lewis, 2003).
Qualitative research poses the why? how? and in what way? questions as it seeks to
understand the meanings which people attach to phenomena (Creswell, 2003;
Hancock et al., 2007; Ritchie and Lewis, 2003). According to the literature (Apulu,
2012; Ritchie and Lewis, 2003; Saglam and Milanova 2013; Sophonthummapharn,
2008), qualitative research is basically used in areas where little or no literature exists.
In such cases, qualitative method is usually used to provide an in-depth insight into the
phenomena being investigated (Apulu, 2012; Ritchie and Lewis, 2003). Table 5.3
summarises the major differences between quantitative and qualitative research
methods.
Table 5.3. The Differences in Emphasis in Qualitative vs. Quantitative

Methods
Qualitative Methods Quantitative Methods

Emphasis on understanding Emphasis on testing and verification
Focus on understanding from Focus on facts and/or social events
respondent’s/informant’s point of view
Interpretation and rational approach Logical and critical approach
Observations and measurements in Controlled measurements
natural settings
Subjective ‘insider view’ and Objective ‘outsider view’ distant from
151
closeness to data data

Explorative orientation Hypothetical-deductive; focus on
hypothesis testing
Process oriented Results oriented
Holistic perspective Particularistic and analytical
Generalisation by comparison of Generalisation by population
properties and contexts of individual membership
organism
Source: Ghauri and Gronhaug (2002)
According to Sophonthummapharn (2008), skills that are needed by the researcher to
successfully conduct a qualitative study include the ability to think abstractly and
critically and to make judgment without bias.
5.6.3 Mixed Methods and Multi Methods
According to the literature (ACAPS, 2012; Creswell, 2003; Saglam and Milanova,
2013; Saunders et al., 2009; Sophonthummapharn, 2008), qualitative and
quantitative research methods can be used either independently or jointly in any
particular study. The literature (Apulu, 2012; Costello, 2009; Creswell, 2003;
Saunders et al., 2009; Sophonthummapharn, 2008) acknowledge that each one of
these methods has its strengths and weaknesses. According to Babbie (2004) more
often, to get a complete understanding of a research topic may require the use of both
methods. Therefore, the two methods can be used to complement each other
depending on the nature of the study (Apulu, 2012; Costello, 2009; Saunders et al.,
2009; Sophonthummapharn, 2008). Sophonthummapharn (2008) note that, when the
two methods are used in complimentary way, that allows researchers to conclude
their findings confidently.

152
In practice, the two methods can be used either as mixed method designs or as
multimethod designs (Esteves and Pastor, 2004). According to Esteves and Pastor
(2004) the mixed methods design is defined as “The incorporation of various
qualitative and quantitative strategies within a single project that may have either a
qualitative or quantitative theoretical drive. The “imported” strategies are
supplemental to the major or core method and serve to enlighten or provide clues
that are followed up within the core method”.
This definition implies that, one method could be used initially and then followed by
the other, depending on the nature of the study. In addition to the mixed method
design, Esteves and Pastor (2004) define multimethod design as “The conduct of two
or more research methods, each conducted rigorously and complete in itself, in one
project. The results are then triangulated to form a complete whole”
From this discussion, it can be concluded that, the selection of a research choice
depends on the nature of the study as suggested by the literature (Apulu, 2012;
Costello, 2009; Creswell, 2003; Esteves and Pastor, 2004; Saunders et al., 2009;
Sophonthummapharn, 2008). According to Creswell (2003), the most important
factors that play in this decision are: the research problem, the personal experiences
of the researcher, and the targeted audience.
5.6.4 The Research Methods Used in the Current Study
This research employed the multiple research design (see section 5.2.5) and that
made the use mixed method design necessary. The research problem demanded for
153
research findings that could be generalised to a large population (about 3million
SMEs in Tanzania). That called for a quantitative study using the survey strategy
(ACAPS, 2012; Costello, 2009; Saglam and Milanova, 2013; Saunders et al., 2009;
Sophonthummapharn, 2008).
However, prior to the major survey, a pilot study which was qualitative in nature was
deemed necessary in order to help the researcher get in depth knowledge of the
research topic before formulating the hypotheses and developing the instrument for
the study. Therefore, while as the core method of the research was the quantitative
method, the qualitative method was used to enlighten on the study as implied by the
definition of mixed method research.
5.7 Research Time Dimension
According to the literature (Kripanont, 2007; Peou, 2009; Saunders et al., 2009;
Sophonthummapharn, 2008), there are two principal options for conducting a
research: a cross-sectional or longitudinal study. A cross-sectional study is based on
observations of a sample, population, or phenomenon that are made at a single point
in time (Kripanont, 2007; Peou, 2009; Saunders et al., 2009; Sophonthummapharn,
2008), while longitudinal study is designed to permit observations of the same
phenomenon over an extended period of time (Kripanont, 2007; Peou, 2009;
Saunders et al., 2009; Sophonthummapharn, 2008). In other words, cross-sectional
study provides a snapshot of the phenomenon at a particular time (Kripanont, 2007;
Saunders et al., 2009; Sophonthummapharn, 2008), while longitudinal study is
mainly intended to provide answers if there are any changes on the observed
phenomenon over a period of time.

154
Generally, research time dimension depends on the intended purpose of the study and
availability of resources needed for the study (Kripanont, 2007; Saunders et al.,
2009; Sophonthummapharn, 2008). Sophonthummapharn (2008) claim that, in
technology adoption studies, cross-sectional study is more appropriate than
longitudinal study (Sophonthummapharn, 2008) and thus, most studies in the context
of technology adoption were conducted through cross-sectional time dimension
(Sophonthummapharn, 2008). This research fall in the category of cross-sectional
study since data was collected once (over a period of nine months) in order to answer
research objectives. This choice was made based on available resources for the
research especially time, budget and staff available. Moreover, the choice is
amenable to the common practise in this research domain as suggested by the
literature (Sophonthummapharn, 2008).
5.8 Data Collection
5.8.1 Development of Research Instrument
In this research, questionnaire served as the primary data gathering instrument. In
order to meet the objectives of the research, this instrument had to be developed and
tested for reliability and validity before it could be used in the major survey. This
section describes the questionnaire development process and explains how the
instrument was tested.

155
5.8.2 Pilot Case Studies
In this work, multiple case studies strategy was used in the exploratory phase of the
research in order to help the researcher get a better understanding of the research area
and the research problem. In this phase, semi structured interviews were used to
collect data from six companies. The main objective of the semi-structured
interviews was to provide the researcher with basic information needed in the design
of a formal survey instrument and not to perform detailed analysis of the case
studies. Thus, all significant variables derived from literature review were combined
with the information gathered from the case studies, with the aim of developing an
effective research instrument that was used in the main survey.
5.8.3 Questionnaire Design
Kripanont (2007) define the questionnaire as “A list of carefully structured
questions, chosen after considerable testing with a view to eliciting reliable
responses from a chosen sample” In this research, the questionnaire design process
was guided by the theories and the conceptual model presented in chapter 4 of this
thesis. According to the literature (Kripanont, 2007; MRS, 2011; Peou, 2009;
Sophonthummapharn, 2008) in order to minimize bias, a researcher needs to consider
the purpose of the research, planning of issues such as how different variables will be
categorised, scaled and coded after receipt of responses, the wording and the general
structure (appearance) of the questionnaire.
In this work, the researcher took all of these points into consideration while
designing the survey questionnaire for the study. In order to remain focused on the
156
purpose of the study, only the questions related to the research objectives were
integrated into the questionnaire. The wording of the questions was kept simple and
avoided the use of jargons and too technical language. This was aimed at enabling
respondents to understand the questions easily and answer them. Also, open ended
questions were kept to the minimum. To maintain a good structure, questions
dealing with similar aspects of the study were grouped together and each group
formed a section. Moreover, each section of the questionnaire was preceded by a
brief sentence or two explaining the purpose of the questions in that section. The
questionnaire had eight sections from section A to section H.
It should be noted that, this research was conducted in mainland Tanzania, a country
whereby Kiswahili and English are widely spoken and they save as the two official
national languages. Bearing that in mind, it was deemed necessary to have the
questionnaire in both English and Kiswahili languages in order to give respondents a
choice. That being the case, the questionnaire was designed first in English, the
official language in Institutions of higher learning and thereafter the questionnaire
was translated from English to Kiswahili. The translation of the questionnaire was
done in consultation with expert translators from the linguistics department of the
University of Dar es Salaam and the author was able to produce a Kiswahili version
of the questionnaire. During the translation, the focus was put on conceptual rather
than literal translation. It should be noted that, the translation process is beyond the
scope of this work. However, it suffices to say that, the translated instrument had the
same conceptual meaning as the original English instrument and thus was able to
157
capture all the wanted information. Different sections of the questionnaire are
further explained in the subsequent sections.
Section A: Company Profile
This section aimed at obtaining information about the general profile of the company
and the respondents. A total 13 questions with different scales were asked in this
section as summarized in Appendix IB section A. Note that question number 13 is an
open-ended question that was meant to measure the entrepreneur value of the owner-
manager. The questionnaire design in this section was purely based on literature
survey. According to the literature (Costello 2009; Sophonthummapharn, 2008; Van
Akkeren and Cavaye, 1999b), many similar studies that contain elements about the
firm, include age, size, sector, legal form, location and ownership. Thus, these
elements were included in this study.
Section B: ICT Infrastructure and Frequency of Use
This section of the questionnaire was designed to elicit the information about what
the company actually has in terms of ICT infrastructure and how often that
technology is used in support of firm activities. A total of six questions were asked
and the respondents were asked to choose all that apply to their firms. This way the
information about computer and the Internet technologies and their frequency of use
in SMEs was gathered. Question six gave the option to the respondent to provide
information about any other ICT the firm has that was not listed. All questions were
established as nominal scale. Appendix IB section B gives a summary of the
questions in this section

158
Section C: Computer and Internet Use
This section was used to gather information about how the computers and the
Internet are used in a firm. The literature (Chung, 2006; Modimogale, 2009) show
that, adoption of technology does not by itself guarantee firms competitiveness and
growth, but rather how the technology is used. There is a total of 13 questions
established as 5- point Likert–type interval scale (1=Strongly Disagree to 5 Strongly
agree). Four questions were used to look into how computers are used and 9
questions are used to look into how the internet is used by the firms. Appendix IB
section C shows the questions in this section.
Section D: ICT Adoption Factors-Personal Factors
The Personal Factors construct is one of the three independent variables of the ICT
adoption model as depicted by figure 4.1. This section focused on obtaining
information about the owner-manager (The Personal Factors) attributes that affect
adoption. In particular, the section seeks to capture the owner-manager perceptions
towards the benefits of using ICTs, since other attributes such as education and
knowledge of ICT has already been measured in section A. Appendix IB section D
shows the questions in this section. A total of twelve questions were asked in section
D and the scale was established as a 5-point Likert–type interval scale (1=Strongly
Disagree to 5=Strongly Agree).
Section E: ICT Adoption Factors-Organisation Factors

159
This section represents Organisation Factors which is the second independent
variable of the ICT adoption model as depicted by Figure 4.1 Within the context of
Organisation Factors, organisation readiness, strategy, and external pressure were
measured and the information obtained used in the ICT adoption modelling.
Appendix IB section E shows the questions used to measure these concepts.
Organisation Readiness was measured by the availability of adequate financial and
material resources within the company. Strategy was measured by looking at the
reasons given for adoption and decision making. External Pressure was measured by
looking at the competitive intensity, suppliers’ pressure, customer’s pressure and
pressure caused by technological advances. A total of 17 questions were asked in
this section of which question 1 to question 15 were established as a 5-point Likert–
type interval scale (1=Strongly Disagree to 5=Strongly Agree). Questions 16 and 17
were open ended questions meant to gather addition information about other
contributing factors.
Section F: ICT Adoption Factors-Economic Factors
Section F was concerned with the identification of Economic Factors that affects ICT
adoption in SMEs. This is the third independent variable of the ICT adoption model
as depicted by Figure 4.1. Within the context of Economic Factors, Access to
Financial Support and Cost of ICTs were measured. Access to financial support was
measured by the lack of capital sources, lack of cash in hand, lack of additional
capital, difficult to find loan, many requirements from financial institutions, banks
require collateral and high interest, and lack of financial institution for SMEs and the
160
need for immediate return on investment. Cost of ICT systems was measured by
acquisition, implementation and operational costs of ICT systems. In total, there was
16 questions in this section and the questions were posed as a 5-point Likert–type
interval scale (1=Strongly Disagree to 5=Strongly Agree). Appendix IB section F
show the questions asked to deduce this information. Note that all questions in this
section are negatively stated statements. Thus, after data collection and prior to any
analysis, data for these questions was transformed so that the entire data set had the
same positive direction.
Section G: Government Policies
This section focused on measuring the moderating effects of government policies on
ICT adoption by SMEs. The section has a total of eleven questions as shown in
Appendix IB section G. The questions were established as a 5-point Likert–type
interval scale (1=Strongly Disagree to 5=Strongly Agree) to measure availability of
support programs, taxes and tariff rates, availability of support infrastructure. The
eleventh question is an open-ended question meant to gather additional information
on other desirable government interventions. Note that all questions in this section
(except question G11) are negatively stated statements. After data collection and
prior to any analysis, data for these questions was transformed so that the entire data
set had the same positive direction.
Section H: SME Performance
This section was concerned with establishing the effect of ICT adoption on SMEs
(Post adoption Evaluation) thus the questions tried to elicit the information on
161
whether the adoption has been a success or not. Appendix IB section H show the
question on this section. Within this context, indicators related to return on
Investment (ROI) and Innovation was measured. ROI was measured by market size,
sale volume, additional capital and profits. Innovation was measured by the
emergence of new processes that enhance efficiency. The scale was established as a
5-point Likert–type interval scale (1=Strongly Disagree to 5=Strongly Agree) and a
total of ten questions were asked.
5.8.4 Questionnaire Pre-test
Prior to the use of the questionnaire in the survey, the instrument was pretested in
order to determine the accuracy and consistency of the responses. The details of this
questionnaire pre-testing are given in this section. According to the literature
(Kripanont, 2007; Peou, 2009; Sophonthummapharn, 2008), “questionnaire pre-
testing is a trial run of the instrument with a group of respondents with the aim of
detecting problems in the questionnaire”. Through pre-testing, the researcher will be
able to detect problems in the questionnaire instructions or design, determine
whether the respondents have any difficulty understanding the questionnaire and
whether there are any ambiguous or biased questions.
Kripanont (2007) suggests that, pre-testing should be administered to a sample that is
expected to respond similarly to the samples on which the scale eventually will be
applied. On the other hand, for the purpose of refining a measuring instrument, pre-
testing may rely on colleagues, respondent surrogates, or actual respondents
(Kripanont, 2007; Peou, 2009; Sophonthummapharn, 2008). The literature

162
(Kripanont, 2007; Peou, 2009; Sophonthummapharn, 2008), suggests that, 20 to 50
subjects may be enough group size for pre-testing of a questionnaire.
In this research, the first pre-test involved two groups. The first group was made up
of two University of Dar es Salaam professors whom are also supervisors of this
work. The second group had five PhD students. The two groups were given the
English questionnaire for pre-testing, and all 7 respondents returned the
questionnaire with their comments. Some questions were also removed from the
questionnaire since they were not adding any value to the study.
In the second pre-test, a group with 30 entrepreneurs were given both the English and
Kiswahili versions of the instrument for evaluation. 21 questionnaires were returned
with comments. This was a 70% return rate. The comments provided by these
individuals, assisted the researcher to improve the questionnaire, particularly on the
layout and the sequencing of the questions. A covering later (See Appendix I-A) that
explains to the respondents the objectives of the research and what the data will be
used for, was prepared to be sent out with each questionnaire. The final pre-test of
the questionnaire was a pilot survey and is covered in the following section.
5.8.5 Pilot Survey
According to Simon (2011), a pilot survey can refer to a small-scale version of the
larger survey. In other words, a pilot survey refers to trial runs done in preparation
for the major study. On the other hand, pilot studies can also refer to trying out or
pre-testing of a particular research instrument or research procedures before a full

163
scale research is carried out (Kripanont, 2007; Simon, 2011). Simon (2011) and
Kripanont (2007) note that, it is a good practise to carry out a pilot survey prior to the
main survey. Simon (2011) notes that, while pilot surveys does not guarantee the
success of the main study (Simon, 2011), they can give advance warnings with
regard to the weaknesses in the proposed study (Kripanont, 2007; Simon, 2011).
Kripanont (2007) asserts that, a pilot survey must always draw subjects from the
target population. It must also simulate the procedures and protocols that have been
designed for data collection (Kripanont, 2007; Simon, 2011). Simon (2011) suggests
that a sample size of between 10% and 20% of the actual sample for the main study
is adequate in many studies. Kripanont (2007) claim that a sample size of 25 to 100
respondents is enough.
In this research, a pilot study was carried out with the intention of collecting data for
further refinement of the research instrument. In particular, the data obtained
assisted in testing the reliability and validity of the instrument before the major
survey was carried out. Moreover, the pilot survey helped the research team to gain
familiarity with respondents, test field work arrangements, estimate questionnaire
completion time and estimate response rate. The researcher also used the pilot study
for training and testing of research assistants.
A total of 50 questionnaires were delivered in person, by the researcher and his
research assistants to respondents in all three districts of the Dar es Salaam region,
which are Ilala, Kinondoni and Temeke. Respondents were given a time frame of
two weeks to complete the questionnaire, after which the researcher or his assistants
164
collected the questionnaires. Out of 50 questionnaires 43 questionnaires were
collected back and out of which 37 were deemed good for analysis and 6 were
discarded. The obtained data was analysed using IBM SPSS version 20 and the
results obtained helped the researcher to further refine the instrument until the
instrument was ready for survey. The tests conducted on the pilot data are explained
in the next sections.
5.8.6 Reliability Analysis of the Research Instrument
According to Kripanont (2007), the reliability and validity test of a research
instrument are all about testing the goodness of data obtained. Reliability can be
defined as “The extent to which research findings would be the same if the research
were to be repeated at a later date, or with a different sample of subjects”
(Kripanont, 2007). In other words, the reliability of a measure indicates the extent to
which the measure is without bias (error free) and hence offers consistent
measurement across time and across the various items in the instrument (Kripanont,
2007; Sekaran, 2003). It helps to assess the goodness of measure, and indicates
accuracy in measurement (Sekaran, 2003).
To test the reliability of the research instrument used in this study, the Cronbach’s
coefficient alpha, the item-to-total correlation and the inter-item correlation were
used. The Cronbach’s coefficient alpha is one of the most popular tests of inter-item
consistency reliability (Kripanont, 2007; Peou, 2009; Sophonthummapharn, 2008).
The Cronbach’s coefficient alpha has a value between 0 and 1 (Kripanont, 2007;
Peou, 2009; Sophonthummapharn, 2008) and according to the literature (Kripanont,

165
2007; Peou, 2009; Sophonthummapharn, 2008) reliabilities in the range of 0 to 0.6
are considered to be poor, those in the 0.7 range, acceptable, and those over 0.8 good.
Generally speaking, the closer the reliability coefficient gets to 1, the better.
Kripanont (2007) and Sophonthummapharn (2008) note that, in most research, the
generally agreed upon lower limit for Cronbach’s alpha is 0.70.
The item-to-total correlation and the inter-item correlation are other measures that
can also be used to assess internal consistency of a research instrument (Hair et al.,
2006; Kripanont, 2007) and thus were also used in this study. According to
Kripanont (2007), correlation, (both positive and negative correlation), in the range
of r = 0.10 to 0.29 is considered small correlation, r = 0.30 to 0.49 is considered
medium correlation and r = 0.50 to 1.00 is large correlation (Hair et al., 2006;
Kripanont, 2007). As a rule of thumb, when used to assess internal consistency of
the instrument in research, the item-to-total correlations should be 0.50 and above,
while the inter-item correlations should exceed 0.30 (Hair et al., 2006; Kripanont,
2007). Table 5.4 presents reliability test results performed on the pilot survey data.
From Table 5.4 it can be seen that, the Cronbach’s alphas for all scales and sub
scales is in the range of 0.701 to 0.987 with only two sub scales registering
Cronbach’s alphas of 0.701 and 0.762. The rest of scale registered Cronbach’s
alphas above 0.800. This indicates that the items in each concept were positively
correlated and thus point out the accuracy in measurement (Hair et al., 2006;
Kripanont, 2007; Peou, 2009; Sophonthummapharn, 2008).

166
Moreover, Table 5.4 shows that, one item related to Organisation Factors, had inter-
item correlation value less than 0.3 and four items, two items in the Organisation
Factors and other two items in the Government Policies had item-to-total correlation
values of less than 0.5. According to Hair et al. (2006) and Kripanont (2007), the
inter-item correlations should exceed 0.30 and that the item-to-total correlations
should exceed 0.50.
Table 5.4: Pilot Survey Reliability and Internal Consistence Analysis Results
Measurement Item Number Cronbach’s Reliability Inter-item Item-to-total

(Interval Scale) of Items alpha Results correlation correlation
Adoption and Use
-Computer 4 0.977 Good 0.825~0.968 0.902~0.988
-Internet 9 0.971 Good 0.568~0.940 0.718~0.952
-All 13 0.983 Good
Personal Factors
-Perceived Benefits 12 0.966 Good 0.373~0.949 0.633~0.892
Organisation Factors
Organisation
Readiness
-Adequate Resources 3 0.701 Acceptable 0.265~0.751 0.338~0.683
Strategy
-Decision making 3 0.760 Acceptable 0.355~0.741 0.458~0.729
-Reason for adoption 5 0.922 Good 0.459~0.924 0.551~0.895
External Pressure 4 0.926 Good 0.589~0.923 0.667~0.922
All 15 0.941 Good

Economic Factors
- Financial Support 8 0.933 Good 0.392~0.963 0.711~0.876
-Costs of ICT systems 8 0.969 Good 0.601~0.907 0.787~0.923
167
All 16 0.957 Good

Government Policies
-Support Programs 4 0.862 Good 0.380~0.848 0.462~0.857
-Taxes and Tariffs 3 0.845 Good 0.606~0.693 0.685~0.756
-Support Infrastructure 3 0.812 Good 0.410~0.762 0.536~0.804
All 10 0.933 Good

SME Performance
-Computer 5 0.980 Good 0.850~0.991 0.912~0.971
-Internet 5 0.930 Good 0.499~0.953 0.697~0.903
All 10 0.976 Good
5.8.7 The Validity of Research Instrument
According to the literature (Golafshani, 2003; Kripanont, 2007; Twycross and
Shields 2000), validity refers to how well a test, measures what it is claim to
measure. In other words, validity can be thought of as “the extent to which the data
collected truly reflect the phenomenon being studied” (Kripanont, 2007). According
to Golafshani (2003), validity test is all about determining whether the research truly
measures that which it was intended to measure or how truthful the research results
are. Several types of validity tests for testing the goodness of measures have been
suggested in the literature (Kripanont, 2007). These include content validity,
criterion-related validity, and construct validity. In this research content validity and
construct validity were carried out.
5.8.7.1 Content Validity
Content validity also referred to as the face validity, is a test that assesses whether a
tool appears to others (expert judges, and pre-tests) to be measuring what it says it
does (Golafshani, 2003; Hair et al., 2006; Kripanont, 2007; Twycross and Shields
168
2000). In this research content validity was done as the instrument was read by two
Professors supervising this work. Moreover, pre-tests involving PhD students were
also carried out. The input provided by both the Professors and pre-tests were used
to refine the research instrument.
5.8.7.2 Criterion-related Validity
Criterion-related validity is a test used to predict future or current performance by
correlating test results with another criterion of interest (Golafshani, 2003; Hair et
al., 2006; Kripanont, 2007; Twycross and Shields 2000). The criterion-related
validity can be either concurrent or predictive (Golafshani, 2003; Hair et al., 2006;
Kripanont, 2007; Twycross and Shields 2000). According to Twycross and Shields
(2000) while as concurrent validity uses an already existing and well-accepted
measure against which the new measure can be compared, predictive validity
measures the extent to which a tool can predict a future event of interest. Criterion
validity is usually measured using a correlation coefficient (Hair et al., 2006;
Kripanont, 2007; Twycross and Shields 2000) whereby, when the correlation is high,
the tool is considered to be valid.
5.8.7.3 Construct Validity
Construct validity tests the link between a measure and the underlying theory
(Golafshani, 2003; Hair et al., 2006; Kripanont, 2007; Twycross and Shields 2000).
Kripanont (2007) claim that, construct validity can be used to testify to how well the
results obtained from the use of the measure fit the theories around which the test
169
was designed. According to Twycross and Shields (2000), construct validity is
usually measured using a correlation coefficient and the rule of thumb is, when the
correlation is high, the tool can be considered valid. Kripanont (2007) suggests that,
construct validity can be established through; correlation analysis (convergent and
discriminant validity); factor analysis, and; the multi-trait, multi-method, method
matrix of correlations. In addition, Kripanont (2007) note that, the three most widely
accepted forms of validity by other researchers are convergent, discriminant, and
nomological validity (Kripanont, 2007). In this research, convergent validity and
discriminant validity were used to establish construct validity. Convergent validity
of research instrument is explained in this section, while discriminant validity is
discussed in chapter seven where the survey data is tested for validity as part of
model development.
Convergent validity determines the degree to which two measures of the same
concept are correlated (Hair et al., 2006; Kripanont, 2007). High correlation
indicates that the scale is measuring its intended concept (Hair et al., 2006;
Kripanont, 2007). Viewed in this light, reliability is also an indicator of convergent
validity (Hair et al., 2006; Kripanont, 2007). Table 5.4 shows that there was one
item with inter-item correlation value less than 0.3, and four items with item-to-total
correlation values of less than 0.5.
Based on these results, that is, the Cronbach’s alphas, the inter-item correlations and
the item-to-total correlations as presented in Table 5.4, the questionnaire was proved
to be a reliable instrument for data collection. Variables that caused the low values
170
in item-total-correlation and the inter-item-correlation were identified and removed
from the questionnaire. These includes 4 items measuring policy awareness in the
Government policies, one item from support programs, three items that measured
adequate resources in Organisation factors, and three items measuring decision
making as part of strategy. After that, the questionnaire was revised before the final
survey was performed. Also, before the final survey was conducted, it was
necessary to determine the sample careful so as to allow generalisation of the results.
The next section gives detailed information on how the sample for this research was
obtained. The introduction letter to respondents, research questionnaire and the
coding sheets for this research are provided in appendix IA
5.8.8 The Need for Sampling
The literature (Costello, 2009; Saunders et al., 2009; Sophonthummapharn, 2008;
Yin, 2009) show that, there are two basic options for a researcher to obtain
information about a population through questioning or testing:
i. Every member of the population can be questioned or tested. This called a
census.
ii. Only selected members of the population are questioned or tested. This is
known as sampling.
From the two approaches above, a census can further be explained as a study using
all available elements (members) of a population. Sampling on the other hand may
be defined as the act, process, or technique of selecting a suitable sample, or a

171
representative part of a population for the purpose of determining parameters or
characteristics of the whole population. In other words, it is the process of obtaining
information about an entire population by examining only a part of it (Costello,
2009; Kothari, 2004; Saunders et al., 2009; Sophonthummapharn, 2008; Yin, 2009).
The choice between the census method and sampling depends on a number of factors
the major ones being the reasons for the research, the size of the population being
studied, economic reasons and time constraints (Costello, 2009; Kothari, 2004;
Saunders et al., 2009; Sophonthummapharn, 2008; Yin, 2009). Generally, when the
population of interest is small, there is no use resorting to sampling. However, when
the target population is very large, conducting a census becomes extremely
expensive, difficult, and time consuming. Thus, census is conducted only when it is
absolutely necessary (Costello, 2009; Kothari, 2004; Saunders et al., 2009;
Sophonthummapharn, 2008; Yin, 2009). On the other hand, sampling is mandatory
in research, when the group of interest is too large, containing too many cases,
elements, or members which make it impossible to collect data from all of them
(Costello, 2009; Saunders et al., 2009; Sophonthummapharn, 2008; Yin, 2009). In
this research sampling procedures were used as it was not possible to collect data
from the entire group of interest given the research time frame and budget.
5.8.8.1 Sampling Procedures
Figure 5.3 gives a general summary of the sampling process. The sampling
procedures include defining the sampling population, development of the sampling

172
frame, specifying sampling methods, determination of the sample size and selecting
the sample. The sampling procedure is further explained in subsequent sections.
Figure 5.3: Sampling Process
5.8.8.2 Study Population
In sample survey literature, study population or the universe is defined as the
collective of study units for which the values of the variates of interest could possibly
be determined (Costello, 2009; Kothari, 2004; Saunders et al., 2009;
Sophonthummapharn, 2008; Yin, 2009). In this research, SMEs in mainland
Tanzania were the target population and the owners or managers of these SMEs were
the sampling units. According to the Tanzania Ministry of Trade and Industry
(MIT), in year 2010 there were about 2,679,626 people in Tanzania mainland
who owned and ran micro, small, or medium size business (URT, 2012). Thus,
this research had a population of 2,679,626.

173
5.8.8.3 Sampling Frame
The sampling frame, also known as the source list or population list refers to the list
from which the study sample is drawn from (Costello, 2009; Kothari, 2004; Saunders
et al., 2009; Sophonthummapharn, 2008; Yin, 2009). The list contains the names of
all items of a universe (in case of finite universe). If the source list is not available,
the researcher has to prepare it (Kothari, 2004; Peou, 2009; Saunders et al., 2009;
Sophonthummapharn, 2008). Peou (2009) note that, the source list should be
comprehensive, correct, reliable and appropriate. In addition, it is extremely
important for the source list to be as representative of the population as possible
(Costello, 2009; Kothari, 2004; Peou, 2009; Saunders et al., 2009;
Sophonthummapharn, 2008; Yin, 2009). According to Peou (2009) it is generally
impractical to find a complete list of the sampling frame if the target population is very
large and dynamic. A source list drawn from such a population always has some
degree of incorrectness because the population of interest changes regularly or is too
large (Peou, 2009). Thus, the researcher must take this fact into consideration when
generalizing the research findings to the entire population because if a sample frame
does not fully represent an intended population, this will result in frame error and
affect the degree of reliability of sample result (Costello, 2009; Kothari, 2004;
Saunders et al., 2009; Sophonthummapharn, 2008; Yin, 2009).
In this study, it was found that there was no formal comprehensive list of all SMEs in
Tanzania. The study population list was thus created from a combined list using data
obtained from District Directors, the Tanzania Revenue Authority (TRA) database and
the Internet. The use of combination list is a common practise in studies that lack a
174
single comprehensive database from which the population frame can be drawn (Peou,
2009). Given the budget and time frame of this study as well as the type of the
population, a source list of more than 2,679, 626 businesses is considered too big and
very dynamic since SMEs start-ups and dysfunctions occur regularly.
5.8.8.4 Sample Size
The sample size refers to the number of items to be selected from the universe to
constitute a sample (Kothari, 2004; Saunders et al., 2009; Sophonthummapharn,
2008; Yin, 2009). The size of the sample should reflect the degree of being
representative of the entire population from which it is drawn and how confidently
researchers can make a generalization of the research findings (Kothari, 2004;
Saunders et al., 2009; Sophonthummapharn, 2008; Yin, 2009). While it can be
proven that the larger the sample size, the better researchers can generalize findings
to the population, when other factors such as time and resource constraints are taken
into consideration, it becomes a reality that the size of sample should neither be
excessively large, nor too small but should be optimum (Kothari, 2004; Saunders et
al., 2009; Sophonthummapharn, 2008; Yin, 2009). An optimum sample is one which
fulfils the requirements of efficiency, representativeness, reliability and flexibility
(Kothari, 2004; Saunders et al., 2009; Sophonthummapharn, 2008; Yin, 2009).
Sample size may be determined by using subjective methods or by using Statistical
methods (Kothari, 2004; Saunders et al., 2009; Costello, 2009;
Sophonthummapharn, 2008; Yin, 2009). Subjective (Ad hock) methods are less
sophisticated than statistical methods and they generally use the following
175
approaches: The rule of thumb approach e.g. take 5% of the population as the sample
size; conventional approach e.g. use of average of sample sizes used in similar or
other studies; the cost basis approach e.g. use the number that can be studied with
the available funds and use of published tables, which provide the sample size for a
given set of criteria (Sophonthummapharn, 2008). In this regard, the literature
shows that a sample size between 30 and 500 is adequate for most research
(Sophonthummapharn, 2008).
Statistical methods on the other hand are more sophisticated than subjective methods
However, when statistical methods are used to determine the sample size, they
enable the researcher to establish a more precise degree of being representative and
to allow researchers to have more confidence when generalizing the findings
According to the literature (Israel, 2013; Sophonthummapharn, 2008) in addition to
the purpose of the study and population size, three criteria usually will need to be
specified to determine the appropriate sample size: the level of precision sometimes
called the sampling error, the level of confidence or risk, and the degree of
variability in the attributes being measured. Israel (2013) suggests that, for
populations that are very large, equation 5.1 can be used to find a representative
sample for proportions.
Where
176
When the sample size is known, and is not very large, the literature (Israel, 2013)
show an alternative simplified formula for the calculation of a representative sample
for proportions as given by Equation 5.2.
Where
When the population is small, the sample size can be reduced slightly (Israel, 2013).
This is because a given sample size provides proportionately more information for a
177
small population than for a large population (Israel, 2013). The sample size can
thus be adjusted using Equation 5.3.
In this study, the formula of Equation 5.1 was used to calculate the minimum sample
size, since the population of about 2,679, 626 sampling units is considered too large.
and in this research refer to the proportion of computer and Internet adopters
and non-adopters in Tanzania mainland SMEs. Since there were no initial estimates
of and , the worst-case scenario of = 0.5 and = 0.5 was selected. Note
that the value 0.5 (maximum variability) is always assumed if and are not
known a priori (Israel, 2013; Saunders, 2009; Sophonthummapharn, 2008). Given
the research time and budget constraints, precision level (margin error) of ± 5% and a
confidence level of 95% which has a corresponding Z value of 1.96 were selected. It
should be noted that, these values are widely used and accepted in academic and
research community (Israel, 2013; Saunders, 2009; Sophonthummapharn, 2008).
Thus, from Equation 5.1 the minimum sample size is found as:
In anticipation of the non-response error in survey research particularly in Least
Developed Countries (Aikaeli, 2013; Peou, 2009) in this study an estimated response
178
rate of 40% was used and this resulted in an actual sample size of 962.5 which was
adjusted to 1000. Thus, a sample size of 1000 was used in this research.
5.8.8.5 Sampling Methods
Sampling methodologies are classified under two general categories: Probability
sampling and non-probability sampling (Israel, 2013; Saunders, 2009;
Sophonthummapharn, 2008). In non-probability sampling, the researcher knows the
exact possibility of selecting each member of the population while as in probability
sampling, the chance of being included in the sample is not known. A probability
sample tends to be more difficult and costly to conduct. However, probability
samples are the only type of samples where the results can be generalized from the
sample to the population (Israel, 2013; Saunders, 2009; Sophonthummapharn, 2008).
In addition, probability samples allow the researcher to calculate the precision of the
estimates obtained from the sample and to specify the sampling error (Aikaeli, 2013;
Israel, 2013; Saunders, 2009; Sophonthummapharn, 2008). Non-probability samples,
in contrast, do not allow the study's findings to be generalized from the sample to the
population (Aikaeli, 2013; Israel, 2013; Saunders, 2009; Sophonthummapharn,
2008). When discussing the results of a non-probability sample, the researcher
must limit the findings to the persons or elements sampled. Detailed explanation of
different types of sampling techniques is found in the literature (Aikaeli, 2013; Israel,
2013; Saunders, 2009; Sophonthummapharn, 2008).

179
Due to the nature of the study design, both methods of sampling were used to collect
data from the Owners-managers of SMEs in Tanzania. Purposive sampling was
made in the initial phase of the study where multiple case studies were used in
exploratory phase of the research. At that point, the aim was to get some helpful
information that could help shape the main study and not to try generalising case
study findings, and thus purposive sampling was ideal for that situation. In the
survey phase of the study, the main objective was to collect data to be used in the
development of an ICT adoption model that could be used by SMEs in Tanzania.
The use of probabilistic sampling was thus mandatory in order to be able to
generalise the results. A two-stage area sampling which is a type of complex
probabilistic sampling (Aikaeli, 2013; Israel, 2013; Saunders, 2009;
Sophonthummapharn, 2008), was applied where four geographical regions namely
Arusha, Dar es Salaam, Mbeya and Mwanza were chosen as large primary sampling
units and within these regions, some districts namely; Kinondoni, Temeke, Ilala,
Nyamagana, Ilemela, Sengerema, Arusha City Council, Meru and Mbeya City Council,
were chosen as secondary sampling units. Within each secondary sampling unit, a
sampling frame was made and then a simple random sampling was used to obtain the
respondents for the survey.
The criterion used in the choice for the regions and districts is the fact that these
regions are home to major urban centres in Tanzania where most SMEs are located
(Mbamba 2003, TRA 2013; TRA, 2015, URT, 2012). This allowed for the SMEs to be
easily reached and this reduced research time and cost (Kripanont, 2007; Mbamba
2003, Sophonthummapharn, 2008).

180
Table 5.5a; show the distribution of the sample into the primary sampling units, that is,
the regions. This distribution was made proportional to the number of SMEs in each
district based on data obtained from TRA and the district director’s offices. According
to the TRA data, Arusha region had about 11, 610 SMEs, Dar es Salaam had about
25,300 SMEs, Mbeya had about 1,146 SMEs and Mwanza had about 4855 SMEs.
Table 5.5a: Sample Distribution in Participating Regions
Region Sample Distribution Sample Size %

Arusha 250 25%
Dar es Salaam 500 50%
Mbeya 100 10%
Mwanza 150 15%
Total 1000 100%
Looking at these figures it can be seen that, Dar es Salaam region had more SMEs than
the rest of regions combined together thus Dar es Salaam was given 50% of the
samples. Arusha region had almost twice the number of SMEs compared to Mbeya
and Mwanza combined thus Arusha region was allocated with a half of the remaining
samples, that is, 25% of the total samples. In a similar way, Mwanza was allocated
with 15% and Mbeya got 10% of the samples.
Table 5.5b, show sample distribution into secondary sampling units that is the different
districts. Proportional distribution was applied again based on TRA data and data
obtained from District/Municipal Council directors.
Table 5.5b: Sample Distribution in Participating Districts
Region District Sample Distribution Sample Size %

Arusha 200 20%
181
Arusha Meru 50 5%
Ilala 250 25%
Dar es Salaam Kinondoni 150 15%
Temeke 100 10%
Mbeya Mbeya 100 10%
Nyamagana 80 8%
Mwanza Ilemela 20 2%
Sengerema 50 5%
Total 1000 100%
5.9 Data Collection Procedures
In this research data were collected from nine districts of four regions in mainland
Tanzania as shown by Table 5.5b. The self-administered questionnaires and a
covering letter were hand distributed by the researcher and his assistants to the unit
of analysis, that is, the owners-managers of SMEs. Participation in the survey was
voluntary. Each district had one research assistant and thus a total of nine research
assistants were employed. The research team explained to the respondents that it
was very important that the questionnaire was completed by the person who was in
charge of the firm and not anybody else. Filled questionnaires were again physically
collected from respondents by the research team. During the collection phase of the
questionnaires, whenever possible, the researcher and his assistants checked the just
received questionnaire for any missing data and politely asked for the respondent to
finish filling in any missing data. Although this tactic did not work all the time, it
nevertheless helped a great deal to reduce the volume of missing data and thus had a
positive impact on the quality of the research.

182
The choice of research instrument administration was motivated by previous research
conducted in Tanzania settings (Mbamba, 2005). Mbamba (2005), used a similar
method due to difficulties associated with data collection. In this research, the same
technique was used with the aim of improving the response rate.
Data collection commenced on 23rd January, 2014 and went on until 23rd September
2014. Initially, data collection was expected to last for six months only. However,
due to problems encountered on the field, in particular the low response rate, data
collection exercise was extended for three months and thus a total of nine months
was spent on the field.
5.9.1 Unique Problems during Data Collection
One notable problem which made it necessary to extend the data collection time was
the dispute between the Tanzania Revenue Authority (TRA) on one hand and the
business community on the other hand, regarding the use of Electronic Fiscal
Devices (EFD) for aiding revenue collection by TRA. In compliance with the
Income Tax Act, through the Income Tax EFD Regulation of 2012, TRA began the
implementation of the so called second phase of EFD in year 2013 (TRA, 2015).
According to TRA (2015), this phase targeted the following business groups:
i. All businesses/persons that were not VAT registered with an annual turnover
ranging from TShs. 14 million and above.

183
ii. Traders trading in the Region’s prime areas, identified on the basis of rent
payable.
iii. Traders dealing with selected business sectors such as Spare Parts, Hardware,
Mini Supermarkets, Petrol stations, Mobile phone shops, Sub wholesale
shops, Bar and Restaurants, Pharmaceutical Stores; Electronic Shops etc.
Soon after the start of implementations, the targeted businesses in many parts of the
country, particularly in Dar es Salaam, Mwanza, Arusha and Mbeya refused to
comply and called for industrial action. Since then, there have been many attempts
by the government, TRA and the business community to resolve the issue through
negotiations, but up until the time of writing this thesis the dispute has remained
unresolved. From time to time, industrial action by traders in the said areas has taken
place.
The impact of this problem in this research has been enormous since businesses in
this dispute were actually the same businesses that made the population of this
research. In some cases, targeted respondents refused to participate in the survey in
a very hostile way to the researcher and/or his team. These people felt that, the
research (in a way that was not known to them at that time) was related to the TRA
exercise which they detested, since EFDs are ICT devices and the research was about
ICTs. In other instances, the targeted respondents took the questionnaire stayed with
it for long periods and did not attempt to fill in any details; no amount of follow ups
could move these people. In other cases, respondents requested to be paid some
184
amount of money (ranging from TShs 5,000 to TShs 10,000) before they could fill in
the questionnaire. Failure of that, they refused to participate in the survey. This had
a great impact on the response rate for the research, since the research had a limited
budget and time.
5.9.2 Response Rate Analysis
Out of the 1000 questionnaires distributed, 599 questionnaires were collected back
out of which, 194 questionnaires were rejected outright as they simply did not
contain any meaningful data. The remaining 405 questionnaires could be used for
one or more type of analyses and thus the initial response rate was 40.5%. The 405
questionnaires were coded and digitised in SPSS and the resulting data set was
named data set1. Data cleaning on the digitised data set1 resulted in a further
removal of 62 entries and data set 2 with 343 cases was realised. This dropped the
response rate to 34.3%. Data set 2 was further analysed in order to get a summary of
useful collected data, the results of which are given in this section. Table 5.6a give a
summary collected survey data from the secondary sampling units (districts) point of
view.
From Table 5.6a it can be seen that the lowest response rate recoded is 8% from
Mbeya City Council while the highest response rate was 62% from Sengerema
District Council. Ilemela District Council had the second lowest response rate of
15% followed by Kinondoni with 26.67%. Temeke District Council had 33%, Meru
District Council had 34% response rate, Ilala 39.6%. Arusha City Council recoded a
185
response rate of 36.5% while Nyamagana District Council recoded the second
highest response rate at 48.75%.
Table 5.6a: Summary of Survey Data from Secondary Sampling Units
District Sample Size Returned Questionnaires Response Rate %

Used Rejected Total
Arusha 200 73 50 123 36.5%
Meru 50 17 27 44 34%
Ilala 250 99 24 123 39.6%
Kinondoni 150 40 45 85 26.67%
Temeke 100 33 33 66 33%
Mbeya 100 8 11 19 8%
Nyamagana 80 39 43 82 48.75%
Ilemela 20 3 8 11 15%
Sengerema 50 31 15 46 62%
Total 1000 343 256 599 34.3%
Table 5.6b give a summary of the primary (Regions) sampling units. From table 5.6b
it can be seen that the highest response rate 48.67% was recorded in Mwanza
Region. The lowest response rate of 8% was recorded in Mbeya. Arusha region had
a response rate of 36 % while Dar es Salaam had a response rate of 34.4%. The
overall survey response rate of 31.3% was recorded. This response rate is lower than
what was expected according to sample size calculations of section 5.9.1.3.
However, according to Kripanont (2007), a response rate of 30% and above of usable
questionnaire data is acceptable in research. Thus, the recoded 34.3% response rate
for this survey was considered adequate for further analyses.

186
Table 5.6b: Summary of Survey Data from Primary Sampling Units
Region Sample Size Returned Questionnaires Response Rate %

Used Rejected Total
Arusha 250 90 77 167 36%
Dar es Salaam 500 172 102 274 34.4%
Mbeya 100 8 11 19 8%
Mwanza 150 73 66 139 48.67%
Total 1000 343 256 599 34.3%
Moreover, in multivariate research, the recommended sample size should be several
times (preferably 10 times or more) as large as the number of variables in the study
(Kripanont, 2007). This research aimed at developing a parsimonious model that
could assist SMEs to successfully adopt ICTs and thus it was not supposed to have
more than 20 variables. In that case, a sample size of 343 cases was further justified.
However, on the other hand, the response rate error is acknowledged.
5.10 Data Analysis Methods
A number of statistical methods were used for data analysis. Each one of these
methods and techniques was meant to individually or collectively address the main
research objective as well as specific research objectives. The methods were
classified in two main groups. The first group was concerned with methods used in
the so called preliminary analysis (Pallant, 2005). These are the kind of analysis that
precedes almost any kind of major statistical analysis. The second group consisted
of the methods used in the development of the ICT adoption model for SMEs. A
summary of the methods for data analysis in this research is given in the next
sections.
187
5.10.1 Preliminary Data Analysis Methods
Preliminary data analysis involved testing for the reliability validity and internal
consistence of the measurements, as well as data characterisation using descriptive
analysis techniques. The Cronbach’s coefficient alpha, the item-to-total correlation
and the inter-item correlation statistics were used to test and inform on the inter-item
consistence reliability and convergent validity of the measurements. Descriptive
analysis of the data included statistics such as frequency, minimum, maximum,
mean, standard deviation, percentage, Skewness, Kurtosis, and correlations.
Pallant (2005) observe that, descriptive statistics have a number of benefits
including: the ability to describe the characteristics of the sample; checking variables
for any violation of the assumptions underlying any of the statistical techniques used
and addressing specific objective. In this research, specific objective number three
was answered by descriptive analysis of the data while specific objective number
four was fulfilled by both descriptive analysis and Structure Equation Modelling
(SEM). The results from tests using methods falling under this category are given in
chapter six.
5.10.2 Structural Equation Modelling
The second group of data analysis methods and techniques involve the methods used
for the development of an ICT adoption model for Tanzania SMEs. This was the
188
main objective of this research. Structure Equation Modelling (SEM) method was
considered to be suitable and thus was chosen for this task.
SEM is a collection of statistical techniques that allow a set of relationships between
one or more Independent Variables (IVs), either continuous or discrete, and one or
more Dependent Variables (DVs), either continuous or discrete, to be examined
(Tabachnick and Fidell, 2013). Both IVs and DVs can be either factors or measured
variables (Tabachnick and Fidell, 2013). Structural equation modelling is also
referred to as causal modelling, causal analysis, simultaneous equation modelling,
analysis of covariance structures, path analysis, or confirmatory factor analysis (Hair
et al., 2010; Kripanont, 2007; Schumacher and Lomax 2010; Tabachnick and Fidell
2013). Kripanont (2007) note that, SEM is becoming a method of choice for
studying dependency relations in multivariate data in behavioural and social
sciences.
SEM is characterised by two components: The structural equation model (also
known as the path model) and the measurement model (Hair et al., 2010). A
structural equation model or path model depicts the structural relationships among
constructs (Hair et al., 2010; Kripanont, 2007). In other words, it relates the
independents to dependent variables (Hair et al., 2010). The measurement model on
the other hand enables the researcher to use several variables (indicators) for a single
IV or a single DV.
189
The choice of SEM as a method for developing the model in this research was
motivated by the advantages SEM has over other techniques. Kripanont (2007)
observe that, when compared to other multivariate techniques, SEM has four
significant benefits over those techniques as follows:
i. SEM takes a confirmatory approach rather than an exploratory approach to
the data analysis, although SEM can also address the latter approach. SEM
lends itself well to the analysis of data for the purposes of inferential
statistics. On the other hand, most other multivariate techniques are
essentially descriptive by nature (e.g. exploratory factor analysis) so that
hypothesis testing is possible but is rather difficult to do.
ii. SEM can provide explicit estimates of error variance parameters, but
traditional multivariate techniques are not capable of either assessing or
correcting for measurement error.
iii. Data analysis using SEM procedures can incorporate both unobserved and
observed variables, but the former data analysis methods are based on
observed measurements only.
iv. SEM methodology has many important features including modelling
multivariate relations, and for estimating point and/or interval indirect effects
whilst there are no widely and easily applied alternative methods for these
kinds of features.
190
Kripanont (2007) identifies three important general strategic frameworks for testing
structural equation models. Strictly confirmatory (SC), Alternative model (AM) and
Model generating (MG) frameworks. This research adopted the Model Generation
framework. According to Kripanont (2007), MG is the most common of the three
scenarios because the researcher could postulate and reject a theoretically derived
model on the basis of its poor fit to the sample data, and could proceed in an
exploratory (rather than confirmatory) fashion to modify and re-estimate the model.
Kripanont (2007), further observes that the primary focus is to locate the source of
misfit in the model and to determine a model that better describes the sample data.
Detailed account of the SC and AM frameworks is found in the literature (Arbuckle
2009; Kripanont, 2007).
According to the literature (Byrne, 2010; Kripanont, 2007), within the MG
framework, there are two approaches for doing SEM analyses: the two-step approach
and the single step approach. In a two-step approach, the measurement models are
first evaluated to ensure that the items used to measure each of the constructs is
adequate (Kripanont, 2007), before the second step which involves the assessment of
the structural model is made. In the single step SEM analysis, measurement and
structural models are simultaneously assessed (Byrne, 2010; Kripanont, 2007).
The literature (Byrne, 2010; Kripanont, 2007) highly recommend the use of the two-
step approach as it overcome the typical problem of not being able to localise the
source of poor model fit associated with the single-step approach (Byrne, 2010;
191
Kripanont, 2007). When a two-step approach is used, the second step is carried out
only after the measurement models have proven to be proper measures of the
constructs (Byrne, 2010; Kripanont, 2007). This research used, IBM SPSS version
21 for preliminary analysis (chapter six) while as IBM AMOS version 20 was used
to generate ICT adoption model for Tanzania SMEs in chapter seven.
5.10.2.1 Structural Equation Modelling Goodness-of-Fit Statistics
Analysis of structural model requires a thorough understanding of the measures of fit
or the goodness-of-fit Statistics. However, Arbuckle (2009) claim that, “model
evaluation is one of the most difficult and unsettled issues related to structural
equation modelling”. The author of this work notes that, there are many statistics
that are used in evaluation of structural equation models (Arbuckle, 2009; Byrne
2010; Hair et al., 2010; Kripanont, 2007; Tabachnick and Fidell, 2013). These
measures of fit are grouped into different types and each type has its specific
capability in model evaluation (Arbuckle, 2009; Byrne 2010Kripanont, 2007).
According to the literature (Arbuckle, 2009; Byrne 2010; Kripanont, 2007) the major
groups include, measures of parsimony, minimum sample discrepancy function,
measures based on the population discrepancy, comparison to a baseline model, and
a goodness of fit index (GFI) and related measures (Arbuckle, 2009; Byrne 2010;
Kripanont, 2007). Detailed account of these measures is found in the literature such
as Arbuckle (2005- 2009) and Byrne (2010). In Arbuckle (2009) the formulae
related to each fit statistic can be found (Arbuckle, 2009; Byrne 2010). This research
adapts the summary of fit measures by Kripanont (2007) since detailed account of all
goodness-of-fit measures related to structural equation modelling is beyond the scope
of this work. Kripanont (2007) claim that, among the many measures of fit, five
192
popular measures are: Chi-square , the normed chi-square ( /df), Goodness of Fit
Index GFI, TLI, and RMSEA. The meaning of this observation is that, a researcher
is not obliged to use all known fit measures in a study (Arbuckle, 2009; Byrne 2010;
Kripanont, 2007). A brief account of fit measures used in this research and their
groups is given in the next sub sections.
Group 1: Measures of Parsimony
According to Arbuckle (2009) structural equation models with relatively few
parameters (and relatively many degrees of freedom) are sometimes said to be high
in parsimony, while as models with many parameters (and few degrees of freedom)
are said to be lacking in parsimony. In this context, a model high in parsimony is
simple model (easy to understand). On the other hand, a model lacking in
parsimony is a complex model. Kripanont (2007) note that, “many fit measures
represent an attempt to balance these two conflicting objectives - simplicity and
goodness of fit”. Degree of Freedom (DF) is one of the fit measures used in
measures of parsimony (Arbuckle, 2009; Kripanont, 2007) and is used in this
research.
Group2: Minimum Sample Discrepancy Function
Minimum Discrepancy abbreviated as CMIN (Byrne 2010), probability value P, the
ratio CMIN/DF and the minimum discrepancy function abbreviated as FMIN are fit
measures based on the minimum value of the discrepancy (Arbuckle, 2009).
According to Kripanont (2007), CMIN (Chi-square statistic ) is the minimum
value of the discrepancy. In the case of maximum likelihood estimation, CMIN

193
contains the chi-square statistic (Kripanont, 2007; Byrne (2010). The chi-square
statistic is an overall measure of how many of the implied moments and sample
moments differ (Kripanont, 2007; Byrne (2010). The more the implied and sample
moments differ, the bigger the chi-square statistic, and the stronger the evidence
against the null hypothesis (Kripanont, 2007; Byrne; 2010).
The P value is the probability of getting as large a discrepancy as occurred with the
present sample under appropriate distributional assumptions and assuming a
correctly specified model (Arbuckle, 2005-2009; Kripanont, 2007). That is to say, P
is a “p value” for testing the hypothesis that the model fits perfectly in the population
(Arbuckle, 2005-2009; Kripanont, 2007). Therefore, this method can be used to
select the model by testing the hypothesis to eliminate any models that are not
consistent with the available data (Kripanont, 2007). According to Kripanont (2007),
a p value greater than 0.05 indicates an acceptable fit.
The CMIN/DF ( / df) is the minimum discrepancy divided by its degrees of
freedom. For correct models, the ratio should be close to 1 (Arbuckle, 2005-2009;
Byrne 2010; Kripanont, 2007). According to Arbuckle (2009), several writers have
suggested the use of this ratio as a measure of fit. However, Arbuckle (2005-2009)
claimed that, it is not clear how far from 1 the ratio should get before concluding that
a model is unsatisfactory. Arbuckle (2009) note that, different researchers have
recommended using ( /df) ratios as low as 2 or as high as 5 to indicate a reasonable

194
fit. Kripanont (2007) recommended the value of 3 which is the one used in this
research.
The literature (Arbuckle, 2009; Byrne 2010; Hair et al., 2010; Kripanont, 2007;
Pallant, 2005; Schumacker and Lomax 2010; Sophonthummapharn, 2008;
Tabachnick and Fidell, 2013) have acknowledged that, the chi-square ( ) statistic
sensitivity to sample size have led to problems of fit that are now widely known
(Byrne; 2010). It was thus necessary for researchers to look at others that also
support goodness of fit. The fit measures covered in the next sections tries to address
the problems associated with the statistics (Byrne; 2010).
Group 3: Measures Based on the Population Discrepancy
Kripanont (2007) note that, under this category, the Root Mean Square Error of
Approximation (RMSEA) is most commonly used fit measure. According to
Arbuckle (2009), when RMSEA is used as a model fit measure the rule of thumb is
that, a value about 0.05 or less indicates a close fit of the model. A value of 0.0
indicates the exact fit of the model (Arbuckle, 2009), while a value of about 0.08 or
less indicates a reasonable error of approximation (Arbuckle, 2009). In addition,
RMSEA value should not be greater than 0.1 (Arbuckle, 2009)
Group 4: Comparison to a Baseline Model
According to Kripanont (2007) there are three significant indices in this group.
Normed Fit Index (NFI), Tucker-Lewis coefficient (TLI), and the Comparative Fit
195
Index (CFI). The rule of thumb when using this goodness of fit statistics is that, the
NFI values range between 0 and 1 with 1 indicating a perfect fit (Arbuckle, 2009).
For the CFI, its value is truncated to fall in the range from 0 to 1 and CFI values
close to 1 indicate a very good fit (Arbuckle, 2009). TLI values lies in the range
between 0 and 1. A TLI values close to 1indicate a very good fit (Arbuckle, 2009).
Group5: GFI and Related Measures
In this group, Kripanont (2007) suggests the use of the Goodness of Fit Index (GFI)
and the Adjusted Goodness of Fit Index (AGFI). The two measures are associated
with Maximum likelihood (ML) and Unweighted Least Squares (ULS) estimation
(Arbuckle, 2009; Kripanont, 2007; Byrne 2010). Byrne (2010) claim that, the GFI
and AGFI can be classified as absolute indices of fit because they basically compare
the hypothesized model with no model at all (Byrne, 2010). Both indices have
values ranging from zero to 1.00, with values close to 1.00 being indicative of good
fit (Arbuckle, 2009; Byrne 2010; Kripanont, 2007). Table 5.7 show the summary of
these fit measures used in this research.
Table 5.7: Summary of SEM Fit Measures Used in this Research
FIT MEASURES BASED ON MINIMUM DISCREPANCY FUNCTION

(CMIN)
Fit Measure Fit Measures’ Indications (Interpretation)

A p value greater than 0.05 indicates an acceptable
CMIN (chi-square) fit.
A value close to 1 and not exceeding 3 indicates a
CMIN/DF ( /df) good fit. A value less than 1 indicates an over fit of
(normed chi-square) the model.
196
FIT MEASURES BASED ON THE POPULATION DISCREPANCY

RMSEA = Root Mean A value about 0.05 or less indicates a close fit of the
Square Error of model.
Approximation A value of 0.0 indicates the exact fit of the model.
A value of about 0.08 or less indicates a reasonable
error of approximation.
A value should not greater than 0.1.
Source: Literature (Arbuckle, 2009; Byrne 2010; Kripanont, 2007).
FIT MEASURES BASED ON GOODNESS OF FIT INDEX

GFI=Goodness of Fit A value always less than or equal to 1 and 1
Index indicates a perfect fit.
AGFI= Adjusted A value is bounded above by 1 and is not bounded
Goodness of Fit Index by 0 and 1 indicates a perfect fit.
FIT MEASRES BASED ON BASELINE COMPARISONS

TLI= Tucker-Lewis A value between 0 and 1, but is not limited to this
Coefficient range; a value close to 1 indicates a very good fit.
A value greater than 1 indicates an over fit of the
model.
CFI= Comparative Fit A value between 0 and 1, a value close to 1 indicates
Index a very good fit.
NFI = Normed Fit Index A value between 0 and 1. 1 indicates a perfect fit.
Source: Literature (Arbuckle, 2009; Byrne 2010; Kripanont, 2007).

197
5.10.2.2 Multiple Group Analysis
One of the important things this research wanted to establish was the effect of
moderators on the influence of predictors towards the dependent variables. To
achieve this, it was necessary to perform multiple-group analysis using the Analysis
of Moment Structures (AMOS) software (Arbuckle 2009; Kripanont, 2007).
Arbuckle (2009) and Kripanont (2007) provide detailed explanations of how AMOS
performs different tests during modelling and how to interpret the output from these
tests. However, at this point, it is important to note that several multivariate
assumptions must be met before SEM using AMOS can take place. For example,
most multivariate analysis assumes normal distribution of the sample and thus it is
necessary to test for normality of data before analysing the data. Under non-
normality situations, a researcher may be justified to use the powerful bootstrapping
procedures in order to get valid results (Arbuckle, 1995-2009; Kripanont, 2007). The
next sections address the critical issues to be considered before and during
multivariate data analysis using the methods that has been discussed so far.
5.10.3 Sample Size Requirements
According to the literature (Graham, 2012; Hair et al., 2010; Kripanont, 2007; Pigott,
2001; Tabachnick and Fidell 2013) different statistical analyses have different
minimum sample size requirements. Table 5.8 give a summary of sample size
requirements for SEM which was the major analyses carried out in this research.
Note that the sample size for this research was 343 cases and thus from the
information of table 5.8 this basic requirement was met.

198
Table 5.8: Sample Size Requirements for Different Statistical Analysis
Statistical Analysis Minimum Sample Size

SEM Sample size as small as 50 found to provide valid
results (Hair et al., 2010).
Recommended minimum sample sizes of 100-150 to
ensure stable Maximum likelihood estimation (MLE)
solution (Hair et al., 2010).
Sample sizes in a range of 150-400 are suggested
(Hair et al., 2010).
Source: Literature (Arbuckle, 2009; Hair et al., 2010; Kripanont, 2007).
5.11 Considerations for Multivariate Data Analysis
The goal of any analysis is to obtain unbiased estimates of population parameters. It
is therefore very important to examine the data, before data analysis commence. The
first step in this respect include checking the data file for errors as suggested in the
literature (Kripanont, 2007; Pallant, 2005) and applying corrective measures to
remove the errors observed. The second step involves the missing data analysis and
the imputation of the missing data. Thereafter, preliminary analyses can be carried
out with the aim of testing for the assumptions underlying the statistical bases for
multivariate analysis. Detailed explanations about these tests are given in the next
sections.
5.11.1 Missing Data
The literatures show that, at some point in their work, all researchers face the
problem of missing quantitative data (Kripanont, 2007; Pigott, 2001; Tabachnick and
Fidell 2013). Pigott (2001) note that, the reasons for missing data are many and
varied. Some of the most common reason for missing data include a case whereby
199
the research respondent refuses or forget to answer a survey question, lost files, or
data are not recorded properly (Pigott, 2001). According to Hair et al. (2010) and
Kripanont (2007), the general impact of missing data (particularly in survey
research) in multivariate analysis is twofold. First, the missing data may result in the
reduction of the sample size available for analysis from an adequate sample to an
inadequate sample if the remedies for missing data are not applied. Second, any
statistical results based on data with a non-random missing data process could be
biased if the missing data lead to erroneous results (Hair et al., 2006; Kripanont,
2007). In addition, Arbuckle (2005) note that, in some cases, fitting some models
with data that has missing values requires extensive computation.
Four imputation methods for dealing with missing data are identified in the literature
(Graham, 2012; Hair et al., 2010; Kripanont, 2007; Pigott, 2001; Tabachnick and
Fidell 2013). Hair et al. (2010) define imputation as the process of estimating the
missing value based on valid values of other variables and or cases in the sample.
The four imputation methods are:
i. Imputation method based on complete cases
ii. Imputation method based on pairwise deletion
iii. Imputation by calculating replacement values based on mean substitution and
regression imputation
200
iv. Model based methods
Hair et al. (2006; 2010) cited by Kripanont (2007) note that, the identification of an
appropriate method for dealing with missing data is guided by a four steps process.
The first step is to determine the type of missing data, the second step is to determine
the extent of missing data, analyse cases and variables, the third step is to diagnose
the randomness of the missing data processes and the fourth step is to select the
appropriate imputation method.
According to the literature (Graham, 2012; Hair et al., 2010; Kripanont, 2007; Pigott,
2001; Tabachnick and Fidell 2013), the application of these methods on a particular
case depends on two things: the nature of missing data and the amount of missing
values (Graham, 2012; Hair et al., 2010; Kripanont, 2007; Pigott, 2001; Tabachnick
and Fidell 2013). Tabachnick and Fidell (2013) further observe that, in missing data
situations, the pattern of missing data is more important than the amount missing.
Tabachnick and Fidell (2013) claim that, missing values scattered randomly through
a data matrix pose less serious problems than non-randomly missing values. On the
other hand, they argue that, non-randomly missing values, no matter how few of
them there are, they pose a serious problem because they affect the generalisability
of results (Tabachnick and Fidell, 2013). Missing data are technically characterised
as Missing Completely At Random (MCAR), Missing At Random (MAR), and
(MNAR) which stands for Missing Not At Random (Graham, 2012; Hair et al.,
2010; Kripanont, 2007; Pigott, 2001; Tabachnick and Fidell 2013). Missing data
201
that is MAR can be ignored while as missing data that is MNAR is considered as
non-ignorable (Tabachnick and Fidell 2013).
The rules for selecting imputation methods (Hair et al., 2010; Kripanont, 2007) are
such that:
i. When the missing data is under 10%, any of the imputation methods can be
applied except the complete case method which is the least preferred.
ii. If the missing data is 10% to 20%, the case substitution, and regression are
the most preferable methods if the missing data is MCAR. However, if the
missing data is MAR, then, model based methods is preferred most.
iii. When the missing data is over 20%, and it is considered that imputation is
necessary, the preferred methods are the regression method for MCAR
missing data and the model-based methods when the missing data is MAR
In this research, missing data analysis was done using SPSS Version 21. SPSS
Version 21 has four techniques for determining the randomness of the missing data
and all of them were used in this research. The techniques are: List wise; Pairwise;
Expectation Maximisation (EM) and Regression. Detailed explanations about these
techniques are found in the literature (Hair et al., 2010; Kripanont, 2007; Tabachnick
and Fidell 2013).

202
5.11.2 Multivariate Outliers
Tabachnick and Fidell (2013) and Pallant (2005) claim that, many statistical methods
are sensitive to outliers. Outliers distort statistics, and thus it is important to identify
outliers and make decisions about what to do with them (Hair et al., 2010;
Tabachnick and Fidell, 2013). A multivariate outlier is an extreme score on one or
more variables (Kripanont, 2007). According to Tabachnick and Fidell (2013) there
are four reasons why outliers may be present in data:
i. Incorrect data entry
ii. Failure to specify missing values in the computer syntax so missing values
are read as real data.
iii. Outlier is not member of population that you intended to sample.
iv. Outlier is representative of population you intended to sample but population
has more extreme scores than a normal distribution.
Tabachnick and Fidell (2013) note that, errors in data entry and missing values
specification can easily be found and remedied. Moreover, if the outlier is due to a
case that should not have been sampled, it can be deleted once detected (Hair et al.,
2010; Tabachnick and Fidell, 2013). However, deciding what to do with outliers
that are representative of population you intended to sample (beneficial outliers) is
difficult (Tabachnick and Fidell, 2013). To retain a beneficial outlier, the researcher
203
must alter the values on the variables so that impact of the case is reduced
(Tabachnick and Fidell, 2013).
The literature (Hair et al., 2010; Kripanont, 2007; Tabachnick and Fidell, 2013)
identify, Mahalanobis distance, leverage, discrepancy, and influence, as statistical
measures used to identify multivariate outliers. In this research, Mahalanobis
distance and the ratio were used to test the data for the presence of multivariate
outliers. Mahalanobis distance is the distance of a case from the centroid of the
remaining cases where the centroid is the point created at the intersection of the
means of all the variables (Tabachnick and Fidell, 2013).
According to the literature (Tabachnick and Fidell, 2013; Peou Chea, 2009),
Mahalanobis distance is evaluated as chi-square ( ) with degrees of freedom ( )
equal to the number of variables, at . A variable is considered a potential
outlier if the maximum Mahalanobis distance exceed the critical chi-squared value
with degrees of freedom equal to number of variables and α = 0.001 (Tabachnick
and Fidell, 2013; Peou, 2009). Alternatively, the ratio can be used to detect
the presence of outliers in data (Hair et al., 2010; Kripanont, 2007). Hair et al.
(2010) show that, at conservative level of significance α = 0.005 or 0.001 (depending
on the nature of statistical test) observations having a value exceeding 2.5 in

204
small samples (80 or fewer observations) and 3 or 4 in large samples can be
considered as possible outliers (Hair et al., 2010; Kripanont, 2007).
Once detected, outliers must be analysed in order to determine their usefulness in the
analysis (Hair et al., 2010). If outliers are found to be useful, they must be retained.
However, if outliers are found not useful, they must be deleted (Hair et al., 2010).
Hair et al. (2010) observe that, although many philosophers offer guidance of how to
deal with outliers, their recommendation is that the outliers be retained if they
represent observation from the intended population (Hair et al., 2010; Peou, 2009;
Tabachnick and Fidell, 2013). It should be noted that, while as deleting outliers
improve the analysis; on the other hand, deleting outliers reduces the sample size and
compromise the generalisability of results (Hair et al., 2010; Tabachnick and Fidell,
2013). Detailed information about the test is presented in chapter six.
5.11.3 Multivariate Normality
In statistics, the normal distribution is regarded as a benchmark for most statistical
methods (Hair et al., 2010). Many statistical techniques assume that the distribution
of scores on the dependent variable is normal. Thus, it is very important to test data
for normality before applying statistical analysis methods. Kripanont (2007) and
Hair et al. (2010) note that, although some robust statistical techniques are less
affected when violating certain assumptions, but in all cases complying with some of
the assumptions critically determines a successful analysis (Hair et al., 2010). On a
similar note, Tabachnick and Fidell (2013) claim that, screening continuous variables
for normality is an important early step in almost every multivariate analysis,
particularly when inference is a goal. They further argue that, although normality of
205
the variables is not always required for analysis, the solution is usually quite a bit
better if the variables are all normally distributed (Tabachnick and Fidell, 2013).
Hair et al. (2010) claim that, multivariate normality is the most fundamental
assumption in multivariate analysis. Multivariate normality refers to the shape of the
data distribution for an individual metric variable and its correspondence to the
normal distribution (Hair et al., 2010). The literature (Hair et al., 2010; Tabachnick
and Fidell, 2013) identify graphical and statistical as methods of assessing normality.
In addition, Skewness and Kurtosis are identified as the two measures used to
evaluate normality (Hair et al., 2010; Tabachnick and Fidell, 2013). Kurtosis refers
to the peakedness or flatness of the distribution compared to the normal distribution.
Skewness on the other hand explains the symmetry of the distribution (Hair et al.,
2010; Tabachnick and Fidell, 2013).
When the data under consideration is normally distributed, the values of skewness
and kurtosis are zero (Hair et al., 2010; Kripanont, 2007; Pallant, 2005; Tabachnick
and Fidell, 2013). If, however a positive skewness value is observed, it indicates a
left skewed distribution whereas a negative skewness value indicates a right skewed
distribution. For the Kurtosis measure, a positive value shows a relatively peaked
distribution while a negative value shows a relatively flat distribution (Hair et al.,
2010; Kripanont, 2007; Tabachnick and Fidell, 2013).
The literature (Hair et al., 2010; Kripanont, 2007; Tabachnick and Fidell, 2013)
show that, skewness values falling outside the range of -1 to +1 indicate a
substantially skewed distribution, while the closer the value is to zero the greater the
206
variables’ normality. For the Kurtosis measure, values less than 1 are negligible;
values from one to ten indicate moderate non-normality while values greater than 10
indicate severe non-normality (Hair et al., 2010; Kripanont, 2007; Tabachnick and
Fidell, 2013)
Byrne (2010) observes that, although sample data may be univariate normal (or
moderately univariate non-normal), the multivariate distribution of the data can still
be multivariate non-normal. In addition, Byrne (2010) note that, of the two measures
for assessing normality of data, that is Skewness and Kurtosis; it is the multivariate
Kurtosis statistic that indicates the extent of departure from multivariate normality.
Thus, the Critical Ratio (C.R) value for the Kurtosis measure is also used to inform
on the multivariate normality of data under test (Arbuckle, 2009; Byrne, 2010).
According to Arbuckle (2009) and Byrne (2010), C.R values above 5 are an
indication that the data is not multivariate normally distributed.
The literature (Arbuckle, 2009; Byrne, 2010), acknowledge the sensitivity of SEM
analyses on non-normal data and in particular the multivariate Kurtosis (Byrne,
2010). According to Byrne (2010), when the data to be analysed is multivariate
kurtotic, interpretations based on the Maximum Likelihood (ML) estimation may be
problematic. ML is the default AMOS estimation method (Arbuckle, 2009; Byrne,
2010; Kripanont, 2007). In such scenarios, the use of bootstrapping methods or the
use of other estimation methods may be justified (Arbuckle, 2009; Byrne, 2010;
Kripanont, 2007). In this work, tests and results about the multivariate normality of
the research data are presented in chapter six.

207
5.11.4 Multicolinearity and Singularity
Some multivariate techniques such as Multivariate ANalysis Of VAriance
(MANOVA) work effectively when the dependent variables are only moderately
correlated (Kripanont, 2007). When two or more explanatory variables are highly
correlated, they lead to a problem technically known as Multicolinearity (Kripanont,
2007; Sophonthummapharn, 2008). When the correlation equals 1, it indicates the
presence of singularity (Tabachnick and Fidell, 2013).
According to the literature (Kripanont, 2007; Pallant, 2005; Sophonthummapharn,
2008; Tabachnick and Fidell, 2013), multicolinearity and singularity cause both
logical and statistical problems. For example, Sophonthummapharn (2008) note that,
multicolinearity leads to difficulty in assessing the effect of the independent variables
on the dependent variable. Thus, it is recommended to test data for multicolinearity
before some statistical analyses such as MANOVA are performed (Kripanont, 2007;
Tabachnick and Fidell, 2013).
In a similar note, Kripanont (2007) recommends testing the data for multicolinearity
before performing discriminant validity analysis. When detected, the literature
(Kripanont, 2007; Pallant, 2005; Sophonthummapharn, 2008; Tabachnick and Fidell,
2013) recommends several methods for dealing with multicollinear variables.
According to Tabachnick and Fidell (2013), when variables are multicollinear or
singular, they contain redundant information and they are not all needed in the same
208
analysis. This means some of the multicollinear variables can be deleted from the
analysis (Kripanont, 2007; Pallant, 2005; Tabachnick and Fidell, 2013).
Tabachnick and Fidell (2013) recommend other ways of dealing with
multicolinearity. They suggest that, if the only goal of analysis is prediction then
multicolinearity can be ignored (Tabachnick and Fidell, 2013). Another option is, to
combine the multicollinear variables and form a composite variable if the variables
are measuring the same construct (Tabachnick and Fidell, 2013). In addition, the
researcher can compute principal components and use the components as the
predictors instead of the original variables (Tabachnick and Fidell, 2013). A final
alternative is for the researcher to centre one or more of the variables, if
multicolinearity is caused by forming interactions or powers of continuous variables
(Tabachnick and Fidell, 2013).
The literature (Kripanont, 2007; Pallant, 2005; Sophonthummapharn, 2008;
Tabachnick and Fidell, 2013) shows that, multicolinearity can be assessed by looking
at the sample correlation. If the value of sample correlation between any two or
more indicators exceeds 0.80, then the variables are said to be multicollinear
(Kripanont, 2007; Tabachnick and Fidell, 2013). Tabachnick and Fidell (2013) note
that, in some cases, correlations from 0.7 and above between variables can be a
matter of concern. In this research data was tested for multicolinearity. Details of
this test are provided in chapter seven, prior to model development.

209
5.12 Chapter Summary
This chapter has presented the research design and methodology used in this study.
All critical issues that have impact on the outcome of the research have been
discussed and explanations given as to what motivated the decisions made. After
determining the type of research and the philosophical standing of the researcher,
appropriate research approaches, strategies and methods were chosen. Moreover,
details of the research instrument development and testing have been provided. Data
collection procedures and the challenges faced during data collection together with
the impact of the challenges on the research outcome have also been explained.
Finally, data analysis methods and their considerations were made. Results of this
research are provided in chapter six and chapter seven.
CHAPTER SIX
PRELIMINARY DATA ANALYSIS AND FINDINGS
6.1 Introduction
This chapter provides results of some of the analyses carried out in this research.
Analyses commence with the so called preliminary data analyses. These are the
analyses that are (almost always/ mandatory) carried out first before other types of
data analyses can take place. For example, in any research, it is always necessary to
perform data cleaning and missing data analysis before other types of analyses can be
done. Preliminary data analyses performed in this research include:

210
i. Missing data analysis
ii. Testing data for multivariate outliers
iii. Testing the reliability of the research instrument
iv. Testing the convergent validity of the constructs.
v. Testing for the multivariate normality of data
vi. Descriptive analyses
At this point it should be noted that, while a lot of analyses were possible with the
obtained survey data, the main objective was to develop a model of ICT adoption and
use by Tanzania SMEs. That being the case, only the analyses that complimented
model development or the ones that addressed fully or partially specific objectives of
the research were carried out. Two specific objectives are addressed in this chapter.
Specific objectives number three is fully addressed here while specific objective
number four is partially addressed. The two specific objectives were respectively
stated as:
i. To establish the status of ICT adoption and use by Tanzania SMEs
ii. To establish the influence of ICT adoption on Tanzania SMEs performance

211
Based on the observation made above, in addition to the preliminary analyses,
descriptive statistics presented in this chapter include the demographic information
of the sample, statistics on ICTs adoption in SMEs, statistics on post adoption use of
ICTs by the SMEs, and the statistics on the influence of ICT adoption on SMEs
performance.
6.2 Missing Data Analysis
Missing data analysis was performed using SPSS version 21. The results for this
analysis show that the missing data were less than 5% for all variables involved (See
Appendix IB). Moreover, using Little’s MCAR test, the pattern of missing data was
found not to be MCAR. (Chi-square =1774.966, DF=1478, Sig=0.000). The low
percentage of missing data and the fact that the missing data was not MCAR justified
the use of multiple imputation method for dealing with missing data. Multiple
imputations method is a model based method of imputation (See Section 5.11.1).
After imputation, a complete data set with 343 cases was realised. Prior to other
analyses, the values of all negatively stated variables (variables F1-F16 and G1-G9)
were transformed in order to ensure that all variables in the data had the same
positive direction (Peou, 2009). After this transformation, other analyses continued.
6.3 Testing Data for Multivariate Outliers
In this research, Mahalanobis distance was used to test the data for the presence of
multivariate outliers, as explained in section 5.11.2. Using SPSS version 21

212
software, all variables in the survey (89 variables) data with a sample size of 343
cases were tested for outliers. In SPSS, Mahalanobis distance values are obtained as
residuals statistics in linear regression analysis. The results of this test are presented
in Table 6.1. Note that, in this regression analysis, for convenience, questionnaire
IDs was used as the dummy Dependent Variable (DV) because multivariate outliers
among Independent Variables (IVs) are unaffected by the DV (Tabachnick and
Fidell, 2013). According to Tabachnick and Fidell (2013), the remaining variables
are considered independent variables.
From Table 6.1 it can be seen that the minimum Mahalanobis distance value is
11.759 and the maximum Mahalanobis distance value is 247.350. The critical Chi-
square value (Obtained from Chi square distribution Tables) given the degrees of
freedom equal to 89 (the number of variables in this analysis), at is
found to be = 135.978. Thus, all variables in the data with Mahalanobis distance
greater than 135.978 were considered to be potential outliers. Based on this
criterion, in this analysis it was discovered that 68 cases had variables that were
possible outliers.
Table 6.1: Mahalanobis Distance Test Results
Residuals Statisticsa
Minimum Maximum Mean Std. Deviation N
Predicted Value -41.14 414.65 202.06 100.536 343
Std. Predicted Value -2.419 2.115 .000 1.000 343
Standard Error of 14.153 62.446 35.615 11.195 343
Predicted Value
213
Adjusted Predicted -67.01 421.23 201.21 106.132 343

Value
Residual -309.261 176.781 .000 63.152 343
Std. Residual -4.220 2.412 .000 .862 343
Stud. Residual -4.638 2.716 .004 .984 343
Deleted Residual -373.573 232.608 .852 84.219 343
Stud. Deleted Residual -4.839 2.751 .003 .993 343
Mahal. Distance 11.759 247.350 87.743 53.738 343
Cook's Distance .000 .050 .004 .007 343
Centered Leverage .034 .723 .257 .157 343
Value
a. Dependent Variable: QUESTIONAIRE ID
Further analysis of the potential outliers revealed that, 2 cases did not meet the SME
criteria and thus were considered not part of the intended population. These were
outright deleted and the sample size was reduced to 341 cases. The remaining 66
observed potential outlier cases were part of the intended population and thus were
retained.
Based on the ratio criterion, there were no suggestion of the presence of outliers
in the data since = 247.350/89 =2.77 did not exceed the threshold of 3 or 4 for
large samples (Hair et al., 2010; Kripanont, 2007). Therefore, although, some cases
demonstrated the characteristics of outliers, they were not extreme cases according to
the value of which did not exceed the threshold value. It was thus not necessary
to delete the 66 observed cases from the sample as recommended by Pallant (2005).
In this way, Outliers in the data were detected analysed and retained or deleted as
214
advocated by the literature (Hair et al., 2010; Peou, 2009; Tabachnick and Fidell,
2013). Finally, a dataset with 341 cases remained for further analyses.
6.4 Reliability Analysis
Results for reliability analysis are given in Table 6.2. It can be observed from Table
6.2 that, all internal consistency reliabilities based on Cronbach’s alphas for
measurement items (all interval scales) were higher than 0.8 except in two cases
within the Government Policies scale where the Cronbach’s alphas were 0.786
(which is acceptable) for the Taxes and Tariffs sub-scale and 0.585 for the support
infrastructure sub-scale.
However, when the two sub-scales were combined the Cronbach’s alphas became
0.844 which was good. According to Kripanont (2007), higher values of Cronbach’s
alphas (0.7 and above) indicate the items in each set (concept) are positively
correlated to one another and that items in each set are independent measures of the
same concept. This in turn indicates accuracy in measurements (Kripanont, 2007).
Therefore, in this case, Table 6.2 results confirm the accuracy of the main survey
measurements for this research.
Table 6.2: Main Survey’s Reliability and Internal Consistence Analysis

Results
Measurement Item Number Cronbach Reliability Inter-item Item-to-

(Interval Scale) of Items ’s alpha Results correlation total
correlation
Adoption and Use
-Computer 4 0.872 Good 0.562-0.751 0.638-0.781
-Internet 8 0.957 Good 0.614-0.826 0.744-0.890
215
All 12 0.963 Good

Personal Factors
-Perceived Benefits 12 0.973 Good 0.585-0.879 0.709-0.905
Organisation Factors
Strategy
-Reason for adoption 5 0.842 Good 0.266-0.800 0.518-0.785
External Pressure 4 0.879 Good 0.495-0.798 0.623-0.806
All 9 0.917 Good

Economic Factors
- Financial Support 8 0.905 Good 0.331-0.848 0.639-0.753
-Costs of ICT systems 8 0.950 Good 0.533-0.851 0.686-0.871
All 16 0.954 Good

Government Policies
-Support Programs 4 0.906 Good 0.600-0.827 0.704-0.847
-Taxes and Tariffs 3 0.786 Accept. 0.505-0.647 0.561-0.666
-Support Infrastructure 2 0.585 Poor 0.416-0.416 0.416-0.416
All 5 0.844 Good 0.396-0.662 0.532-0.747

SME Performance
-Computer 5 0.932 Good 0.548-0.900 0.677-0.897
-Internet 5 0.944 Good 0.638-0.902 0.764-0.905
All 10 0.966 Good
For the internal consistence measures, this research used the inter-item-correlation
and the item-total-correlation. From Table 6.2, one item in the Organisation Factors
recoded a value of 0.226 for the inter-item-correlation, which is less than the
threshold of 0.3 (Kripanont, 2007). One variable E4 was removed from the scale and
the minimum inter-item-correlation for the strategy sub scale improved to 0.38. For
the item-total-correlation measure a value below the threshold of 0.5 was recorded
for the Government Policies sub-scale. Kripanont (2007) recommends that, the item-
total-correlation exceeds 0.5. However, when the sub-scales were combined into a
single scale, the minimum acceptable item-total-correlation was achieved. Thus, the
216
overall result for reliability test as presented in Table 6.2 prove that, the
questionnaire used in the main survey was a reliable measurement tool.
6.5 Convergent Validity Analysis
Convergent validity is determined by Correlational analysis (See section 5.8.6).
Table 6.2 shows the convergent analysis results for this study. From Table 6.2, only
one case has inter-item-correlation of less than the threshold value of 0.3. The
remaining cases have inter-item-correlation above 0.3. For the item-total-correlation,
only one case has a value below the threshold value of 0.5. The remaining cases
have high values than the threshold. These results ascertain the convergent validity
of the instrument used in this study. The two cases with low values were omitted in
model development.
6.6 Testing Data for Multivariate Normality
In this research, Skewness and Kurtosis were used to test survey data for normality.
The background details of the two measures were covered in section 5.11.3. In
section 5.11.3, it was explained that multivariate normality is the most fundamental
assumption in multivariate analysis. Thus, bearing this in mind and the sensitivity of
SEM analyses on the multivariate non-normality of data, it was necessary to test the
survey data for multivariate normality.
Appendix IIB Table 1 shows the results from measurement model 3 variables where
it is found that the variables have a Kurtosis value of 124.053 and Critical Ratio
217
value of 50.719. These values suggested that the survey data is highly multivariate
nonnormal. Appendix IIB Table 2 reports Kurtosis value of 71.106 and Critical
Ratio value of 33.245 for variables of the developed model. The non-normality of
data in this research justified the use of Bollen-Stine bootstrap method during model
development (see chapter seven).
6.7 Descriptive Analyses
6.7.1 Demographic Statistics
A summary of the demographic information of this survey is given in Table 6.3. It is
important to note that, this research focused on owner-managers of SMEs in the
service and general trading sectors and thus at this point, all information that was not
given by the owner-manager were already filtered out. Further analysis such as the
test for outliers helped to prune out all cases that did not fit the rest of research
criteria. From Table 6.3, Figure 6.1 and Figure 6.2 it can be observed that, the
surveyed firms fit Tanzania’s definition of SME (see section 1.2.5) by 100% both in
terms of the capital investment criteria and by the number of employees’ criteria.
Table 6.3: Demographic Information
Demographic Information Sample Size=341

Measurement Items Frequenc Percent
y
Business/Company Age (Number of years)
 1-5 101 29.62%
 6-10 114 33.43%
 11-20 97 28.45%
 21-30 27 7.92%
 31-45 2 0.58%
Business/Company Size (Number of Employees)
218
 1-4 130 38.12%

 5-49 200 58.65%
 50-99 11 3.23%
 100+ 0 0%
Business size (Capital Investment in Million TShs.)
 Up to 5 88 25.80%
 5 to 200 230 67.45%
 200 to 800 23 6.75%
Business Sector
 Service 206 60.41%
 General Trading 135 39.59%
Legal Form of Company
 Sole Proprietor 203 59.53%
 Partnership 95 27.86%
 Limited Liability 43 12.61%
Business Status
 Starting 32 9.38%
 Surviving 76 22.29%
 Growing 184 53.96%
 Maturing 44 12.90%
 Declining 5 1.47%
Respondents Time with Company (Years)
 1-4 93 27.27%
 5-10 133 39.00%
 11-20 96 28.15%
 21-30 17 4.99%
 Above 31 2 0.59%
Table 6.3: (Continued)
Demographic Information Sample Size=341

Measurement Items Frequenc Percent
y
Respondent Level of Education
 Postgraduate 25 7.33%
 Degree 132 38.71%
 Diploma 114 33.43%
 A-Level 37 10.85%
 Other 33 9.68%
Respondent Knowledge of IT
 Knowledgeable 159 46.63%
 Not Knowledgeable 182 53.37%
Annual Turnover
 Up to 12, 000,000 126 36.95%
 12,000,000 to 50, 000,000 129 37.83%
 50,000,000 to 100,000,000 53 15.54%
 100,000,000 to 500,000,000 29 8.51%
219
 500,000,000 to 1,000,000,000 4 1.17%
Figure 6.1: Company Size-Number of Employees

220
Figure 6.2: Company Size-Capital Investment
The surveyed SMEs were from the service sector 60.64% and general trading sector
39.36%. Most respondents were reasonably educated as 38.78% had university
degrees and 33.53% had college diplomas. The rest of the respondents were
distributed as: 7.29% had Postgraduate degrees, 10.78% had education up to
advanced level secondary school. The remaining 9.62% of the respondents were
below the A-level. In terms of respondent’s knowledge of IT, 46.65% of the
respondents claimed to be IT knowledgeable while the remaining 53.35%
acknowledged that they did not possess any knowledge of IT.
Looking at the age of the surveyed firms, it was observed that 29.45% had been in
business for a period of between 1 and 5 years (start-ups). 33.53% had been in
business for a period between 6 and 10 years, 28.57% between 11 and 20 years,
221
7.87% had been doing business for between 21 to 30 years, while 0.58% has been in
business for between 31 to 45 years. The rest of demographic information can be
obtained from Table 6.3
6.7.2 ICT Adoption
In this section and analysis of ICT adoptions in the surveyed firms is made. The
ICTs studied in this research are computers and the Internet, which include
connectivity here simply referred to as the Internet, the website and email
technologies. Appendix II-C gives a summary of adoption and frequency of use of
these technologies in this research. Further analysis is given the next sections.
6.7.2.1 Computer Adoption
From the survey data (Appendix II-C Table 1) it was observed that, 75.4% of the
SMEs have computers that are used on daily basis. 4.1% have computers that are
used on weekly basis and 0.6% of the surveyed firms had computers used at least
once in a month. Non-adopters of the computer technology included 18.2% of
respondents (62 cases) who know about the technology but have not adopted and
1.8% (6 respondents) who do not know about the technology at all. Figure 6.3 give a
pictorial overview of computer adoption and frequency of use in the surveyed SMEs.
Overall there were 273 (80.1%) adopters and 68 (19.9%) non-adopters of the
computer technology as depicted in Table 6.4. In addition, it was also observed that
34.6% of computer adopters had their computers networked as shown in Table 6.5.
222
Table 6.4: Summary of Computer Adoption
Frequency Percent Valid Percent Cumulative

Percent
Adopters 273 80.1 80.1 80.1
Valid Non-Adopters 68 19.9 19.9 100.0
Total 341 100.0 100.0
Table 6.5: Networked Computers (LAN)

Percent
Adopters 118 34.6 34.6 34.6
Total 341 100.0 100.0
Figure 6.3: Adoption and Frequency of Use of Computers

223
6.7.2.2 Internet Adoption
Table 2 of appendix II-C shows adoption and frequency of use of the internet. Out of
341firms, 224 (65.7%) were reported to have adopted and used the Internet on daily
basis, 9 firms (2.6%) had adopted and used the Internet on weekly basis while 2
respondents (0.6%) claimed to use the Internet at least once every month. 104 firms
which is equivalent to 30.5% were non-adopters of the Internet technology. 2
respondents (0.6%) claimed not to know the technology at all. Overall 235
respondents (68.9%) could be categorised as Internet adopters while 106 respondents
(31.1%) could be categorised as non-Internet adopters. Table 6.6 and Figure 6.4 give
a summary of Internet adoption in Tanzania SMEs.
Figure 6.4: Adoption and Frequency of Use of the Internet

224
Table 6.6: Summary of Internet Adoption

Percent
Adopters 235 68.9 68.9 68.9
Total 341 100.0 100.0
6.7.2.3 Website Adoption
Analysis results for website adoption are provided by Appendix IIC Table 4, Figure
6.5. and in Table 6.7. From appendix II-C Table 3, 181 firms which is equivalent to
53.1% have a company website that is used on a daily basis, 11 firms (3.2%) have a
website that is used at least once in a week and 1 firm (0.3%) that has a website that
is used at least once in a month. Non-adopters of website technology include 143
firms (49.5%) with owners-managers who know about the technology but
nonetheless have not adopted and 5 firms (1.5%) that has owners-managers that do
not know about the website technology. Overall, there were 193 firms (56.6%) that
have adopted the website technology and 148 (43.6%) non-adopting firms as
depicted in Table 6.7
Table 6.7: Summary of Website Adoption
Frequency Percent Valid Cumulative

Percent Percent
Adopters 193 56.6 56.6 56.6
Total 341 100.0 100.0
225
Figure 6.5: Adoption and Frequency of Use of Website
6.7.2.4 Email Adoption
Email adoption in Tanzania SMEs follow a similar pattern like the Internet and
computer technologies in that it appears there are more ICT adopters in Tanzania
SMEs than previously thought. The summary of Email adoption is given in
Appendix II-C Table 4, Figure 6.6 and Table 6.8. According to Appendix II-C Table
4, 237 (69.5%) firms out of 341 have Email that is used on daily basis, 9 firms
(2.6%) have Email that is used on weekly basis and 2 firms (0.6%) have Email that is
used at least once in a month. 88 firms (25.8%) have owners-managers that are
aware of the Email technology although they have not adopted the technology and
5firms (1.5%) have owners-managers that are not even aware of the technology.
From Table 6.6 it can be seen that overall there are a total of 248 (72.7%) firms that
226
have adopted the Email technology while 93 firms (27.3%) can be categorised as
Email non-adopting firms. Figure 6.4 give a summary of adoption and frequency of
use of the Email technology in Tanzania SMEs.
Table 6.8: Summary of Email Adoption

Percent
Adopters 248 72.7 72.7 72.7
Total 341 100.0 100.0
Figure 6.6: Adoption and Frequency of Use of Email

227
6.7.3 ICT use by Tanzania SMEs
In this section, the analysis of how the surveyed firms use adopted ICT is made. In
particular, the analysis seeks to investigate and explain if Tanzania SMEs use ICT to
gain competitive advantage over their competitors. The rationale for this type of
analysis is based on the literature (Galloway and Mochrie, 2005; Harindranath et al.,
2008; Modimogale, 2009) which has shown that, adoption of ICTs is not prolific by
itself. According to Modimogale (2009) the advantages of ICTs such as
competitiveness of SMEs actually depends on the way in which ICTs are used to
support the business process (Modimogale, 2009). To clarify the issue further,
Modimogale (2009) categorised ICT adoption and use into three groups: The general
user ICT group; Production Integrating ICT group and the Market Oriented ICT
group. In this work, the three ICT grouping by Modimogale (2009) were used as a
benchmark for determining the competitiveness of Tanzania SMEs. The next
paragraphs summarise group information before the analysis.
The general user ICT group is a one that has implemented basic ICTs such as stand-
alone computers, the email and the Internet (Modimogale, 2009). The adoption rate
in this group is usually very high and is uncoordinated (Modimogale, 2009).
According to Modimogale (2009), this is considered as the entry point into to the
new economy and ICTs are mostly used to meet specific business needs such as
financial planning, keeping of customer records and development of marketing
material (Galloway and Mochrie, 2005; Modimogale, 2009).

228
The Production Integrating ICT group is a more advanced group than the general
user ICT group (Modimogale, 2009). Modimogale (2009) note that, ICTs in this
group would form part of the strategic plans of a business for achieving business
objectives and enhancing or changing business processes (Modimogale, 2009). This
leads to Product innovations and fast delivery of services (Modimogale, 2009).
The Market Oriented ICT group relates to marketing aspect of the company and is
represented by the firm’s web presence (Modimogale, 2009). Using web
technologies, company information and goods can be displayed (Ashrafi and
Murtaza, 2010; Jannex et al., 2004; Modimogale, 2009; Pavic et al., 2007) and some
websites can even allow e-commerce to take place, meaning, orders can be made and
payments be made online. This allows companies to do business 24 hours a day
almost anywhere in the world (Ashrafi and Murtaza, 2010; Jannex et al., 2004;
Modimogale, 2009; Pavic et al., 2007).
6.7.3.1 Use of Computers
Survey results on the use of computers are presented in this section. Data analysis
reveal that, use of computers which is consistent with good practise as explained in
section 4.1.5.1 of this thesis, is above 44% of surveyed firms. To be more specific,
58.1% of surveyed firms (Table 6.9a) show that, use computers for all accounting
activities. 52.8% firms (Table 6.9b) use computers for market research, while 48.4%
of surveyed firms (Table 6.9d) use computers for business transactions using
software packages. Finally, 44.9% of all surveyed firms claimed to use computers for
all business transactions.

229
Table 6.9a: Use of Computers for all Accounting Activities
Frequency Percent Valid Percent Cumulative Percent

No 89 26.1 26.1 26.1
Yes 198 58.1 58.1 84.2
Valid
Neutral 54 15.8 15.8 100.0
Total 341 100.0 100.0
Table6.9b: Use of Computers for Market Research

No 114 33.4 33.4 33.4
Yes 180 52.8 52.8 86.2
Valid
Neutral 47 13.8 13.8 100.0
Total 341 100.0 100.0
Table 6.9c: Use of Computers for Business Transactions Using Software

Packages

No 121 35.5 35.5 35.5
Yes 165 48.4 48.4 83.9
Valid
Neutral 55 16.1 16.1 100.0
Total 341 100.0 100.0
Table 6.9d: Use of Computers for all Business Transactions

Valid No 126 37.0 37.0 37.0
Yes 153 44.9 44.9 81.8
230
Neutral 62 18.2 18.2 100.0

Total 341 100.0 100.0
6.7.3.2 Use of the Internet
According to the literature (Chung, 2006; Mehrtens, 2001; Modimogale, 2009), the
Internet can be used as a communication tool as well as a business tool. Survey
results for Internet usage in this research show that, 60% of the respondents (Table
6.10a) use the Internet (email) for external communications, 55.7% of respondents
(Table 6.10b) use the Internet for obtaining information from suppliers, 52.2% of
respondents (Table 6.10c) use the Internet for sending purchase orders to suppliers,
and 53.7% of respondents (Table 6.10d) use the Internet for product and market
research. In addition, 53.4% of the respondents (Table 6.10e) use the website
technology to offer information to customers while 53.1% of respondents (Table
6.10f) use the company website to receive orders from customers. It is also seen
from the results (Table 6.10g) that, 36.1% of the respondents were able to offer
online payments options using the Internet.
Table 6.10a: Use of Email for external Communication

No 95 27.9 27.9 27.9
Yes 206 60.4 60.4 88.3
Valid
Neutral 40 11.7 11.7 100.0
Total 341 100.0 100.0
Table 6.10b: Use of Internet for obtaining Information from Suppliers

Valid No 102 29.9 29.9 29.9
231
Yes 190 55.7 55.7 85.6

Neutral 49 14.4 14.4 100.0
Total 341 100.0 100.0
Table 6.10c: Use of Internet for Sending Purchase Orders to Suppliers

No 120 35.2 35.2 35.2
Yes 178 52.2 52.2 87.4
Valid
Neutral 43 12.6 12.6 100.0
Total 341 100.0 100.0
Table 6.10d: Use of Internet for Product and Market Research

No 113 33.1 33.1 33.1
Yes 183 53.7 53.7 86.8
Valid
Neutral 45 13.2 13.2 100.0
Total 341 100.0 100.0
Table 6.10e: Use of Websites to Offer Information to Customers

No 116 34.0 34.0 34.0
Yes 182 53.4 53.4 87.4
Valid
Neutral 43 12.6 12.6 100.0
Total 341 100.0 100.0
Table 6.10f: Use of Websites to Receive Orders from Customers

Valid No 122 35.8 35.8 35.8
Yes 181 53.1 53.1 88.9
Neutral 38 11.1 11.1 100.0
232
Total 341 100.0 100.0
Table 6.10g: Use of Internet for Offering Online Payments

No 152 44.6 44.6 44.6
Yes 123 36.1 36.1 80.6
Valid
Neutral 66 19.4 19.4 100.0
Total 341 100.0 100.0
6.7.4 Post Adoption Performance Evaluation
In this section, SMEs performance evaluation after adoption is made. Performance
measures looked here includes profit, market size and the ability of firms to innovate.
Analysis result show that, 49.6% of the surveyed firms attribute their profit to
computer adoption (Table 6.11a) and 44.9% of the surveyed firms attribute their
profit rise to Internet adoption (Table 6.11b).
Table 6.11a: Increase of Profit Due to Computer Adoption

No 101 29.6 29.6 29.6
Yes 169 49.6 49.6 79.2
Valid
Neutral 71 20.8 20.8 100.0
Total 341 100.0 100.0
Table 6.11b: Increase of Profit Due to Internet Adoption

Valid No 103 30.2 30.2 30.2
233
Yes 153 44.9 44.9 75.1

Neutral 85 24.9 24.9 100.0
Total 341 100.0 100.0
In terms of increase of market share due to ICT adoption, 47.5% of the firms
attributed their market share increase to computer adoption (Table 6.11d) while as
48.1% of the surveyed firms attributed their market share increase to Internet
adoption (Table 6.11c). In addition to the market share, 53.4% of the firms recoded
the emergence of new processes that enhance efficiency (Process Innovation) due to
computer adoption (Table 6.9e) while 48.7% of the surveyed firms attributed their
process innovation to Internet adoption (Table 6.10h)
Table 6.11c: Increase of Market Size Due to Internet Adoption

No 105 30.8 30.8 30.8
Yes 162 47.5 47.5 78.3
Valid
Neutral 74 21.7 21.7 100.0
Total 341 100.0 100.0
Table 6.11d: Increase of Market Size Due to Computer Adoption

No 105 30.8 30.8 30.8
Yes 164 48.1 48.1 78.9
Valid
Neutral 72 21.1 21.1 100.0
Total 341 100.0 100.0
Table 6.11e: Process Innovations Due to Computer Adoption

Valid No 81 23.8 23.8 23.8
234
Yes 182 53.4 53.4 77.1

Neutral 78 22.9 22.9 100.0
Total 341 100.0 100.0
Table 6.11f: Process Innovations Due to the Internet

No 94 27.6 27.6 27.6
Yes 166 48.7 48.7 76.2
Valid
Neutral 81 23.8 23.8 100.0
Total 341 100.0 100.0
6.8 Discussion of Descriptive Analyses Findings
Contrary to the data presented by the literature prior to commencement of this
research (Behitsa et al., 2010; Isaya, 2009; Mbamba, 2004; Muttayabalwa, 2007;
Naul, 2012; URT, 2012; Souter et al., 2007) which showed very low usage of ICTs
in SMEs, the findings of these analyses show that, by year 2014, computer adoption
by SMEs in the service and trading sector had reached 80%, (See Table 6.4); Internet
adoption had reached 68.9%, (Table 6.6); website adoption had reached 56.6%,
(Table 6.7); and email adoption had reached 72.7% as depicted by Table 6.8. These
findings point to remarkable and rapid growth in adoption and usage of ICTs in
Tanzania SMEs sector in the last decade.
The research findings regarding ICT use in Tanzania SMEs show that, about 53% of
the surveyed firm’s respondents were shown to use the ICTs in a manner consistent
with the Market Oriented ICT group as suggested by Modimogale (2009). This is
supported by the fact that 53.7% of these firms were able to use company websites to
235
conduct product and market research, (Table 6.10d), 53.4% were able to offer
product information to their customers, (Table 6.10e) and 53.1% were able to receive
orders from customers (Table 6.10f). On a similar note, it is observed that (Table
6.10g), about 36.1 % of the surveyed SMEs could offer online payment options.
These observations confirm the emergence of e-commerce within the SME sector in
Tanzania as suggested by Yonazi (2012) and explained in section 2.3.3 of this thesis.
Post adoption performance evaluation of the surveyed firms present evidence to the
effect that by early 2014, about 48% of surveyed SMEs were using ICTs in a manner
consistent with the Production Integrating ICT group as suggested by Modimogale
(2009) since the firms managed to achieve process innovation as depicted by Table
6.11e and Table 6.11f. Further post adoption analyses show that, by early year
2014, about 47% of SMEs in Tanzania were using ICTs in a way that allowed the
firms to gain competitive advantage. These firms acknowledged improved
performance in terms of increase in profit and market size as a result of ICT adoption
and use. (See Table 6.11a, Table 6.11b Table 6.11c and Table 6.11d). This
observation suggests that, for firms that are competitive, there is a positive and direct
relationship between ICT adoption and SMEs performance.
On the other hand, it is also imperative to point out that, more than a half of the
SMEs that had adopted (See Table 6.11a, Table 6.11b Table 6.11c and Table 6.11d)
were not using ICTs in a productive way since they did not report any improvement
in their performance despite having adopted ICTs. These observations augur well
with the literature (Modimogale, 2009). According to Modimogale (2009), adoption

236
of ICTs by firms does not necessarily give a firm competitive advantage. He claims
that, what ICTs does is to give firms the ability to be competitive (Modimogale,
2009). In other words, competitiveness is a product of how the firms use ICTs. The
literature (DBIS, 2014; Gunday et al., 2009; OECD, 2005) show that, innovation is
key for firms to become competitive. This notion is supported by this research
finding as we are seeing about 48% of firms that are innovative and about 47% of
firms that are competitive. Thus, it can be said that, the majority (53%) of SMEs in
Tanzania are not using ICTs to gain competitive advantage over their competitors
because they are not innovative.
In view of the analyses made and the descriptive statistics presented in section 6.7.3
section 6.7.4, and section 6.8, the requirements of the two specific objectives were
met. The statistics provided give a full and clear picture about the status of ICTs
adoption and use in Tanzania SMEs. This address fully specific objective number
three which sought to establish the status of ICT adoption and use in Tanzania SMEs.
The specific objective number four of this research sought to establish the influence
of ICT adoption has on Tanzania SMEs performance. In partial fulfilment of this
specific objective, it can be seen from the analysis results that, about 47% of
surveyed firms demonstrated a positive relationship between ICTs and firm
performance. The author notes that, the observed relationship is based on how ICTs
are used by the SMEs. That is, SMEs that managed to be innovative using ICTs,
became competitive and recorded good performance. Therefore, it can be said that,
adopted ICTs when used in the right way has a direct positive relationship with
firm’s performance. In other words, the more the firm innovate using ICTs, the
237
higher the firm performance is expected to be (DBIS, 2014; Gunday et al., 2009;
OECD, 2005). This relationship is further analysed and concluded in chapter seven
during hypotheses testing.
6.9 Chapter Summary
This chapter has presented results from preliminary analysis and descriptive analysis.
Preliminary analysis addressed the missing data issue. In addition, the reliability and
internal consistence of the instrument and constructs were tested and found to meet
the criteria acceptable in research of this nature. Moreover, the data was tested for
normality and outliers were detected and dealt with. All this was done so as to
ensure validity and reliability of the research results.
Descriptive analyses carried out were all aimed at fulfilling the requirements of
specific objectives number three and specific objective number four of this research.
Thus, reported statistics inform on the status of ICT adoption and use by Tanzania
SMEs. IT is noted that in the time frame between year 2004 and year 2014, there has
been a remarkable and rapid growth of ICT adoption and use by Tanzania SMEs. In
addition, the statistics also explain the effect ICT adoption and use has on SMEs
performance. Findings suggest a positive relationship between ICTs and firm
performance, when ICTs are used in the right way. It should be noted that, many
other analyses were possible with the survey data. However, since those analyses
were not the focus of the study, they will be addressed in other publications related to
this research. Development of the model is addressed in chapter seven.

238
CHAPTER SEVEN
THE ICT ADOPTION MODEL DEVELOPMENT
7.1 Introduction
In chapter six, the characteristics of SMEs from the service sector and the general
trading were explored. In particular, the analysis on the current status of ICT
adoption and use in Tanzania SMEs was made. In so doing, specific objective
number three was fulfilled. Specific objective number one and specific objective
number four were also partially achieved in chapter six. In this chapter, the research
main objective which is to develop a model of ICT adoption by SMEs together with
remaining specific objectives is addressed. The theoretically derived conceptual
model for ICT adoption as depicted by Figure 4.1 is implemented in SEM.
Thereafter, using survey data, confirmatory and exploratory factor analyses are
performed. That way, the ICT adoption model for SMEs in Tanzania is generated
and validated. It is anticipated that, the generated model together with its
interpretation will encourage SMEs in Tanzania to adopt and use ICTs in the manner
that allows them to become competitive. In addition, by analysing the moderation
effect of government policies on ICT adoption and the post adoption performance
evaluation of the SMEs, government interventions for development of the SME

239
sector can be prepared and executed based on informed decisions. Structural
Equation Modelling (SEM) was used in this phase of model generation. The
justifications on the use of SEM were explained in chapter five.
Modelling commence by a thorough description of the model constructs. This is
followed by construct reliability analysis, test for multicolinearity and discriminant
validity analysis. These tests are performed in order to determine the suitability of
different variables in the model. There after the model is developed.
In model development, a two-step approach to SEM analyses (which is highly
recommended by the literature) was used (See section 5.10.2). Three measurement
models were separately evaluated first before the evaluation of the final model.
Measurement model one evaluated the exogenous latent constructs, measurement
model two evaluated the endogenous latent constructs and measurement model three
evaluated all latent variables. Finally, the structural model was evaluated and testing
of the hypotheses ensued. In the next sections, detailed information about the
different steps of the development of the ICT adoption model is given.
7.2 Constructs of the Research Model
As depicted by Figure 4.1, the proposed model has a total of six latent constructs (or
factors). According to the literature (Byrne, 2010; Hair et al., 2010; Kripanont, 2007;
Schumacker and Lomax 2010; Tabachnick and Fidell 2013) latent constructs
(factors) refer to variables that cannot be measured directly. Instead, latent factors are
240
represented by measurable variables (or indicators) which are specific item or
questions obtained either from respondents in response to questions in a
questionnaire or from some type of observation (Byrne, 2010; Hair et al., 2010;
Kripanont, 2007; Schumacker and Lomax 2010; Tabachnick and Fidell 2013). In
other words, measured variables are used as the indicators of latent constructs.
Kripanont (2007) observe that, indicators are associated with each latent construct
and are specified by the researcher (Byrne, 2010; Hair et al., 2010; Kripanont, 2007;
Schumacker and Lomax 2010; Tabachnick and Fidell 2013).
Out of six latent constructs, four constructs were exogenous constructs and two
constructs were endogenous constructs. According to the literature (Byrne, 2010;
Hair et al., 2010; Kripanont, 2007), an exogenous construct is a latent, multi-item
equivalent of an independent variable, and as such it is not affected by any other
construct in the model (Byrne, 2010; Hair et al., 2010; Kripanont, 2007). The
exogenous latent constructs of this research were: Personal Factors, Organisation
Factors, Economic Factors, and Government Policies that are studied as moderators
in this research. For the endogenous constructs, these are latent, multi-item
equivalents to dependent variables (Byrne, 2010; Hair et al., 2010; Kripanont, 2007;
Sharma, 1996; Tabachnick and Fidell 2013). They are thus constructs that are
affected by other constructs in the model (Byrne, 2010; Hair et al., 2010; Kripanont,
2007; Sharma, 1996; Tabachnick and Fidell 2013). Kripanont (2007) note that, the
consideration of what item belongs to a specific latent construct as well as the
number of observed variables (indicators) is made by the researcher based on the
literature.
241
Generally, the literature (Byrne, 2010; Hair et al., 2010; Kripanont, 2007; Sharma,
1996; Tabachnick and Fidell 2013) acknowledge that, the number of observed
variables has an impact on many statistical analyses. That is, the bigger the number
of observed variables, the bigger the sample size will have to be (Stevens, 2002).
Stevens (2002) noted that, in a standard ordinary least squares multiple regression
analysis, 15 cases per predictor is considered adequate. Based on the argument that
SEM is closely related to multiple regressions in some respects, Stevens (2002)
recommends that, 15 cases per measured variable in SEM should be taken as the rule
of thumb.
Table 7.1 give a summary of the model latent constructs and the number of indicators
used for both exogenous constructs and endogenous constructs. At this point it
should be noted that, all factors that did not meet the requirements for reliability and
internal consistence analysis (See section 6.1.1 and Table 6.1) were omitted from the
analyses of this stage. In addition, other factors from the questionnaire that were
meant for descriptive analyses only were also not considered in SEM analysis. Table
7.1 show that there is a total of 64 observed variables, with 41 variables measuring
the exogenous constructs and 23 variables measuring the endogenous constructs.
Table 7.1: Constructs of the Model
Construct Name Construct Type Number Items Code

of Items
Personal Factors Exogenous 14 A7, A11, D1-D12 PF
Organisation Exogenous 8 E5-E12 OF
Factors
242
Economic Factors Exogenous 16 F1-F16 EF

Government Exogenous 3 G1, G5, G8 GP
Policies
ICT Adoption Endogenous 13 C1-C13 AD
SME Performance Endogenous 10 H1-H10 SMEP
Taking into consideration for the rule of thumb of 15 cases for each observed
variable as recommended by Stevens (2002) it is evident that a bigger sample (usable
dataset) than what was available (see chapters 5 and 6) was needed for this
modelling. However, prior to model generation by using SEM and AMOS, all
factors were analysed for their suitability in the final model and thus the number of
factors in the final model met the sample size requirements. This analysis fulfilled
the requirements of specific objective number one, which sought to analyse the
factors influencing ICT adoption by Tanzania SMEs.
In chapter six it was noted that the research data is not multivariate normal. The non-
normality of data made the use of Bollen-Stine bootstrap method during model
development necessary. When used, this bootstrap method produces the Bollen-
Stine p-value to be used as an alternative p-value in analysis, since the normal p-
value may become unreliable. In the next two sections, detailed account of other
tests that are a pre-requisite to model generation by SEM technique that were not
covered in chapter six is given. In particular, these tests aim to establish the
construct reliability and discriminant validity of the measures involved in this
research. In the process, multicolinearity and or singularity are also tested.

243
7.3 Construct Reliability
Hair et al. (2010) define reliability as “the extent to which a variable or a set of
variables is consistent in what it intended to measure” They argue that, if multiple
measurements are taken, the reliable measures will all be consistence in their values
(Hair et al., 2010). Thus, reliability is about the consistency of measurements
(Kripanont, 2007). According to Kripanont (2007), construct reliability measures
the internal consistency of a set of measures rather than the reliability of a single
variable (Kripanont, 2007). Construct reliability captures the degree to which a set
of measures indicate the common latent construct (Kripanont, 2007). According to
Kripanont (2007) an advantage of construct reliability is that it is based on estimates
of model parameters. Kripanont (2007) further note that, the construct reliability
measure has wide applicability because it can be computed for the construct (s) in a
model regardless of whether the researcher is estimating a congeneric measurement
model, confirmatory factor analysis or path model with latent variables (Kripanont,
2007). Kripanont (2007) also note that, model based estimates of reliability can be
found by looking at either the Squared Multiple Correlations (SMC) for the observed
variables, or the variance extracted estimate. In this research, the SMC (also referred
to, as an item reliability coefficient) was used to measure construct reliability.
According to Tabachnick and Fidell (2013) SMC is the squared multiple correlation
of a variable where it serves as a dependent variable with the rest as independent
variables in multiple correlation. Kripanont (2007) note that, when SMC is used as a
measure of construct reliability, the cut off point for an acceptable indicator variable
is a one that has SMC of 0.3. Kripanont (2007) further recommends that a good
observed variable should have a SMC of 0.50 and above.

244
Table 7.2 show the SMC for all the 41 observed variables for the exogenous
constructs. From table 7.2 it can be observed that a total of 11 (as highlighted in
Table 7.2) variables measuring the three exogenous latent constructs had a SMC less
than the recommended 0.5. These variables were (deleted) excluded from further
SEM analysis in order to improve model fit to the data. Observable variables that
remained after the removal of factors that failed the SMC test included variables D1-
D12 measuring the Personal Factors (PF) latent construct, variables E7-E12
measuring the Organisation Factor (OF) latent construct and variables F7-F15
measuring the Economic Factors (OCF) latent construct. The three moderating
variables G1, G5 and G8 also remained.
Table 7.2: The SMC for the 41 Indicators Measuring Exogenous Constructs
VARIABLE SMC ESTIMATE VARIABLE SMC ESTIMATE

E12 .535 F4 .268
E11 .632 F5 .262
F16 .497 F6 .429
F15 .506 F7 .718
F14 .799 F8 .725
F13 .777 A7 .051
F12 .790 A11 .052
F11 .807 D1 .500
F10 .754 D2 .658
F9 .763 D3 .701
D12 .741 D4 .741
D11 .789 D5 .760
E10 .596 D6 .850
D10 .824 D7 .833
D9 .850 D8 .835
E9 .742 E5 .383
E8 .704 G1 .611
E7 .787 G5 .615
E6 .462 G8 .577
245
F1 .363 F3 .187
F2 .354
The SMC for the indicators measuring the two endogenous latent constructs are
shown in Table 7.3. From Table 7.3 it can be seen that, out of 23 observed variables
for the two endogenous latent constructs only one variable had SMC less than 0.5.
This variable (C1) was also deleted from further SEM analysis for the same reason of
improving model fit to data. The remaining variables for the endogenous constructs
included variables C2-C13 measuring Adoption (AD) construct and variables H1-
H10 used to measure the SME Performance (SMEP) construct.
Table 7.3: The SMC for the 22 Indicators Measuring Endogenous Constructs
VARIABLE SMC ESTIMATE

C1 .439
C2 .729
C3 .667
C4 .542
C5 .719
C6 .653
C7 .771
C8 .749
C9 .739
C10 .848
C11 .801
C12 .599
C13 .536
H10 .624
H9 .857
H8 .855
H7 .694
H6 .778
H5 .516
H4 .830
H3 .845
H2 .680
246
H1 .749
At the end of the construct reliability analysis, a total of 52 indicator variables
remained for SEM analysis. Out of the 52 indicator variables, 27 were measuring the
three exogenous latent constructs while 22 were measuring the two endogenous
latent constructs and 3 exogenous variables were used as moderators.
7.4 Tests for Multicolinearity
The background theory on multicolinearity is provided in section 5.11.4 of this
thesis. Using AMOS version 20 sample correlations, standardised residual
covariance, and implied (all variables) correlations of the survey data after deletion
of variables with low SMCs, was determined and recorded as shown in appendix
IIIA for investigation of multicolinearity and discriminant validity of model
constructs.
Appendix III-A Table 1 show the sample correlation for the 27 variables used to
measure the three exogenous latent constructs. From appendix III-A Table 1 it is
seen that only variables D1, E9-E12 and F15 were not highly correlated with other
variables. The rest of observed variables were highly correlated and this proved the
presence of multicolinearity in the data. Table 7.4 give a summary of multicollinear
variables for the three exogenous latent constructs.

247
Appendix III-B Table 4 show the sample correlation for the 22 variables used to
measure the two endogenous latent constructs after removal of variables with low
SMC. Like for the exogenous variables, there was evidence of multicolinearity for
endogenous variables as summarised by Table 7.5. However, variables, C3-C6, C9,
C12, C13, H5 and H10 were not highly correlated with any variable.
The presence of multicolinearity in the data made it necessary for further analysis of
the variables in order to determine their usefulness in the SEM analysis. This was
achieved during Discriminant Validity analysis and fitting of the measurement
models.
Table 7.4: Summary of Highly Correlated Variables for the Three Exogenous
Latent Constructs
VARIABLE NAME VARIABLES HIGHLY CORRELATED

D1 -
D2 D4
D3 D4
D4 D2, D3, D6
D5 D6, D7, D8, D9
D6 D4, D5, D7, D8, D9, D10, D11, D12
D7 D5, D6, D8, D9, D10
D8 D5, D6, D7, D9, D10, D12
D9 D5, D6, D7, D8, D10, D11
D10 D6, D7, D8, D9, D11, D12
D11 D6, D9, D10, D11
D12 D6, D8, D10, D11
E7 E8
E8 E7
E9 -
E10 -
E11 -
E12 -
F7 F8
F8 F7
F9 F10, F11
248
F10 F9, F11, F12

F11 F9, F10, F12, F13, F14
F12 F10, F11, F13, F14
F13 F11, F12, F14
F14 F10, F11, F12, F13
Table 7.5: Summary of Highly Correlated Variables for the Two Endogenous
Latent Constructs
VARIABLE NAME VARIABLES HIGHLY CORRELATED

C2 C10
C3 -
C4 -
C5 -
C6 -
C7 C10
C8 C11
C9 -
C10 C2, C7, C11
C11 C8, C10
C12 -
C13 -
H1 H3, H4, H6
H2 H7
H3 H1, H4, H8, H9
H4 H1, H3, H8, H9
H6 H1, H8, H9
H5 -
H7 H2
H8 H3, H4, H6, H9
H9 H3, H4, H6, H8
H10 -
249
7.5 Discriminant Validity of Constructs
Hair et al. (2010) define validity as “the degree to which a measure accurately
represents what it is supposed to”. Kripanont (2007) note that, validity is the
accuracy of a measure, and exists when the measure is a perfect representation of the
variable intended to measure. In statistical modelling, it is desired that measures of
constructs that are theoretically different should in reality be different from each
other (Guo et al., 2008; Kripanont, 2007). Discriminant validity reflects the extent to
which the different constructs in a model are different (Kripanont, 2007). In other
words, discriminant validity refers to the distinctiveness of different constructs (Guo
et al., 2008). Therefore, discriminant validity test is performed in order to show that
measures or concepts that are supposed to be unrelated are, in fact, not related (Guo
et al., 2008; Kripanont, 2007).
Kripanont (2007) note that, Structural equation modelling techniques can be used to
estimate discriminant validity. In typical SEM analysis, large correlations (0.8 or
0.9) between latent constructs suggest a lack of discriminant validity (Kripanont,
2007). In addition to the correlation analysis, pattern and structure coefficients can
also be used in determining whether constructs in the measurement models are
empirically distinguishable (Kripanont, 2007). These two methods were used in this
research to determine the discriminant validity of the latent constructs. It should be
noted that, the structure coefficients are generated in AMOS output in the “all
implied moments”. According to Kripanont (2007) when performing validity

250
analysis in SEM, there should be no more than five constructs under investigation at
a time. In this research, due to the large number of observed variables involved,
discriminant validity analysis was done in two phases. Phase one involved the
analysis for the exogenous latent constructs (here referred to as the measurement
model one) and endogenous latent constructs (here referred to as the measurement
model two) separately. In phase two, all five constructs for the model were analysed
together and this model is here referred to as the “measurement model three”. The
next three sections give detailed explanations on these analyses.
7.5.1 Measurement Model One
In section 5.11.4 the fact that some statistical analyses are sensitive to
multicolinearity was explained. It was also recommended that, when
multicolinearity is detected in data, one of the multicollinear indicators is deleted
from SEM analysis, that is, if the sample correlation between the two indicators
exceeds 0.8. In addition to this observation, the literature (Arbuckle, 2009;
Kripanont, 2007) also recommends the investigation of standardised residual
covariance between two indicators, in order to determine the correct model. The
literature (Arbuckle, 2009; Kripanont, 2007) note that, with correct model, most
standardised residuals should have an absolute value less than 2. Byrne (2010) claim
that, the maximum allowable absolute value for standardised residuals covariances is
2.5 (Byrne, 2010). Taking into account the level of multicolinearity between
exogenous variables (See Table 7.4 and Appendix III-A Table 1) and the number of
variables that had standardised residual covariance above 2 (Appendix III-A Table 2)
251
the correct measurement model one for the exogenous latent constructs was
determined by deleting one of the offending variables. Moreover, since many
variables were highly correlated with more than one variable (See Table 7.4 and
Table 7.5) it was also necessary to choose a pair of factors that described the
measurement model better than others. After identification of the measurement
model one, discriminant validity analysis was carried out for the exogenous latent
variables.
Table 7.6 show the correlation between the three latent constructs. The construct
correlations are provided by AMOS provides the construct correlations whenever
standardised estimates are requested as part of the output (Arbuckle, 2009; Hair et
al., 2010; Kripanont, 2007). From this output, it can be seen that the three latent
variables were different since the correlation between them was less than 0.8. The
highest correlation is 0.6 between OF (organisation factors) and PF (personal
factors).
Table 7.6: Correlation between Latent Constructs for Investigating

Discriminant Validity
Latent Constructs’ Correlation Estimates

PF <--> ECF -.179
OF <--> ECF -.509
PF <--> OF .684
Table 7.7 depict the pattern and structure coefficients (implied correlations for all
variables) of the three constructs in measurement model one. The pattern and
structure coefficients are calculated and generated by AMOS when all implied
252
moments are ordered as part of AMOS output (Arbuckle, 2009; Hair et al., 2010;
Kripanont, 2007). In this case, it can be seen that observed variables load strongly
on the latent construct they are supposed to measure and less on the other constructs.
Thus, the pattern and structure coefficients confirm discriminant validity of the three
exogenous latent constructs in the model.
Analysis of the standardised residual covariances showed that, the absolute value of
standardised residuals between all variables (after deleting offending variables) was
less than 2 as depicted in Table 7.8. This suggested that the measurement model was
correct.
Table 7.7: Implied (for all remaining variables) Correlations for Three
Exogenous Latent Constructs for Investigating Discriminant
Validity
ECF OF PF F7 F10 F13 E7 E9 E11

ECF 1.000
OF -.509 1.000
PF -.179 .684 1.000
F7 .848 -.432 -.152 1.000
1.00
F10 .847 -.431 -.152 .718
0
F13 .874 -.445 -.157 .742 .741 1.000
E7 -.437 .858 .587 -.370 -.370 -.382 1.000
E9 -.461 .906 .620 -.391 -.391 -.403 .778 1.000
E11 -.394 .774 .529 -.334 -.333 -.344 .664 .701 1.000
D2 -.153 .585 .856 -.130 -.130 -.134 .502 .530 .453
D3 -.155 .592 .866 -.132 -.131 -.136 .508 .537 .458
D9 -.157 .599 .876 -.133 -.133 -.137 .514 .543 .463
253
Table 7.7: (continued)
F7 F10 F13 E7 E9 E11 D2 D3 D9

ECF
OF
PF
F7 1.000
F10 .718 1.000
F13 .742 .741 1.000
E7 -.370 -.370 -.382 1.000
1.00
E9 -.391 -.391 -.403 .778
0
E11 -.334 -.333 -.344 .664 .701 1.000
D2 -.130 -.130 -.134 .502 .530 .453 1.000
D3 -.132 -.131 -.136 .508 .537 .458 .741 1.000
D9 -.133 -.133 -.137 .514 .543 .463 .749 .758 1.000
Table 7.8: Standardized Residual Covariance’s for Measurement Model One
F7 F10 F13 E7 E9 E11 D2 D3 D9

F7 .000
F10 -.112 .000
F13 -.017 .107 .000
E7 -.107 1.088 .830 .000
E9 -.200 .266 .541 .028 .000
E1
-1.784 -1.722 -1.523 -.120 .026 .000
1
D2 .562 1.066 .206 .274 .097 -1.086 .000
D3 .445 .247 -.038 .772 -.291 -.981 -.003 .000
D9 -.476 -.468 -1.010 .298 .285 -.589 -.015 .016 .000
The measurement model one for the three exogenous latent constructs is depicted in
Figure 7.1. The literature (Arbuckle, 2009; Byrne 2010; Hair et al., 2010; Kripanont,
2007; Tabachnick and Fidell, 2013) have detailed information on how AMOS and
254
SEM outputs are interpreted. A summary of fit measures for this research are
provided in section 5.10.2.1.
For the model presented by Fig 7.1, fit statistics CMIN (Chi-square) = =44.627,
degree of freedom (DF) = 24, p-value=0.006, (which is significant at the level of
0.05), Bollen-Stine bootstrap p-value=0.160 (which is not significant at the level of
0.05) and CMIN/DF ratio =1.859, indicated a very good fit of model to data. In
addition, taking into account the fact that chi-square statistic is very sensitive with
samples size (Kripanont, 2007; Arbuckle, 2009; Hair et al. 2010, Tabachnick and
Fidell, 2013; Byrne 2010) other fit measures were considered as recommended by
the literature (Kripanont, 2007; Arbuckle, 2009; Hair et al. 2010, Tabachnick and
Fidell, 2013; Byrne 2010). These statistics RMSEA = 0.050, TLI=0.985, CFI =
0.990, NFI = 0.979, GFI = 0.972, AGFI = 0.947 also indicated a very good fit of
model to data. (See section 5.10.2.1 for meaning of fit measures)

255
Figure 7.1: Measurement Model One with Standardised Estimates for Three
Exogenous Latent Constructs
Summary of fit measures For Model of Figure 7.1
CMIN (Chi-square) = =44.627, DF = 24, p-value=0.006,
Bollen-Stine bootstrap p-value = .160 CMIN/DF =1.859 RMSEA = 0.050,
TLI=0.985, CFI = 0.990, NFI = 0.979, GFI = 0.972, AGFI = 0.947
7.5.2 Measurement Model Two
Analysis of measurement model two were similar to the analyses carried out for the
measurement model one. Taking into account that some variables were
256
multicollinear (See Table 7.5 and Appendix III-B Table 4) and that some of the
variables had standardised residual covariance above 2 (See Appendix III-B Table 5)
the correct measurement model for the two endogenous latent variables was
determined.
The measurement model for the two endogenous latent variables is shown in Figure
7.2. The model yielded =21.085, degree of freedom=19, p-value 0.332 which is
not significant at the level of 0.05, Bollen-Stine bootstrap p-value =0.850 which is
also not significant at the level of 0.005 and CMIN/DF =1.110. All these measures
indicated a very good fit of model to data (Kripanont, 2007; Arbuckle, 2009; Hair et
al. 2010, Tabachnick and Fidell, 2013; Byrne 2010). Other fit measures included
RMSEA = 0.018, TLI=0.999, CFI = 0.999, NFI = 0.990, GFI = 0.984, AGFI = 0.970.
These measures also indicated a very good fit of model to data. (See section 5.10.2.1
for meaning of fit measures)
Discriminant validity analysis for measurement model two showed that the latent
constructs were different since the correlation between the two latent constructs was
0.63 a value which is below the threshold of 0.8 (Kripanont, 2007; Arbuckle, 2009;
Hair et al. 2010, Tabachnick and Fidell, 2013; Byrne 2010). Moreover, the pattern
and structure coefficients as shown in Table 7.9 indicated that, the two endogenous
latent constructs in the measurement models were empirically distinguishable.
In addition, the analysis of the standardised residual covariances showed that, the
absolute value of standardised residuals between all variables (after deleting

257
offending variables) was less than 2 as depicted in Table 7.10. This suggested that
the measurement model was correct.
Table 7.9: Implied (for all variables) Correlations for two Endogenous
Latent Constructs for Investigating Discriminant Validity
AD SMEP C5 C8 C9 C10 H1 H5
AD 1.000
SMEP .623 1.000
C5 .873 .544 1.000
1.00
C8 .867 .540 .757
0
C9 .875 .545 .764 .759 1.000
C10 .898 .559 .784 .779 .786 1.000
1.00
H1 .524 .841 .457 .454 .458 .470
0
H5 .435 .699 .380 .377 .381 .391 .588 1.000
H7 .508 .816 .444 .441 .445 .456 .687 .571
H9 .584 .938 .510 .507 .511 .524 .789 .656
Table 7.10: Standardized Residual Covariance’s for Measurement Model

Two
C5 C8 C9 C10 H1 H5 H7 H9
C5 .000
-.02
C8 .000
4
C9 .067 -.019 .000
-.05
C10 -.021 .054 .000
6
-.62 -
H1 -.079 -.232 .000
6 1.302
H5 .661 .299 -.252 -.097 .099 .000
-.42 -
H7 -.430 -.451 .370 -.092 .000
7 1.122
-.06 -.04
H9 .444 .634 -.024 .391 -.023 .000
3 7
258
Figure 7.2: Measurement Model Two with Standardised Estimates for Two
Endogenous Latent Constructs
CMIN=21.085, DF=19, p-value 0.332,
Bollen-Stine bootstrap p-value =0.850 CMIN/DF =1.110, RMSEA = 0.018,
TLI=0.999, CFI = 0.999, NFI = 0.990, GFI = 0.984, AGFI = 0.970.

259
7.5.3 Measurement Model Three
A measurement model for all five constructs involved in this study is depicted in
Figure 7.3. This model was arrived at by taking into account the multicolinearity of
variables as well as looking at the standardised residual covariances. One variable
from each pair that were highly correlated (0.8 or above) was deleted. Likewise, any
pair of variables that had standardised residual covariance of more that 2 (in absolute
terms) were deleted.
The model of Figure 7.3 fit the data very well as attested by the fit measures: Chi-
square = 158.471, Degree of Freedom = 80, p-value =0.000, Bollen-Stine bootstrap
p-value=0.114, CMIN/DF=1.981, RMSEA= 0.054, TLI = 0.973, CFI=0.980,
NFI=0.960, GFI = 0.942 and AGFI = 0.912. Note that the Bollen-Stine bootstrap p-
value =0.114 is not significant at 0.05.
Discriminant validity analysis for measurement model three indicated lack of
discriminant validity between the Personal Factor (PF), an exogenous latent construct
and Adoption (AD) an endogenous latent construct as well as for Economic Factors
(ECF) which is an exogenous latent construct and SME Performance which is an
endogenous latent construct. The remaining constructs were different as attested by
Table 7.11 showing the correlations between different constructs of the model.
260
Figure 7.3: Measurement Model Three with Standardised Estimates for all
Five Latent Constructs
CMIN = 158.471, DF = 80, p-value =0.000, CMIN/DF=1.981,
Bollen-Stine bootstrap p-value=0.114, RMSEA= 0.054, TLI = 0.973, CFI=0.980,
NFI=0.960, GFI = 0.942 and AGFI = 0.912.

261
Table 7.11: Correlation between Five Latent Constructs for Investigating

Discriminant Validity
Latent Constructs Correlation Estimates

PF <--> ECF .692
AD <--> SMEP .609
OF <--> AD -.259
ECF <--> AD .634
ECF <--> SMEP .856
OF <--> ECF -.508
OF <--> SMEP -.447
PF <--> OF -.190
PF <--> AD .877
PF <--> SMEP .632
Analysis of pattern and structure coefficients as shown in Appendix III-C Table 7
indicated that, the constructs in the measurement model three were empirically
distinguishable. Moreover, the analysis of the standardised residual covariances
showed that, the absolute value of standardised residuals between all variables (after
deleting offending variables) was less than 2 as depicted in Appendix III-C Table 8.
This suggested that the measurement model was correct, except for the highly
correlating latent constructs.
A comparison between measurement model three (Figure 7.3) with measurement
model two (Figure 7.2) and measurement model one (Figure 7.1) respectively show
that model fit was achieved using different combinations of variables. For example,
the PF latent construct in measurement model one used the variables D2, D3 and D9
while for the measurement mode three PF used the variables D3, D5 and D10 to
achieve model fit. This was taken into consideration while fitting the overall
262
research model. That is, different pairs of variables were tried and the ones that
resulted in the best model fit were used in the ICT Adoption model.
To address the issue of high correlation between PF and AD and that between ECF
and SMEP (see Table 7.11) the variables C6, C9, C10, H1, H5 and H8 were recorded
from 5 points to three points (Tabachnick and Fidell, 2013). The resulting model
using these recoded values is shown in Figure 7.4. The fit measures for this model
are summarised as: Chi-square =139.152, Degree of Freedom =80, p-value=0.000,
Bollen-Stine bootstrap p-value = 0.134, CMIN/DF =1.739, RMSEA=0.047,
TLI=0.975, CFI=0.981, NFI=0.956, GFI =0.950, AGFI=0.925 which indicated a
very good fit of model to data. In addition to model fit, the problem of high
correlation between the constructs PF and AD as well as ECF and SMEP was
addressed.
Table 7.12 show the correlations between constructs after re-specification of
measurement model three. It can be seen from Table 7.11 that the highest correlation
between constructs is 0.692 recoded between PF and ECF constructs. Thus,
discriminant validity between constructs was achieved.

263
Figure 7.4: Measurement Model Three with Standardised Estimates for all
Latent Constructs using recorded Variables
CMIN=139.152, DF=80, p-value=0.000, CMIN/DF =1.739,
Bollen-Stine bootstrap p-value = 0.134, RMSEA=0.047, TLI=0.975, CFI=0.981,
NFI=0.956, GFI =0.950, AGFI=0.925

264
Table 7.12: Correlation between Five Latent Constructs (after recording

variables) for Investigating Discriminant Validity
Latent Constructs Correlation Estimates

PF <--> ECF .692
AD <--> SMEP .399
OF <--> AD -.318
ECF <--> AD .578
ECF <--> SMEP .392
OF <--> ECF -.510
OF <--> SMEP -.278
PF <--> OF -.190
PF <--> AD .637
PF <--> SMEP .192
7.6 Model Estimation.
7.6.1 Unstandardised and Standardised Estimates
After model identification, the next logical step in SEM analysis process is to
estimate model parameters. Model parameters for this research were obtained using
AMOS version 20, which uses Maximum Likelihood as its default method of
estimation (Arbuckle, 2009). According to Arbuckle (2009), AMOS produces
estimates with very desirable properties including both Unstandardised and
standardised model solutions (Arbuckle, 2009; Byrne 2010; Kripanont, 2007).
Kripanont (2007) observed that, when an Unstandardised model is estimated, the
regression weights, covariance’s, intercepts (only when mean structures are
analysed) and variances will be displayed in the path diagram (Byrne 2010;
Kripanont, 2007). Regression weights represent the influence of one or more
variables on another variable (Byrne 2010; Kripanont, 2007). On the other hand,
when a standardised model is estimated, standardised regression weights, correlation

265
and squared multiple correlations will be displayed. Furthermore, Kripanont (2007)
note that, the standardized regression weights and the correlations are independent of
the units in which all variables are measured, and will not be affected by the choice
of identification constraints (Arbuckle 2005). In this research, when performing path
analysis, both unstandardised and standardised model solutions will be presented.
7.6.2. Squared Multiple Correlations (SMC)
The literature (Arbuckle, 2009; Byrne 2010; Hair et al., 2010; Kripanont, 2007;
Schumacker and Lomax 2010; Tabachnick and Fidell, 2013) show that, the fit
measures discussed in section 5.10.2.1 provide information about how well the
model fits the data. However, according to Kripanont (2007), the strength of the
structural paths in the model is determined by Squared Multiple Correlations. It is
therefore important that the SMC of each dependent variable is considered together
with fit measures in order to best describe the structural model (Arbuckle 2009;
Kripanont, 2007). Kripanont (2007) note that, the interpretation of SMC is
analogous to the R2 statistic in multiple regression analysis (Kripanont, 2007). In
addition, Arbuckle (2009) observe that, SMC is a useful statistic that is also
independent of all units of measurement (Arbuckle, 2009). Thus, apart from the fit
measures, in the analyses, SMC will also be reported.

266
7.7 The ICT Adoption Model for Tanzania SMEs
The proposed research model shown in Fig 4.1 was implemented in AMOS for SEM
analysis using variables that had passed all initial tests for their usefulness in SEM
analyses. Figure 7.5 show the Initial ICT Adoption model for SMEs in Tanzania
with Unstandardised Estimates while Figure 7.6 shows the initial ICT Adoption
model for SMEs in Tanzania with Standardised Estimates. Variables used in
measurement model one and measurement model two were used in coming up with
the initial ICT Adoption model. This enabled for the initial ICT Adoption model to
consider a total of 17 observed variables. Looking at the fit measures for the Initial
ICT Adoption model in particular the p-value = 0.000 and the Bollen-Stine bootstrap
p-value = 0.005 both of which are significant at the level 0.05 it can be concluded
that the initial model did not fit the data very well. Other fit measures such as the
normed chi-square, CMIN/DF =4.563 and RMSEA = 0.102 also confirmed the
inadequacy of the model fit to data (see fit measures in section 5.10.2.1). It was
therefore necessary to re-specify the model as suggested by the literature (Arbuckle,
2009; Byrne 2010; Hair et al., 2010; Kripanont, 2007; Schumacker and Lomax 2010;
Tabachnick and Fidell, 2013) in order to obtain a model that fits the data well.
The first steps in model respecification involved the use of model variables used in
measurement model three rather than using model variables based on measurement
model one and measurement model two respectively. The variables of measurement
model one and measurement model two were the ones used in the initial ICT
Adoption model with poor results. Different combination of variables based on
measurement model three were tested in order identify a model with best statistics.
267
After these steps, the initial ICT adoption model still did not fit the data well. It was
therefore necessary, to analyse the sample correlations, standardised residual
covariances, and modification indices in order to determine the source of model
misfit. In this phase, it was necessary to delete some variables that contravened the
underlying statistical assumptions for the SEM analysis. Finally, an ICT Adoption
model which fitted the data well was achieved.
Figure 7.5: The Initial ICT Adoption Model with Unstandardised Estimates
Summary of fit measures for the model of Figure 7.5
CMIN = 511.112, DF = 112, P-value = 0.000, CMIN/DF=4.563
Bollen-Stine bootstrap p-value=0.005, RMSEA = 0.102, NFI=.898, TLI = .901,
CFI=.918, GFI=.859, AGFI=.807

268
Figure 7.6: The Initial ICT Adoption Model with Standardised Estimates
CMIN = 511.112, DF = 112, P-value = 0.000, CMIN/DF=4.563
Bollen-Stine bootstrap p-value=0.005, RMSEA = 0.102, NFI=.898, TLI
= .901, CFI=.918, GFI=.859, AGFI=.807
Table 7.13 gives a summary of parameter for the ICT Adoption model. The model
had a total of 53 parameters estimates. Figure 7.7 show the ICT Adoption model
with unstandardised estimates which includes regression weights, covariances and
variances (Kripanont, 2007; Arbuckle, 2009; Byrne 2010).

269
Table 7.13: Parameter Summary for the Developed ICT Adoption Model
Weights Covariances Variances Means Intercepts Total

Fixed 20 0 0 0 0 20
Labelled 0 0 0 0 0 0
Unlabelle
12 3 18 0 0 33
d
Total 32 3 18 0 0 53
Figure 7.8 show the same ICT Adoption model but this time with standardised
estimates. The estimates for the standardised model include standardised regression
weights, correlations, and square multiple correlations (Arbuckle, 2009; Byrne 2010;
Kripanont, 2007). It should be noted that, standardised estimates are independent of
the units measured (Arbuckle, 2009; Byrne 2010; Kripanont, 2007) and thus they are
not affected by the choice of identification constraints (Arbuckle, 2009; Byrne 2010;
Kripanont, 2007).
The developed ICT Adoption model had the following fit statistics: Chi-
square=108.370, Degree of Freedom=58, CMIN/DF=1.868, p-value=0.000, Bollen-
Stine bootstrap p-value=0.085, which is not significant at the level of 0.05 and thus
indicating that the model fits the data very well. In addition, other fit measure that
are not based on the Chi-square, RMSEA=0.051, TLI=0.970, CFI=0.978,
NFI=0.954, GFI=0.955, AGFI=0.929 also showed that the model fits the data well
(see Table 5.15 for reference of fit measures).

270
Figure 7.7: The ICT Adoption Model with Unstandardised Estimates
CMIN=108.370, DF=58, p-value=0.000, CMIN/DF=1.868,
Bollen-Stine bootstrap p-value=0.085, RMSEA=0.051, TLI=0.970, CFI=0.978,
NFI=0.954, GFI=0.955, AGFI=0.929

271
Figure 7.8: The ICT Adoption Model with Standardised Estimates
NFI=0.954, GFI=0.955, AGFI=0.929

272
7.7.1 Variables of the Developed ICT Adoption Model
A summary of developed model variables is given in Table 7.14. From Table 7.14 it
is seen that, out of a total of fourteen observed variables measuring the PF latent
construct, only three variables were found to be most significant. These factors were
the perceptions (Perceived benefits) of the manger-owner that: The Internet reduces
organisation operational costs; the Internet is an efficient way of communicating with
customers; and the Internet is an effective way to gather market and competitor
related information. In the research instrument, these variables were coded D3, D5,
and D10 respectively.
Table 7.14: Summary of developed ICT adoption Model Variables
Construct Name Construct Type Number Items Code

of Items
Personal Factors Exogenous 3 D3, D5, D10 PF
Organisation Factors Exogenous 2 E7 E11 OF
Economic Factors Exogenous 2 F9, F14 EF
ICT Adoption Endogenous 3 C6 C9 C10 AD
SME Performance Endogenous 3 H1, H5, H8 SMEP
For the OF construct, out of eight observed variables measuring this construct, only
two variables that is the marketing strategy and external pressure from customers
were found to be the most important measures of the construct. These variables were
coded in the research instrument as E7 and E11 respectively.
For the ECF construct, out of 16 variables measuring this construct, two variables,
that is acquisition costs of computer systems and operational costs for the internet
273
were found to be the most important measures of the construct. These variables
were coded in the research instrument as F9 and F14 respectively.
For the AD construct, out of 13 variables measuring this construct, use of the internet
for randomly looking for information, use of the internet for sending purchase orders
to suppliers and use of the internet for product and market research were found to be
the most important variables for measuring ICTs adoption and use by Tanzania
SMEs. These variables were coded in the research instrument as C6, C9 and C10
respectively.
For the SMEP construct, out of 10 variables measuring this construct, only three
variables, that is, increased market size as a result of computer adoption, emergence
of new processes that enhance company efficiency and effectiveness as a result of
computer technology adoption and increased sales volume in the business as a result
of internet adoption were found to be the most important variables for measuring
SME performance by Tanzania SMEs. These variables were coded in the research
instrument as H1, H5 and H8 respectively.
The next step carried out after the identification of the ICT adoption model was to
test the hypotheses as put forward in section 4.3. Hypotheses’ testing was done in
two phases. The first phase involved testing of the hypothesis without the effect of
moderator variables (The direct path hypotheses). These were the hypotheses that
studied the effect of indicator variables on the endogenous variables. In this
category, the hypotheses tested include the hypotheses H1a, H1b, H1c, H2a, H2b,
274
H2c, H3 and H4 as explained in section 4.3. Detailed account of hypotheses testing
is given in the next section.
7.8 Hypotheses Testing-Direct Path Hypotheses
Testing of the hypotheses in this phase involved the analysis of the regression
weights, Squared Multiple Correlation, standardised regression weights and the
covariances of the ICT Adoption Model. These estimates were calculated by AMOS
software.
Testing commenced by testing the hypotheses H1a, H1b and H1c. These hypotheses
tested the individual effect of predictor variables (exogenous constructs) to the
dependent variables (endogenous constructs) of the model. Results (See Table 7.26)
show that the three hypotheses H1a, H1b and H1c were accepted. This was
followed by testing the hypotheses H2a, H2b, and H2c. These hypotheses tested the
combined effect of two predicting variables (exogenous constructs) to the dependent
variables (endogenous constructs) of the model. Results (See Table 7.26) show that
the hypotheses H2a and H2b were accepted while hypothesis H2c was rejected.
Finally, the direct hypothesis H3 and H4 were tested. Contrary to the testing of
hypotheses from H1a to H2c, analysis of hypothesis H3 and hypothesis H4 involved
analysis of the entire ICT adoption structural model developed rather than testing
only the constructs involved with a particular hypothesis. In order to avoid repetition
of analysis procedures, only detailed analysis of hypotheses H3 and H4 are reported
in this section. Overall hypothesis testing results are given in Table 7.26.
275
Table 7.15 report the regression weights of the ICT Adoption model. The results
show that, the paths between OF and AD is statistically significant at probability
level = 0.05, while the paths between PF and AD and the path between AD and
SMEP are statistically significant at the probability level of 0.01. The path between
ECF and AD had a p value of 0.252. According to the literature (Kripanont, 2007;
Arbuckle, 2009; Byrne 2010) based on the probability level of 0.05, parameters that
are not significant, with the exception of error variances, can be considered
unimportant to the model. Furthermore, Byrne (2010) recommend that non-
significant parameters should be deleted from the model in the interest of scientific
parsimony. Thus, with a p-value of 0.252, the effect of Economic Factors (ECF) on
ICT adoption (AD) was unimportant and thus the construct was not considered
during moderating hypotheses testing.
Table 7.15: Regression Weights for ICT Adoption Model
Estimat
S.E. C.R. P Label
e
AD <--- OF .117 .045 2.581 .010* par_12
AD <--- PF .141 .036 3.915 *** par_14
-
AD <--- ECF -.032 .028 .252 par_15
1.145
SMEP <--- AD .575 .095 6.042 *** par_13
Note: *** A p value is statistically significant at the 0.01 level (two-tailed)

* A p value is statistically significant at the 0.05 level (two-tailed)
Arbuckle (2005-2009) and Kripanont (2007) note that, the square multiple
correlations of a variable is the proportion of its variance that is accounted for by its
predictors (Arbuckle, 2009; Kripanont, 2007). The meaning of this is, if for example
276
a dependent variable Y has SMC of 0.5, then determinants (predictors) of Y account
for 50% of the variance of Y. Table 7.16 record the squared multiple correlations for
the ICT Adoption model. From Table 7.16 is can be deduced that the observed
variables D3, D5 D10, E7 and E11accounted for 47.8% of the variance of (adoption)
AD. In addition, the variables D3, D5 D10, E7 and E11 also accounted for 16.5% of
the variance of (SME Performance) SMEP.
Table 7.16: Squared Multiple Correlations for the ICT Adoption Model
SMC Estimate
AD .478
SMEP .165
According to the literature (Hair et al., 2010; Kripanont, 2007), standardised
regression weights allows for direct comparison of the relative effect of each
independent variable on the dependent variable (Hair et al., 2010; Kripanont, 2007)
and thus they are used in hypothesis testing. Table 7.17 show the standardized
regression weights for the ICT Adoption model. From Table 7.17 it can be observed
that, the standardized regression weights between OF and AD, PF and AD and AD
and SMEP were statistically significant while the standardized regression weight
between EFC and AD was not significant. Therefore, hypothesis H3 was rejected
while hypothesis H4 was accepted. It was also observed that, when the model is
evaluated as a whole the hypotheses H1a, H1b, H2a and H4 were accepted, while the
hypotheses H1c, H2b, H2c and H3 were rejected. The implication of this
observation is that, Personal Factors and Organisational Factors influenced ICT
Adoption-AD significantly and AD significantly influenced SME Performance. On

277
the other hand, the rejected hypotheses H2b, H2c and H3 involved testing the
combined effect of ECF with one or two other predictors on AD. As can be seen
from Table 7.15 and Table 7.17, the ECF construct was not significant (p value =
0.252 and standardised regression weight -0.086). This explains why all hypotheses
that involved ECF were rejected.
Table 7.17: Standardized Regression Weights
Estimate
AD <--- OF .316
AD <--- PF .391
AD <--- ECF -.086
SMEP <--- AD .407
Analysis of covariances showed that, all factors were statistically significant. Also,
the factors were interrelated as expected as shown in Table 7.18. It was also noted
that, the factors PF and OF were positively correlated while the factors ECF and PF
and the factors ECF and OF were negatively correlated. The positive correlation
between PF and OF (0.944) suggests that, the higher the Personal Factors, (in this
case perceived benefits) the higher Organisation Factors (in this case the external
pressure and strategy). In addition, a higher value of ECF has a negative effect on
both PF and OF and the opposite is also true.
Table 7.18: Covariances of the ICT Adoption Model
Estimate S.E. C.R. P Label

-
PF <--> ECF -.242 .083 .003 par_5
2.936
PF <--> OF .944 .112 8.429 *** par_6
O <--> ECF -.685 .102 - *** par_7
278
Estimate S.E. C.R. P Label

F 6.722
In conclusion of phase one of hypothesis testing, it was observed that
determinants/predictor variables PF and OF as measured by the observed variables
D3, D5, D10, E7 and E11 accounted for 47.8% of the variance of (adoption) AD and
16.5% of the variance of (SME Performance) SMEP. This indicated that the
variables were able to explain ICT adoption well, while at the same time giving
reasonable explanation for SME performance. In addition, standardised regression
weights between PF and AD (0.391) and between OF and AD (0.316) proved that
there is a causal link between PF and AD and also between OF and AD. The causal
link between AD and SMEP was also confirmed by a standardised regression weight
of 0.407.
7.9 Hypotheses Testing-Moderating Hypotheses
The second phase of hypothesis testing, involved tests for the moderating effect of
the government policies on adoption and use of ICTs in SMEs. However, prior to
that, predictor variable ECF had to be removed from the ICT Adoption model as it
was proved not to be significant (See section 7.7). The ICT adoption model after
removal of the variable ECF is shown in Figure 7.9 and is hereby referred to as the
Final ICT Adoption Model. Table 7.19 give a summary of the Final ICT adoption
Model parameters.
279
It should be noted that after the removal of ECF the model fitted the data much better
than before: CMIN=61.266, CMIN/DF=1.532, p-value=0.017 which is significant at
the level of 0.05, Bollen-Stine bootstrap p-value=0.289 which is not significant at the
level of 0.05 and thus indicating that the model fits the data very well. Other fit
measures include RMSEA=0.04, TLI=0.984, CFI=0.989, NFI=0.968, GFI=0.969 and
AGFI=0.949. All these measures indicated a very good fit of model to the data. In
addition to the fit measures provided for the final ICT adoption model, analysis of
pattern coefficients, standardised residual covariances and the sample correlation did
not reveal any problem.
Table 7.19: Parameter Summary for the Final ICT Adoption Model
Weights Covariances Variances Means Intercepts Total

Fixed 17 0 0 0 0 17
Labelled 0 0 0 0 0 0
Unlabelle
10 1 15 0 0 26
d
Total 27 1 15 0 0 43
In order to test for the moderating effect of government policies on adoption and use
of ICTs in SMEs, multiple group analysis was conducted using AMOS. The
moderators had three variables namely, support programs, taxes and tariffs and
support infrastructure. Hypotheses H5a, H5b and H5c were tested with three
moderating variables and thus the impact of moderators on the influence of
determinants PF and OF toward ICT adoption AD and consequently on the post
adoption SME performance was established. Hypothesis testing was conducted by

280
investigating whether there were any significant differences across groups. If any
significant differences were found across the groups, then, that suggested the
moderator had an effect on the predictors (Arbuckle, 2009; Byrne, 2010; Kripanont,
2007). If, however it was found that there were no significant differences between or
among the groups, this suggested the moderator had no effect on the predictor
(Arbuckle, 2009; Byrne, 2010; Kripanont, 2007).
Figure 7.9: The Final ICT Adoption Model with Unstandardised Estimates

281
NFI=0.968, GFI=0.969, AGFI=0.949.
According to the literature (Kripanont, 2007; Byrne, 2010) testing of a moderating
hypothesis in a multi group analysis involve three steps as follows:
i. First the parameter estimates are computed separately for both groups.
ii. Second, the paths in the model are simultaneously estimated for both groups
iii. The third and last step is to estimate the constrained model. This is done by
constraining the parameter estimates in both groups to be equal. The
parameter estimates across both groups are specified as invariant. If the chi-
square difference tests reveal a significant difference across the baseline and
constrained models, then it might be concluded that the moderating
hypothesis is supported (accepted). Kripanont (2007) note that, this initial
test provides evidence that at least one or more of the direct effects differs
significantly across the subgroups. If differences across the groups are
detected, Kripanont (2007) further recommends for estimation of a series of
models in order to identify the specific paths that differ significantly across
the groups.
282
7.9.1 Effect of Support Programs on ICT Adoption
In this section, an investigation is made in order to establish if support programs
provided by the Government to SMEs has any significant effect on the influence of
PF and OF on AD. The hypotheses to be tested (H5a) was stated as:
“The influence of personal factors and organisation factors (predictors) on
ICT adoption in SMEs is moderated by support programs provided by the
government to SMEs”.
Two groups were created from the dataset. One group supported the notion that
there were enough support programs by the government to support ICT adoption by
SMEs in Tanzania. The second group were of the opinion that there were no enough
support programs by the government to support SMEs adopt ICTs. Due to the low
number of respondents (37 cases) who supported the notion that there were enough
support programs, respondents who were neutral were regarded as supporting the
idea. This was done in order to meet the minimum sample size for analysis (Hair
et.al. 2010; Kripanont, 2007; Tabachnick and Fidell 2013). Table 7.20 shows the
frequency distribution of respondents to the question regarding support programs by
the government.
Table 7.20: Frequency Distribution for Government Support Programs
G1-There are no enough support programs by the government to support ICTs

adoption in SMEs
Percent
Valid Strongly disagree 18 5.3 5.3 5.3
283
Disagree 19 5.6 5.6 10.9

Neutral 65 19.1 19.1 29.9
Agree 91 26.7 26.7 56.6
Strongly agree 148 43.4 43.4 100.0
Total 341 100.0 100.0
In addition to the sample size considerations for the two groups, it should be noted
that, the estimation of the baseline model for the two groups made it necessary to re-
specify the Final ICT Adoption model in order to make sure the baseline model fit
the data for both groups (Byrne, 2010).
Figure 7.10 depict the path diagram of the baseline model (unconstrained model)
with unstandardised estimates for a group of respondents that disagree to the
question that “there are no enough support programs by the government to support
ICT adoption by SMEs in Tanzania”. This group had a total of 102 respondents
which is a summation of strongly disagree + disagree + neutral = 18+19+65 (see
Table 7.20) Figure 7.11 show the path diagram of the baseline model Unstandardised
estimates for a group of respondents (group 2). This group had a total of 239 cases,
which is a summation of those who strongly agree + agree (see Table 7.20). The
respondents of this group agreed with the question that “there are no enough support
programs by the government to support ICT adoption by SMEs in Tanzania”.

284
Group Analysis-Adequate Support Programs
CMIN=123.725, DF=78, CMIN/DF=1.586, p-value=0.001,
NFI=0.941, GFI=0.941, AGFI=0.900.

285
Group Analysis-Inadequate Support Programs
Bollen-Stine bootstrap p-value=0.234, RMSEA=0.042, TLI=0.968,
CFI=0.977, NFI=0.941, GFI=0.941, AGFI=0.900.

286
The baseline model for this analysis had the following fit measures: CMIN=123.725,
CMIN/DF=1.586, p-value=0.001, which is significant at the level of 0.05. Bollen-
Stine bootstrap p-value=0.234, which is not significant at the level of 0.05 and thus
indicating that the model fits the data very well for both groups. Other fit measures
include RMSEA=0.042, TLI=0.968, CFI=0.977, NFI=0.941, GFI=0.941 and
AGFI=0.900. All these fit statistics indicated that the model fitted the data for both
groups very well.
These results indicated that the two groups use the same path diagram albeit with
possible different parameter estimates, as suggested by Figure 7.10 and Figure 7.11.
However, the differences in parameter estimates from the baseline model (as
depicted by Figure 7.10 and Figure 7.11) were not enough by themselves to justify
the acceptance or rejection of the hypothesis (Arbuckle, 2009; Byrne, 2010;
Kripanont, 2007). Therefore, additional tests were necessary in order to determine
the statistical significance of the differences (Arbuckle, 2009; Byrne, 2010;
Kripanont, 2007).
Structural weights models and measurement models (constrained models) for both
groups were generated by using multiple group analysis. According to the literature
(Kripanont, 2007; Arbuckle, 2009; Byrne, 2010), constrained models presents the
parameter estimates in measurement and structural weights that were constrained to
be equal in both groups.

287
Figure 7.12 and Figure 7.13 show structural weights model for the two groups. It is
seen from these estimates that the two groups had the same structural weights
estimates as depicted by Figure 7.12 and Figure 7.13. Fit statistics for the structural
weights model were recorded as: CMIN=140.669, CMIN/DF=1.599, p-value=0.000,
(which is significant at the level 0.05), Bollen-Stine bootstrap p-value =0.154 which
is not significant at the level of 0.05 and thus indicating that the model fits the data
very well for both groups. Other fit measures were RMSEA=0.042, NFI=0.933,
TLI=0.967, CFI=0.973, GFI=0.933 and AGFI=0.899. These additional fit statistics
also indicated that the model fits the data for both groups very well.
The comparison of the baseline model and the measurement model (constrained
model) did not show any significant differences across the models as indicated by the
increase in fit measures. That is, DF increase = 7, CMIN increase = 9.670 and p-
value = 0.208 which is not significant at the level 0.05. In addition, the chi-square
difference test also revealed a non-significant difference across the baseline model
and the constrained (structure weight) model: DF increase = (88-78) = 10, CMIN
increase = (140.669-123.725) = 16.944, p-value = 0.000 (which is significant at the
level of 0.05) (See nested model comparisons in Appendix IV-A). Finally, looking
at the regression weights for the two groups as depicted in Table 7.21a and Table
7.21b it was observed that the regression weights in all direct paths in the model was
the same for both groups.

288
Multiple Group Analysis-Adequate Support Programs
Bollen-Stine bootstrap p-value =0.154, RMSEA=0.042, NFI=0.933, TLI=0.967,
CFI=0.973, GFI=0.933, AGFI=0.899

289
Multiple Group Analysis-Inadequate Support Programs
Bollen-Stine bootstrap p-value =0.154, RMSEA=0.042, NFI=0.933, TLI=0.967,
CFI=0.973, GFI=0.933 and AGFI=0.899
Consequently, based on this analysis, the hypothesis (H5a) that support programs
provided by the government to SMEs, has a moderating effect on the influence of PF
and OF on ICT adoption AD was rejected.

290
Table 7.21a: Regression Weights (Structural weights)-Inadequate Support
Estimat
S.E. C.R. P Label
e
**
AD <--- OF .102 .029 3.558 b2_1
*
**
AD <--- PF .178 .028 6.416 b3_1
*
SME **
<--- AD .591 .088 6.688 b1_1
P *
*** A p value is statistically significant at the 0.01 level (two-tailed)
Table 7.21b: Regression Weights (Structural weights)-Adequate Support
Estimat
S.E. C.R. P Label
e
**
AD <--- OF .102 .029 3.558 b2_1
*
**
AD <--- PF .178 .028 6.416 b3_1
*
SME **
<--- AD .591 .088 6.688 b1_1
P *
*** A p value is statistically significant at the 0.01 level (two-tailed)
7.9.2 Effect of Taxes and Tariffs on ICT Adoption
To test for the moderating effect of taxes and tariffs on the adoption and use of ICTs
in SMEs, the sample was split into two groups. The first group had 115 cases.
These were the respondents that do not believe that the taxes and tariffs charged by
the government on businesses was an obstacle in their adoption of computers and
internet technologies. The second group had 226 cases. These were respondents
who acknowledged that high taxes and tariffs charged by the government on their
businesses was an obstacle in their ICT adoption situations. Note that, in order to
291
meet the minimum sample for analysis, respondents who were neutral to the question
“High tariffs’ and taxes hinder computer and Internet adoption in Tanzania SMEs”
were assumed to belong to the group of respondents who disagreed with the question
(See Table 7.22). The hypothesis (H5b) to be tested was started as:
ICT adoption in SMEs is moderated by the taxes and tariffs charged by the
government on the SMEs.
Table 7.22: Frequency Distribution regarding Taxes and Tariffs
G5-High tariffs’ and taxes hinder computer and Internet adoption in Tanzania SMEs.
Percent
Strongly disagree 18 5.3 5.3 5.3
Disagree 29 8.5 8.5 13.8
Neutral 68 19.9 19.9 33.7
Valid
Agree 85 24.9 24.9 58.7
Strongly agree 141 41.3 41.3 100.0
Total 341 100.0 100.0
Figure 7.14 and Figure 7.15 show the path diagrams of the baseline model with
unstandardised estimates for the two groups respectively. The overall fit statistics for
the baseline (unconstrained) model were: CMIN=117.928, CMIN/DF=1.512, p-
value=0.002, which is significant at the level of 0.05, Bollen-Stine bootstrap p-
value=0 .279 which is not significant at the level of 0.05. This indicated that the
model fitted the data very well for both groups. With RMSEA=0.039, NFI=0.941,
TLI=0.970, CFI=0.979, GFI=0.943 and AGFI=0.903, the additional fit measures also
indicated that the model fitted the data very well for both groups. In addition, the fit
292
measures and the estimates of shown in Figure 7.14 and Figure 7.15 indicated that
the two groups used the same path diagram but possibly with different parameter
estimates. To confirm this further analysis was carried out.
Group Analysis-Taxes not an Obstacle
CMIN=117.928, CMIN/DF=1.512, p-value=0.002,
Bollen-Stine bootstrap p-value=0 .279, RMSEA=0.039, NFI=0.941, TLI=0.970,
CFI=0.979, GFI=0.943 and AGFI=0.903

293
Group Analysis-Taxes an Obstacle
CMIN=117.928, CMIN/DF=1.512, p-value=0.002, , Bollen-Stine bootstrap p-
value=0 .279, RMSEA=0.039, NFI=0.941, TLI=0.970, CFI=0.979, GFI=0.943,
AGFI=0.903
294
In confirmation, constrained models were generated and evaluated. Figure 7.16 and
Figure 7.17 show the structural weight models for the two groups respectively. The
fit measures for the structural model was found to be: CMIN=132.624,
CMIN/DF=1.507, p-value=0.001, which is not significant at the level of 0.05,
Bollen-Stine bootstrap p-value=0.214, which is not significant at the level of 0.05,
RMSEA=0.039, NFI=0.933, TLI=0.970, CFI=0.976, GFI=0.937, AGFI=0.905.
These fit statistics indicated that the structural model fitted data very well for both
groups.
Multiple Group Analysis-Taxes not an Obstacle
295
Bollen-Stine bootstrap p-value=0.214, RMSEA=0.039, NFI=0.933, TLI=0.970,
CFI=0.976, GFI=0.937, AGFI=0.905.
Multiple Group Analysis-Taxes an Obstacle

296
CFI=0.976, GFI=0.937, AGFI=0.905.
The comparison between the baseline model (unconstrained model) and the
measurement weights model revealed no significant difference across the models.
Note the increase in degree of freedom =7, CMIN increase=8.546 and the p-
value=0.287 which is not significant at the level of 0.05. On a similar note, the chi-
square difference test also revealed a non-significant difference across the baseline
model and the constrained (structure weight) model, as suggested by the comparison
of the baseline model with the structural weights model: DF increase = (88-78) = 10,
CMIN increase = (132.624 - 117.928) = 14.696 and p-value = 0.144 which was not
significant at the level 0.05. (For more details see nested model comparisons in
Appendix IV-B). In addition to this analysis, it was also observed from Figure 7.16
and Figure 7.17 as well as Table 7.23a and Table 7.23b that, direct paths of the
model had the same regression weights for two groups. Therefore, the hypothesis
(H5b) that “The influence of personal factors and organisation factors (predictors) on
ICT adoption in SMEs is moderated by the taxes and tariffs charged by the
government on the SMEs” was rejected since there was no statistical difference
across the two groups.
Table 7.23a: Regression Weights (Structural weights)-Taxes not an Obstacle
Estimat
S.E. C.R. P Label
e
**
AD <--- OF .104 .029 3.612 b2_1
*
**
AD <--- PF .162 .028 5.753 b3_1
*
SME **
<--- AD .582 .097 5.984 b1_1
P *
297
Table 7.23b: Regression Weights (Structural weights)-Taxes an Obstacle
Estimat
S.E. C.R. P Label
e
**
AD <--- OF .104 .029 3.612 b2_1
*
**
AD <--- PF .162 .028 5.753 b3_1
*
SME **
<--- AD .582 .097 5.984 b1_1
P *
7.9.3 Effect of Infrastructure on ICT Adoption
The hypothesis (H5c) regarding the impact of support infrastructure provided by the
government was stated as:
ICT adoption in SMEs is moderated by the support infrastructure provided by
the government to SMEs”
To test this hypothesis, the sample was split into two groups. The first group had
126 cases. These were the respondents who were of the opinion that the
infrastructures that support businesses in ICT adoption situations were adequate. The
second group had 215 cases. The respondents in this group were of the opinion that
the infrastructures that support businesses in ICT adoption situations were inadequate
298
In order to meet minimum sample for analysis, respondents who were neutral were
considered to be part of the first group, like in previous cases in section 7.8.1 and
section 7.8.2. Thereafter the groups were analysed in order to establish if there were
any significant differences across the groups.
Table 7.24: Frequency Distribution on Support Infrastructure
G8-Lack of enough infrastructures to support ICTs adoption by SMEs

Percent
Strongly disagree 25 7.3 7.3 7.3
Disagree 24 7.0 7.0 14.4
Neutral 77 22.6 22.6 37.0
Valid
Agree 90 26.4 26.4 63.3
Strongly agree 125 36.7 36.7 100.0
Total 341 100.0 100.0
Figure 7.18 and Figure 7.19 show the path diagrams of the baseline models with
unstandardised estimates for the two groups respectively. Model fit statistics for the
baseline model were recorded as: CMIN=120.871, CMIN/DF=1.550, p-value=0.001,
which is significant at the level of 0.05, Bollen-Stine bootstrap p-value=0.289, which
is not significant at the level 0.05. This indicated a good fit of the model to the data
for both groups. Additional fit measures, RMSEA=0.040, NFI=0.943, TLI=0.970,
CFI=0.979, GFI=0.940, AGFI=0.898, also indicated that the model fitted the data
very well for both groups.

299
The fit measures and the estimates as shown in Figure 7.18 and Figure 7.19 indicated
that the two groups use the same path diagram but with different parameter
estimates. Since this information was not conclusive for the purpose accepting or
rejecting the hypothesis (Kripanont, 2007; Arbuckle, 2009; Byrne, 2010) analysis of
the constrained models was made.
Group Analysis-Adequate Infrastructure

300
CFI=0.979, GFI=0.940, AGFI=0.898
Group Analysis-Inadequate Infrastructure

301
CFI=0.979, GFI=0.940, AGFI=0.898
Figure 7.20 and Figure 7.21 show the structural weight models for the two groups
respectively. The fit measures for the structural model was found to be:
CMIN=132.495, CMIN/DF=1.506, p-value=0.002, which is significant at the level
0.05, Bollen-Stine bootstrap p-value=0.289, which is not significant at the level 0.05,
indicating a good fit of data to the model for both groups. Other fit measures,
RMSEA=0.039, NFI=0.938, TLI=0.972, CFI=0.978, GFI=0.936, and AGFI=0.904
also indicated that the model fitted the data very well for both groups.
302
Multiple Group Analysis-Adequate Infrastructure
CFI=0.978, GFI=0.936, and AGFI=0.904
Multiple Group Analysis-Inadequate Infrastructure

303
CFI=0.978, GFI=0.936, and AGFI=0.904
Like in the other moderating hypotheses tests, the comparison between the baseline
model (unconstrained model) and the measurement weights model revealed no
significant difference across the models. Note the increase in degree of freedom =7,
CMIN increase = 8.119 and the p-value=0.322 which is not significant at the level of
0.05. On a similar note, the chi-square difference test also revealed a non-significant
difference across the baseline model and the constrained (structure weight) model, as
suggested by the comparison of the baseline model with the structural weights
model: DF increase = (88-78) = 10, CMIN increase = (132.486 - 120.871) = 11.615
and p-value = 0.312 which was not significant at the level 0.05. (For more details see
nested model comparisons in Appendix IV-C).
Moreover, it was observed that, the direct paths of the model had the same regression
weights for two groups as depicted by Table 7.25a and Table 7.25b respectively.
Table 7.25a: Regression Weights (Structural weights)-Adequate Infrastructure
Estimat
S.E. C.R. P Label
e
**
AD <--- OF .121 .031 3.946 b2_1
*
**
AD <--- PF .164 .028 5.910 b3_1
*
SME **
<--- AD .525 .091 5.774 b1_1
P *
304
Table 7.25b: Regression Weights (Structural weights)-Inadequate

Infrastructure
Estimat
S.E. C.R. P Label
e
**
AD <--- OF .121 .031 3.946 b2_1
*
**
AD <--- PF .164 .028 5.910 b3_1
*
SME **
<--- AD .525 .091 5.774 b1_1
P *
Thus, as a result of this analysis, the hypothesis (H5c) that support infrastructure has
a moderating effect on the influence of PF and OF on ICT adoption was rejected.
7.10 Interpretations of the Developed ICT Adoption Model
After model development and hypothesis testing, it was necessary to interpret the
model findings and show how the model can guide Tanzania SMEs in ICT adoption
process. It should be pointed out that all points raised in section 4.1 are important
and therefore must be taken into consideration in adoption situations by SMEs.
However, the developed model informs on the critical factors of ICT adoption in the
context of Tanzania. Model interpretations can be viewed from two perspectives: the
owner-manager perspective and the government perspective.
From the owner-manager perspective, generally the literature shows that the owner-
manager of small business make all key decisions on how the firm conducts its
business including the decisions for acquiring new technologies for the firm. It is
therefore correct, to point out that, the rate of ICT adoption and effective utilisation
305
in SMEs depends on the commitment and support of owner-manager. Short of that,
ICT adoption and use in SMEs remains low.
The developed model specifically identifies critical factors as the perceptions of the
owner-manager towards the benefits of ICTs to his/her business, business strategies
external pressures and costs of ICT systems. These factors alone explain ICTs
adoption by 47.8% and SMEs performance by 16.5%. Therefore, based on these
observations, for successful adoption, owners-managers must approach adoption of
ICTs in SMEs as explained in the subsequent paragraphs.
Owners-managers must have the right education needed to manage their firms since
adoption of ICTs is not a substitute for poor management skills. In additional to the
managerial skills, owners-managers must strive to develop their ICT knowledge and
skills. That will enable them to make informed decisions in adoption situations like
what type of ICTs are needed by their firms, the skills needed by their employees for
success and the actual benefits to their firms. It should be noted that, education
impacts on the perceptions which have been shown by research findings to be of
great importance in adoption situations.
Owners-managers must ensure that ICTs adoptions are aligned with the business
strategy of their firms. Research findings show that, adoption of ICTs with the aim
of supporting the firm’s strategy is vital for successful adoption and use of ICTs.
306
Owners-managers must be smart enough to realise external pressures to adopt when
they happen and thus take initiatives that will ensure successful adoption. More
often, SMEs fail to identify the external pressure to adopt and thus fail to use it to
their advantage. Costello (2009) argues that SMEs should view external pressure as
strategic method or a driver.
Owners-manager must make sure that their firms have enough skilled labour force
that is capable of using ICTs effectively. This can be achieved by employing ICT
knowledgeable personnel and or by promoting the awareness of ICTs potentials by
using staff training programs to develop the ICT skills of employees. Effective
utilisation of ICTs enables the identification of gaps/problems thus triggering
innovations. Research findings show that innovativeness is key to competitiveness
and success of a firm.
From the government point of view, good policies were found from the literature to
be important for successful adoption of ICTs in SMEs. However, empirical findings
suggest that government policies aimed at helping the SMEs to adopt ICTs are not
effective. Given the importance of the SMEs sector in the economic development of
the country, the government of Tanzania needs to address the issues raised by
research findings by developing new policies with enough initiatives and incentives
that promote ICT adoption and use within the SME sector. Based on research
findings, specific areas that need attention are explained in subsequent paragraphs.
307
The policy regarding government support to SMEs need to be reviewed since
research findings show the current one is not effective. SMEs often need assistance
in improving managerial and technological skills, financial assistance and access to
market information. Therefore, the government must look into how support
institutions that offer necessary training and support programs to SMEs are created
and coordinated in order to achieve their objectives.
The tax policy needs to be redesigned such that a tax system with enough incentives that
stimulates ICT adoption, competitiveness and growth of SMEs is introduced. Research
findings show that SMEs consider the present tax system an inhibitor of ICT adoption
and growth. It is important that the level of taxation set by the new tax system is
friendly and does not stifle the running of SMEs. For example, laws that results in
heavy costs of compliance by SMEs must be revised or abolished. Tax rebates may
also be considered as incentives for performing SMEs.
Research findings show that support infrastructures are not adequate. This may have
made ICT solutions unaffordable to many SMEs. Therefore, based on this, the
government of Tanzania must continue to develop and maintain critical
infrastructures that support use of ICTs in SMEs. That way service provision of ICT
solutions will be affordable to many SMEs and thus stimulate adoption and use of
ICTs.
Research findings suggested that, lack of laws and regulatory frameworks inhibits
the adoption of e-commerce by SMEs in Tanzania. The government of Tanzania

308
needs to enact laws and regulatory frameworks that to guide the SME sector in the
positive direction. Generally speaking, the author noted that, it is the duty of the
government to create and promote favourable environment that supports business
growth and the wellbeing of its people.
7.11 Chapter Summary
In this chapter, the ICT adoption model for Tanzania SMEs was developed, tested
and the results of SEM data analyses were presented and discussed along with the
results of testing hypotheses. Thereafter, the interpretation of the developed model
was carried out and summarised in section 7.9. Model development process
involved several stages. In each stage, analysis of all research variables that had
passed preliminary analysis was done in order to determine the usefulness of each
variable in the model. To begin with, construct reliability was carried out and the
SMC of all variables was determined. All variables with SMC values less than 0.5
were removed from further analysis since they were considered useless for any
meaningful analysis (Kripanont, 2007). Construct reliability was followed by
discriminant validity of latent constructs. In this test, all variables that contravened
critical statistical assumptions for SEM analysis were omitted (deleted) from the final
analysis (Arbuckle, 2009; Byrne 2010; Hair et.al. 2010; Kripanont, 2007;
Schumacker and Lomax 2010; Tabachnick and Fidell, 2013). For example, for all
variables that were highly correlated, (correlation from 0.8 and above) one variable
was deleted from the pair. The highly correlated variables indicated the presence of
multicolinearity in data (Arbuckle, 2009; Byrne 2010; Hair et al., 2010; Kripanont,
2007; Schumacker and Lomax 2010; Tabachnick and Fidell, 2013). In a similar
309
way, all variables were tested and appropriate measures were taken to ensure that
they conformed to SEM requirements. This way, the reliability and validity of the
model constructs and variables was firmly established before the model could be
fitted to data and tested.
Model fitting was done in two stages as recommended by the literature (Kripanont,
2007; Byrne 2010). First the measurement models were fitted and analysed before
the structural model was fitted and analysed. This helped to easily identify the
source of poor fit in the final model. During model fitting, it was necessary to
remove and or to respecify model parameters and variables in order to achieve a
model that fitted the data well. Thereafter the model was used to test the hypotheses.
Two sets of hypotheses were tested in two stages. In the first case, direct path
hypotheses were tested without the effect of moderating variables. Standardised
regression weights of direct paths were used to judge the hypotheses. That is,
hypotheses with significant standardised regression weights were accepted while
those with non-significant standardised regression weights were rejected. In addition
to the standardised regression weights, SMC and covariances were also used in the
description of the structural model. The results of this testing confirmed the final ICT
adoption model.
The second stage of hypothesis testing involved the test of moderating hypotheses.
The effect of moderating variables on the predictor variables was tested using
simultaneous multiple group analysis in AMOS. The chi-square difference test

310
between the baseline model and structure weights model was used as the major
criteria for rejecting or accepting the hypotheses.
Table 7.26 and Table 7.27 give a summary of the hypothesis testing results. From
Table 7.26 it is seen that five direct path hypotheses were accepted and three
hypotheses were rejected. The rejected hypotheses were all related to the economic
factors. Table 7.27 show the summary of moderating hypotheses. All of the
moderating hypotheses in this research were rejected. Table 7.28 gives a summary
of the implication of the rejected or accepted hypotheses. Detailed account of
research findings is provided in chapter eight.
Table 7.26: Summary of Direct Hypotheses Results
SN Hypothesis Statement Standardised P -Value Result

Regression
Weight
H1a Personal factors have a direct
positive significant relationship 0.632 0.000 Accepted
with ICT adoption in SMEs.
H1b Organizational factors have a
direct positive significant 0.644 0.000 Accepted
relationship with ICT adoption in
SMEs.
H1c Economic factors have a direct
positive significant relationship 0.326 0.000 Accepted
H2a Personal factors and
Organizational factors have a 0.390 0.000 Accepted
direct positive significant 0.345 0.000
relationship with ICT adoption in
SMEs.
H2b Personal factors and Economic
311
factors have a direct positive 0.591 0.000 Accepted

significant relationship with ICT 0.210 0.000
adoption in SMEs.
H2c Organizational factors and
Economic factors have a direct 0.699 0.000 Rejected
positive significant relationship 0.045 0.539
H3 Personal factors, Organizational
factors and Economic factors have 0.391 0.000 Rejected
a direct positive significant 0.316 0.010
relationship with ICT adoption in -0.086 0.252
SMEs.
H4 ICT adoptions and use have a
direct positive significant 0.407 0.000 Accepted
relationship with SMEs
performance.
Table 7.27: Summary of Moderating Hypotheses Results
SN Hypothesis Statement Result

difference test
H5a The influence of personal factors and
organisation factors (predictors) on ICT Non-significant Rejected
adoption in SMEs is moderated by the
taxes and tariffs charged by the
government on the SMEs.
H5b The influence of personal factors and Non-significant Rejected
organisation factors (predictors) on ICT
adoption in SMEs is moderated by
support programs provided by the
government to SMEs.
H5c The influence of personal factors and Non-significant Rejected
organisation factors (predictors) on ICT
adoption in SMEs is moderated by the
support infrastructure provided by the
government to SMEs.
312
Table 7.28: Summary of the Implication of the Tested Hypotheses
Hypothesis Result Explanation and Implications

H1a Accepted Personal Factors Construct: The implication of this
hypothesis is that, the higher the perceived benefits of
ICTs by the owner-Manager of the firm the more likely
the adoption and use of ICTs by that firm.
H1b Accepted Organisation Factors: The implication of this
hypothesis is that, in most cases SMEs adopt and use
ICTs as a result of external pressure from customers as
well as due to strategic decisions.
H1c Accepted Economic Factors Construct: The implication of this is
that, high acquisition and operation costs prevent SMEs
from adopting and using ICTs.
H2a Accepted Personal Factors and Organisation Factors Constructs:
The implication of this hypothesis is that perceived
benefits, customer pressure (external pressure) and
marketing strategy contribute significantly and
positively to the adoption and use of ICTs in SMEs.
H2b Accepted Personal factors and Economic Factors Constructs:
The implication of this hypothesis is that, perceived
benefits, together with low acquisition and operation
costs factors contribute significantly and positively to
the adoption and use of ICTs in SMEs.
Hypothesis Result Explanation and Implications

H2c Rejected Within the model framework, Economic Factors
Construct was not significant.
H3 Rejected Within the model framework, Economic Factors
Construct was not significant.
H4 Accepted Adoption and use of ICTs (Computers and Internet)
have a direct significant positive influence on SME
Performance
H5a Rejected The fact that “Support Programs” failed to moderate the
influence of predictors (PF and OF) on ICT adoption by
SMEs suggests that, support programs provided by the
government to the SME sector are inadequate and or
ineffective.
H5b Rejected The fact that Taxes and Tariffs failed to moderate the
influence of predictors (PF and OF) on ICT adoption by
SMEs suggests lack of effective tax incentives to SMEs
(See section 5.9.1)
313
H5c Rejected The fact that “Support Infrastructure” failed to

moderate the influence of predictors (PF and OF) on
ICT adoption by SMEs suggest that ICT support
infrastructure is still inadequate. (see section 2.3)
CHAPTER EIGHT
CONCLUSIONS AND RECOMENDATIONS
8.1 Introduction
Small and Medium Sized Enterprises are documented as the basis of most economies
worldwide. However, the SMEs sector in Tanzania is characterised by torpid micro
enterprises that have little incidence of growth. In addition, there is proof to the
effect that SMEs in Tanzania are not exploiting ICTs to their advantage despite the fact
that, in today’s knowledge economy, ICT adoption and use has become essential for any
business survival and growth. Based on the observed problems, i t was theorised that,
314
with an adoption model that informs on key adoption issues in the context of
Tanzania, many SMEs will adopt ICTs, innovate, and become competitive in today’s
digital economy.
Thus, this research sought to develop and test an ICT adoption model for SMEs in
Tanzania. The theories relevant to ICT adoption in SMEs and related fields were
examined in order to determine pertinent issues for the development of the ICT
adoption model. A conceptual model based on the literature was then proposed.
Using multiple research design, methodological triangulation and a survey as the
main data collection strategy, data was collected from the service and general trading
sectors from the mainland Tanzania. Structural Equation Modelling (SEM)
techniques were then used to develop, test and validate an ICT adoption model in
line with the research objectives.
This chapter concludes this research work and is arranged as follows: First,
conclusions based on key research findings are briefly explained. Secondly, the
contributions of this work to the knowledge body are mentioned. Thirdly, based on
research findings, recommendations for successful adoption of ICTs in SMEs are
given to key stakeholders. Finally, limitations of the research and future research are
outlined.
315
8.2 Conclusions
8.2.1 The ICT Adoption Model
The main objective of this research was to develop a model for guiding Tanzania
SMEs in the ICT adoption process. Therefore, a Structural Equation Model that
identified six constructs and their critical factors was developed. The developed
model constructs include: (i) Three exogenous latent constructs, namely Personal
Factors, Organisation Factors and Economic Factors which were identified as the
direct determinants of ICTs adoption in SMEs; (ii) One moderating construct namely
Government Policies that explains the moderating influence of government policies
on ICT adoption by Tanzania SMEs and (iii) Two endogenous constructs namely
ICT Adoption and SMEs Performance. ICT adoption construct explains ICT
adoption and use while SMEs Performance construct explains the effect ICT
adoption has on SMEs performance.
The study established the sequence of model operations as follows: First the most
important factors that influence ICT adoption in SMEs were identified and
categorised. This was then followed by theoretical analysis to establish the
relationship between the identified categories (constructs) of factors and the effect
they have on ICT adoption by SMEs. Finally, using Structural Equation Modelling
(SEM) analysis techniques, the strength and influence of constructs and variables
were determined. This process informed on the critical successful adoption issues
for consideration prior and after adoption of ICTs. On the other hand, government
policies and regulations for guiding the SME sector were also tested for their
effectiveness.
316
Initially 64 different variables were identified for model development. However,
model testing and validation using research data resulted in a parsimonious model
with 13 variables as follows: 3 variables for measuring Personal Factors; 2 variables
for measuring Organisation Factors; 2 variables for measuring Economic Factors; 3
variables for testing the effectiveness of Government Policies; 3 variables that
explain ICT Adoption and 3 variables for measuring post adoption SMEs
Performance. The research further established that, the Personal Factors,
Organisation Factors and Economic Factors predictor variables accounted for 47.8%
of the variance of ICT adoption and 16.5% of the variance of SME Performance.
This indicates that the identified variables explain ICT adoption very well, while at
the same time giving reasonable explanation for SME performance. Thus, the model
was proved to be a reliable tool for guiding ICTs adoption process within the context
of Tanzania. Conclusions based on specific findings from the model testing and
descriptive analyses are presented in subsequent sections.
8.2.2 Factors Influencing ICT Adoption by Tanzania SMEs
Theoretical analysis of the factors influencing ICT adoption by Tanzania SMEs was
the first specific objective and the initial step in ICT adoption model development.
This study established four main categories (constructs) of factors that influence ICT
adoption by Tanzania SMEs as follows:
i. Personal Factors: These are the factors related to the owner-manager of the
SMEs. The study established that the education level of the owner-manager,
317
owner-manager knowledge of IT and perceptions of the owner-manager
(perceived benefit) towards the benefit of ICTs to his/her business are the
most important factors in this category.
ii. Organisation Factors: These are factors that are directly related to the
organisation. The study established that, the firm age, the firm size, business
sector, organisational readiness, external pressure to adopt and the SME’s
business strategy are the most important factors in this category.
iii. Economic Factors: These are factors that are related to post start up firm
financing. The study established that access to financial support and cost of
ICT systems, are the most important factors in this category.
iv. Government Policies: These are government interventions and incentives
aimed at helping SMEs. The study established that support programmes,
taxes and Tariffs and support infrastructure are the most important factors in
this category.
Specific ICTs studied in this research were the computer and Internet technologies.
The internet technologies included access (connectivity) to the Internet, adoption and
use of company websites and adoption and use of email. In the developed model,
ICT Adoption is identified as an endogenous construct. In addition, a construct that
measures post adoption SMEs performance was identified. The study established
318
that return on investment measures and innovations are the most important factors in
this category.
8.2.3 The Relationship between the Relevant Factors and their Effect on ICT
Adoption by Tanzania SMEs
The second specific objective was to establish the relationship between the relevant
factors and their effect on ICT adoption. Theoretical analysis of research variables
suggested that Personal factors, Organisation factors and Economic factors are direct
determinants (predictors) of ICT adoption in SMEs while Government Policies were
theorised to have a moderating effect on ICTs adoption by SMEs. Further analysis
of model variables using survey data and SEM techniques revealed that:
i. Personal Factors have a direct positive significant relationship with ICT
adoption in SMEs.
ii. Organisation Factors have a direct positive significant relationship with ICT
adoption in SMEs.
iii. Economic factors have a direct positive significant relationship with ICT
adoption in SMEs.
iv. Within the Personal Factors construct, the perceptions of the owner-manager
(Perceived benefits) towards the benefits of ICTs in his/her business were

319
found to be the most significant determinants/predictors of ICT adoption by
Tanzania SMEs. Specific ICTs benefits identified include the perception that:
ICTs can reduce firm’s operational costs, ICTs are an efficient
communication tool with customers and that ICTs are an efficient way of
doing market research.
v. Within the Organisation Factors construct, business strategy and external
pressures were found to be the most significant determinants/predictors of
ICT adoption by Tanzania SMEs. Specific variables identified are marketing
strategy and customer pressures.
vi. Within the Economic Factors construct, access to financial support and cost
of ICT systems were found to be the most significant determinants/predictors
of ICT adoption by Tanzania SMEs. Specific variables identified are
acquisition and operational costs.
vii. Personal Factors and Organisation Factors together have a direct positive
significant relationship with ICT adoption in SMEs.
viii. Personal factors and Economic factors together have a direct positive
significant relationship with ICT adoption in SMEs
ix. Organizational factors and Economic factors together do not have a direct
positive significant relationship with ICT adoption in SMEs.

320
x. Personal factors, Organizational factors and Economic factors together do not
have a direct positive significant relationship with ICT adoption in SMEs.
xi. Personal Factors, Organisation Factors and Economic Factors predictor
variables, that is, perceived benefits, business strategy, external pressure
acquisition and operational costs together accounted for 47.8% of the
variance of ICT adoption.
xii. Taxes and Tariffs charged by the government on the SMEs do not have any
positive influence on the predictors (Personal Factors and Organisation
Factors) of ICT adoption in SMEs.
xiii. Support Programs provided by the government to SMEs do not have any
positive influence on the predictors (Personal Factors and Organisation
Factors) of ICT adoption in SMEs.
xiv. Support Infrastructure provided by the government to SMEs does not
have any positive influence on the predictors (Personal Factors and
Organisation Factors) of ICT adoption in SMEs.
8.2.4 The Status of ICT Adoption and Use by Tanzania SMEs
The third specific objective was to establish the status of ICT adoption and use by
Tanzania SMEs. Research findings show that:

321
i. Adoption and use of ICTs in the Tanzania SMEs sector has rapidly grown
from almost nonexistence in 2004 to very significant levels in year 2014 as
shown by 80.1% of computer adoption, 68.9% of Internet access adoption,
56.6% website adoption and 72.7% of e-mail adoption.
ii. About 55% of SMEs use their ICTs as a strategic communication tool with
their suppliers.
iii. About 58% of SMEs use ICTs as a strategic financial tool.
iv. About 53% of SMEs use their ICTs as a strategic marketing tool.
v. About 36% of SMEs use their ICTs to conduct e-business.
8.2.5 The Influence of ICT Adoption on Tanzania SMEs Performance
The fourth and last specific objective was to establish the influence of ICT adoption
on Tanzania SMEs performance. In this research, it was established that:
i. About 48% of SMEs were innovative as a result of ICT adoption and use.
ii. About 47% of SMEs were competitive as a result of ICT adoption and use.
iii. ICT adoptions have a direct positive significant relationship with SMEs
performance.
322
iv. Personal Factors, Organisation Factors and Economic Factors predictor
variables, that is, perceived benefits, business strategy, external pressure
acquisition and operational costs together accounted for accounted for 16.5%
of the variance of SME Performance.
v. Post adoption use of ICTs to improve firm performance is noted as a big
problem since 53% of SMEs that had adopted were not competitive and 52%
were not innovative.
8.3 Research Contributions
8.3.1 Contribution to Knowledge
One of the areas that were highlighted in knowledge gaps analysis from the literature
was the lack of comprehensive and systematic studies on ICT adoption and use in
SMEs in LDC’s countries south of Sahara including Tanzania. Generally, this
research has added to the existing body of literature in the field of IS research with
particular emphasis on Tanzania. This contributes towards the filling of the observed
knowledge gap.
Specifically, this research identified and addressed the inadequacies of earlier studies
regarding ICT adoption in Tanzania and contributes to the knowledge body of IS
research as follows:
323
i. Most of previous research had concentrated on identifying drivers and
barriers to adoption of ICT and none of them offered a comprehensive
framework to help with strategic intent of SMEs in their adoption situation as
suggested by the literature. The developed model fills in this knowledge gap
and thus it is a major contribution to knowledge and the IS body of literature.
ii. None of the previous studies in Tanzania settings has attempted empirically
to establish in a broad sense the status of how Tanzania SMEs use ICTs after
adoption. As such, lack of scholarly articles on the level of ICT utilisation
amongst Tanzanian SMEs is highly noted. This research contributes to the
existing body of IS literature by providing insights on the post adoption status
of ICT utilisation or use amongst Tanzanian SMEs.
iii. None of the previous research in Tanzania had attempted empirically to
holistically validate the theory that ICT adoption and use can enable small
firms to become competitive as suggested by the literature. This research
managed to show empirically the causal relationship between the use of ICTs
and the competitiveness of the firms and therefore contributing to the existing
body of IS literature within the context of Tanzania.
iv. Literature review and results of this work show that, most of the issues
relating to SMEs are homogeneous. This means that the developed model
can be used for testing adoption issues in other countries with similar socio-
324
economic conditions to Tanzania. This also contributes to knowledge in IS
research.
8.3.2 Methodological Contributions
This research employed multiple research design which involved exploratory,
descriptive and explanatory research. As a result of this research design,
triangulation methodology was employed in different phases of the research.
Methodological contributions of this work are clarified as follows:
i. The author noted that, many of the previous researches that have examined
ICT adoption issues in developing countries have focused on the use of a
single method of data collection. However, use of data and method
triangulation are encouraged in research whenever possible since they are
known to increase the validity of research findings. Therefore, the use of
method triangulation in this research is considered a methodological
contribution to research.
ii. The use of multiple case studies strategy in the exploratory phase of the
research is yet another methodological contribution to research. Multiple
case studies can be used to achieve replication of a single type of incident in
different settings or to compare and contrast different cases. In this research,
the multiple case studies assisted the researcher get a better understanding of
the problems and anxieties faced by SMEs in ICT adoption situations and
post adoption issues from different perspectives. That increased the

325
reliability of the obtained information and helped in the development of an
effective research instrument for the main survey.
iii. Structural Equation Modelling has many advantages over other multivariate
statistical methods. However, the author noted that most of the previous
research used different methods of data analysis other than SEM. The
methodology used in this research incorporates the use of SEM for ICT
adoption model development. This provides alternative guidelines for further
research in this area of study and is thus considered a major contribution.
8.3.3 Practical Contributions
The practical contributions of this research are brought to light by the insights
provided by key research findings to both policy and practise. These are summarised
as follows:
i. Research findings on the impact of the policies aimed at helping the SMEs
revealed the ineffectiveness of government interventions on SMEs in the
name of support programs, taxes and tariffs and support infrastructure.
Therefore, it can be argued that based on the findings of this research, some
areas that need government interventions has been identified.
ii. Research findings inform the owner-manager of SMEs on the key issues to
consider for successful adoption and utilisation of ICTs. Results show that,
the perceptions of the owner-manager towards the benefits of ICTs together
with marketing strategy and customer pressure are the most significant ICT
326
adoption factors in Tanzania SMEs. Post adoption results inform the owner-
manager that innovation using ICTs is vital for the firm to become
competitive.
iii. Research findings informs on the current status of ICT adoption and use in
Tanzania SMEs. This information was lacking in the literature and therefore
these results are a practical contribution.
8.4 Recommendations for Successful ICT Adoption by SMEs
The literature review identified the owner-manager of SMEs, the government and
policy makers as the key stakeholder of the SMEs sector. Therefore, results of this
research have implications that may be very useful to the key stakeholders in their
attempts to facilitate successful adoption and use of ICTs in SMEs. Based on
research findings, recommendations for successful implementations of ICTs in SMEs
are given in the subsequent sections.
8.4.1 General Recommendations for Practice/use
The implications for practice from this research are based on the developed ICT
adoption model and its interpretations. The model informs on the key ICT adoption
factors and therefore its use can result in many SMEs adopting ICTs, innovate and
become competitive. In particular, the model can be used to:
i. Inform those in practice on the key issues to consider prior to ICTs adoption
and after adoption in order to make ICT adoption a success.

327
ii. Test the effectiveness of government efforts aimed at the development of the
SME sector. This will enable the government to make informed decisions
when reviewing its policies.
iii. Study SMEs adoption issues in other countries with similar social economic
conditions. That way the model may provide some very useful information
and further contribute to the IS research knowledge body. Neighbouring
countries to Tanzania like Uganda, Kenya, Malawi, Zambia and Mozambique
can be used for this purpose.
8.4.2 Recommendations for Owner-Managers
The owner-managers of small businesses are known to make all key decisions on
how the firm should conduct its business including the decisions for acquiring new
technologies for the firm. Therefore, the support and commitment of the owners-
manager is very important in any adoption situation. Based on this observation and
other research findings the author recommends the following to the owner-manager:
i. Owner-manager must have the right education and skills needed to manage
their firms. Education impacts on the perceptions of owners-managers which
have been shown by research findings to be of great importance in adoption
situations.
328
ii. Owner-managers must make every effort to acquire or improve their ICTs
knowledge and skills as this will enable them make successful and beneficial
informed decisions in adoption situations.
iii. Prior to adoption, the owner-manager must ensure that his/her firm have
adequate skilled personnel that can use ICTs effectively. Effective use of
ICTs gives a firm the ability to become innovative and competitive.
iv. The owner-managers must ensure that ICTs adoptions are aligned with
business strategies of their firms. Research findings show that marketing
strategy is a critical factor for successful ICT adoption in SMEs.
v. Owner-manager must realise external pressures to adopt when they happen
and thus take initiatives that will ensure successful adoption. Research
findings show that customer pressure is a critical factor for successful ICT
adoption in SMEs.
vi. Owners-managers must offer continuous support to initiatives that will
increase the use of ICT in their firms since the success of the firms depends
on their decisions and visions.
8.4.3 Recommendations to the Government
The SME policy of Tanzania has been in operation for more than a decade. Review
of the literature suggested that that the policy has failed to address many of the
constraints it aimed to deal with. The empirical evidence from this research confirms
329
these observations made by previous work regarding policy failure. Based on
research findings, it is evident that there is a need to develop new policies with
enough initiatives and incentives that promote ICT adoption and use within the SME
sector. In particular, the government (policy makers and support agencies) needs to
do the following:
i. Review the policy regarding government support to SMEs since the present
one has been shown to be ineffective.
ii. Redesign and introduce a new tax system with more incentives for ICT
adoption by SMEs since the present one is considered by SMEs to be an
inhibitor of ICT adoption.
iii. Test the effectiveness of policies from time to time in order to allow for
timely policy reviews.
iv. Continue to develop and maintain critical infrastructures that support use of
ICTs in SMEs and business growth in general.
v. Enact laws and regulatory frameworks that guide the SME sector in the
positive direction. Lack of laws and regulatory frameworks were implied as
inhibitors of e-commerce adoption by SMEs in Tanzania.

330
vi. Through support programs, ensure that SMEs adopt and use ICTs in line with
their business process. This in turn will help to improve the competitiveness
of the SMEs and thus become a part of the present digital economy.
8.5 Research Limitations
Just like in any other research, this research also had a number of limitations and
constraints that need to be addressed.
To begin with, it should be noted that ICT is a broad term covering a number of
different technologies. However, computer and Internet technologies were the only
ICTs considered in this study. It may be interesting to explore for example the impact
of smart phones on small businesses. Smart phones can give small business a quick,
easy and portable access to the Internet, email and other applications using voice,
data and text albeit at a much higher cost. It will be interesting to find out how SME
owners-managers manage to take advantages of this technology.
Another limitation in this research is the fact that only two sectors, the service and
general trading were considered in the study. Inclusion of SMEs in other sectors
might have brought more insights and understanding on ICT adoption issues faced
by SMEs.
Prior to data collection, the minimum sample size was determined and found to be
385 cases. However, only 341 cases of collected data were found to be useful in
analysis. The author noted from the literature that, reduction of minimum sample
331
size has effect of reducing the expected accuracy and generalisability of research
findings. This reduction in the sample size is therefore considered a limitation in this
research.
Different statistical methods and analysis tools have different sample size
requirements. In this research, due to sample size requirements as dictated by
analysis methods and tools, the SMEs from the two sectors were treated as a single
homogeneous group of businesses. This may have had an impact on the
generalisability of research results and thus is a limitation in this research.
The challenges faced during data collection, affected the response rate of the survey
in a way that the requirements of the calculated sample size were not met. Moreover,
without the researcher’s knowledge, respondents could also have hidden some
important information which could possibly have improved the research outcome
and this is also a limitation.
Budget and time constraints prohibited the use of more data analysis tools such as the
Satorra-Bentler robust method to further validate the results. This is also considered
a limitation.
8.6 Direction for Future Research
Taking into consideration the findings of the research, the scope of this study and the
observed limitations, the areas of potential future research were identified and are
outlined below:
332
More research with adequate sample size and other estimation tools is needed to
further validate the developed model so that generalisation of results can be done
with much higher accuracy. The sample may also include SMEs from more sectors
such as manufacturing and construction sectors.
The developed model can be used to test the impact of smart phone’s use in SMEs
and thus provide extra insights on how the SME sector interacts with ICTs to gain
competitiveness. The author noted that, smart phones can provide quick access to
the internet, email and other useful applications using mobile phone infrastructures.
Available data show that, by 2014 the teledensity (penetration) had reached 68% of
the population and that most of that growth is attributed to mobile operators and thus
they offer a huge potential for SMEs growth.
There is a need to research and clearly explain the emergence of the e-commerce
amongst SMEs in Tanzania as there seem to be some contradictions. For instance,
research findings show the inadequacy of support infrastructure and lack of
regulatory framework for supporting e-business. Yet it seems that some SMEs have
managed to defy the odds and have taken advantage of what is available effectively.
Finally, further research is needed on how to help effective utilisation of ICT by
SMEs. Research findings show that about 53% of SMEs that had adopted ICTS
(Computer and Internet) were not able to achieve any remarkable success. This
333
needs to be addressed since the 53% represents a big potential for the development of
the SME sector.

Final Cam PHD Error Free Thesis

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Final Cam PHD Error Free Thesis

Uploaded by

Copyright:

Available Formats

i

the degree of Doctor of Philosophy (Engineering Management) of the University of

University for a similar or any other degree award.

on intellectual property. It may not be reproduced by any means, in full or in part,

scholarly review or discourse with an acknowledgement, without the written

the University of Dar es Salaam.

important goal in my life.

Second, I am especially indebted to Professor E. Mjema and Prof. B. A. T Kundi of

and appear in a coherent manner.

My appreciation also goes to University of Dar es Salaam Management for the

sponsorship, I would not have been able to pursue my studies.

access to an E-Library with vast collection of reference books. I am also indebted to

This work is dedicated to my father Amos W. Msuya, my mother Naojwa Herini, my

AGFI Adjusted Goodness-Of-Fit Index

AMOS Analysis of Moment Structures

CFI Comparative Fit Index

DBERR Department of Business, Enterprise and Regulatory Reform

DOI Diffusion of Innovation

DTPB Decomposed Theory of Planned Behaviour

EDI Electronic Data Interchange

GDP Gross Domestic Product

GFI Goodness of Fit Index

ICT Information and Communication Technology

IDT Innovations Diffusion Theory

ISPs Internet Service Providers

LANs Local Area Networks

LDC Least Developed Countries

MAR Missing At Random

MCAR Missing Completely At Random

NFI Normed Fit Index

NTP National Telecommunication Policy

OECD Organisation for Economic co-operation and Development

RMSEA Root Mean Square Error of Approximation

ROI Return on Investment

SCT Social Cognitive Theory

SEM Structural Equation Modelling

SMC Squared Multiple Correlations

SME Small and Medium Scale Enterprise

TAM Technology Acceptance Model

TAM2 Technology Acceptance Model 2

TCP/IP Transmission Control Protocol/Internet Protocol

TCRA Tanzania Communications Regulatory Authority

TLI Tucker-Lewis coefficient Index

TPB Theory of Planned Behaviour

TPC Tanzania Posts Corporation

TPTC Tanzania Posts and Telecommunications Corporation

TRA Tanzania Revenue Authority

TRA Theory of Reasoned Action

TTCL Tanzania Telecommunication Company Limited

UDSM University of Dar es Salaam

ULS Unweighted Least Squares

URT United Republic of Tanzania

UTAUT Unified Theory of Acceptance and Use of Technology

WWW World Wide Web

adoption model to aid Tanzania SMEs in the adoption process.

sample of 341 respondents, a parsimonious explanatory Structural Equation Model

Research resulted as follows: (i) Perceived benefits, customer pressures and

of the SMEs were competitive due to ICT adoption and use.

countries with similar social and economic conditions to Tanzania.

LIST OF PAPERS PUBLISHED IN ASSOCIATION WITH THIS THESIS

Use in Tanzania SME”, Proceedings of the 4th International Conference on

Mechanical and Industrial Engineering, September, 2016, AICC, Arusha

SMEs”, Accepted by Tanzania Journal of Engineering and Technology, May